PUBLIC VERSION

UNITED STATES INTERNATIONAL TRADE COMMISSION WASHINGTON, D.C.

BEFORE THE HONORABLE THEODORE R. ESSEX ADMINISTRATIVE LAW JUDGE

In the Matter of: Investigation No. 337-TA-753 CERTAIN SEMICONDUCTOR CHIPS AND PRODUCTS CONTAINING SAME

COMPLAINANT RAMBUS INC.’S PETITION FOR REVIEW

PUBLIC VERSION

TABLE OF CONTENTS

I. II. III.

INTRODUCTION ..................................................................................................................... 1 THE ALJ’S INVALIDITY DETERMINATIONS SHOULD BE REVIEWED ...................... 3 THE DALLY PATENTS ARE NOT INVALID ...................................................................... 4 A. The Widmer Article Does Not Invalidate the Asserted Dally Claims .......................... 4 1. 2. 3. 4. The Dally Patents and the Asserted Dally Claims............................................ 4 The Widmer Article’s Disclosure ..................................................................... 5 The ALJ Erred in Finding the Widmer Article Enabling ................................. 5 The ALJ Erred in Finding that the Widmer Article Discloses the Claimed On-Chip Circuitry ............................................................................ 10 a. The Widmer Article’s Disclosure of a Chip Does Not Mean that Circuitry for Asserting DHPEAK and DLPEAK Is OnChip .................................................................................................... 10 Not Requiring External Components Does Not Mean that Circuitry for Asserting DHPEAK and DLPEAK Is On-Chip............ 12 Respondents’ Expert’s Testimony Cannot Replace the Widmer Article’s Disclosure ............................................................. 13

b. c. 5. B.

The ALJ Erred in Finding that Additional Features from Ewen Would Have Been Combined with the Widmer Article............................................. 14

The SL500 Art Does Not Anticipate the Asserted Claims of the Dally Patents ......... 17 1. 2. 3. 4. 5. The ALJ Erred By Finding It Was Irrelevant To Determine What Products Were Prior Art ................................................................................. 17 The Evidence Shows that the SL500 Fibre Channel Transceiver With No Emphasis Was the Focus of LSI’s Relationship with Seagate ................. 20 Respondents Did Not Show What SL500 Product Was Reduced To Practice, Used, or Offered for Sale ................................................................. 21 Respondents Did Not Meet Their Burden Regarding an Offer For Sale to Cabletron .................................................................................................... 25 The SL500 Art Does Not Meet the Speed Limitations of the Dally Claims ............................................................................................................. 26

i

PUBLIC VERSION

C. IV.

The ALJ Erred in Construing “Output Frequency” to Mean “Output Data Rate” ........................................................................................................................... 26

THE BARTH I PATENTS ARE NOT INVALID .................................................................. 28 A. B. Background of the Technology of the Asserted Barth I Patents ................................. 29 The Prior Art Does Not Anticipate the Barth Claims ................................................. 30 1. The NextBus Specification (RX-4265C) Does Not Anticipate Any Claims of the ’353 and ’109 Patents............................................................... 30 a. b. The ALJ Erred in Finding that the NextBus Specification Is Prior Art ............................................................................................. 30 The ALJ’s Construction of “Memory Device” is Wrong .................. 32 i. The Intrinsic Evidence Supports Rambus’s Construction .......................................................................... 33 (a) (b) (c) ii. The Claims Support Rambus’s Construction ............ 33 The Specification Supports Rambus’s Construction .............................................................. 34 The ALJ’s Construction, Inconsistent With the Intrinsic Evidence, Must Be Rejected ....................... 35

The Extrinsic Evidence Supports Rambus’s Construction .......................................................................... 36 (a) (b) The Prior Art Supports Rambus’s Construction .............................................................. 37 Testimony from Parties Adverse to Rambus Supports Rambus’s Construction .............................. 38

c. d. 2. 3. 4.

NextBus Does Not Anticipate the ’353 or ’109 Claims..................... 39 NextBus Does Not Anticipate Claims 12 and 13 ............................... 39

Harriman (RX-4277) Does Not Anticipate the Asserted Claims of the ’353 and ’109 Patents ..................................................................................... 40 Hayes (RX-4268) Does Not Anticipate Claim 11 of the ’353 Patent ............ 40 Dan and Yano Do Not Anticipate Claim 11 of the ’353 Patent ..................... 42

ii

PUBLIC VERSION

a.

The Barth I Patents Are Directed to Synchronous Memory Systems Whereas Dan and Yano Describe Asynchronous Memory Systems ............................................................................... 42 Dan Does Not Anticipate ’353 Patent Claim 11 ................................ 43 Dan Does Not Disclose the Claimed “Write Command” .................. 44 Yano Does Not Anticipate ’353 Patent Claim 11 .............................. 46 Yano Does Not Disclose the Claimed “Write Command” ................ 47

b. c. d. e. C.

The ALJ Committed Errors of Law, Fact, and Policy in His Obviousness Analysis of the Barth I Claims .................................................................................... 47 1. 2. None of the References Disclose or Suggest the Claimed Strobe Signal/Signal Features .................................................................................... 49 Combinations 1-4 Do Not Render the Asserted Barth I Claims Obvious .......................................................................................................... 49 a. b. SCI Is Not Prior Art ........................................................................... 50 The ALJ Erred in Finding that It Would Have Been Obvious to Combine Synchronous and Asynchronous Memory Systems .............................................................................................. 50 The ALJ Erred in Finding that It Would Have Been Obvious to Incorporate Double Data Rate Features into the Systems of Yano and Dan .................................................................................... 51 The ALJ Erred in Finding it Obvious to Incorporate Precharge, Bank Selection Information, and Sense Operation Features into Dan and Yano ............................................................... 52

c.

d.

3. 4. 5. V.

Combinations 5-6 Do Not Render the Barth I Claims Obvious ..................... 53 Combinations 7-12 Do Not Render the Barth I Claims Obvious ................... 54 Secondary Considerations Support a Finding of Non-Obviousness .............. 55

THE BARTH I PATENTS ARE NOT UNENFORCEABLE FOR UNCLEAN HANDS ................................................................................................................................... 56 A. The ALJ Erred in Conflating the Determination of Whether Rambus Anticipated Litigation of the Barth I Patents with Rambus’s Anticipated Enforcement of the Farmwald Portfolio ..................................................................... 57 The ALJ Erred in Finding that Litigation on the Barth I Patents Was Reasonably Foreseeable at Least by July 1998 ........................................................... 58 iii

B.

PUBLIC VERSION

1. 2.

The ALJ’s Broad Interpretation of Anticipated Litigation Contradicts Established Authority ..................................................................................... 59 No Evidence Supports the ALJ’s Finding that Rambus Adopted a “Long Term Strategic Litigation Plan” Based on the Barth I Patents ............ 60 a. b. That Rambus Sought to Create a “Patent Minefield” Does Not Evidence an Intent to Litigate the Barth I Patents ............................. 60 That Rambus’s Document Retention Policy Was Not Patent Specific Does Not Show an Intent to Litigate Barth I Patents ........... 61

3.

The Totality of the Circumstances Regarding the Barth I Portfolio Differs Substantially from Those of the Farmwald Portfolio ......................... 62 a. The ALJ Clearly Erred in Dismissing the Importance of the Fact the Asserted Barth I Patents Were Years from Issuing When Rambus Adopted Its Document Retention Policy ................... 62 The ALJ Clearly Erred in Equating Considered Enforcement of the Farmwald Patents with Anticipated Litigation of the Barth I Patents .................................................................................... 63 i. ii. c. The ALJ Ignored Unrebutted Evidence Demonstrating a Distinction Between Rambus Patent Families .................... 63 The ALJ Clearly Erred in Attributing Activities Involving SPP1 to the Barth I Patents ................................... 64

b.

The ALJ Clearly Erred in Confusing Notice of Potential Infringement of the Farmwald Patents with Potential Infringement of the Barth I Patents .................................................... 66 The ALJ Clearly Erred in Ignoring Evidence of Rambus’s Efforts to License and Resolve Disputes Without Litigation ............ 68

d. C.

The ALJ Erred in Finding Bad Faith Destruction of Barth I Documents ................... 70 1. The ALJ Legally Erred in Concluding that Supposed Bad Faith in the Farmwald Litigations Justified Finding Bad Faith Here ................................ 71 a. b. 2. The ALJ Clearly Erred in Minimizing Outside Counsel’s Role in Implementing Rambus’s Document Retention Policy ................... 71 Rambus’s Farmwald Prosecution Activities Have No Bearing on Bad Faith ....................................................................................... 74

The ALJ Clearly Erred in Assessing the Implementation of Rambus’s Document Retention Policy ............................................................................ 75

iv

PUBLIC VERSION

3.

The ALJ Erroneously Dismissed Relevant Evidence as Irrelevant and Mischaracterized Relevant Arguments as “Misrepresentations” ................... 76 a. b. Arguments and Evidence Presented Regarding the Scope of Destroyed Documents Were Not Misleading .................................... 76 Rambus’s Witnesses Provided Relevant Testimony Demonstrating Lack of Prejudice ...................................................... 77

4. 5. D.

Demonstrably False Determinations of “Fact” Infected the ALJ’s Credibility Determinations ............................................................................. 78 The ALJ’s Commentary on Dr. Przybylski Has No Bearing on Bad Faith and Should Be Vacated ......................................................................... 79

Respondents Were Not Prejudiced by Any Document Destruction ........................... 81 1. The ALJ Made Clearly Erroneous Findings and Misconstrued the Record in Finding Prejudice Based on Alleged Destruction of Mr. Barth’s Emails and JEDEC Materials ............................................................ 81 a. Clear Legal and Factual Error Infected the ALJ’s Prejudice Determination .................................................................................... 81 i. ii. b. 2. Written Description ............................................................... 82 Anticipation and Obviousness ............................................... 82

The ALJ Grounded His Finding of Prejudicial Destruction on Documents that Demonstrably Still Exist .......................................... 84

No Prejudice Resulted from the Destruction of Other Documents ................ 87 a. b. Conceptual Tree ................................................................................. 87 Vincent Files ...................................................................................... 88

3. 4. E. VI.

Respondents’ Other Defenses Were Not Prejudiced ...................................... 90 The ALJ Legally Erred by Relieving Respondents of Their Burden and Placing Too High a Burden on Rambus .................................................. 91

The Harsh Sanction of Dismissal Is Unwarranted and Contrary to Law .................... 92

THERE IS NO PATENT EXHAUSTION .............................................................................. 95 A. The ALJ Erred in Finding that Authorized Sales from Samsung to Take Place in the U.S. ................................................................................................. 95 1. The SamsungSales Take Place in v ....................................... 95

PUBLIC VERSION

2. B. VII.

The Samsung/Elpida Sales Are Not Authorized Sales ................................... 97

The ALJ Erred in Finding that the Samsung/Elpida Memory Devices Substantially Embody the Barth Patents ..................................................................... 98

CONCLUSION ....................................................................................................................... 99

vi

PUBLIC VERSION

TABLE OF AUTHORITIES Page(s) FEDERAL CASES Adv. Display Sys., Inc. v. Kent State Univ., 212 F.3d 1272 (Fed. Cir. 2000) ..................................................................................................... 32 Anderson v. Cryovac, Inc., 862 F.2d 910, 925 (1st Cir. 1988) .................................................................................................. 91 Ariad Pharms., Inc. v. Eli Lilly & Co., 598 F.3d 1336 (Fed. Cir. 2010) ..................................................................................................... 82 Arthur A. Collins, Inc. v. N. Telecom Ltd., 216 F.3d 1042 (Fed. Cir. 2000) ..................................................................................................... 37 Arthur Andersen LLP v. United States, 544 U.S. 696 (2005)..................................................................................................... 58, 59, 62, 71 Deere & Co. v. Int’l Trade Comm’n, 605 F.3d 1350 (Fed. Cir. 2010) ............................................................................................... 70, 80 Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579 (1993)....................................................................................................................... 23 Doe v. Menefee, 391 F.3d 147 (2d Cir. 2004) .......................................................................................................... 78 Fujifilm Corp. v. Benun, 605 F.3d 1366 (Fed. Cir. 2010) ..................................................................................................... 95 Genentech, Inc. v. Int’l Trade Comm’n, 122 F.3d 1409 (Fed. Cir. 1997) ..................................................................................................... 94 Graham v. John Deere Co., 383 U.S. 1 (1966)........................................................................................................................... 47 Gumbs v. Int’l Harvester, Inc., 718 F.2d 88 (3d Cir. 1983) ............................................................................................................ 70 HTC Corp. v. IPCom GmbH & Co., 667 F.3d 1270 (Fed. Cir. 2012) ............................................................................................... 33, 34 Hynix Semiconductor, Inc. v. Rambus Inc., 591 F. Supp. 2d 1038 (N.D. Cal. 2006) ................................................................................... 86, 87 Hynix Semiconductor Inc. v. Rambus Inc., 645 F.3d 1336 (Fed. Cir. 2011) ................................................................................... 35, 73, 74, 90

vii

PUBLIC VERSION

In re Lister, 583 F.3d 1307 (Fed. Cir. 2009) ......................................................................................... 30, 31, 50 In re NTP, Inc., 654 F.3d 1279 (Fed. Cir. 2011) ..................................................................................................... 48 In re Rambus, 2004 WL 390647 (FTC Feb. 23, 2004) ................................................................................... 74, 86 In re Swanson, 540 F.3d 1368 (Fed. Cir. 2008) ..................................................................................................... 29 In re Turlay, 304 F.2d 893 (C.C.P.A. 1962) ....................................................................................................... 25 Jiminez v. Mary Washington College, 57 F.3d. 369 (4th Cir. 1995) .......................................................................................................... 78 Kay v. Fed. Commc’ns Comm’n, 396 F.3d 1184, 1189 (D.C. Cir. 2005) ........................................................................................... 70 Kingsdown Med. Consultants, Ltd v. Hollister, Inc., 863 F.2d 867 (Fed. Cir. 1988) ....................................................................................................... 74 Kronisch v. United States, 150 F.3d 112 (Fed. Cir. 1998) ....................................................................................................... 57 KSR Int’l v. Teleflex, 550 U.S. 398 (2007)................................................................................................................. 47, 48 Kyocera Wireless Corp. v. ITC, 545 F.3d 1340 (Fed. Cir. 2008) ..................................................................................................... 32 Leon v. IDX Sys. Corp., 464 F.3d 951 (9th Cir. 2006) ......................................................................................................... 94 Litecubes, LLC v. N. Light Prods., Inc., 523 F.3d 1353 (Fed. Cir. 2008) ..................................................................................................... 96 McCook Metals L.L.C. v. Alcoa Inc., 192 F.R.D. 242 (N.D. Ill. 2000)............................................................................................... 84, 89 Micron Tech., Inc. v. Rambus, Inc., 255 F.R.D. 135 (D. Del. 2009) ............................................................................................... passim Micron Tech., Inc. v. Rambus Inc., 645 F.3d 1311 (Fed. Cir. 2011) .............................................................................................. passim Microsoft Corp. v. i4i Ltd. P’ship, 131 S. Ct. 2238 (2011) ..................................................................................................................... 3 viii

PUBLIC VERSION

Motorola, Inc. v. Interdigital Tech. Corp., 121 F.3d 1461 (Fed. Cir. 1997) ..................................................................................................... 14 Mycogen Plant Sci., Inc. v. Monsanto Co., 243 F.3d 1316 (Fed. Cir. 2001) ......................................................................................... 17, 18, 20 Netscape Commc’ns. Corp. v. Konrad, 295 F.3d 1315 (Fed. Cir. 2002) ......................................................................................... 17, 18, 20 Network Commerce, Inc. v. Microsoft Corp., 422 F.3d 1353 (Fed. Cir. 2005) ............................................................................................... 35, 36 Orion IP, LLC. v. Hyundai Motor Am., 605 F.3d 967 (Fed. Cir. 2010) ................................................................................................. 30, 32 Pfaff v. Wells Elecs., Inc., 525 U.S. 55 (1998)............................................................................................................. 17, 18, 20 Pfizer, Inc. v. Apotex, Inc., 480 F.3d 1348 (Fed. Cir. 2007) ....................................................................................................... 3 Phillips v. AWH Corp., 415 F.3d 1303 (Fed. Cir. 2005) (en banc) ..................................................................................... 36 Quanta Computer, Inc. v. LG Elecs., Inc., 553 U.S. 617 (2008)................................................................................................................. 97, 98 Rambus Inc. v. Hynix Semiconductor, Inc., 2008 WL 5771128 (N.D. Cal. Nov. 21, 2008) .............................................................................. 37 Rambus Inc. v. Infineon Techs. Ag, 318 F.3d 1081 (Fed. Cir. 2003) ............................................................................................... 74, 90 Sanofi-Synthelabo v. Apotex, Inc., 550 F.3d 1075 (Fed. Cir. 2008) ....................................................................................................... 6 Schmid v. Milwaukee Elec. Tool Corp, 13 F.3d 76 (3d Cir. 1994) .............................................................................................................. 92 Sud-Chemie, Inc. v. Multisorb Techs. Inc., 554 F.3d 1001 (Fed. Cir. 2009) ..................................................................................................... 49 Symbol Techs., Inc. v. Lemelson Med., Educ. & Research Found., 422 F.3d 1378 (Fed. Cir. 2005) ..................................................................................................... 90 Therasense, Inc. v. Becton, Dickinson & Co., 649 F.3d 1276 (Fed. Cir. 2011) (en banc) ..................................................................................... 90 TransCore, L.P. v. Elec. Transaction Consultants Corp., 563 F.3d 1271 (Fed. Cir. 2009) ..................................................................................................... 95 ix

PUBLIC VERSION

United States v. Univis Lens Co., 316 U.S. 241 (1942)........................................................................................................... 97, 98, 99 FEDERAL STATUTES 35 U.S.C. § 102 ................................................................................................................................ 5, 31

x

PUBLIC VERSION

TABLE OF ABBREVIATIONS CPB RPB SPB CIB RIB SIB CRB RRB SRB HTr. CXRXSXJXID Comm’n Op. Complainant’s Corrected Prehearing Brief Respondents’ Prehearing Brief Staff’s Prehearing Brief Complainant’s Initial Posthearing Brief Respondents’ Initial Posthearing Brief Staff’s Initial Posthearing Brief Complainant’s Posthearing Reply Brief Respondents’ Posthearing Reply Brief Staff’s Posthearing Reply Brief Hearing Transcript Complainant Exhibit Respondent Exhibit Staff Exhibit Joint Exhibit Initial Determination Commission Opinion

xi

PUBLIC VERSION

I.

INTRODUCTION Complainant Rambus Inc. petitions for review of the March 2, 2012 Initial Determination

which found no violation of Section 337 as to all five of the asserted patents, i.e., U.S. Patent Nos. 6,470,405 (the ’405 patent), 6,591,353 (the ’353 patent), and 7,287,109 (the ’109 patent) (collectively the Barth I patents) and U.S. Patent Nos. 7,602,857 (the ’857 patent) and 7,715,494 (the ’494 patent) (collectively the Dally patents). The ID found all of the asserted claims invalid as anticipated by or obvious in view of alleged prior art. The ID also found the Barth I patents unenforceable due to unclean hands arising out of Rambus’s alleged spoliation of evidence. The ID further found that Rambus had exhausted its rights in the Barth I patents directed to memory controllers as to a subset of the accused products from downstream Respondents of licensed memory devices in the United States. Rambus petitions for review of the ALJ’s holdings of invalidity (anticipation and obviousness) of all 61 asserted Barth I and Dally claims. For Dally, the ALJ legally erred in: (1) shifting the burden of providing rebuttal evidence to Rambus; (2) finding that the Widmer Article anticipates or renders obvious the asserted Dally claims despite that article’s lack of enablement and its failure to disclose the claimed on-chip equalization mechanism; (3) finding that the Widmer Article in combination with Ewen renders obvious certain asserted Dally claims when the Widmer Article states that it already incorporated the relevant teachings of Ewen and expressly teaches away from combining the specific features the ALJ found it obvious to combine; (4) finding that the SL500 Art anticipates the asserted Dally claims based on prior invention, public use, and on-sale activity without first determining what product was allegedly prior art and whether that product contained all of the claimed features; and (5) construing “output frequency” to mean “output data rate” as opposed to “half the value of the output data rate” when the technology mandates that latter construction and the Dally patent specification supports it. Many of these legal errors were based on or compounded by clearly erroneous factual findings. 1 based on alleged purchase

PUBLIC VERSION

With respect to Barth I, the ALJ reached erroneous legal conclusions in: (1) shifting the burden of providing rebuttal evidence to Rambus; (2) finding that the NextBus specification is prior art when there is no evidence of the document’s public availability before the Barth I 1995 filing date and the indicia of such availability on the document are from 2005, a decade after Barth I was filed; (3) finding sub silentio that SCI is prior art without addressing Rambus’s contrary arguments; (4) construing “memory device” and “write command” based on extrinsic dictionary definitions while not addressing the intrinsic and extrinsic evidence refuting those constructions and stating that Rambus did not present that evidence; and (5) failing to perform the mandated Graham obviousness analysis in summarily concluding that 12 different combinations of prior art render Barth I claims obvious. These legal errors were compounded by clearly erroneous findings of fact in concluding that the Barth I claims are invalid based on prior art. Rambus petitions for review of the ALJ’s holding of unenforceability of the Barth I patents for unclean hands based on spoliation. The ALJ erred as a matter of law and fact in finding that Rambus anticipated litigation at least by July 1998, and therefore had a duty to preserve documents, based on prosecution, and licensing/litigation activities unrelated to the Barth I patents. In so doing, the ID defies Federal Circuit precedent in holding that Rambus reasonably foresaw litigation on its asserted Barth I patent applications years before: (1) those applications were amended to claim features incorporated into Respondents’ accused products; (2) those applications issued as patents, (3) the accused products were marketed, and (4) Rambus considered litigating the Barth I patents. Fundamental errors of law and fact made in assessing anticipation of litigation infected the ALJ’s bad faith determination, which also depended on adverse credibility determinations grounded on clearly erroneous findings of fact. Moreover, the ALJ erred as a matter of law in effectively establishing an irrebutable presumption of prejudice in the face of considerable evidence demonstrating that Respondents’ defenses were not prejudiced by any document destruction. At the very least, the degree of fault and degree of prejudice associated with the document destruction does 2

PUBLIC VERSION

not warrant the harsh sanction of dismissal. Rambus also petitions for review of the ALJ’s holding on exhaustion based on legal and factual errors. The ALJ did not identify any authorized sale of a memory controller, but instead erroneously relied on the sale of DRAMs, i.e. memory devices, an error of law. Even if a sale of a memory device could result in exhaustion, there is insufficient evidence to establish that an authorized sale occurred in the United States. Finally, the ALJ made both legal and factual errors in concluding that the Samsung/Elpida memory devices substantially embody the Barth I Patents. The ALJ’s conclusions are based on the wrong legal standard, as well as a fundamental misunderstanding of the applicable facts. II. THE ALJ’S INVALIDITY DETERMINATIONS SHOULD BE REVIEWED In holding all 61 claims of the Dally and Barth I claims invalid, the ALJ applied an incorrect legal standard. While noting that Respondents bear the burden of persuasion, the ALJ incorrectly shifted the burden to Rambus of providing evidence that the patents are not invalid. ID at 85-86. In support, the ALJ quoted a statement from Pfizer, Inc. v. Apotex, Inc., 480 F.3d 1348, 1360 (Fed. Cir. 2007), that “once a challenger has presented a prima facie case of invalidity, the patentee has the burden of going forward with rebuttal evidence.” Pfizer, however, in an analysis not addressed by the ALJ, clarifies that this does not mean a patentee is required to present rebuttal evidence: But, all that means is that even though a patentee never must submit evidence to support a conclusion by a judge or jury that a patent remains valid, once a challenger introduces evidence that might lead to a conclusion of invalidity—what we call a prima facie case—the patentee “would be well advised to introduce evidence sufficient to rebut that of the challenger.” Id. (emphasis in original). Consistent with that premise, the Supreme Court recently confirmed that a challenger always bears the burden of proving invalidity by clear and convincing evidence. Microsoft Corp. v. i4i Ltd. P’ship, 131 S. Ct. 2238 (2011). The ALJ’s erroneous holding that Rambus was required to provide rebuttal evidence infected his entire validity analysis. Moreover, 3

PUBLIC VERSION

while improperly placing a burden of presenting rebuttal evidence on Rambus, the ALJ compounded that error by failing to address the evidence Rambus presented establishing that the prior art does not invalidate the Dally and Barth claims. See Section III (Dally) and Section IV (Barth). The

Commission should review the ALJ’s invalidity holdings. III. THE DALLY PATENTS ARE NOT INVALID A. The Widmer Article Does Not Invalidate the Asserted Dally Claims

The Commission should review and reverse the ALJ’s finding that the Dally claims are invalid based on the Widmer Article (RX-4109) because Respondents have not established that it is either enabling or invalidating. The ALJ relies on erroneous facts and conclusions of law in finding that it contains an enabling disclosure that teaches the claimed on-chip equalization mechanism, and it would have been combined with aspects of Ewen (RX-4125) to achieve the claimed “output frequency of at least 1 GHz” or to have a differential output. 1. The Dally Patents and the Asserted Dally Claims

Rambus asserts two Dally patents: U.S. Patent Nos. 7,602,857 (JX-120, the ’857 patent) and 7,715,494 (JX-121, the ’494 patent). The Dally patents share a common specification and are directed to, among other things, improving signal integrity and data transmission for high speed, chip-to-chip communication. See JX-120.0013 at 1:54-60. For instance, they teach a particular form of transmitter equalization that is used on a chip to mitigate the negative effects a transmission channel can have on the transmitted signal. See JX-120.0014 at 4:54-61; CX-9542C.0007-16. By disclosing specific on-chip techniques to mitigate those negative effects, the receiver is more likely to properly interpret the signal leading to more reliable and faster chip-to-chip communication. JX120.0014 at 3:1-8; CX-9542C.0014-16. Rambus asserts claims 1, 2, 4-6, 9-13, 24-28, 31-36, 39-44, 47, and 49-53 of the ’857 patent and claims 1-3, 6, 8, 25, 26, 30, 39, 40, and 42 of the ’494 patent. Claim 1 of the ’857 patent is exemplary of the independent claims, each of which recites, among other things, a semiconductor 4

PUBLIC VERSION

chip or integrated circuit having circuitry for performing a particular type of equalization. ’857 patent claim 2 is exemplary of the claims that require high data-rate transfer capability, reciting “[t]he component as claimed in claim 1 wherein the transmitter circuit is operable to send the output signal with an output frequency of at least 1 GHz and a bandwidth greater than 100 MHz.” 2. The Widmer Article’s Disclosure

The Widmer Article (the first three pages of RX-4109)1 is a paper that was part of the International Solid State Circuits Conference (“ISSCC”) in 1996. It briefly discloses a

semiconductor transceiver with four transmitter circuits, each of which outputs data through a singleended serial output at 500 MBaud (or 500 megabits per second). See, e.g., RX-4109.0002 at Fig. 1. The transmitter circuits each include a pre-distorting driver that “boosts high-frequency content” by allegedly peaking the low and high levels of a signal by 50% in the intervals immediately following transitions. RX-4109.0001. The left side of Figure 4 shows the pre-distorting driver, which has four control inputs. According to the Widmer Article, two of the inputs (DHBASE and DLBASE) set the base driver levels, while the other two inputs (DHPEAK and DLPEAK) allegedly turn on the extra drive required for peaking. RX.4109.0001; CX-10764C.0022. It is this peaking and the related circuitry that Respondents contend is the equalization mechanism called for in the Dally claims. 3. The ALJ Erred in Finding the Widmer Article Enabling

The ALJ found the Widmer Article to be enabling (ID at 131-37), but because the Widmer Article does not disclose a mechanism for generating the critical DHPEAK and DLPEAK signals, Respondents did not present clear and convincing evidence that the Widmer Article enables the The ALJ correctly found that the slides appearing at the end of RX-4109 are not prior art because they are not part of the Widmer Article and are not a printed publication. ID at 130 (in a typographical error, the word “not” was omitted from “Respondents have not presented clear and convincing evidence that the slides are a printed publication within the meaning of § 102 in part because Mr. Widmer, who was to verify that the slides were presented and publicly disseminated, failed to provide such testimony at the evidentiary hearing”). 5
1

PUBLIC VERSION

claimed on-chip equalization circuitry. Sanofi-Synthelabo v. Apotex, Inc., 550 F.3d 1075, 1082 (Fed. Cir. 2008) (“An anticipating reference must be enabling; that is, the description must be such that a person of ordinary skill in the field of the invention can practice the subject matter based on the reference, without undue experimentation.”). Instead, the evidence demonstrates that the Widmer Article is nothing more than another “hard-to-decode” ISSCC paper (see RX-2772C.0103, discussing another ISSCC paper), leading Dr. Singer to explain that “one of ordinary skill in the art would have found it impossible to make and use the claimed invention without undue experimentation.” CX10764C.0023. The ALJ’s finding to the contrary was based on a number of incorrect grounds. One is that the Widmer Article is allegedly enabled because a different reference, the Widmer Patent (RX-2870), “would have been an obvious addition to the disclosure of the Widmer Article.” ID at 133. According to the ALJ, “a person of ordinary skill in the relevant art would have found the Widmer Article to contain an enabling disclosure, with the Widmer Patent being exemplary of how equalization can be done using the circuits disclosed and described in the Widmer Article.” Id. The Widmer Patent, however, is a 1970s-era reference that does not enable one of ordinary skill in the art to perform the claimed on-chip equalization without undue experimentation. Instead, it discloses a system implemented mostly with discrete circuit components (as opposed to integrated circuit components on a chip), such as those connected by wires on a circuit board or similar structure. CX-10764C.0075; RX-2870.0003 at Fig. 5. In addition to giving no indication that all of its transmitter components are located on a chip, (CX-10764C.0075), the Widmer Patent discloses components, such as inductors, that are difficult to place on a chip even today, much less when the Dally patents were filed in 1997. Id.; RX-2870.0003 (showing inductors as part of transformer 33 on the output of the transmitter); HTr. 2416:20-21 (“Even today it is very difficult to put inductors or transformers on a chip.”). The equalization mechanism of the Widmer Patent also differs substantially from the Widmer Article’s equalization mechanism. For example, it outputs data at 1.39 Mbps (RX-2870.0008 at 6

PUBLIC VERSION

6:17-20), which is over 350 times slower than the 500 Mbps rate contemplated by the Widmer Article.2 There is no indication that the Widmer Patent’s equalization mechanism would scale to the rates contemplated by the Widmer Article, much less do so on a chip. CX-10764C.0081. Another significant distinction is that the Widmer Patent applies its equalization to signals that are bifrequency or biphase-encoded. See, e.g., RX-2870.0001 at Abstract. These encoding schemes mean that two signal levels are output for each bit, and the Widmer Patent discloses that its emphasis operates on only one of those two signal levels (i.e., half of the bit period). Id.; CX-10764C.0074-75, 78-79. The Widmer Article, in contrast, is alleged to disclose a different emphasis mechanism—one that emphasizes a signal for an entire bit period. ID at 144-45. These different emphasis

mechanisms are not interchangeable, so the Widmer Patent cannot make up for the lack of enablement of the Widmer Article. CX-10764C.0027, 78. The ALJ’s finding of enablement through the Widmer Patent ultimately rests on the erroneous assumptions that one of ordinary skill in the art would have understood that its 1970s-era technology, operating orders of magnitude slower than the Widmer Article’s disclosed rate, implemented in discrete components as opposed to on a chip, and using an entirely different equalization mechanism than the Widmer Article, would somehow make up for the Widmer Article’s lack of enablement and enable the claimed on-chip equalization mechanism. The Widmer Patent, however, cannot save the Widmer Article. The ALJ also erroneously relies on Respondents’ expert’s flippant and conclusory testimony that a sophomore or junior in electrical engineering would have been able to make and use the Widmer Article’s emphasis circuit. ID at 135. To the contrary, it is, in fact, impossible to

understand the Widmer Article as written because it discloses a nonfunctional design. HTr. 2577:172

It is also over 700 times slower than 1 Gbps, which is the rate required under the ALJ’s construction of “an output frequency of at least 1 GHz,” and it is over 1400 times slower than 2 Gbps, which is the rate required under Rambus’s construction. HTr. 1502:9-1503:13. 7

PUBLIC VERSION

2579:19. In particular, the Widmer Article states that “DHPEAK and DLPEAK turn on the extra drive required for peaking.” RX-4109.0001. This means that the circuitry for generating DHPEAK and DLPEAK is the very circuitry that provides the emphasis mechanism alleged to satisfy many of the features of the asserted Dally claims. Figure 4, however, shows that the transistors driven by DHPEAK and DLPEAK are providing the same voltage as the transistors driven by DHBASE and DLBASE. When two transistors connected across the same voltage have their outputs combined, the voltage is not increased. Instead, the same voltage is output on the line as when either transistor is enabled. HTr. 2577:17-2579:19. Thus, an output with DHBASE asserted will have the same voltage as when both DHBASE and DHPEAK are asserted. As Dr. Singer explained, “[s]imply asserting DH base or DH peak would give exactly the same output.” HTr. 2578:8-11. Since the same output results regardless of whether DHPEAK and DLPEAK are asserted, no peaking actually occurs in the circuit shown in the Widmer Article. Because the circuitry depicted in the Widmer Article cannot function the way the text says it should, “there must be more sophisticated circuitry involved” to accomplish the type of peaking mentioned in the Widmer Article. HTr. 2579:18-19. The Widmer Article does not disclose that circuitry, and the circuitry it does disclose cannot possibly perform the on-chip equalization of the asserted Dally claims. HTr. 2577:172579:19.

3

The ALJ correctly held that Odysseus was not proven to be prior art. ID at 138. 8

PUBLIC VERSION

10764C.0026.4 Because Respondents’ expert, whose qualifications place him above the level of ordinary skill, had to seek this confirmation, it is unreasonable to conclude that one of ordinary skill in the art would have understood how the pre-distorting driver worked, let alone been capable of making and using it. The ALJ’s finding that the Widmer Article contains an enabling disclosure of the claimed invention is clearly erroneous because the Widmer Patent cannot make up for the Widmer Article’s incomplete disclosure and the record provides strong affirmative evidence that one of ordinary skill in the art would not have been able to make and use the claimed invention based on the Widmer Article’s brief and nonfunctional disclosure. 4. The ALJ Erred in Finding that the Widmer Article Discloses the Claimed On-Chip Circuitry

The Widmer Article does not show that the circuitry for generating DHPEAK and DLPEAK is on-chip. The ALJ found to the contrary, primarily basing his finding on three facts: (1) the Widmer Article’s disclosure of a chip; (2) the Widmer Article’s statement that “[n]o external components are required” (RX-4109.0001); and (3) Respondents’ expert’s testimony that the Widmer Article’s equalization circuitry was implemented on a chip. ID at 133-136, 139. None of these facts, either alone or collectively, provide clear and convincing evidence that the Widmer Article’s equalization mechanism was on-chip, as required by all of the asserted Dally claims. Each fact is discussed below. a. The Widmer Article’s Disclosure of a Chip Does Not Mean that Circuitry for Asserting DHPEAK and DLPEAK Is On-Chip

Regarding the first fact relied on by the ALJ, it is undisputed that the Widmer Article discloses a chip. The question is whether the chip contains the claimed equalization logic and Dr. Singer discussed Respondents’ expert’s testimony in one of his witness statements, but the ALJ incorrectly excluded this portion of the witness statement because the testimony was not in his expert report. But Ground Rule 10.5.6 permits witness statements to include evidence adduced at deposition, so the Commission should reverse the ALJ’s exclusion of it. 10
4

PUBLIC VERSION

circuitry. The ALJ’s citations to the Widmer Article’s general chip image and other portions that mention chips do not answer this important question. Since “DHPEAK and DLPEAK turn on the extra drive required for peaking,” (RX-4109.0001), the circuitry that determines when to activate DHPEAK and DLPEAK is a key component of the equalization mechanism.5 See CX-10764C.002225. Without it, peaking will not occur as allegedly intended, if at all. The Widmer Article, however, does not disclose how to implement this circuitry, and therefore does not disclose or suggest that it is on-chip. Id. Instead, several facts lead to the opposite conclusion, indicating that the circuitry for asserting DHPEAK and DLPEAK is located off-chip. Id. For instance, the Widmer Article describes DHPEAK and DLPEAK as “inputs,” much the same way as data and other signals that are not generated within the chip are provided to the chip as inputs. RX-4109.0001 (“Control inputs DHBASE and DLBASE set the base driver levels, and DHPEAK and DLPEAK turn on the extra drive required for peaking,” emphasis added). Providing these signals as off-chip inputs (as opposed to generating them on-chip) was consistent with how chips described in ISSCC papers were often designed. As Dr. Singer explained, ISSCC publications disclose “cutting edge chips” where it is often not known how well portions of the chip will function. HTr. 2580:3-5. For that reason, “all the time, a chip that will appear at ISSCC will have numerous inputs” so the signals on the inputs can be changed to test the chip. Id. at 2580:2-11. The Widmer Article states that its chip was designed consistent with these principles. In discussing one of “[s]everal features [that] provide testability,” the Widmer Article says that “an onchip monitor with an interface to an external processor provides control capabilities and collects diagnostic data.” RX-4109.0001 (emphasis added). The Widmer Article also specifies that its chip The ALJ criticized the testimony of Rambus’s expert regarding which circuitry is part of the transmitter and which circuitry is not. See ID at 141-143. While Rambus disagrees with this criticism, it is beyond dispute that DHPEAK, DLPEAK, and their associated circuitry are part of the transmitter, as they are the exact features that allegedly “turn on the extra drive required for peaking” on the transmitter’s output. RX-4109.0001. Therefore, if they are not on-chip, as claimed, the Widmer Article does not invalidate the asserted Dally claims. 11
5

PUBLIC VERSION

has 361 pins that can be used for power, accepting inputs, and providing outputs, with 140 pins used for ground and power, eight used for the “4 serial, duplex links” (two pins per duplex), and 32 for receiving the four parallel words of eight-bit data shown in Figure 1. Id. That leaves 181 pins for use as inputs to “provide testability” and perform other functions, such as providing inputs DHPEAK and DLPEAK. See CX-10764C.0023. In light of the Widmer Article’s disclosure of an off-chip processor providing control inputs and a multitude of pins that can accept inputs, and because “DH peak and DH base are referred to as inputs, one of ordinary skill sitting at ISSCC would have no reason not to assume that they could be off-chip inputs.” HTr. at 2580:11-15 (Singer). The ALJ dismisses these facts, contending that “a person of ordinary skill in the relevant art would interpret the ‘input’ language as referring to the signals being input into the Predistorting Driver portion of the chip as explicitly shown in Figure 4.” ID at 134 (emphasis in original). This finding is incorrect for the reasons discussed above, as one of ordinary skill in the art would have expected a chip being presented at ISSCC to have number of off-chip inputs and there is no indication that DHPEAK and DLPEAK are generated on-chip. The ALJ also contends that the Widmer Article refers to off-chip signaling as “in” and “out” as opposed to “input” and “output.” Id. at 135-36. This finding is clearly erroneous because the Widmer Article states that an “input” to one of its circuits can be provided by an “external receiver clock” that is indisputably located off-chip. RX-4109.0001. Thus, contrary to the ALJ’s finding, the Widmer Article does use the term “input” to refer to signals provided off-chip. Finally, the ALJ contends that certain off-chip signals are

described as being optional and not related to DHPEAK and DLPEAK. Id. at 136. Even if this were true, it is beside the point. The Widmer Article’s description of those off-chip signals as being “inputs” establishes that one cannot assume that “inputs” DHPEAK and DLPEAK are on-chip as opposed to off-chip. b. Not Requiring External Components Does Not Mean that Circuitry for Asserting DHPEAK and DLPEAK Is On-Chip 12

PUBLIC VERSION

The second fact the ALJ relied on to support his finding that the equalization circuitry is onchip is the Widmer Article’s statement that “[n]o external components are required.” ID at 134-36. This statement does not mean that the chip generates all signals, including DHPEAK and DLPEAK, internally on-chip. Instead, the Widmer Article makes clear that a number of its signals are

generated by and provided from sources external to its chip. Indeed, the very purpose of the chip’s many inputs is to accept externally-generated, off-chip signals. One example of an off-chip input is data, which the ALJ acknowledges comes from an off-chip source. Id. at 135-36. Another example is an “external controller” used to load a parameter related to packet size. RX-4109.0001. Yet another example is the “reference input” for one of the Widmer Article’s PLL circuits, which may be provided by “an external [i.e., off-chip] receiver clock.” Id. The Widmer Article also discloses “a dummy off-chip driver (OCD) and receiver (OCR) to compensate for I/O circuit delay variations so output timing is controlled more precisely.” Id. (emphasis added). It further discloses “an external [i.e., off-chip] processor” that “provides control capabilities,” among other things. Id. Without these critical, externally-generated signals (like data), the chip would not function and would serve no purpose. Accordingly, one of ordinary skill would not have understood the Widmer Article’s statement that “[n]o external components are required” to mean that all signals, including DHPEAK and DLPEAK, are generated on-chip. CX-10764C.0024-25. c. Respondents’ Expert’s Testimony Cannot Replace the Widmer Article’s Disclosure

The final fact the ALJ relied on to support his finding that the Widmer Article’s equalization circuitry is on-chip is Respondents’ expert’s conclusory testimony. ID at 135, 143-44. Given the Widmer Article’s brief disclosure, the expert’s testimony is essentially based on the same facts that the ALJ relied on (i.e., the Widmer Article’s disclosure of a chip and its statement that no external components are required). See id. Because these facts do not establish that the equalization circuitry is on-chip for the reasons discussed above, Respondents’ expert’s testimony is entitled to no weight 13

PUBLIC VERSION

and cannot be used to supplant the Widmer Article’s actual disclosure. Motorola, Inc. v. Interdigital Tech. Corp., 121 F.3d 1461, 1473 (Fed. Cir. 1997) (“An expert's conclusory testimony, unsupported by the documentary evidence, cannot supplant the requirement of anticipatory disclosure in the prior art reference itself.”). In summary, the Widmer Article’s brief disclosure suggests that DHPEAK and DLPEAK are generated off-chip, and none of the facts cited by the ALJ establishes otherwise. 5. The ALJ Erred in Finding that Additional Features from Ewen Would Have Been Combined with the Widmer Article

The ALJ combined Ewen (RX-4125) with the Widmer Article to render obvious the claims reciting “an output frequency of at least 1 GHz” (’857 claims 2, 31, and 49; ’494 claim 3) and differential signaling (’494 claims 25, 26, 30, 39, 40, and 42). ID at 147-54. The ALJ’s findings are flawed because one of ordinary skill would not have combined those references as proposed and, even if one had, it would not have resulted in the claimed inventions. One of ordinary skill in the art would not have made the proposed combination because the Widmer Article (published in 1996) states that its technology already “adopts techniques used” in Ewen (published in 1995). RX-4109.0001. As Dr. Singer explained, “[t]he authors of the Widmer Article already claim to have incorporated the relevant features of [Ewen] into their design, so one of ordinary skill in the art would not have considered combining other features from that paper.” CX10764C.0034. Not only does the Widmer Article already adopt the relevant Ewen techniques, but the authors of Ewen are all authors of the Widmer Article. Compare the author list at RX-4125.0001 with the author list at RX-4109.0001. Thus, the authors of the Widmer Article were intimately familiar with the teachings of Ewen when they designed the technology of the Widmer Article, yet they only achieved an output data rate of 500 Mbps (half the rate of the claimed “at least 1 GHz” under the ALJ’s construction, and only 1/4 the rate under Rambus’s construction). This is not 14

PUBLIC VERSION

surprising, as serial links are sensitive designs where small changes can have significant consequences. See CX-10764C.0035, 130. Simply swapping technology from a system without emphasis, like Ewen’s, into a system attempting to provide a form of emphasis, like the Widmer Article’s, is not a straightforward task and one would not expect the combined system to operate at the highest rates disclosed in either reference. Id. Moreover, as Respondents’ expert Dr. Hassoun acknowledged, at least one of the authors of the Widmer Article was a person of exceptional skill in the art. HTr. 1500:18-23. Despite this exceptional skill and full knowledge of Ewen, the authors of the Widmer Article did not achieve the 1062 MBaud output that the ALJ contends one of ordinary skill in the art would have achieved by combining the references. Even if one of ordinary skill in the art had attempted to combine Ewen with the Widmer Article, the resulting combination would not have provided an output frequency of at least 1 GHz. Instead, Ewen discloses a system that has an on-chip signaling rate of 1062 Mbaud. RX-4125.0001. This on-chip signaling rate is not the same as the rate of the signal once it is sent off-chip. CX10764C.0033. Typically, off-chip signaling rates are half of the on-chip rates, so Ewen’s output data rate is 531 Mbps, which falls far short of the claimed rate under both the ALJ’s construction (1 Gbps) and Rambus’s construction (2 Gbps).6 Id. Even if one were to incorrectly assume that Ewen’s onchip rates were comparable to its off-chip rates, Ewen’s signaling rate is approximately half of the claimed rate under Rambus’s construction of “output frequency.” Id. The ALJ’s obviousness analysis regarding the claims that recite “an output frequency of at least 1 GHz” is two paragraphs long and does not address these significant problems with the proposed combination. ID at 153-54. In that analysis, the ALJ merely states that the Widmer Article refers to Ewen, provides a quote from Ewen, and states that it would have been obvious to modify In addition, under Rambus’s construction of “a rate of at least 400 megahertz”, which requires a data rate of at least 800 Mbps, Ewen’s 531 Mbps rate and the Widmer Article’s 500 Mbps rate are both too slow to meet the claimed rate. Accordingly, taken separately or when combined, Ewen and Widmer do not anticipate or render obvious ’494 patent claims 30 and 42. 15
6

PUBLIC VERSION

the Widmer Article to achieve an output data rate of 1.062 Gbps “in light of the need for increasingly fast data transmission techniques in certain implementations.” Id. at 153. Nowhere, however, does the ALJ explain or cite any evidence explaining how one of ordinary skill in the art could have accomplished such a combination or how the proposed combination would have been constructed. He also does not explain why one of ordinary skill in the art would have attempted to incorporate features from Ewen into the Widmer Article when the Widmer Article says that it had already incorporated the relevant aspects of Ewen. He also does not address the fact that those of exceptional skill achieved only half the data rate he contends one of ordinary skill would have achieved through the combination. The ALJ’s conclusory analysis cannot withstand scrutiny. Regarding the differential signaling claims, the ALJ’s analysis of the proposed Ewen/Widmer Article combination is more extensive but equally flawed. See ID at 148-50. There, the ALJ contends that one of ordinary skill in the art would have incorporated Ewen’s differential signaling into the Widmer Article’s single-ended system despite the fact that the Widmer Article teaches away from using differential signaling by stressing the need to reduce wire count to avoid wire congestion. RX-4109.0001; CX-10764C.0043. Converting to differential signaling would have increased wire congestion, contrary to the Widmer Article’s teachings. Id. As the ALJ recognized, “[g]enerally, a prior art reference that teaches away from the claimed invention does not create [a] prima facie case of obviousness,” ID at 210 (citing cases), so the Widmer Article/Ewen combination was not obvious. And even if one had been motivated to use differential signaling, it would have required a complete redesign of the single-ended pre-distorting driver disclosed in the Widmer Article. CX-

10764C.0043-44. Given the lack of detail as to how the Widmer Article’s driver is structured redesigning it to support differential signaling would not have been possible for one of ordinary skill in the art. Id. For these reasons, the ALJ erred in finding that the Widmer Article, either alone or in combination with Ewen, anticipates or renders obvious any asserted Dally patent claim. Rambus 16

PUBLIC VERSION

requests that the Commission review the ALJ’s invalidity findings and reverse them. B. The SL500 Art Does Not Anticipate the Asserted Claims of the Dally Patents

Respondents contend that the asserted Dally claims are invalid based on prior invention, public use, and offers for sale. To prevail, they had to show an actual reduction to practice occurred, an actual product was in public use,7 or an actual product was offered for sale. Mycogen Plant Sci., Inc. v. Monsanto Co., 243 F.3d 1316, 1332 (Fed. Cir. 2001); Pfaff v. Wells Elecs., Inc., 525 U.S. 55, 67 (1998); Netscape Commc’ns. Corp. v. Konrad, 295 F.3d 1315, 1321 (Fed. Cir. 2002). They did not. The record reflects that LSI was attempting to design various chips, some of which did not include the claimed emphasis features and therefore did not meet any of the asserted Dally claims. Thus, it was Respondents’ burden to show that a product that was actually reduced to practice, in public use, or offered for sale included the claimed emphasis features. The ALJ erred by finding that it was irrelevant to determine whether a product with the claimed emphasis features was actually reduced to practice, in use, or offered for sale. ID at 171-72. The ALJ further erred by not considering evidence, including testimony by Respondents’ own expert, which showed that the product that was at issue did not contain the emphasis features claimed by the asserted Dally patents. Due to these errors, the Commission should review the ALJ’s invalidity findings and reverse them. 1. The ALJ Erred By Finding It Was Irrelevant To Determine What Products Were Prior Art

While Respondents alleged that a chip called “SL500” was “conceived, reduced to practice, used, and offered for sale” to Seagate in 1995 to 1996, the ALJ correctly noted that LSI was attempting to design multiple versions of its SL500 chips in this time frame. Id. at 160; 171-72; 17680. The ALJ found that at least one of these designs, had its

Although the ALJ did not explicitly address Respondents’ public use defense, it has been addressed here for completeness. 17

7

PUBLIC VERSION

emphasis features disabled. Id. at 171-72. RX-4171.0001-2. The ALJ specifically noted that “Rambus correctly observes that some of the documents on which LSI relies refer to applications in which the equalization circuitry has been disabled.” ID at 171. Despite the fact that at least some SL500 designs did not even include the claimed emphasis features, the ALJ found that it was “not relevant” for him to determine whether a product that actually included those features was reduced to practice, in public use, or offered for sale. Id. at 17172. But only those particular products which were reduced to practice, in public use, or offered for sale would constitute prior art. Mycogen Plant Sci., 243 F.3d at 1332; Pfaff, 525 U.S. at 67; Netscape Commc’ns., 295 F.3d at 1321. Thus, in other words, the ALJ held that it was irrelevant for him to determine which products were even prior art. ID at 171-72. This was legal error. Mycogen Plant Sci., 243 F.3d at 1332; Pfaff, 525 U.S. at 67; Netscape Commc’ns., 295 F.3d at 1321 Rather than making this determination, the ALJ sidestepped the issue by arguing that even though the emphasis features were disabled in the they “could be

enabled.” ID at 172. But in making that determination, the ALJ did not refer to the documentation for the Id. That documentation explicitly states that RX-4171.0006-7. The ALJ also ignored testimony by Respondents’ own expert, who testified that HTr. 1618-19. Referring to the Respondents’ expert explicitly stated “there was no emphasis in the application.” Id. at 1618:18-22 (emphasis added). Thus, it was of critical importance for the ALJ to determine what SL500 design was reduced to practice, in use, or offered for sale to Seagate, as certain designs did not contain the claimed emphasis features. But the ALJ simply did not address either of these crucial pieces of evidence showing that emphasis was removed from the product. Had he done so, he would have found that determining what products actually constituted prior art was far from irrelevant. Rather than referring to documentation regarding the 18 the

PUBLIC VERSION

ALJ instead referred to documentation regarding the CWSL500 core.8 ID at 172. He noted that this documentation describes that could be used to enable the

equalization circuitry. Id. But as described above, the documentation for the specifically states that the RX-4171.0006-7 (emphasis added). Thus, the inputs relied on by the ALJ could not be changed or enabled, but were instead “fixed” such that any emphasis was removed from the product.9 Id. Because of this, even Respondents’ own expert agreed that “there was no emphasis.” HTr. 1618:18-22. The ALJ also referred to simulations as evidence that the emphasis in the SL500 Fibre Channel Transceiver “could be enabled.” ID at 172. But the simulations are just that, simulated results that were not run on actual products. Indeed, documentation relied on by Respondents confirms this, stating that the document “contains data derived from functional simulations and performance estimates. LSI Logic has not verified [] the functional descriptions.” RX-4130.0062 (emphasis added). Further, Respondents’ expert clarified that the simulations were run in reference to the CWSL500 core. RX-5430C at Q/A 388. They show how LSI hoped the core could operate in a differently designed chip where the inputs were not “fixed” so that the emphasis functionality could be turned “off” or “on.” RX-5430C at Q/As 488-89. But as explained above, the documentation and even Respondents’ own expert confirm that had its inputs “fixed”

in the “off” position. RX-4171.0006-7; HTr. 1618:18-22. Thus, emphasis could not be activated in the SL500 Fibre Channel Transceiver.
8

As the ALJ found, the CWSL500 core is meant to be incorporated into the SL500 chips. ID at 160. But, as the ALJ notes, certain chips have their emphasis functionality disabled even though they are based upon the CWSL500. Id. at 171-72. 9 The ALJ analogizes the emphasis settings in the SL500 to those in the infringing products. ID at 171; 190-91. But documents and testimony show the settings can be enabled in the infringing products, unlike in the where they are “fixed” to have no emphasis and cannot be enabled. CX-9542C at Q/As 446-457; RX-4171.006-7. 19

PUBLIC VERSION

Accordingly, the ALJ clearly erred by dismissing as irrelevant whether the product at issue even contained the claimed emphasis features. He further erred by determining that the SL500 “could” have included emphasis, contrary to that product’s documentation and testimony by Respondents’ own expert. ID at 171-72. It was critically important that the ALJ determine which product was reduced to practice, publicly used, and offered for sale and actually used emphasis. By failing to perform that analysis, the ALJ committed legal error. Mycogen Plant Sci., 243 F.3d at 1332; Pfaff, 525 U.S. at 67; Netscape Commc’ns., 295 F.3d at 1321 2. The Evidence Shows that the SL500 Fibre Channel Transceiver With No Emphasis Was the Focus of LSI’s Relationship with Seagate

The evidence shows LSI’s relationship with Seagate was focused on and in particular • with emphasis removed:

In a presentation by LSI to Seagate, LSI indicated that the CWSL500 based product “for Seagate” would be RX-4149.0004. and says that RX-4156C.0007, 7.

A Seagate presentation details its Seagate is creating a

Bruce Johnson, a Seagate employee who allegedly met with LSI regarding these products, was in the See RX-4175C.0001.

A third party email refers to a chip developed by LSI and says and RX-4176C.0001. Respondents stated that this email refers to the SL500 that was reduced to practice. RIB at 185-86.

Respondents’ expert confirmed that the

is an LSI

specification that was produced by Seagate. HTr. 1615:4-9. It states that there is no emphasis in the application. Id. at 1618:14-22; RX-4171.0006-2, 7. • Respondents’ expert testified that the documentation showing the 20

PUBLIC VERSION

transceiver had emphasis removed is “documentation” for “LSI’s SL500.” RX5430C at Q/A 51. • The SeriaLink 500k Technology Core Specification produced by Seagate does not mention emphasis. See RX-4150. • The only two product specifications related to the SL500 from Seagate either 1) do not mention emphasis or 2) affirmatively state that there is no emphasis in the product. RX-4171.0006-7; RX-4150. • Rambus’s expert testified that testing of the SL500 by Seagate contained a large amount of intersymbol interference, indicating that emphasis was likely not used in the product. HTr. 2587:4-89:14; RX-4175C.0067, 77. • Rambus’s expert testified that the simulations performed on the preliminary CWSL500 design show that the circuit was not functioning when emphasis was enabled, possibly indicating why the feature was removed. CX-10764C.0129-30; HTr. 2584:9-86:12. This evidence shows that if any product was prior art, it was the of the

SL500. That product did not meet the asserted Dally claims because the emphasis functionality, which is alleged to satisfy multiple features of the asserted Dally claims, was permanently disabled. ID at 187-92. The ALJ, however, did not consider this evidence, as he erroneously found he need not determine what products were actually prior art. Id. at 171-72. 3. Respondents Did Not Show What SL500 Product Was Reduced To Practice, Used, or Offered for Sale

To prevail on their invalidity defense, Respondents had to show by clear and convincing evidence that a product incorporating the claimed emphasis features was reduced to practice, publicly used, or offered for sale. Respondents have not done this, as they have provided scant information on which chip was actually produced. Indeed, the ALJ noted that the final documentation or 21

PUBLIC VERSION

circuitry for the SL500 was not of record, noting that

Id. at 170. Thus, the ALJ has acknowledged that definitive evidence that would allow him to determine what products were reduced to practice is not even of record.10 For this reason alone, the Commission should reverse the ALJ’s finding. Respondents have also provided no testimony by any witness with personal knowledge of what was actually reduced to practice, used, or offered for sale. Respondents recognized that such evidence would be necessary to prove what was actually prior art and attempted to submit testimony from two current or former LSI employees and two Seagate employees. But due to Respondents’ questionable discovery tactics, this evidence was excluded. HTr. 34:6-36:5. Thus, with no testimony based on personal knowledge, Respondents could not meet their burden of showing what was prior art. Instead, Respondents relied entirely on the testimony of their expert, who admitted that he was not at LSI at the relevant time and so has no personal knowledge of what was offered for sale. HTr. 1612:14-18. Indeed, when asked whether it was the of the SL500 with no

emphasis that was offered for sale, Respondents’ expert floundered, stating “I’m not sure if this was exactly what was offered for sale or not, this exact document . . . [p]robably not,” “I am not sure if this particular product was, what was being offered for sale,” and “[i]t is not clear to me that this particular product, this one was the only thing that was being offered that that point.” HTr. 1616:620. “Not sures” and “probably nots” are not clear and convincing evidence of invalidity. Respondents attempt to salvage this lack of evidence by relying on other inadequate and circumstantial evidence. For example, the ALJ relies on passing reference to the word “emphasis” or “pre-emphasis option” in a Seagate employee’s notebook (RX-4173C.0030-31) to argue that LSI The ALJ attempts to state that the lack of circuitry would not matter because “such circuitry was well known in the prior art as exemplified by the Widmer Article, the Widmer Patent, and other art of record.” ID at 170. But there were no obviousness allegations related to the SL500 Art and the ALJ provided no obviousness analysis. 22
10

PUBLIC VERSION

“was informing its customers” that the products would “include[] the emphasis functionality.” ID at 163, 181. This is speculative at best. There is no evidence regarding what was meant by these passing references. And it is not evidence that the product being offered to Seagate included emphasis. Indeed, to the contrary, as discussed above the weight of the evidence shows that LSI was offering its which had no emphasis. Speculation-based, isolated, and

uncorroborated notebook entries simply cannot carry Respondents’ burden of clear and convincing evidence, especially in light of the evidence showing that what was developed had no emphasis. The ALJ also relies on the publication of a paper in February 1997 to conclude that a chip must have been fabricated before February 1997.11 Id. at 165. He refers to testimony by

Respondents’ expert who states as a general matter that the journal requires documentation related to a fabricated chip be accompanied with submissions. HTr. 1677:9-22. But this expert admitted that he was not at LSI in the relevant time frame. HTr. 1612:14-18. Thus, he has no knowledge of what, if anything, was submitted with this paper. His testimony regarding the supposed general practice of a publication cannot substitute for evidence of what, if anything, actually accompanied the submission. Further, he was proffered as an expert on technical issues related to the Dally

technology, not the general practice of the ISSCC. Thus, his testimony should not be given weight. Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579, 597 (1993). Respondents’ argument is at best speculation, not clear and convincing evidence. Indeed, this same speculative argument was rejected by the ALJ with reference to the Widmer Article and Odysseus. ID at 137-38. There, the ALJ stated that: Dr Hassoun speculates that because the Widmer Article and the Odysseus document list the same authors, the Odysseus device “was the working chip that permitted the Widmer Article to be accepted and presented at ISSCC Respondents do not rely on the paper as a printed publication. And in any event, the paper was provided to the Patent Office during prosecution of the Dally patents and the asserted claims were allowed over it. See JX-120.003. Indeed, the PTO found that Dally conceived of his invention before the date of the paper. See JX-122.067-69. 23
11

PUBLIC VERSION

1996.” However, besides Dr. Hassoun’s speculation, Respondents provide no hard evidence to support this conclusion. Id. at 138. Likewise, there is no hard evidence of what, if anything, was submitted with the ISSCC article other than Dr. Hajimiri’s speculation. The ALJ also noted that there were differences between the Widmer Article and the Odysseus documentation, stating that “while the technology of the Widmer Article and Odysseus may have been similar in many ways, the evidence demonstrates that Odysseus was not necessarily the chip disclosed in the Widmer Article.” Id. Here, too, there are differences between the ISSCC paper and the SL500 documentation relied upon by the ALJ. For example, the ISSCC paper states that the chip’s “TX rise/fall time” is 100-385 ps. RX-4107.0002. Yet the documentation relied upon by the ALJ merely lists for entries related to the transmitter’s rise and fall

time. RX-4130.0055. Indeed, many of the entries throughout the document are labeled “TBD.” See, e.g., id. at 39, 55, 56. This is because the design of these products was just that -- to be determined. It was incomplete and changing. Respondents essentially attempt to bridge the gap between

preliminary documentation and an article nearly a year and a half later to argue that a chip must have been created some time in between, but they provide nothing more than speculation. See RX4130.0007 (dated September 1995). Thus, while there may be similarities in documentation,

Respondents simply have provided no concrete evidence of what, if anything, was submitted with the ISSCC paper. Thus, Respondents’ evidence that an SL500 chip with emphasis was reduced to practice, in public use, and offered for sale, amounts in total to passing references to “emphasis” in a notebook with no explanation of what was meant or discussed and a paper that itself is not prior art with no hard evidence of what, if anything, was submitted with it. This is not clear and convincing evidence, especially in light of the overwhelming evidence that LSI was focused on reducing to practice and offering for sale to Seagate the 24 of the SL500 without emphasis.

PUBLIC VERSION

Given that: 1) Respondents rely on speculation; 2) the ALJ found that Respondents did not provide final documentation regarding the SL500; 3) Respondents provided no testimony based on personal knowledge; and 4) Respondents’ own expert admitted he was not at LSI at the time and is not sure if the product offered for sale to Seagate was one which had emphasis removed, Respondents cannot meet their heavy burden of clear and convincing evidence. 4. Respondents Did Not Meet Their Burden Regarding an Offer For Sale to Cabletron

The ALJ also found that the Dally patents are invalid based on an offer to sell a chip called to Cabletron. ID at 174-76; 181-82. But like the SL500, the ALJ never determined whether the chip that was allegedly offered for sale actually contained the claimed emphasis features. Id. Instead, he merely presumed that the chip must have included that functionality because it was meant to be based on the CWSL500 core. But as described above with reference to the the mere incorporation of the CWSL500 did not guarantee that the chip contained emphasis. See RX-4171.0006-7. The incorporated the

CWSL500 core, yet it is clear that it contained no emphasis. Id. When confronted with this conflict at trial, even Respondents’ own expert admitted that the emphasis circuitry in the have been “fixed” to be “turned off” just as in the chip could HTr. 1624:7-

17. That evidence alone, not addressed by the ALJ, justifies reversal. In re Turlay, 304 F.2d 893, 899 (C.C.P.A. 1962) (“It is well established that an anticipation rejection cannot be predicated on an ambiguous reference.”). The ALJ erred again by failing to determine what product was actually prior art, and whether that product even contained the claimed emphasis features. The ALJ further erred by not addressing testimony of Respondents’ expert who stated he “can’t say” whether the circuitry in Shemp could be “fixed and turned off” just like the fibre channel version of the SL500. Id. The ALJ’s finding should be reversed. 25

PUBLIC VERSION

Dally claims,12 each high signal level represents one bit of data (e.g., a 1) and each low signal level also represents one bit of data (e.g., a 0). HTr. 1472:22-1473:5. Thus, as the figure above

demonstrates, two bits of data are represented within a single Hertz. Because two bits of data are provided in each Hertz, the plain and ordinary meaning of the term “output frequency” is one-half the value of the output data rate. CX-9542C.0054-55; HTr. 705:24-706:6. The Dally patent specification confirms this construction. CX-9542C.0054-55. For example, in describing Figures 2A and 2B, the Dally patents state that the highest frequency of interest for a 4 Gb/s signal is 2 GHz (i.e., half the data rate). JX-120.0014 at 3:59-63. Further confirming this relationship between output frequency and data rate, in describing another set of figures, the Dally patents state that “our operating frequency of 2 GHz correspond[s] to a bit rate of 4 Gb/s.” Id. at 4:16-19. These passages confirm that an NRZ signal with an output frequency of 2 GHz has a data rate of 4 Gb/s, just as an NRZ signal with the claimed “output frequency of at least 1 GHz” has a data rate of at least 2 Gb/s. In spite of these facts, the ALJ found that “output frequency” has a one-to-one relationship with data rate, contending that Rambus’s construction is based on a number of mistaken assumptions. ID at 24-34. These “assumptions,” however, are not assumptions at all—they are simply statements of how the technology operates. The ALJ’s analysis of these “assumptions” contains several

fundamental errors. One example that permeates the ALJ’s analysis is the focus on actual data patterns as opposed to the rate at which a signal is capable of sending data. For example, on page 26 The ALJ contends that Rambus’s construction assumes NRZ signaling but NRZ signaling is not required by the asserted Dally claims. ID at 27. The ALJ has misinterpreted the claims. The asserted Dally claims require NRZ signaling by reciting that each bit of data is represented by one signal level, where that signal level may vary based on certain claimed conditions. See, e.g., ’857 claim 1, reciting “one signal level” under a first set of conditions and “another signal level” under a second set of conditions. The ALJ attempts to refute this fact by citing the Staff’s figure of a Manchester-encoded signal. ID at 27-28. Manchester-encoded signals, however, provide two signal levels per input bit, not one. CX-10764C.0009-10, 78-79. Thus, the figure is not covered by the claims, which expressly require one output signal level per input bit, and the figure does not support the ALJ’s conclusions. 27
12

PUBLIC VERSION

of the ID, the ALJ relies on the Staff’s figure of a signal having signal levels representing successive ones followed by signal levels representing successive zeros. The figure labels this four-bit sequence as spanning one Hertz. This is incorrect as a technical matter because the signal is capable of transmitting a 1010 data pattern in that same span. See CX-9542C.0055. Using a 1010 data pattern clarifies that the span identified as one Hertz in the Staff’s figure actually represents two Hertz. Accordingly, when properly understood, the Staff’s figure supports Rambus’s construction. Another example of this same type of error appears on page 27 of the ID, where the ALJ illustrates Figure 2A of the Dally patents. That figure shows a data pattern with varying numbers of consecutive high and low signal levels representing zeros and ones. Using this figure, the ALJ erroneously concludes that “Rambus’s assumption that one cycle necessarily includes one high signal and one low signal” is in error. ID at 27. But once again the ALJ focused on the highs and lows in a particular data pattern as opposed to the numbers of highs and lows that the signal is capable of providing. See CX-9542C.0055; HTr. 705:24-706:6. The critical question is how fast the signal is capable of transitioning, not how fast it actually transitioned when transmitting a particular data pattern (as the ALJ analyzed). Id. Due to these and other errors, the Commission should review the ALJ’s construction of “output frequency” and adopt Rambus’s construction. IV. THE BARTH I PATENTS ARE NOT INVALID The ALJ found that Respondents met their heavy burden of proving all Barth I claims invalid by clear and convincing evidence. But in reaching that conclusion, the ALJ legally erred in finding that the NextBus specification is prior art despite no evidence of its public availability and the exclusion of testimony from “the one person” Respondents said “that can really authenticate for us [its] public availability.” HTr. 286:5-7. The ALJ also erred in construing the claimed “memory device” so broadly as to include memory controllers, memory modules, and memory systems based on a dictionary definition of “device” when the intrinsic and extrinsic evidence refute that construction, evidence the ALJ did not address and said Rambus did not even present. ID at 108. 28

PUBLIC VERSION

And the ALJ erred in concluding that prior art asynchronous systems anticipate the claimed synchronous control operations and that aspects of asynchronous and synchronous prior art can be combined to improve speed and efficiency, deficits not remedied by the proposed combinations. The ALJ also erred in relying on the on-going reexamination proceedings. At the Hearing, the ALJ correctly noted the different standard of the PTO and the fact that the PTO does not apply the statutory presumption of validity. HTr. 2712:18-22. This was not disputed by Respondents and cannot be. HTr. 2712:18-2713:15; see also In re Swanson, 540 F.3d 1368, 1377-78 (Fed. Cir. 2008). Given that, the ALJ stated: “I am not going to follow what the Board does. I am not going to cite what the Board does.” HTr. 2713:7-10. Nonetheless, the ALJ repeatedly followed and cited the Board and PTO. ID at 94, 108, 109, 118, 120, 212. Not only was that legal error, the PTO’s analysis contradicts the Commission’s findings in the 661 Investigation and the ALJ’s findings here that Farmwald ’037 does not invalidate the Barth I claims. 661 ID at 49-56; CX-10765C.0088-89; ID at 121-22. The ALJ’s invalidity holdings on the Barth I patents should be reviewed. A. Background of the Technology of the Asserted Barth I Patents

The asserted Barth I patents are directed to synchronous systems and methods for increasing the efficiency of communication between a memory controller and memory devices. JX-3 at 1:1417. This is achieved by decoupling the timing of data control from the timing of data transfer. Id. at 3:24-25. They teach, for example, that after a memory controller issues a read command or a write command to a memory device, the memory controller separately sends a signal to indicate when sampling or data transfer is to begin. Id. at 7:28-30; 8:63-67. Barth I calls that signal a strobe signal. Id. at 8:63-67. Barth I explains that there are several benefits to decoupling command control information, such as a read or write command, from the data transfer start information, or strobe signal. One is to “specify a transfer operation for any amount of data.” JX-3 at 9:20-22. A second is to reduce the latency between a data transfer request and the beginning of data transfer can be reduced. Id. at 9:35-45; 661 Order No. 12 at 21. Decoupling also makes the system more flexible. 29

PUBLIC VERSION

JX-3 at 10:21-33. B. The Prior Art Does Not Anticipate the Barth Claims

The ALJ concluded that a NextBus specification, Harriman, Hayes, Dan, and Yano, anticipate certain claims of the ’353 and ’109 Barth I patents. ID at 89-121. The ALJ’s conclusions on NextBus, Harriman and Hayes are based on a legally erroneous construction of “memory device,” while his conclusions on Dan and Yano are premised on ignoring what Respondents’ expert characterized as the “significant departure” between the claimed synchronous control operations and the asynchronous Dan and Yano systems. HTr. 1133:7-21. 1. The NextBus Specification (RX-4265C) Does Not Anticipate Any Claims of the ’353 and ’109 Patents

The ALJ’s conclusions on the NextBus specification should be reversed because it is not prior art and because the ALJ erroneously interpreted the “memory device” as including a memory board containing a memory controller. ID at 105-109. a. The ALJ Erred in Finding that the NextBus Specification Is Prior Art

Whether the NextBus specification is prior art is a question of law based on underlying facts. Orion IP, LLC. v. Hyundai Motor Am., 605 F.3d 967, 974 (Fed. Cir. 2010). While the ALJ noted that “‘public accessibility’ has been called the touchstone in determining whether a reference constitutes a ‘printed publication’ bar under 35 U.S.C. § 102(b)” and that “[a] reference is considered publicly accessible if it was ‘disseminated or otherwise made available to the extent that persons interested and ordinarily skilled in the subject matter or art exercising reasonable diligence can locate it,’” ID at 105 (quoting In re Lister, 583 F.3d 1307, 1311 (Fed. Cir. 2009)), he did not apply those principles. There is no evidence, let alone clear and convincing evidence, that the NextBus specification was publicly accessible before the October 1995 filing date of the Barth I patents. The ALJ found that “the 1990 copyright date, a specified reorder number . . . and technical support telephone number within the NextBus Specification collectively suggest that the document 30

PUBLIC VERSION

was indeed publicly available.” ID at 105. The ALJ’s analysis is akin to finding that the placement of a copyright date on this brief (©2012) along with a telephone number (555-1212) and something called a “reorder number” (12345) means that this document is publicly available. Those indicia, however, do not show that a document was “disseminated or otherwise made available to the extent that persons interested and ordinarily skilled in the subject matter or art exercising reasonable diligence can locate it.” Lister, 583 F.3d at 1311. Instead, the ALJ’s conclusion assumes the NextBus specification was publically available, someone obtained a copy of it, and that a person therefore could find the information contained in it. None of this is supported by the evidence. To the contrary, the only evidence showing publicly accessibility is a website from which Respondents obtained the document, which was not created until 2005, a decade after the 1995 Barth I filing date. RX-4265.0081; HTr. 1167:4-25. In fact, Respondents recognized this problem when they

unsuccessfully tried to introduce evidence, arguing that a witness should testify because “he is the one person that can really authenticate for us the public availability of NextBus documentation.” HTr. 286:5-7. Because no one authenticated the public availability of the NextBus specification, the ALJ erred in finding that it is prior art. The ALJ also stated that “other publicly available documents in the 1991 timeframe, such as Harriman, specifically incorporate by reference a NextBus Specification,” showing public accessibility. ID at 105-06. Harriman, however, does not “incorporate by reference” any NextBus specification let alone the NextBus document relied on by the ALJ, stating only that “bus 209 is a bus known as the ‘NextBus’ and is described in NextBus Specification, published by NeXT Computer, Inc., Redwood City, Calif.” RX-4277 at 15:43-46.13 “Whether and to what extent material has been incorporated by reference into a host
13

While referring to “other publicly available documents,” the ALJ addresses only Harriman. No party has alleged that any other document incorporates by reference the NextBus specification. ID at 105-06. 31

PUBLIC VERSION

document is a question of law.” Adv. Display Sys., Inc. v. Kent State Univ., 212 F.3d 1272, 1283 (Fed. Cir. 2000). To incorporate material by reference, “the host document must identify with detailed particularity what specific material it incorporates and clearly indicate where that material is found in the various documents.” Id. at 1282. The reference in Harriman to a NextBus specification does not: (1) use the words “incorporate by reference;” (2) identify with detailed particularity the alleged material being incorporated; or (3) clearly indicate where that material is found. Such a “vague referencing practice is hardly sufficient to meet [the] legal requirements for incorporation.” Kyocera Wireless Corp. v. ITC, 545 F.3d 1340, 1351-52 (Fed. Cir. 2008). Further, there is no evidence that any NextBus specification was provided to the PTO in conjunction with the Harriman patent, that the public could have obtained a copy, or that it is the same as the NextBus document relied on by the ALJ. The ALJ also relied on the Federal Circuit’s Orion decision as supporting the NextBus specification being prior art. ID at 106. There, however, a manager testified that a document was a direct-mail promotional, “was produced in 1987 and that it was used, by upwards of 150 to 200 salespersons,” and that he had used it in hundreds of sales pitches prior to the critical date. Orion, 605 F.3d at 974-75. That testimony was corroborated by multiple third-party witnesses. Id. at 97576. Contrary to Orion, there is no evidence14 that the NextBus specification was used by or publically available to anyone before the October 1995 filing date of the Barth I patents. b. The ALJ’s Construction of “Memory Device” is Wrong

The ALJ also erred in construing “memory device,” a term used in every asserted Barth I claim. ID at 108-109. But for that error, Respondents agree that NextBus does not anticipate any
14

The ID also cites expert testimony. ID at 105-06. Dr. Przybylski testified that the document had a 1990 copyright. HTr. 2647:14-22. Dr. Jacob testified what the documents “suggested” to him, HTr. 1152:25-1154:25, not based on direct knowledge. Given another ALJ’s finding that in the mid-1990’s Dr. Jacob’s DRAM-related experience was limited to reviewing some DRAM data sheets, this is not surprising. In re Rambus, 2004 WL 390647 at FOF 1128 (FTC 2004). 32

PUBLIC VERSION

Barth I claim. HTr. 1121:15-1122:3. The same is true for Harriman and Hayes. ID at 104, 117-121; RX-4268 at FIG. 7; RX-4266 at 15:43-16:2; CX-10765C.0062, 75. The ALJ’s construction of “memory device” is inconsistent with the intrinsic evidence (the claim language and specification) and the extrinsic evidence (writings of Respondents’ expert, prior art, and prior testimony) and Federal Circuit precedent. This is not an instance, however, in which a losing party simply disagrees with a judge’s conclusion after the evidence has been addressed. Here, the ALJ stated that “Rambus fails to point to anything in the Barth Patents that supports their construction of memory device.” ID at 108. Likewise, the ALJ stated that “Rambus fails to offer any extrinsic evidence (not a single dictionary definition) to support their proposed construction that that the term ‘memory device’ has to be limited to a single monolithic substrate.” ID at 108. Neither statement is true. In its briefing and at the Hearing, Rambus explained at length why the intrinsic evidence supports its construction, CPB at 56-59, CIB at 13-15, CRB at 37, CX-9543C.0139-40 (Przybylski), CX-10765C.0049, 78-80 (Przybylski), HTr. 1084:9-1095:3 (Jacob), and why the extrinsic evidence supports its construction and refutes the construction adopted by the ALJ that the claimed “memory device” can include a module, board, or system with a memory controller. CPB at 59-63; CIB at 15-18; CRB at 38; HTr. 1095:4-1112:9 (addressing CX-9731 (Jacob text book), CX9784 (Jacob patent), RX-4489 (Liu textbook), and RX-4276 (JEDEC standard)). Indeed, at the Hearing, the ALJ admonished Rambus’s counsel during cross-examination of Respondents’ expert that another example showing “modules as different than devices” was not necessary. HTr. 1112:1222. The ALJ’s contradictory findings in the face of that admonishment justify review. i. The Intrinsic Evidence Supports Rambus’s Construction (a) The Claims Support Rambus’s Construction

The ALJ did not address the Barth I claims, which confirm that a memory device does not include a memory controller. HTC Corp. v. IPCom GmbH & Co., 667 F.3d 1270 (Fed. Cir. 2012) (finding legal error in failure to “examine adequately the claims themselves”). Claim 11 of the ’353 33

PUBLIC VERSION

patent recites a “method of controlling a memory device,” including “issuing a first write command” and “issuing the strobe signal” to the memory device. JX-4 at claim 11. The method of controlling is performed by a memory controller distinct from the claimed memory device, which receives the write command and strobe signal. CX-10765C.0079-80. The other Barth I claims are similar, reciting methods of controlling including sending signals to a memory device. JX-3 at claim 11; JX5 at claims 1, 12, and 20. Claims of other patents in the Barth I family confirm that a memory controller is distinct from a memory device. Claim 6 of U.S. Patent No. 5,748,914, parent to the asserted Barth I patents, recites a method “for maximizing usage of a bus that connects the memory controller to one or more memory devices,” making clear that a memory controller and a memory device are distinct things. CX-3065C at claim 6; CX-10765C.0080. Claim 1 of the ’914 patent is similar, reciting “causing a controller to” perform some steps and “causing a memory device” to perform other steps. CX-3065C at claim 1. The Barth I claims themselves thus support construing memory device as distinct from a memory controller. (b) The Specification Supports Rambus’s Construction HTC Corp., 667

The ALJ also committed legal error by not addressing the specification.

F.3d at 1270 (finding legal error in failure to refer to the specification to understand the claims). The specification, however, is clear that a memory device does not include a memory controller. CX9543C.0140; CX-10765C.0079. Figure 6 shows memory controller 601 separate from DRAM

memory device 603. JX-4 at Fig. 6. Figures 2, 20A, 20B all show a controller distinct from a memory device. Respondents’ expert agreed. HTr. 1090:7-1093:15. No figure shows a memory device differently and no figure shows a memory device including a memory controller. The written description also distinguishes a memory controller from a memory device, explaining that a controller “transmit[s] control information on a bus” while a “memory device reads the control 34

PUBLIC VERSION

information on the bus.” Id. at 3:65-66, 4:6-7; see also JX-4 at 1:23-26, 2:45-52, 4:35-40, 6:21-33, 6:49-8:9, 8:25-27, 8:41-43, 9:5-10, 11:18-29, 12:67-13:2, 13:39-43, 15:42-45, 17:54-57, 17:66-18:21, 19:53-55, 21:40-46, App. A-C; CX-9543C.0139-40. The specification supports Rambus’s

construction and demonstrates that at the very least the claimed memory device is distinct from a memory controller. (c) The ALJ’s Construction, Inconsistent With the Intrinsic Evidence, Must Be Rejected

In construing memory device, the ALJ did not address the intrinsic evidence, relying instead on a statement by Respondents’ expert that “the Barth patents do not provide any special meaning to the term memory device.” ID at 109. But the claims and specification at least implicitly define “memory device” as distinct from a memory controller. Hynix Semiconductor Inc. v. Rambus Inc., 645 F.3d 1336, 1350 (Fed. Cir. 2011) (patentee may implicitly define terms). Moreover, the ALJ’s reliance on testimony from Respondents’ expert was improper because it cannot trump the claims’ and specification’s confirmation that a memory device is distinct from a memory controller. Network Commerce, Inc. v. Microsoft Corp., 422 F.3d 1353, 1361 (Fed. Cir. 2005) (expert conclusion at odds with intrinsic evidence must be disregarded). Even ignoring that the specification and claims at the very least distinguish a memory controller from a memory device, the ALJ’s construction leads to the absurd result that a memory device can be virtually anything, including a memory controller, a memory module, and a memory system. That construction is inconsistent with other portions of the ID indicating that a memory device is different than a memory controller. See, e.g., ID at 54 n.6 (“In a write operation the memory controller is the data transferring device and the memory device is the data receiving device, such that the memory controller ‘writes’ data to the memory device”). The ALJ also erred by violating the rule against elevating dictionaries above the specification, focusing on the definition of “device” from an IEEE dictionary, ID at 108, rather than 35

PUBLIC VERSION

the meaning of “memory device” in the context of the patent. Phillips v. AWH Corp., 415 F.3d 1303, 1321 (Fed. Cir. 2005) (en banc). Not only was that legal error, the definition the ALJ relied on is irrelevant because it pertains to “696 interface devices,” not to the claimed “memory device.” See RDX-3425. The ALJ also erred in pulling apart the term “memory device” to construe it. The claim term “memory device” is a compound noun, like “fire station,” which means something different than its constituent parts. It requires construction as a singular term, not as definitions of the words “memory” and “device” separately. HTr. 2657:13-2659:16. Indeed, the Federal Circuit has rejected pulling apart a term like “memory device” to construe it. In Network Commerce, the court rejected individual dictionary definitions of “download” and “component” when construing “download component” as “not a tenable theory in light of the specification.”15 422 F.3d at 1359-60. The same is true here. Nothing justifies splitting up the claim term and just considering a definition of “device.” Instead, “memory device” should be construed as the whole claim term it is and, in accordance with the claim language and specification, as distinct from a memory controller. ii. The Extrinsic Evidence Supports Rambus’s Construction

Contrary to the ALJ’s assertion that Rambus failed to offer any extrinsic evidence supporting its construction, ID at 108, such evidence, including the writings of Respondents’ expert, the prior art, and witness testimony, confirm that Rambus’s proposed construction is correct. Figure 7.6 from Dr. Jacob’s textbook is illustrative, showing a memory controller (left) and DRAM memory devices (right) as separate components. HTr. 1095:4-1096:18.

The ALJ similarly erred in construing the claimed “write command” by relying on separate extrinsic definitions of “write” and “command.” ID at 94. 36

15

PUBLIC VERSION

Figure 7.6 shows a memory system with individual chip DRAM memory devices that are separate and distinct from the memory controller. CX-9731.0349. The memory devices are on a DIMM (dual in-line memory module) connected over a data bus to the memory controller. This is not an isolated example in Dr. Jacob’s textbook. His Figure 12.25 also distinguishes a memory device from a module, a memory controller, and a memory system. CX-9731.0463 at Figure 12.25; HTr. 1102:1-1104:2. In fact, outside of the context of this litigation, and in his purported 20-plus years of experience, Dr. Jacob has consistently and repeatedly used memory device to mean a DRAM or memory chip, an integrated circuit, and as distinct from a memory controller and a memory system. CX-9731.0403, 494; RX-4421; CX-9784; HTr. 1104:6-1109:3. This non-

litigation-related evidence from Respondents’ expert refutes the position he took here and that the ALJ adopted, namely that a memory device can be a board containing a controller. ID at 109 (citing RX-5429C (Jacob direct) at Q/A 306, 310 and Jacob trial testimony); ID at 118 (citing RX5429C.0123). (a) The Prior Art Supports Rambus’s Construction

The prior art supports Rambus’s construction of memory device. Arthur A. Collins, Inc. v. N. Telecom Ltd., 216 F.3d 1042, 1044-45 (Fed. Cir. 2000) (prior art may assist in determining the meaning of a claim term). For example, a district court construed memory device in connection with Rambus’s Farmwald disclosure as on the “order of a single chip.” Rambus Inc. v. Hynix

Semiconductor, Inc., 2008 WL 5771128, at *2 (N.D. Cal. Nov. 21, 2008). Importantly, the court concluded that a memory device is different from a memory board and does “not include a 37

PUBLIC VERSION

microprocessor like a CPU or memory controller.” Id. at *1-2. Likewise, Release 4 of JEDEC Standard No. 21-C (RX-2108C) from November 1993, relied on by the ALJ, ID at 210, refers to a memory device as a chip and differentiates it from a module, which includes memory devices. It describes a “SIMM, Single-in-line-Memory-Module” as “[a] multichip module in which the body has a SIP form and is made primarily of memory devices.” RX2108C.0024. Similarly, a 1991 Liu textbook (below), RX-4489.0004 and 313; HTr. 1110:9-1112:9, distinguishes a memory module from a memory device.

The prior art thus refutes the ALJ’s construction. (b) Testimony from Parties Adverse to Rambus Supports Rambus’s Construction

Witnesses for Broadcom and LSI testified in this Investigation that a memory device is a memory chip, consistent with Rambus’s construction. JX-127C at 62:2-18; JX-75C at 25:21-23; JX72C at 36:18-37:2. In addition, Desi Rhoden, chairman of the Board of Directors of JEDEC in the 1990s, appeared as an adverse witness to Rambus before the FTC and testified that a memory module is a board that contains multiple memory devices and that memory devices are integrated circuits or chips. JX-105C.0024 at 272:1-274:7, JX-105C.0032 at 1190:2-11. This testimony, not addressed by the ALJ, supports reversing his claim construction. In sum, as a matter of law and policy, the Commission should review the ALJ’s construction of “memory device,” premised on wrong assertions that Rambus did not provide intrinsic or extrinsic support for its construction and which is inconsistent with that evidence.

38

PUBLIC VERSION

c.

NextBus Does Not Anticipate the ’353 or ’109 Claims

If “memory device” is construed as a chip, as it should be, there is no anticipation. Indeed, even if “memory device” is construed just so it does not include a memory controller, Respondents agree there is no anticipation. HTr. 1121-22. This is because NextBus discloses sending either DSTB* or DRQ* (the alleged strobe signals16) to a board that includes a memory controller. RX4265.0014-15; HTr. 1120. It does not disclose that those signals are issued to a memory device, as the claims require, and thus also cannot disclose that those signals initiate sampling at the memory device. RX-4265.0014-15; CX-10765C.0050-51. Because NextBus discloses only connecting

memory boards (and not a memory device) to a bus, it does not disclose the claimed “method of controlling a memory device” or the steps of “issuing a first write command to the memory device” and “issuing the strobe signal to the memory device,” of claim 11 of the ’353 patent. The ALJ’s analysis of claims 1, 20, and 24 of the ’109 patent is identical to his analysis of claims 11-13 of the ’353 patent. ID at 104-14. He erred in finding the NextBus specification anticipates those claims for the same reasons. d. NextBus Does Not Anticipate Claims 12 and 13

The ALJ erred in finding that MCLKSEL* is the “external clock signal” of claims 12 and 13 of the ’353 patent because that signal is not a clock signal. ID at 116-17. Instead, it is a qualification signal used to create an internal signal, MCLK*. RX-4265.0025; CX-10765C.0052-53, 56. It is called a “signal,” a “clock qualification signal,” and a “major clock select signal,” as distinguished from “BUSCLK,” which is called a “clock.” RX-4265.0022, 69; CX-10765C.0052-53, 56. As a qualification signal, not a clock signal, MCLKSEL* is not the “external clock signal” of claims 12 and 13. CX-10765C.0052-53, 56. The ALJ stated that “MCLKSEL* is described as a clock, even

Rambus also disagrees with the ALJ’s finding that the DRQ* signal initiates sampling of data because it only indicates “that at least two words of data are available” and does not initiate sampling. RX-4265.0049; CX-10765C at 50. 39

16

PUBLIC VERSION

containing the abbreviation ‘CLK’ in its label.” ID at 117. But MCLKSEL* also includes “SEL” in its label which the NextBus specification explains means a qualification signal rather than a clock signal. 2. Harriman (RX-4277) Does Not Anticipate the Asserted Claims of the ’353 and ’109 Patents

The ALJ did not separately analyze Harriman, stating instead that he “incorporates by reference” his analysis of NextBus for Harriman because “setting forth the analysis for Harriman would be a duplicative effort.” ID at 104.17 Harriman and NextBus, however, are different

documents and the ALJ’s “incorporation by reference” cannot be clear and convincing evidence of invalidity. Thus, the ALJ’s invalidity holding should be reversed. The ALJ also erred in not addressing Rambus’s arguments distinguishing Harriman. CIB at 72-75. Harriman does not disclose issuing a strobe signal to a memory device, as required by the claims of the ’353 and ’109 patents. Respondents pointed to the DSTB* signal of Harriman, but that signal is not issued to a memory device. Instead, it travels on a bus connecting a controller to a multiplexer and cannot initiate sampling in a memory device, as required in the claims. RX-4266 at 15:43-16:2; CX-10765C.0061-62. Harriman also does not disclose the external clock signal and its functionality, recited in ’353 patent claims 12 and 13. CX-10765C.0064-65. Harriman’s Figure 8 labels “MCLK” as “(internal),” while MCLKSEL* is described as a signal that selects BusClk 401 and not a clock signal. RX-4266 at 15:55-57; RX-4266 at Fig. 8; CX-10765C.0063. And Figure 8 shows a single portion of data, not a first and second portion of the data, as claimed. RX-4266 at Fig. 8. 3. Hayes (RX-4268) Does Not Anticipate Claim 11 of the ’353 Patent

The ALJ’s erroneous construction of “memory device” also led him to wrongly conclude that To the extent the ALJ’s rationale on Harriman is based on the supposed incorporation by reference of NextBus, that document was not incorporated by reference into Harriman, as discussed above in Section IV.B.1.a. 40
17

PUBLIC VERSION

Hayes anticipates claim 11 of the ’353 patent. ID at 117-21. Like the memory board in the NextBus specification, the memory board of Hayes includes a memory controller (i.e., RAM control logic 62) and multiple memory devices (i.e., DRAM chips). RX-4268.0009. Hayes discloses that a signal called DS (data strobe) is issued to RAM control logic 62 and is not forwarded to the DRAMs on the memory board. RX-4268.0009 at FIG. 7; CX-10765C.0075. Thus, although there is something called a “strobe” in Hayes, it is issued to a memory controller, not a memory device, as claimed, and thus cannot initiate sampling in a memory device, as also claimed. Id.; HTr. 2713-14. At trial, Respondents’ expert agreed that the DRAM chips 64 in Hayes do not receive the Hayes DS signal. HTr. 1126:16-23. Even if ALJ’s erroneous construction of “memory device” is accepted, Hayes still fails to disclose issuing the claimed strobe signal to the memory device “to initiate sampling,” as recited in claim 11. RX-4268 at 7:61-65; CX-10765C.0076. Rather than initiating sampling, DS is used to indicate whether the shared data and address lines (DAL) “contain valid write data” or address data. RX-4268 at 7:61-63; CX-10765C.0076. The ALJ states that indicating that valid write data is on the DAL lines is initiating sampling. ID at 120-21. As Hayes states, however, DS acts as an indicator of what type of information is on the bus, RX-4268 at 7:61-65, 9:49-65, and does not initiate sampling of data, which is capturing of data at a discrete point in time. RX-4268 at 7:61-65; CX-

10765C.0076. Hayes in fact confirms that DS does not initiate sampling. If it did, there would be no need for the bus master to “deassert[] DS to indicate that it is about to remove the write data,” since sampling would necessarily already have occurred. Id. The ALJ reasoned that the “data is valid” language in Hayes is similar to language Rambus relied on in proving infringement. ID at 120-21. The ALJ quoted deposition testimony stating that the DQS signal is “used to tell the receiver that data is valid on the bus, so now you can clock those data into . . . its registers.” ID at 121 (quoting CX-9543C.0249). This testimony, however, is very different from the Hayes disclosure because it not only states that “data is valid,” it states that you 41

PUBLIC VERSION

can “clock those data into . . . its registers,” referencing a memory chip, something nowhere disclosed in Hayes with respect to a strobe signal. In addition, in proving infringement, Rambus relied on other evidence demonstrating that the accused memory controller products issue a strobe signal to initiate sampling. ID at 49-56. Hayes, on the other hand, discloses that the DS signal merely differentiates between data and address information to a memory controller. In fact, in Hayes, the DS signal is not sent to the same place as the data, further indicating it does not initiate sampling of data. Indeed as shown in Figure 7, the DAL lines that carry data are inputs to the DATA transceiver, but DS is never sent to the DATA transceiver. Therefore, DS does not indicate when to sample the data.18 RX-4268 at Fig. 7; CX-10765C at 81. For at least these reasons, even under the ALJ’s incorrect interpretation of memory device, Hayes does not disclose the claimed “strobe signal” features. 4. Dan and Yano Do Not Anticipate Claim 11 of the ’353 Patent a. The Barth I Patents Are Directed to Synchronous Memory Systems Whereas Dan and Yano Describe Asynchronous Memory Systems

The ALJ erred because he did not differentiate between the asynchronous memory systems of Dan and Yano and the synchronous memory systems of the Barth I patents. ID at 96-97, 101-102. This distinction is a very important one, since synchronous systems such as in the Barth I patents operate very differently than the asynchronous systems of Dan and Yano. Most strikingly,

asynchronous memory devices in Dan and Yano do not receive a clock signal that governs the timing of operations. RX-4261.0007; RX-4261.0010; CX-10765C at 109-11. The Dan and Yano memory devices also do not sample a command or data at a discrete point in time. Instead, operations are controlled with a sequence of signals and data simply flows into the memory device over the duration

Like Dan and Yano, Hayes discloses an asynchronous memory system and explicitly references an asynchronous bus. RX-4268 at 6:3-17, Fig. 5. Thus, Hayes does not disclose the claimed “write command” for the same reasons as Dan and Yano. 42

18

PUBLIC VERSION

of the write operation. CX-10765.0029, 35-37. In contrast, the Barth I patents teach and claim synchronous memory systems where the memory device receives a clock signal that governs the timing of operations. JX-3 at Fig. 6 (showing a line that carries a clock signal to the memory device). The specification explains that the command is “encoded” into a code, i.e., one or more bits that indicate which operation to perform. Id. at 3:3132, 3:50-52, 3:62-64. Claim 11 of the ’353 patent requires such a “write command,” i.e., one or more bits that specify that the memory device receive and store data. Such a command is sampled, i.e., captured at a discrete point in time. Claim 11 also requires “strobe signal . . . to initiate sampling of a first portion of the data,” where sampling requires capturing data at a discrete point in time. JX-04 at claim 11. Despite Respondents’ expert characterizing later synchronous systems as a “significant departure” from earlier asynchronous systems (like Dan, Yano, and Hayes), the ALJ did not credit these fundamental differences. Instead, he appears to have concluded that these distinctions do not exist or do not matter. ID at 90-104, 211-12. The ALJ thus erred in finding anticipation. b. Dan Does Not Anticipate ’353 Patent Claim 11

The ALJ erred in finding that Dan discloses the claimed strobe signal. ID at 102. The “external write strobe” of Dan does not initiate sampling as required in claim 11, i.e., capturing data at a discrete point in time.19 RX-4262.0010; CX-10765C.0019-23, 32. Rather than capturing data at a discrete point in time, Dan discloses allowing data to flow into its asynchronous memory device20 over the duration of the write operation. Id. For example, Figure 3 of Dan shows data must be held for the duration of the external write strobe, which would not be needed if data was sampled, i.e., The ALJ asserts that Rambus offers a “narrow construction of sampling.” ID at 102. But the Commission confirmed that construction in the 661 Investigation. See 661 ID at 20-28 (the accused products sample data by capturing data at a discrete point in time). 20 The “synchronous random access memory” title is a typographical/translation error. It should read “static random access memory,” as used throughout Dan (e.g., RX-4262 at 1:11-14) and as is the understood meaning of SRAM. 43
19

PUBLIC VERSION

it “is broad enough to encompass a combination of signals asserted through the entire write operation.” ID at 94-95. As with “memory device,” the ALJ’s construction is based on dictionary definitions of “write” and “command” and not the intrinsic evidence. Id. The ALJ relied on a dictionary defining “command” as “instruction.” ID at 94 (citing SX-5).21 But “instruction” is defined as “a set of bits” defining an operation and comprising an “operation code.” This latter definition is consistent with the specification, which discloses that the controller transmits “command control information” to the memory device, which includes an operation code comprising one or more bits to specify the type of operation (e.g., a write or read). JX-3 at 6:51-7:15. The command is “encoded,” meaning one or more bits form a code that represents the operation to be performed. Id. at 3:31-32, 3:50-52, 3:62-64. At no point does the specification consider an asynchronous control signal to be this sort of code. One of ordinary skill would thus understand write command as “one or more bits which specify that the memory device receive and store data.” CX-10765C.0030-31. The ALJ’s statement that the specification “at most confirms that different techniques can be used to instruct the memory device to store data associated with a write operation,” ID at 95, is inconsistent with the specification. Properly construed, no such bits, codes, or instructions are used in Dan to specify a write operation and thus Dan does not disclose the claimed write command. In asynchronous systems like Dan, signals must be transitioned up and down and in a particular sequence to perform an operation. The state of those signals at a particular moment in time cannot be used to determine what operation is to be performed. RX-4262.0013; RX4261.0008; CX-10765C.0030-31, 35-37. This distinction is important because the claimed “write command,” as one or more bits and that are sampled to indicate a write operation, allows lines conveying the command to be used for other purposes, such as sending other commands, enabling multiple outstanding operations and permitting higher throughput. The ALJ also relied on the testimony from a Barth patent inventor. ID at 94-95. That testimony related to “any memory in general” rather than in the context of the Barth I patents. 45
21

PUBLIC VERSION

(captured at a discrete point in time) as claimed. Indeed, if data were sampled on an edge of DS, in Yano there would be no need to hold data throughout the entire duration of DS, as shown in Figure 2. RX-4261 at 4:8-21, 0008; CX-10765C.0028. e. Yano Does Not Disclose the Claimed “Write Command”

The ALJ erred in finding that Yano discloses the claimed write command. Like Dan’s output enable signal, Yano’s R/W signal is controlled over a period of time rather than as one or more bits (or an instruction) that are sampled at a discrete point in time. RX-4261 at 4:8-21, 0008; CX10765C.0028. C. The ALJ Committed Errors of Law, Fact, and Policy in His Obviousness Analysis of the Barth I Claims

In just thirteen pages of the ID, the ALJ found that 12 different prior art combinations render obvious all of the asserted claims of the Barth I patents. ID at 210-23. In that short discussion, which largely repeats and references the anticipation analysis and then concludes that combinations are obvious because of “improved speed” and “predictable results,” the ALJ committed errors of law. In particular, the ALJ failed to address the inquiries the Supreme Court has mandated in an obviousness analysis. Graham v. John Deere Co., 383 U.S. 1, 17-18 (1966) (identifying the required factual inquiries of establishing the scope and content of prior art, the differences between the prior art and the claims at issue, and the level of ordinary skill in the art). The ALJ erred by grouping references without discussing or distinguishing what they disclose. ID at 212-23. In the lengthy ID, there is also no discussion at all of SCI and Release 4, though the ALJ relies on them in four obviousness combinations. ID at 210-11. This cannot comply with the Graham mandate of

establishing the scope and content of the prior art. The ALJ also erred by grouping claim features (ID at 213 and 220, “double data rate;” ID at 214 and 220-21, “precharge;” ID at 214 and 221, “multiple memory bank;” and ID at 216 and 221 “sense amplifier) without addressing the claimed inventions as a whole. KSR Int’l v. Teleflex, 550 47

PUBLIC VERSION

U.S. 398 (2007) (a determination of obviousness is made with respect to the claimed subject matter as a whole.) The ALJ’s failure to consider the claimed inventions as a whole is especially apparent with respect to Farmwald ’037. The ALJ said it was “important to note” that the only claim element not taught in Farmwald ’037 is the strobe signal and that that claim element is taught by Dan, Yano, NextBus or Harriman. ID at 121, 217-219, 222-23. The isolation of that claim element was legal error. The ALJ also improperly relied on hindsight. Aided by the Barth I claims, he pointed to where the alleged prior art supposedly discloses individual claimed features. In re NTP, Inc., 654 F.3d 1279, 1299 (Fed. Cir. 2011) (reversing an obviousness finding “based on improper hindsight reasoning.”) The ALJ selected isolated features from different pieces of prior art and concluded it would be obvious to modify other prior art with those features. Even if each claimed feature is disclosed in the prior art, however, a conclusion of obviousness does not follow. Rather, there must be a reason prompting a person having ordinary skill in the pertinent art to combine allegedly known features in the way the claimed invention does. See Unigene Labs., 655 F.3d at 1360. Here, the ALJ’s rationale for the combinations is “improved speed” and “faster data access.” ID at 213, 214, 218, 221, 222. That is the same reason the ALJ used to find Dally claims obvious. ID at 153 (“fast data transmission techniques”). It was the same reason the ALJ used to find the Ware claims obvious in the 661 Investigation. 661 ID at 85 (addressing increased operating speeds). In the ALJ’s apparent view, the generic benefit of “increased speed” would likely make any prior art combination in the electronics field obvious. That cannot be correct. And the ALJ’s “improved speed” rationale for the Barth I claims is refuted by the evidence, including the references themselves. For example, NextBus and Harriman disclose slow systems; the alleged strobe there would not increase speed even if it was obvious to use it in a different system. CX-10765C.0131. Likewise, nothing in Dan or Yano suggests that the alleged strobes increase speed even if it was obvious to use them in a different system. CX-10765C.0118-19. The ALJ, however, failed to address these and most of Rambus’s 48

PUBLIC VERSION

arguments and evidence refuting obviousness. Compare CIB at 84-99 and CRB at 43-50, identifying Rambus’s arguments, with ID at 211-212, 219, 223 (addressing Rambus’s arguments). The ALJ’s cavalier approach in finding all Barth I claims obvious, without addressing the claims as a whole or even referencing claim language, by repeating a rationale that would make everything obvious, and without addressing Rambus’s arguments and evidence, should be reversed as contrary to law, fact, and Commission policy. 1. None of the References Disclose or Suggest the Claimed Strobe Signal/Signal Features

None of the NextBus specification, Harriman, Dan, or Yano, disclose the claimed strobe signal/signal features.22 See Section IV.B., supra. Farmwald ’037, SCI (called Kristiansen in Rambus’s prior briefing), and Release 4 do not either. ID at 121-22, 210-23. As a result, the ALJ’s obviousness conclusions are erroneous. 2. Combinations 1-4 Do Not Render the Asserted Barth I Claims Obvious

The ALJ concluded that the Dan or Yano combined with Farmwald ’037, SCI, and/or Release 4 renders obvious certain claims of the Barth I patents. ID at 210-19. The ALJ’s conclusion, however, is unsupported at least because of the fundamental differences between Dan or Yano and the Barth I claims. The Federal Circuit has held that when the prior art and the claimed invention “are different in a way that the [asserted] patent treats as important to the invention” an obviousness holding is improper. Sud-Chemie, Inc. v. Multisorb Techs. Inc., 554 F.3d 1001, 1008 (Fed. Cir. 2009). Here, the Barth I patents treat as important decoupling the timing between a command and

The ALJ stated that “even assuming Rambus is correct that Yano’s data strobe (DS) does not initiate sampling as claimed, . . . it would at least have been obvious to modify Yano so as to achieve the functionality of the claimed strobe signal.” ID at 97 n.13. The evidence cited does not support this conclusion. The ALJ’s reason for modifying Yano’s DS signal—that “an alternative, improved approach to initiate sampling by a memory device, particularly as synchronous systems started overtaking asynchronous systems and the speeds of memory systems increased”—is unsupported and based on hindsight. None of the references suggest that their strobe signal is an “improved approach” to initiate sampling by a memory device. 49

22

PUBLIC VERSION

data transfer through use of a strobe signal. The purported strobe signals of asynchronous interfaces in Dan and Yano do not. Instead, Dan and Yano disclose asynchronous signals that are controlled throughout the duration of the write operation while data flows into memory. This distinction, by itself, demands a conclusion that Dan and Yano cannot properly form the basis of an obviousness finding. The ALJ’s conclusions are also unsupported because SCI has not shown to be prior art and one of ordinary skill would not have combined Yano or Dan with Farmwald ’037, SCI and Release 4 as the ALJ did. a. SCI Is Not Prior Art

Rambus has repeatedly asserted that Respondents have not demonstrated that SCI is prior art. CIB at 90; RIB at 47. The ALJ failed to address Rambus’s arguments. See ID at 212-17. Although SCI has a purported date in 1989, Respondents provided no evidence that it was publicly available or disseminated before the October 1995 filing date of the Barth I patents. See Lister, 583 F.3d at 1311. And SCI does not indicate whether it was published, and if so, where and when. See RX-4278C. The ALJ erred as a matter of in relying on SCI as prior art and as a matter of Commission policy by not addressing Rambus’s arguments. b. The ALJ Erred in Finding that It Would Have Been Obvious to Combine Synchronous and Asynchronous Memory Systems

The ALJ rejected Rambus’s argument that it would not have been obvious to take features from asynchronous systems (such as Dan and Yano) and put them in synchronous systems (such as Farmwald ’037) and vice versa to arrive at the claimed inventions. Relying exclusively on the ongoing reexaminations and testimony from Respondents’ expert about the reexaminations, the ALJ stated that “the proposed asynchronous/synchronous combinations would have been obvious to a person of ordinary skill in the art” because “synchronous systems . . . began overtaking asynchronous systems prior to the October 1995 filing date of the Barth I patents.” ID at 211-12, 214. The ALJ, however, did not address the evidence refuting those combinations, evidence not considered in the 50

PUBLIC VERSION

reexaminations. Respondents’ expert described the transition from asynchronous to synchronous memory devices as a “significant departure.” CX-9731.0494; see also HTr. 1133:17-21. He agreed it was an “important distinction” that synchronous memory devices are “driven directly by a system clock” whereas asynchronous memory devices are not, HTr. 1129:12-24, and agreed that in the timeframe of the Barth I patents, it “would have been simpler” to develop asynchronous memory devices rather than synchronous ones, and that asynchronous systems were cheaper and had better performance. Id. at 1131:2-1133:6. Given these distinctions and the benefits of asynchronous systems at the time, Respondents’ expert refuted the ALJ’s conclusion that “it would have been obvious to combine aspects of asynchronous and synchronous art.” ID at 212. The ALJ also ignored the JEDEC transition from asynchronous to synchronous systems, which took years, supporting a conclusion that combining aspects of asynchronous and synchronous systems was not easy or obvious in 1995 when the Barth I application was filed. Creation of the earliest synchronous JEDEC specification was a long process by many individuals of greater than ordinary skill in the art. CX-10765C.0113. That specification was not published until 1999,

demonstrating that combining aspects of Dan or Yano with aspects of a synchronous system was not obvious in 1995. c. The ALJ Erred in Finding that It Would Have Been Obvious to Incorporate Double Data Rate Features into the Systems of Yano and Dan

The ALJ found that although the “double data rate, or ‘DDR,’ elements of the asserted Barth claims may not be disclosed in Yano or Dan,” they would have been obvious in view of Farmwald ’037 and SCI. ID at 213-14. In finding these individual claim elements obvious, the ALJ asserted that “Yano, Dan, Farmwald ’037, and SCI each sought to address the well-known need in the industry for faster data access operations, and creating faster data speeds without increasing the number of input/output pins by sampling two pieces of data during a single cycle of a clock or other 51

PUBLIC VERSION

periodic signal would be a desirable result.” Id. at 213. But the memory devices in Yano and Dan are asynchronous and do not receive a clock signal. See RX-4261.0007; RX-4262.0010; CX10765C.0029, 34. It would not have made sense to incorporate synchronous DDR clock features from Farmwald ’037 or SCI into the memory devices of Yano and Dan when they do not even receive a clock signal. CX-10765C.0109-114 at 108-13. That would have required a fundamental redesign of memory interfaces and memory devices not addressed by the ALJ. CX-10765C at 11112. Switching from an asynchronous to a synchronous memory device was no simple matter. It was, as Respondents’ expert said, a “significant departure.” CX-9731.0494; HTr. 1133:17-21. The ALJ’s analysis of SCI also fails because it is not prior art, something not addressed by the ALJ. ID at 212-19; CX-10765C.0113-14. Regardless, one of ordinary skill would not have combined SCI with Yano or Dan because SCI does not describe sending data to memory devices. It describes sending data to computers and processors. Id. The ALJ did not explain how to modify SCI to be used with a memory device. He did not explain why one would be motivated to combine a “radically different interconnect system” of SCI (RX-4278C.0001) with Yano or Dan. He did not substantively address SCI at all. ID at 212-27. The ALJ thus erred in concluding that it would have been obvious to incorporate DDR clock features into the memory of Yano and Dan, which use no clock. d. The ALJ Erred in Finding it Obvious to Incorporate Precharge, Bank Selection Information, and Sense Operation Features into Dan and Yano

The ALJ found that the claimed precharge, bank selection information, and sense operation would have been obvious in view of Farmwald ’037 and Release 4, stating that it would have been obvious to incorporate those features into Yano or Dan to improve “efficiency” and/or “speed” and because synchronous systems began overtaking asynchronous systems. ID at 214-17. This

reasoning fails because it would not have been obvious to combine aspects of asynchronous and synchronous systems. Moreover, nothing in the record supports that the claimed precharge, bank 52

PUBLIC VERSION

selection information, and sense operation features would solve problems Yano and Dan were trying to solve or make them more efficient or faster. The ALJ also overlooked evidence showing that it would make no sense to incorporate those features into Yano or Dan. Asynchronous SRAMs of Yano and Dan have no need for precharge. CX-10765C.0115. At the time of Yano and Dan, their asynchronous signals could only access one memory bank at a time, making the addition of multiple banks and bank selection information meaningless.23 CX-10765C.0116-17. The ALJ thus erred in finding that it would have been obvious to incorporate precharge, bank selection, and sense operation features into Yano and Dan. 3. Combinations 5-6 Do Not Render the Barth I Claims Obvious

The ALJ found that it would have been obvious to incorporate the strobe signals from Yano and Dan into Farmwald ’037 to enable “faster data access operations” and to overcome “the inherent skew associated with various clocking schemes.” ID at 217-19. But one of ordinary skill would not have incorporated the alleged strobe signals from the slower asynchronous systems of Yano and Dan into the already faster systems of Farmwald ’037 to enable “faster data access operations.” Farmwald ’037 teaches systems that operate at 500 MHz and was already much faster. CX-

10765C.0119; RX-4269 at 6:15-18. Farmwald ’037 also had a mechanism for dealing with skew, and the ALJ does not explain why or how this mechanism would have been replaced by anything in Yano or Dan. CX-10765C.0119. Further, nowhere do Dan and Yano disclose or suggest using their purported strobes as a solution for increasing speed or dealing with skew. RX-4264.0014; CX10765C.0118-19. No reason exists for the ALJ’s combinations than impermissible hindsight.

The ALJ also relies on inherency for the claimed sense operation features. But Yano and Dan do not inherently disclose the claimed codes. RX-4261; RX-4262; CX-10765C.0117. One of ordinary skill would understand these codes to be “one or more bits” in the context of the Barth I patents. Id. The assertion of asynchronous signals, such as the AS signal in Yano, is not one or more bits for the same reasons explained regarding the R/W signal of Yano. 53

23

PUBLIC VERSION

4.

Combinations 7-12 Do Not Render the Barth I Claims Obvious

The ALJ also concluded that various combinations of NextBus, Harriman, Farmwald ’037 and Release 4 render obvious claims of the ’405 and ’109 patents. ID at 219-23. He based these conclusions on the same reasons provided for the combinations based on Yano and Dan. Id. As discussed above, however, those reasons are unsupported. In addition, the ALJ did not address evidence and arguments supporting non-obviousness. CIB at 91-100; CRB at 48-51. For example, the ALJ failed to address why it would have been obvious to implement the alleged DDR features in NextBus or Harriman or even that such a modification was possible. Indeed, the fact that the NextBus and Harriman systems ran at 25 MHz (at most) is strong evidence that they would not have functioned at twice that rate. CX-

10765C.0131. If they could have operated twice as fast, the clock frequency simply would have been 50 MHz instead of 25 MHz. And if confronted with a system that could have operated at 50 MHz, but was limited because it only had a 25 MHz clock, one of ordinary skill would have simply increased the clock frequency to 50 MHz and would not have looked to modify the system to use both edges of the clock signal (add DDR functionality) since there would have been no impediment to increasing the clock frequency. Id. at 123. Further, NextBus involved communications over a backplane bus between a CPU and a board. Id. at 46-47. The CPU does not keep track of low-level information necessary for issuing a precharge operation, providing bank selection information, or issuing a code specifying a sense operation. Id. at 79, 123-26. The claimed precharge operation, bank selection information, and sense operation are irrelevant to such a backplane bus or bus 209 of Harriman. Id. at 124-27. In addition, under the ALJ’s construction, a “memory device” in NextBus and Harriman is a board with a memory controller. ID at 109. That board would issue the codes or bank selection to itself. CX-10765C.0124-27. Dr. Jacob agreed this is non-sensical, testifying that it “doesn’t make sense” for the structure that is “performing the methods described in the claims [to] reside[] within 54

PUBLIC VERSION

the memory device.” HTr. 1099:20-24. Finally, the ALJ’s reasons why the purported strobe signal of NextBus or Harriman would have been incorporated into Farmwald ’037 are unsupported. He asserted that it would have been obvious to do so because “all the references sought to address the well-known need in the industry for faster data access by overcoming the inherent skew associated with various clocking schemes.” ID at 222. But the ALJ ignores that Farmwald ’037 operates at 500 MHz—twenty times faster than NextBus and Harriman. CX-10765C.0130-31; RX-4269 at 4:44-46; RX-4265.0013. He does not explain why one of ordinary skill would have looked to such slow systems to improve the speed of the much faster of Farmwald ’037 systems. CX-10765C.0131 The ALJ also asserted that combinations with Farmwald ’037 would “improve the efficiency and speed . . . by indicating to the target memory device that data is on the bus, thereby enabling higher-frequency data access rates.” ID at 222. But Farmwald ’037 already knows when data is on the bus, for example, through the use of a delay value stored in an access time register, enabling speeds well beyond NextBus and Harriman. RX-4269.0022 at 6:40-45; CX-10765C.0130-31. The ALJ ignored this point. ID at 223. He also did not address that Respondents’ expert acknowledged that he is unaware of any memory devices at the time that could implement the protocol of the NextBus specification. HTr. 1121-22. Simply put, the ALJ’s reasons for the combinations of the NextBus specification, Harriman, Farmwald ’037 and Release 4 are contrary to the evidence and do not make sense. 5. Secondary Considerations Support a Finding of Non-Obviousness

The ALJ also failed to properly consider the evidence of secondary considerations. ID at 226-27. He asserted that the evidence of secondary considerations does not overcome the strong showing of obviousness. Id. But as discussed above, the ALJ’s obviousness conclusions are legally and factually incorrect. The ALJ also asserted that the secondary considerations evidence was just testimony of Dr. Przybylski. ID at 226-27. This is untrue; specific examples were provided, CX55

PUBLIC VERSION

10765C at 132-33, in addition to the infringement findings in the 661 Investigation and the Staff’s statements there that “the evidence adduced at trial shows that secondary considerations of nonobviousness, including at least commercial success and praise by others, further demonstrates that the asserted ’353 patent claims are not obvious.” 661 SIB 31; CIB at 101-02; RIB at 51. That remains true now. V. THE BARTH I PATENTS ARE NOT UNENFORCEABLE FOR UNCLEAN HANDS The ALJ’s unclean hands decision distorts the spoliation doctrine, rests on legal error and clearly erroneous findings of fact, and establishes an irrebuttable presumption of prejudice that precluded Rambus from addressing the degree of prejudice, much less the complete lack of prejudice, stemming from the destruction of documents in 1998 and 1999. In particular, the ID broadens the scope of anticipated litigation for spoliation purposes far beyond the bounds of any precedent, holding that Rambus reasonably foresaw litigation on its asserted Barth I patent applications years before: (1) those applications were amended to claim features incorporated into Respondents’ accused products; (2) those applications issued as patents, (3) the accused products were marketed, and (4) Rambus considered litigating the Barth I patents. The ALJ’s misconceptions regarding the scope of Rambus’s duty to preserve fueled a flawed analysis of bad faith that ignores objective evidence and rests on adverse credibility determinations grounded on “facts” that are demonstrably wrong. For example, the ALJ stated that “there is no written evidence that Rambus relied on advice from outside counsel to form the document retention policy” when there are numerous documents showing that the outside attorneys who helped formulate Rambus’s licensing and litigation strategy also recommended, provided the template for, and helped implement Rambus’s document retention policy. The ID contains no analysis of how Respondents were prejudiced by the alleged destruction of Barth I documents. The one-sentence explanation given—that internal Barth emails were

destroyed that could have supported invalidity contentions necessarily rooted in publicly available 56

PUBLIC VERSION

prior art or the four corners of the specification—is nonsensical, legally flawed, and factually baseless. Likewise, the ALJ erred when he rejected Rambus’s efforts to demonstrate the lack of prejudice through the enormous amount of preserved documents as “misdirection” and “irrelevant.” Those documents demonstrate a lack of prejudice and are, at the very least, relevant to mitigate the degree of prejudice suffered as they contain the kinds of information that would be expected to be reflected in the allegedly destroyed documents. An objective analysis of the evidence shows that Rambus did not have a duty in 1998 or 1999 to preserve the Barth I related documents that the ALJ identified as supporting prejudice and dismissal. And even if such a duty existed, the minimal degree of prejudice suffered as a result of the alleged destruction, the mitigated degree of bad faith once errors are corrected, and the availability of adequate lesser sanctions do not warrant the ultimate sanction of dismissal. A. The ALJ Erred in Conflating the Determination of Whether Rambus Anticipated Litigation of the Barth I Patents with Rambus’s Anticipated Enforcement of the Farmwald Portfolio

Spoliation must be determined on a case by case basis. “[S]poliation refers to the destruction or material alteration of evidence or to the failure to preserve property for another’s use as evidence in pending or reasonably foreseeable litigation.” Micron Tech., Inc. v. Rambus Inc., 645 F.3d 1311, 1320 (Fed. Cir. 2011) (citation omitted). “Thus, ‘a party can only be sanctioned for destroying evidence if it had a duty to preserve it.’” Id. (citation omitted). The duty to preserve arises when litigation is pending or “reasonably foreseeable,” a “flexible fact-specific standard.” Id. The duty extends only to documents that the party knows or reasonably should know could be used as evidence in the foreseeable litigation. Id.; Kronisch v. United States, 150 F.3d 112, 126 (Fed. Cir. 1998). The ALJ erred in failing to assess whether Rambus had a duty to preserve Barth I related documents during the period of alleged destruction. The ALJ presumed that the allegedly destroyed documents at issue in this case would have been “relevant to both its Farmwald-Horowitz litigations 57

PUBLIC VERSION

on DRAMs and this subsequent Barth litigation on DRAM controllers,” and thus within the scope of the Farmwald litigation preservation duty. ID at 256. But these litigations fundamentally differ. The Farmwald and Barth I patents name different inventors, involve different technologies, and are distinct in time. The parent Barth I application was filed five years after the original Farmwald application. No evidence suggests that documents related to the Barth I patents fell within the scope of Rambus’s duty to preserve documents for the Farmwald litigations against DRAM manufacturers. Here, the ID’s dismissal sanction rests entirely on the supposed prejudice suffered from the alleged loss of emails concerning the Barth I patents and documents relating to the prosecution of the Barth I patents. ID at 312-14. But Rambus had no duty to preserve such Barth I documents simply because Farmwald litigation might have been foreseen. Respondents did not argue, and the ALJ did not find, that such documents were relevant to the Farmwald litigations. No basis for such a finding exists. Accordingly, that Rambus had a duty to preserve documents for the Farmwald litigations does not support a finding of spoliation here and the dismissal sanction cannot stand absent an independent finding that Rambus had a duty to preserve the specific documents upon which the ALJ grounded his finding of prejudice. B. The ALJ Erred in Finding that Litigation on the Barth I Patents Was Reasonably Foreseeable at Least by July 1998

The ALJ’s holding “that Rambus not only could foresee, but was planning litigation involving the Barth Patents as early as July 1998” obliterates the distinction between anticipating litigation generally, which does not give rise to a duty to preserve, and anticipation of a specific litigation, which does give rise to such a duty. ID at 257. Arthur Andersen LLP v. United States, 544 U.S. 696, 704 (2005), set that distinction and the Federal Circuit’s ruling in Micron, 645 F.3d at 1319-20, reinforced it. The ALJ’s holding of anticipated litigation rests on the notion that Rambus should have preserved documents relating to the Barth I patent applications simply because they were pending and part of Rambus’s “long-term” portfolio at a time when Rambus was considering 58

PUBLIC VERSION

licensing and enforcement of a different family of patents. This was clear legal error, antithetical to Andersen and Micron. Arthur Andersen, 544 U.S. at 704 (“‘Document retention policies,’ which are created in part to keep certain information from getting into the hands of others . . . are common in business” (citation omitted)); Micron, 645 at 1322 (“[L]itigation is an ever-present possibility in American life.” (citation omitted)). 1. The ALJ’s Broad Interpretation of Anticipated Litigation Contradicts Established Authority

The ALJ’s finding that Rambus anticipated litigation at least by July 1998 defies the Federal Circuit’s holding in Micron and establishes a standard for reasonable foreseeability of litigation that subjects virtually all patent holders to a duty to preserve commencing with prosecution. The Micron court declined to find that Rambus anticipated litigation on its Farmwald patents by December 1998. Micron, 645 F.3d at 1321-22. The ALJ’s finding that Rambus anticipated litigation across all patents and pending applications by July 1998 (ID at 257-58) is clearly erroneous. Indeed, Rambus had no “duty to preserve documents from the mere existence of a potential claim or the distant possibility of litigation.” Id. at 1320. Rather than affirm a finding that Rambus anticipated litigation on the Farmwald patents in December 1998, Micron framed the issue as a “binary” question: “the question this court must answer is whether the district court clearly erred when it determined that, at some time before the second shred day in August of 1999, litigation was reasonably foreseeable.” Id. at 1322. In answering no, the Micron court noted that “[i]n June 1999, the first [Farmwald] patent in suit issued” providing Rambus a “loaded gun,” that “[t]he first steps toward this litigation were spelled out on June 27, 1999,” and that “on July 8, 1999, Rambus’s outside litigation counsel . . . prepared a timeline for the proposed patent infringement suits showing that Rambus planned to file complaints on October 1, 1999.” Id. at 1318, 1323-24. Thus, the court relied heavily on events occurring between December 1998 and August 1999 to find Rambus anticipated litigation on the Farmwald 59

PUBLIC VERSION

patents by August 26, 1999. Id. at 1323. The ALJ erred in using the exact same evidence to find that Rambus anticipated litigation more than a year earlier on the Barth I patents when they were not even part of Rambus’s licensing and litigation strategy at that time. 2. No Evidence Supports the ALJ’s Finding that Rambus Adopted a “Long Term Strategic Litigation Plan” Based on the Barth I Patents a. That Rambus Sought to Create a “Patent Minefield” Does Not Evidence an Intent to Litigate the Barth I Patents

The ALJ clearly erred in concluding that “[a]lthough the Barth Patents were not mature during the relevant time period, it is evident that Rambus intended to use the Barth Patents as part of its long term litigation plan: the Barth Patents were part of Rambus’ ‘patent minefield.’” ID at 265; see also id. at 261, 263, 298. Having or desiring a patent minefield does not equate to an intent to litigate, let alone reasonable anticipation of litigation. In stark contrast to the ALJ’s heavy reliance on a “patent minefield,” the Federal Circuit makes no mention of Rambus’s “patent minefield” in concluding that Rambus reasonably foresaw litigation before August 26, 1999, despite the district court having repeatedly cited the existence of the “minefield” as support for that conclusion. Compare Micron, 645 F.3d 1311, with Micron Tech., Inc., v. Rambus, Inc., 255 F.R.D. 135, 138, 150 (D. Del. 2009). The ALJ focused on what he deemed a “long term litigation plan” as supposedly reflected in the “minefield” while accepting that “during the relevant time period, [Rambus] did not foresee near term litigation for the Barth Patents.” ID at 265. Without the flawed “minefield” linchpin, the ID fails to establish anticipated litigation of the Barth I patents during the relevant period. While any patent could ultimately be enforced, the documents show that Rambus relied on the minefield for defensive purposes. The ALJ incorrectly states that the “natural reading of ‘prepare a patent minefield’ is that Rambus wanted to prepare multiple patents for offensive use,” equating “offensive” with litigation. ID at 261 (referring to RX-448C and RX-468C). But the cited

documents show that Rambus intended its patent position to serve as leverage in securing industry 60

PUBLIC VERSION

wide adoption of Rambus’s flagship RDRAM technology and as a deterrent to Intel’s considering alternatives to RDRAM. In 1996, Intel embraced Direct RDRAM as the memory interface for its next generation of microprocessors. CX-3710C (1996 Intel Letter of Intent); JX-108C.045-46. And the 1997 email cited by the ALJ reflects this deterrent role: “we should meet with intel . . . to get them aware that IF they were to consider a DDR chipset that there is a minefield of 60+ Rambus patents that would have to be avoided.” RX-468C. Even if a “patent minefield” were pertinent to the anticipation of litigation inquiry, the ALJ erroneously held that “[t]his patent minefield was clearly applicable to Barth controller patents.” ID at 261. In the time of the patent minefield discussions (1996 and 1997), there were no Barth I patents, much less Barth I controller patents, only applications. And in citing RX-2115C for the proposition that “Rambus’s patents focused on expanding its claim scope to cover ‘auto-precharge’ that ‘would have high harassment value’ against controller manufacturers,” the ALJ again conflated events involving Farmwald patents and applications with Barth I patents. ID at 261. RX-2115C (sent on April 3, 1995, forwarding an email date June 16, 1994) predates the original Barth I application (filed on October 19, 1995) and relates to a potential “overlooked patent claim” in “the original Rambus patent application” (i.e., the 1990 Farmwald application). RX-2115C focuses on obtaining patent claims for “leverage over [competing technologies]” in the competition for industry adoption of next generation memory systems, not litigation. b. That Rambus’s Document Retention Policy Was Not Patent Specific Does Not Show an Intent to Litigate Barth I Patents

The ALJ also misapplied Micron in holding that litigation on the Barth I patents was reasonably foreseeable because “it is clear that Rambus’s wide-spread document destruction was not patent specific.” ID at 264; see also id. at 258-59, 272. That makes no sense. While implementing a patent specific policy could evidence an intent to litigate those patents, implementing a general policy that treats all categories of documents in the same manner, regardless of content, demonstrates 61

PUBLIC VERSION

a desire to implement a standard content-neutral document retention policy, not an intent to conduct litigation involving all documents impacted by the policy. That alone does not evidence a specific intent to litigate and “frustrate the fact-finding efforts of parties adverse to Rambus” (ID at 257), absent other evidence to the contrary. Arthur Andersen, 544 U.S. at 704. The ALJ concluded that “Rambus’ implementation of a document retention policy as an important component of its general (i.e. not patent-specific) litigation strategy makes it more likely that litigation was reasonably foreseeable. See Micron, 645 F.3d at 1322.” ID at 259. But to read Micron as holding that Rambus must have anticipated litigation on all of its patents and pending and future patent applications, when it adopted its document retention policy in July 1998, distorts Micron and contravenes Arthur Andersen. 3. The Totality of the Circumstances Regarding the Barth I Portfolio Differs Substantially from Those of the Farmwald Portfolio

The ALJ clearly erred in holding that Rambus anticipated litigation on whatever claims might ultimately issue from the nascent Barth I patent applications based on products the Respondents might produce years later that might incorporate features of a JEDEC standard that did not exist in July 1998. Indeed, key facts that the Micron court relied on to hold that it was not clear error to find that Rambus reasonably foresaw litigation of the Farmwald patents by August 1999, do not exist for the Barth I patents. a. The ALJ Clearly Erred in Dismissing the Importance of the Fact the Asserted Barth I Patents Were Years from Issuing When Rambus Adopted Its Document Retention Policy

The ALJ erred in dismissing the importance of the nonexistence of the asserted Barth I patents in July 1998. ID. at 256 (“[T]he Federal Circuit did not set forth the narrow requirement reasonable foreseeability to litigation based on issued patents.”). Whether issued patents upon which Rambus could assess infringement existed is an important factor in evaluating the totality of the circumstances in determining whether Rambus anticipated litigation on the Barth I patents. Micron, 62

PUBLIC VERSION

645 F.3d at 1323. The Federal Circuit declined to affirm the December 1998 anticipation of litigation date and considered only whether litigation was anticipated by August 26, 1999, enabling the analysis to include issuance of the first asserted Farmwald patent and Rambus’s “first steps toward [Farmwald] litigation.” Id. Instead of conducting a similar analysis for the Barth I patents, the ALJ erroneously couched the “spoliation inquiry” as “broader” and assessed “whether Rambus reasonably foresaw litigation” in the “long term.” ID at 256, 260-61, 264-65. In turn, the ALJ repeatedly cited facts and actions pertinent only to the Farmwald litigations to justify holding that Rambus had a duty to preserve Barth I documents, while ignoring or dismissing without justification key facts demonstrating that litigation over the Barth I patents was not reasonably foreseeable. That constitutes clear error. b. The ALJ Clearly Erred in Equating Considered Enforcement of the Farmwald Patents with Anticipated Litigation of the Barth I Patents

The unrebutted documentary evidence demonstrates that the Barth I patents were not considered among its strategic patent portfolio for enforcement in 1998 or 1999, which then included only non-Farmwald patents. i. The ALJ Ignored Unrebutted Evidence Demonstrating a Distinction Between Rambus Patent Families

The ALJ erred in concluding that Rambus had a “strategic plan” to litigate all its patents and future patents because “[t]he distinction between using one family of patents over another is not one that was ever mentioned in the surviving documents of Rambus.” ID at 272-73. The documents reflect fundamental differences between Rambus’s activities involving its original strategic patent portfolio, which related only to the Farmwald patents, and its second strategic patent portfolio, which began in 2000 and focused on non-Farmwald patents and applications, including the Barth I patents. Rambus was founded based on the pioneering inventions of Dr. Michael Farmwald and Dr. Mark Horowitz embodied in an application filed April 18, 1990. Throughout the 1990s, Rambus saw 63

PUBLIC VERSION

competitors and others incorporate Rambus’s inventions into their products or proposed standards. Rambus became concerned that the issued Farmwald patents were not capturing the full scope of their inventions. HTr. 1328. Rambus thus undertook an initiative called strategic patent prosecution where it sought to ensure that it filed claims covering its inventions described in the 1990 specification, including the inventions being taken by others. CX-10771C.0038; RX-535C.0024. The strategic patent portfolio, called SPP (and later SPP1), is the subset of patents and applications that arose from this initiative. CX-10771C.0038. SPP1 only involved applications in the family of the 1990 Farmwald application. CX-10771C.0038-39; HTr. 2025-26; JX-131C.0114, 226. In

January 2000, it was reported that Rambus’s patent prosecution had “Gone Through Major Quality Upgrade In 1999” due to the “Success of SPP1.” CX-10504C.0009. On January 18, 2000, it was recommended that Rambus “Start Prosecution Efforts for SPP2.” CX-10504C.0010; CX-10771C.0039; HTr. 2026. SPP2 involved analyzing and classifying patent applications not included in SPP1, and strengthening those portfolios through “market-directed prosecution.” CX-10504C.0011; CX-10771C.0039-40. SPP2 included “applications filed after the initial—the original filing of the Farmwald application,” including certain Barth I patents. JX131C.0226; CX-10771C.0039-40. The ALJ erred in overlooking the distinction between SPP1 and SPP2, which demonstrates that Rambus did not contemplate strategic prosecution, let alone enforcement, of the Barth I Patents until well after the alleged destruction. ii. The ALJ Clearly Erred in Attributing Activities Involving SPP1 to the Barth I Patents

The ALJ disregarded the fundamental distinction between Rambus’s efforts involving SPP1 and SPP2 in holding that Rambus planned a “litigation campaign” that included the Barth I patents. ID. at 261-63. The ALJ cited Joel Karp’s hire and documents from the 1998 and 1999 time periods as evidence of Rambus’s intent to litigate the Barth I patents. Id. None of those documents mention Barth I patents or applications. Instead, key evidence from that time reflecting the scope of 64

PUBLIC VERSION

Rambus’s licensing and enforcement strategy shows that Rambus was concerned about licensing or enforcing only the Farmwald patents that formed SPP1: • Rambus’s efforts to strategically prosecute patents to cover the full scope of its inventions that were being considered for use by JEDEC during the relevant period were limited to the Farmwald patents (see Section V.A.3.C, supra); • Multiple documents, including the October 1998 Strategy Update upon which the ALJ relies (ID at 261), show that Rambus grounded its non-compatible licensing strategy on its original strategic portfolio, i.e., those Farmwald patents and applications claiming priority to the original 1990 application (see, e.g., RX-517); • None of the claim charts prepared by Mr. Karp evaluated patent claims outside the Farmwald portfolio; • When Rambus eventually instituted litigation against DRAM manufacturers in 2000 it asserted only Farmwald patents, even though other patents in Rambus’s portfolio had already issued, including patents in the Barth family (not asserted here); • When Rambus initiated licensing discussions with a controller manufacturer (nVidia) it included only Farmwald patents in its presentation (RX-417C); • Rambus did not begin strategically prosecuting patents outside the Farmwald portfolio until 2000; Respondents’ admitted that Rambus did not seek to cover the DDR standard until May 2000, when it filed the application that became the asserted ’353 patent. RIB 223 (filing of the ’353 patent claims “was prompted by the impending final release in June 2000 of JEDEC Standard No. 21-C Ver. 9”). Thus, all of the documentary evidence reflects Rambus’s singular focus on the Farmwald patents as the basis of its licensing and litigation strategy in 1998 and 1999. The ALJ’s statement that “[w]hile the Barth Patents may not have been mature, they were a

65

PUBLIC VERSION

part of a long term strategic litigation plan” rests on clearly erroneous factual findings. ID. at 260. The only cited support is RX-517C, a presentation from October 1998 entitled “Strategy Update,” and Mr. Karp’s testimony regarding that presentation. ID at 261. The document itself refutes any connection between this presentation and the Barth I patents. On its face, the presentation relates only to Rambus’s “Strategic Portfolio” (i.e., those application based on the original 4/18/90 Farmwald application) and the “top priority” involving claims covering SDRAM, DDR, SLDRAM, and controllers referred to by the ALJ falls under the heading of “File continuation cases based on 4/18/90 filing.” RX-517C.001-002. Thus, this “Strategy Update” does not relate to the Barth I patents, much less support the notion that those Barth I patents were part of a “strategic litigation plan” at that time. Mr. Karp’s responsibility for overseeing Rambus’s entire patent portfolio does not, as the ALJ seems to suggest (ID at 262, 272), demonstrate that Rambus intended to litigate every patent and patent application in its portfolio. That is especially true where the documents discussing Rambus’s enforcement strategy in 1998 and 1999 do so solely in the context of the Farmwald portfolio and where Rambus revealed the scope of its enforcement strategy in the original DRAM litigations by asserting only Farmwald patents, even though it had other issued patents in its portfolio. Indeed, the ID is devoid of any evidence that Mr. Karp prepared a licensing or litigation strategy involving the Barth I portfolio, evaluated potential competitor products against the pending claims of the Barth I applications, or otherwise considered that portfolio for strategic enforcement. Overseeing every day patent prosecution cannot evidence anticipated litigation, particularly with respect to patent applications that did not then exist. c. The ALJ Clearly Erred in Confusing Notice of Potential Infringement of the Farmwald Patents with Potential Infringement of the Barth I Patents

The ALJ erred in finding that “Rambus was on notice of potentially infringing activities by particular manufacturers,” an important factor in assessing whether Rambus anticipated litigation of 66

PUBLIC VERSION

the Barth I patents. ID at 259; Micron, 645 F.3d at 1323. Rather than find actual notice of potential infringement of the Barth I patents, the ALJ’s finding of spoliation hinges on Rambus’s notice of potential infringement of the Farmwald patents and gross speculation as to what might occur in the “long term” future regarding the Barth I patents. ID at 256-65. Indeed, the ALJ framed what he deemed the “proper question” as whether “Rambus knew that particular parties were engaging in activities that Rambus could ultimately use to show infringement.” ID at 259-60. That broad inquiry departs from Micron, which held that the reasonable foreseeability “standard does not trigger the duty to preserve documents from the mere existence of a potential claim or the distant possibility of litigation.” 645 F.3d at 1320. Several key facts demonstrate that Rambus did not have notice of potential infringement of the Barth I patents before 2000. First, none of the asserted Barth I patents had even been filed (the earliest one was filed on May 1, 2000). Second, Rambus had not begun any strategic prosecution of any Barth I application to cover features incorporated into standardized memory in 1998 or 1999, as Respondents admit. See RIB at 21-22, 219-24; HTr. 237-38. Third, the infringing memory

controller products were not yet on the market in 1998 and 1999. CX-8060; CX-8061; CX-8062. Fourth, the JEDEC standard upon which those products are based did not issue until June 2000. The ALJ’s acknowledgement that “the activity may not have been infringing at the time,” ID at 260, vastly understates the situation. There was no activity at the time—infringing or otherwise—given the nonexistence of the products, standard, and patents. Rambus was not on notice of Barth I infringing activity in 1998, and speculation about possible future infringement and enforcement is insufficient to trigger a duty to preserve. Micron, 645 F.3d at 1320. The ALJ clearly erred in finding that Rambus invited infringement of the Barth I patents based on prosecution activities in the mid-1990’s related to JEDEC. First, the ALJ overlooked Respondents’ concession that Rambus did not file Barth I patent applications to cover the JEDEC DDR standard at issue until May 2000. RIB at 21, 223. Second, the ALJ erroneously stated that 67

PUBLIC VERSION

“[f]rom Mr. Barth’s testimony alone, is clear that the information from JEDEC was used to draft the claims of the Barth Patents. (Barth, Tr. 407:17-408:15.)” ID at 260. The cited testimony does not relate to any Barth patent, much less the Barth I patents. Contrary to the ALJ’s assertion that “Mr. Barth was not involved with [the Farmwald patents]” (ID at 277), Mr. Barth had responsibility for coordinating prosecution of all of Rambus’s patents. JX014C.0016, 44, 80-81, 90, 156; CX-7242C; CX-7243C; RX-2469C. It was wrong for the ALJ to assume that every interaction involving Mr. Barth necessarily concerned the Barth I patents. In fact, that testimony concerned RX-4288, an email string from June 1995, that the ALJ cites as evidence that information from JEDEC was used to draft the Barth I claims. ID at 260, 295. But RX-4288 predates any Barth I application (the earliest was filed in October 1995) and relates to reviving prosecution of a Farmwald divisional application (P015D) that may have gone abandoned. RX-4288; RX-3957C.0021 (showing that file P015D relates to a divisional application filed on March 5, 1992, claiming priority to the original Farmwald application).24 Thus, Rambus did not admit to “intentionally broaden[ing] the Barth Patents’ claims to cover JEDEC standard-compliant products” in June 1995. ID at 260. The undisputed evidence shows that such prosecution first occurred in 2000, well after the ALJ found anticipation of litigation and after Mr. Barth left Rambus in 1999. Thus, the ALJ’s conclusion that “[t]he Barth claims read on JEDEC compliant products because Mr. Barth sat with the JEDEC standards, took notes, and drafted claims” (ID at 272) is erroneous. d. The ALJ Clearly Erred in Ignoring Evidence of Rambus’s Efforts to License and Resolve Disputes Without Litigation

The ALJ’s finding that “[t]here is no evidence that Rambus’ efforts focused on preparing a The ALJ’s conclusion that because RX-4288 refers to filing a “divisional” application, it necessarily does not concern the already filed Farmwald patents reflects a fundamental misunderstanding of patent prosecution. ID at 296. Indeed, a “divisional application” is “[a] later application for an independent or distinct invention, carved out of a pending application and disclosing and claiming only subject matter disclosed in the earlier or parent application.” MPEP 201.06. Because this email predates the original Barth I application by several months, it cannot possibly be referring to any Barth I patent or application. 68
24

PUBLIC VERSION

licensing team, evaluating negotiating strategies, or considering ‘peaceful’ action” defies the record. ID at 263. Indeed, it is inconsistent with the ALJ’s findings on domestic industry and Respondents’ admission that “Rambus’ principal business consists of licensing.” ID at 355. Further, Rambus did not abandon its commitment to licensing and developing industry partnerships when infringing products were introduced. For instance, peacefully resolving an infringement dispute with Mosys in 1996, Rambus made its position and desire to avoid litigation clear: “our position is they need a license and we want to meet asap to work out a reasonable deal . . . our strategy is to avoid any legal battle, . . . have them take a license . . . . [and] make the point that rambus has a lot of IP on high bandwidth memory devices . . . that rambus is willing to license under reasonable terms.” CX-3703. Rambus’s desire for peaceful resolution is evident in its dealings with controller companies in particular. Rambus negotiated Technology Licenses with Respondents nVidia and LSI during the relevant time period. CX-3649C (1998 nVidia license); CX-55C (1994 LSI license). And Rambus devoted substantial resources over many years attempting to develop mutually beneficial business relationships and license non-compatible controller products before first filing suit on the Barth I patents in 2008. RX-417C (April 2000 licensing presentation to nVidia covering only Farmwald patents); CX-3775C; CX-3778C; CX-3791C; CX-3792C; CX-3795C; CX-3798C through CX3829C; CX-3831C through CX-3835C; CX-3840C through CX-3848C; RX-806C.001-02; RX797C; CX-3850C; CX-3851C; CX-3852C. Accordingly, the “delay in initiating litigation on the Barth Patents” was not “a strategic part of [Rambus’s] litigation plan.” ID at 264. Ten years elapsed between the ALJ’s anticipated

litigation date of July 1998 and first litigation on the Barth I patents in 2008 because Rambus was committed to developing business partnerships through licensing and joint product development. Indeed, during that time, Rambus successfully negotiated non-compatible license agreements with at least eight controller companies—agreements that the ALJ and the Commission agree establish a domestic industry on the Barth I patents. ID at 347-48. The ALJ’s assessment that in 1998 Rambus 69

PUBLIC VERSION

was intent on litigating patents that were years away from issuing against products that were years away from being introduced and had no interest in licensing or “peaceful action” is clearly erroneous. C. The ALJ Erred in Finding Bad Faith Destruction of Barth I Documents

The ALJ’s bad faith determination similarly hinges on multiple errors of law and fundamental misunderstandings of fact. Document destruction is in “bad faith” only if it is

established that “the spoliating party ‘intended to impair the ability of the potential defendant to defend itself,’” that is the destruction is “intentional, and indicates fraud and a desire to suppress the truth.” Micron, 645 F.3d at 1326 (citation omitted); Gumbs v. Int’l Harvester, Inc., 718 F.2d 88, 96 (3d Cir. 1983)). Assessing bad faith requires an “exacting standard.” Micron, 645 F.3d at 1327. A bad faith finding “must be made with caution” and cannot be based on an inference of what Rambus “should have known” about the potential materiality of destroyed documents. Id. The ALJ failed to proceed with “caution,” or apply the requisite “exacting standard,” in finding bad faith based on (1) the same fundamental errors of law and fact made in assessing anticipation of litigation, (2) demonstrably wrong assertions that “there is no written evidence that Mr. Karp relied on advice from outside counsel to form the document retention policy” (ID at 294), (3) a misunderstanding of patent prosecution that compelled an erroneous conclusion that Mr. Barth was a copyist, and (4) adverse credibility determinations infected by these erroneous views of law and fact, including an irrelevant critique of the testimony of Rambus’s technical expert. These and similar errors of law and fact formed the backbone of the ALJ’s bad faith determination. On review, the Commission “need not accept any of the ALJ’s findings, ‘even if the ALJ’s findings rested on his evaluation of the credibility of witnesses.’” Deere & Co. v. Int’l Trade Comm’n, 605 F.3d 1350, 1358 (Fed. Cir. 2010) (quoting Kay v. Fed. Commc’ns Comm’n, 396 F.3d 1184, 1189 (D.C. Cir. 2005)).

70

PUBLIC VERSION

1.

The ALJ Legally Erred in Concluding that Supposed Bad Faith in the Farmwald Litigations Justified Finding Bad Faith Here

The ALJ legally erred in concluding that “bad faith can be found even if Rambus did not necessarily intend to prejudice these parties in this investigation.” ID at 265. The proper inquiry is “whether Rambus intended to impair the ability of the potential defendant to defend itself” Micron, 645 F.3d 1326. Bad faith in other cases does not justify imposing a “heavy burden” on Rambus to demonstrate lack of prejudice when no evidence exists suggesting Rambus intended to prejudice a future defense to claims of infringement of the Barth I patents. Id. at 1327. The ALJ erroneously found bad faith because adoption of a document retention policy was “part of a litigation plan.” ID at 266-68. Even if true, that “litigation plan” related to the Farmwald patents and did not involve the Barth I patents. See Section V.A-B., supra. Rambus had no duty to preserve Barth I related documents for the Farmwald litigations and was free to adopt a document retention policy and discard documents like the Barth emails cited by the ALJ to establish prejudice. Arthur Andersen, 544 U.S. at 704. a. The ALJ Clearly Erred in Minimizing Outside Counsel’s Role in Implementing Rambus’s Document Retention Policy

The ALJ’s unequivocal assertion that “there is no written evidence that Mr. Karp relied on advice from outside counsel to form the document retention policy nor do his notes or slides indicate the policy came from outside” defies the record. ID at 294. In early 1998, Cooley Godward attorneys, including Dan Johnson, Peter Leal and John Girvin, helped Rambus to develop a licensing and litigation strategy. JX-58C.015, 19-21. On January 13, 1998, Mr. Leal noted in his first meeting with Karp and Tate, Rambus’s CEO, that Rambus requested a licensing strategy and a “litigation strategy” should licensing efforts fail. RX-1107C; JX-77C.004-07. To fulfill that request, Johnson and Girvin met with Karp on February 12, 1998. RX-1107C; RX-484C. During that meeting, Mr. Johnson, in accordance with his standard practice, advised Rambus that it and that it 71

PUBLIC VERSION

needed to JX58C.020-21; see also JX58C.019, 22-26, 64-66; JX60C.476-81; RX-484C. Cooley counsel Diane Savage sent Mr. Karp a detailed memorandum on the subject of document retention policies, which Rambus adopted in large part. RX-489C; JX-60C.277, 321-22; 508-09; HTr. 2010-12, 2016-17; JX-13C.061-62; RX-506C. On multiple occasions Mr. Johnson reviewed draft presentations for the rollout of the document retention policy. CX-3865; CX3866; CX-3867; HTr. at 2019:18-2023:13; JX58C.102. He further participated in the rollout of the policy and gave a presentation of his own directly to Rambus’s managers. HTr. 2017:24-2019-17; RX-493; HTr 2023:14-2024:24; RX-505 (Johnson slides are those without a Rambus logo); JX58C.034-39, 96-102. Mr. Johnson indisputably knew Rambus was developing a licensing and litigation strategy at the time he advised Rambus to adopt a document retention policy and counseled them through development and rollout of that policy. Rambus’s request for a “litigation strategy” should licensing efforts fail prompted Mr. Johnson’s involvement. RX-1107C.002; JX58C.007, 27-29. Those present during Johnson’s initial meeting with Rambus on February 2, 1998, discussed the need to make Rambus “battle ready” and that RX-

484C; HTr. 2009:17-2012:24; JX-058C.052-53, 58. Mr. Johnson opined on the merits of a breach of contract action versus a patent infringement action. RX-484C Moreover, outside counsel, including Johnson, was to Id. Mr. Johnson was aware of each of the primary factors the ALJ relied on in holding that Rambus was anticipating litigation, and Johnson was discussing potential legal actions, all while recommending adoption of a document retention policy. Indeed, Mr. Johnson and Mr. Girvin prepared a RX-485C; HTr. 2012:25-2014:16; JX-58C.107-08. 72

PUBLIC VERSION

Despite the detailed documentary record of Mr. Johnson’s role in both the development of the licensing and litigation strategy and the document retention policy, the ALJ found “Rambus’ claim of reliance on Mr. Johnson misleading” because “Mr. Karp never informed Mr. Johnson that Mr. Tate directed Mr. Karp in 1998 to prepare a litigation strategy for the Board.” ID at 267-68. But from Rambus’s very first meeting with Cooley attorneys an objective was to develop a as noted by Cooley attorney Mr. Leal. RX-1107C; Hynix, 645 F.3d at 1343. Mr. Johnson and Mr. Girvin fulfilled that request by formulating a plan to get “battle ready.” RX-484C. Mr. Johnson drafted the tiered litigation strategy that Karp presented to the Board and participated in the discussion of RX-

485C.; HTr. 2012:25-2014:16. The ALJ cites these very same facts as evidencing anticipation of litigation. ID at 257-58, 262-63. Moreover, Mr. Johnson testified regarding a document evidencing Tate’s directive to what I would have expected.” JX-58C.051. Mr. Johnson and Rambus’s other outside attorneys were fully informed while counseling Rambus to implement a document retention policy and assisting in the rollout of that policy. Neither Mr. Johnson nor Rambus believed that Rambus had a duty to preserve documents in July 1998. Indeed, Mr. Johnson entitled his presentation to Rambus during the rollout of the document retention policy “The Eve of Litigation - The Need for an Effective Document Retention Policy.” RX505C.032. That Rambus relied on Mr. Johnson’s and his colleagues’ counsel must be considered as evidence of Rambus’s subjective good faith, not rejected as “misleading.” That Rambus continued to follow the advice of counsel and instituted a litigation hold in December 1999 before filing the first Farmwald litigation on January 18, 2000, further evidences good faith and should not be ignored. Micron, 255 F.R.D. at 146, 150. that “what I see here is not a whole lot different than

73

PUBLIC VERSION

b.

Rambus’s Farmwald Prosecution Activities Have No Bearing on Bad Faith

The ALJ erred in finding bad faith based on his conclusion that Rambus acted improperly in drafting the Barth claims in 1995 based on JEDEC activities. ID at 271-72, 295-97. First, the JEDEC activity that the ALJ cites relates only to Farmwald patents and could not be indicative of bad faith in this case. See Section V.B.3.C, supra. Second, Rambus acted properly. Indeed, as the ALJ acknowledged in rejecting Respondents’ prosecution laches defense, “the Federal Circuit has twice confirmed that Rambus did nothing improper in pursuing claims that cover a JEDEC-standard or products intended to comply with various JEDEC standards even while Rambus was itself participating in JEDEC as a member.” ID at 253 (citing Hynix, 645 F.3d at 1349); Rambus Inc. v. Infineon Techs. Ag, 318 F.3d 1081, 1096-1105 (Fed. Cir. 2003). While paying lip service to the “per se” law, the ALJ found that “the manner in which Rambus went about this process was, in fact, improper” because “Mr. Barth sat with the JEDEC standards, took notes, and drafted claims,” which made Mr. Barth “less an inventor and more a copyist.” ID at 272, 295. No authority supports the ALJ’s holding that otherwise lawful prosecution activity can demonstrate bad faith. The referenced activity involved filing a divisional application claiming priority to the Farmwald patent application filed in 1990, and had nothing to do with the October 1995 filing of the original Barth I application or any other Barth I application. See Section V.B.3. Thus, the ALJ erred in citing this activity as supporting a bad faith finding. ID at 271-72, 295-97 Moreover, it was JEDEC that copied Rambus’s technology disclosed in the Farmwald application, which published in a PCT application on October 31, 1991. In re Rambus, 2004 WL 390647, *36 (FTC Feb. 23, 2004). To the extent JEDEC was using technology invented by Rambus and disclosed in the original Farmwald application, Rambus was entitled to draft claims to cover it. Kingsdown Med. Consultants, Ltd v. Hollister, Inc., 863 F.2d 867, 874 (Fed. Cir. 1988). That noninventors may have drafted claims based on the original work of the Farmwald inventors is of no 74

PUBLIC VERSION

moment, contrary to the ALJ’s express concerns. ID at 296. Non-inventors (e.g., patent attorneys, patent agents, and technical employees charged with prosecution duties) often draft claims based on what is disclosed in the specification. Thus, the ALJ’s conclusion that this “is not a process of invention, but of entrapment of competitors” is wrong as a matter of law. ID at 296. The ALJ also erred in stating that “Mr. Barth is forthcoming on the use of competitive products as the basis for forming claims in the Barth patents, even on the ideas that were not in the specification at the time.” ID at 298; see also ID at 304-05. The cited testimony relates to whether SyncLink was disclosed to the patent office, which has no bearing on claim drafting. Even the ALJ concluded that SyncLink was not material to the prosecution of the Barth I patent applications, was not withheld with an intent to deceive, and was not anticipatory (ID at 123, 246-48), belying the unsupported notion that claims were specifically drafted to cover SyncLink. 2. The ALJ Clearly Erred in Assessing the Implementation of Rambus’s Document Retention Policy

The ALJ’s erroneous views of Rambus’s prosecution activities, the scope of Rambus’s licensing and litigation strategy, and the role of outside counsel, infected his interpretation of Rambus’s actions in implementing the policy. For example, the ALJ cited that “Mr. Karp instructed Rambus’ engineers to look for things to keep that would help establish conception and prove that Rambus had intellectual property” as supporting selective destruction and bad faith. ID at 268. Any standard document retention policy would require inventors to preserve their invention records. Mr. Johnson was “very strong” on preserving conception documents, and each of the draft slide presentations sent to Mr. Johnson for review included a bullet point that “Information Relating To Patent Disclosures and Proof of Invention Dates Kept Permanently.”25 HTr. 2020:18-2021:3; CX25

Mr. Hampel, one of the Barth I inventors, confirmed that they were supposed to retain proof of invention documents under the retention policy. HTr. 329:1-3. While the ALJ says that “Mr. Hampel also testified that Mr. Karp told Rambus personnel not to keep evidence that would cast doubt on the patentability of an invention” (ID at 268), Mr. Hampel later clarified his testimony on 75

PUBLIC VERSION

3865.009; CX-3866.011; CX-3867.011. That Mr. Karp “sweated bullets” while searching a backup tape for an electronic conception document evidences not “selective execution of the document retention policy,” ID at 268-69, but merely the retention of conception documents required by the document retention policy. RX506.001. The ALJ cites no evidence that Rambus discarded any conception documents, much less that it selected a subset of those documents to discard based on their content (i.e., whether they would ultimately prove “helpful” or “adverse” for Rambus in subsequent litigation). Similarly unwarranted is the ALJ’s conclusion that Rambus sought to “cover up” its policy and document destruction based on preserved documents the ALJ concluded demonstrated an ill intent. ID at 269 (citing RX-493C and RX-539C). Keeping distribution of emails regarding the document retention policy “narrow,” as requested by Mr. Karp, and shifting the task of organizing the second shred day from the heading of “Licensing/Litigation Readiness” to “Database Maintenance” does not evidence ill intent. Indeed, that these supposedly damaging documents still exist demonstrates the opposite. Had Rambus implemented its policy for an improper purpose and covered its tracks, these documents would have been discarded. Instead, these, and myriad other documents reflecting Rambus’s

proposed licensing and litigation strategy, remain for scrutiny and mischaracterization after two shred days the ALJ believes were designed to rid Rambus of adverse documents. 3. The ALJ Erroneously Dismissed Relevant Evidence as Irrelevant and Mischaracterized Relevant Arguments as “Misrepresentations” a. Arguments and Evidence Presented Regarding the Scope of Destroyed Documents Were Not Misleading

Rambus’s arguments regarding “pizza boxes and phone books” in the 661 Investigation and employment records in this investigation were relevant, contrary to the ALJ’s assertions. ID at 310

this point, indicating that, while people were told not to create these kinds of documents, they were not told to destroy them (CX-10767C.0002-3). And no other Rambus employee has said that they were told to destroy documents questioning patentability. 76

PUBLIC VERSION

(“Rambus’s counsel’s decision to mention the irrelevant shredding is itself irrelevant and presents no legitimate point.”). In the 661 Investigation, Respondents offered estimates of the number of

documents destroyed based on the weight of materials discarded into shredding trucks. Rambus referenced testimony about junk being discarded to counter those irrelevant and misleading estimates of the volume of documents destroyed, not as a plea for absolution. In this Investigation, the ALJ characterized as misleading Rambus’s references to employment records. ID at 309-10. Rambus agrees that such records are irrelevant. Rambus addressed employment records only because

Respondents accused Rambus of spoliating performance evaluations. For instance, Respondents raised Rambus’s alleged failure to produce performance evaluations during discovery in a conference with the ALJ. EDIS ID 455502 (7/25/11) at 6-9. In their prehearing brief, Respondents asserted that Rambus had not produced a performance evaluation that would “shed light on the nature of Rambus’s inventions.” RPB at 371-72. Far from trying to “mislead” or “misdirect” the ALJ, Rambus’s arguments regarding the treatment of employment records under the document retention policy was relevant to refute Respondents’ contentions. b. Rambus’s Witnesses Provided Relevant Testimony Demonstrating Lack of Prejudice

The ALJ’s credibility assessment of Mr. Steinberg and Mr. Moniz rests on clearly erroneous facts and cannot support a finding of bad faith. The ALJ faulted “Rambus [for] providing two witnesses, who were not with Rambus at the relevant time and place, to testify on the lack of destruction of relevant evidence. Such an approach certainly damages its credibility and the ability of the Commission and ALJ to rely on anything that Rambus has presented.” ID at 285. But it was Respondents who read in Mr. Steinberg’s deposition testimony from different cases during their case in chief. Similarly, the ALJ sharply criticized Mr. Steinberg as “untrustworthy” and “dishonest” because he had a duty as a Rule 30(b)(6) witness to answer questions he could not answer. ID at 287. But Respondents read in testimony from five Steinberg depositions, only one of which (August 77

PUBLIC VERSION

1, 2001) constituted Rule 30(b)(6) testimony. The particularly lengthy testimony cited by the ALJ on pages 285-87 as evidence of Mr. Steinberg’s lack of preparedness is not 30(b)(6) testimony, it is personal testimony from his April 2004 deposition given years after he left Rambus. JX-131C.21921. The ALJ’s treatment of all of Mr. Steinberg’s testimony as Rule 30(b)(6) testimony was erroneous and compromised both his credibility determination and bad faith finding. ID at 283-87. Likewise, the ALJ’s criticism of Mr. Moniz’s testimony as “at best another misdirection play for Rambus” is unwarranted. ID at 288-89, 311. Mr. Moniz provided relevant testimony on the preservation of the Barth I files while at Rambus. That testimony directly rebutted Respondents’ allegations of destruction and prejudice. RIB at 252-54; RPB at 363-66. As the Staff recognized during a hearing, the ALJ suggested to Rambus that it should show the chain of custody of the prosecution files. EDIS ID 455502 (7/25/11) at 29. Rambus did that through Mr. Moniz who managed the prosecution of the Barth I patents from the time he joined Rambus in 1999. No basis exists for the ALJ’s conclusion that “Mr. Moniz provided no relevant or pertinent testimony.” ID at 311. The general credibility of Rambus’s arguments cannot be legitimately impugned by testimony prompted by Respondents’ allegations and sought by the ALJ. 4. Demonstrably False Determinations of “Fact” Infected the ALJ’s Credibility Determinations

The ALJ’s bad faith finding hinges on erroneous assessments of credibility grounded on demonstrably false factual findings. Numerous errors in the ALJ’s assessment of Mr. Karp’s, Mr. Barth’s, Mr. Steinberg’s and Mr. Moniz’s testimony are well documented above. As a result, those purported credibility determinations should be set aside and given no weight. See, e.g., Doe v. Menefee, 391 F.3d 147, 164 (2d Cir. 2004) (finding “the trial court’s credibility determinations vulnerable where they were founded on factual inferences that the evidence did not permit”); Jiminez v. Mary Washington College, 57 F.3d. 369, 380 (4th Cir. 1995) (reversing credibility determinations that were based on speculation and inconsistent with record evidence). 78

PUBLIC VERSION

In addition to the errors addressed above, the ALJ further erred in accusing Mr. Karp of misrepresenting the scope of his duties as including only the Farmwald patents. ID at 272-75, 277, 297-98. Mr. Karp testified that his duties as Vice President of Intellectual Property involved

prosecution of all Rambus patents: “Among other things, I was responsible for making sure that Rambus’s patents and patent applications appropriately protected Rambus’s inventions. I was also responsible for developing a licensing program for non-compatible products.” CX-10770C.002. The testimony that the ALJ cites is limited to the scope of the “non-compatible licensing program” (which at that time was restricted to Farmwald patents), not the scope of Mr. Karp’s overall patent prosecution and other duties and activities. The ALJ further criticized Mr. Karp and Mr. Tate for lack of recall. But the events in question occurred more than a decade ago and both Mr. Tate and Mr. Karp have provided sworn testimony numerous times on these same subjects. That these witnesses would refer to their prior testimony or the documentary evidence when their recall has been clouded by the passage of time and infected by intervening attorney argument about “the evidence” is understandable. As Mr. Tate explained, in testimony that the ALJ faulted, “I have been deposed or testified 25 or plus times, so I have a hard time remembering what I would have recalled without all those intermediate events. So I don’t recall specifically sitting here today. I know it [degaussing backup tapes] happened.” HTr. 1318:6-20 (emphasis added). And many of Respondents’ counsel’s questions were extremely

specific, requiring the witnesses to admit to dates or the specific use of certain words. That is particularly true for the criticized testimony of Mr. Tate. See, e.g., ID at 290 (criticizing Tate for not recalling what exactly occurred on January 7, 1998). 5. The ALJ’s Commentary on Dr. Przybylski Has No Bearing on Bad Faith and Should Be Vacated

The ALJ’s erroneous and irrelevant assessment of the testimony of Rambus’s technical expert further undermines the ALJ’s bad faith finding. ID at 305-309. Dr. Przybylski did not offer a 79

PUBLIC VERSION

word of testimony on unclean hands and the nature and quality of his testimony could not provide any support for finding bad faith. Indeed, the ALJ’s reliance on “credibility findings” that are so divorced from the relevant law or facts render the ALJ’s unclean hands conclusions unworthy of deference and require reversal. Moreover, the ALJ’s commentary on Dr. Przybylski should be vacated because it is irrelevant to unclean hands and is incorrect. See Deere, 605 F.3d at 1358.26 For example, the ALJ stated that he “provides no useful testimony to the ALJ or the Commission” (ID at 306), even though the ALJ cited extensively to his testimony in support of other findings, such as infringement. ID at 47-67. In reality, Dr. Przybylski testified consistently with his testimony from the 661 Investigation (which resulted in the ALJ and Commission finding the Barth I patents valid and infringed). See 661 ID; see also CIB at 30-31; CRB at 13, 31. And all of the examples provided by the ALJ simply illustrate his confusion. ID at 306-308. First, Dr. Przybylski correctly testified that Yano’s memory controller is part of a microprocessor that uses a system clock, and that the microprocessor is therefore synchronous. RX4261.003, .008 (Fig. 2); HTr. 2613:19-2614:13. He also correctly testified that the system clock is not sent to the memory device (RX-4261.008 (Fig. 1); HTr. 2613:19-2614:13) and therefore the memory controller controls that memory device asynchronously. CX-10765C at 109. These are not conflicting statements; indeed Respondents’ expert agreed that Figure 2 of Yano does not show a clock signal going from the microprocessor to the memory. HTr. 1137. The ALJ’s confusion is apparently premised on not understanding that a synchronous microprocessor can contain a memory controller that operates asynchronously with a memory device (i.e., without providing a clock signal to the memory device), which is precisely what Yano shows in Figures 1 and 2. ID at 306-307; RXThe Commission could vacate the ALJ’s demonstrably flawed credibility determinations for Rambus’s fact witnesses as well. The documentary evidence alone is sufficient to conclude that Rambus did not anticipate litigation of the Barth I patents or conduct any activity with respect to the Barth I patents in bad faith. 80
26

PUBLIC VERSION

4261.008. Second, Dr. Przybylski’s testimony that the claimed “write command” is limited to synchronous memory systems is consistent with his understanding that a “command” in the context of the Barth I patents requires one or more bits, which in turn connote values sampled at particular times, as is done in a synchronous system. HTr. 2637:2-21. After acknowledging this testimony, the ALJ incorrectly states that “Dr. Przybylski does not remember the claim construction” (ID at 308), citing testimony where Dr. Przybylski was simply asking whether he should look to see where he defined it within his witness statement—a reasonable request given that his witness statement was over 2000 pages. HTr. 2638:11-2639:22. In sum, the ALJ’s discourse on Dr. Przybylski in the unclean hands section was unwarranted, incorrect, and unrelated to unclean hands. Consequently, it should be vacated. D. Respondents Were Not Prejudiced by Any Document Destruction 1. The ALJ Made Clearly Erroneous Findings and Misconstrued the Record in Finding Prejudice Based on Alleged Destruction of Mr. Barth’s Emails and JEDEC Materials a. Clear Legal and Factual Error Infected the ALJ’s Prejudice Determination

The ALJ’s analysis of prejudice is a single sentence: “Email correspondence, from Mr. Barth alone, could have provided support for Respondents’ invalidity contentions.” ID at 313. The ALJ applies both an erroneous legal standard—“could have provided support”—and insufficient explanation. Prejudice requires that spoliation “materially affect[s] the substantial rights of the adverse party and is prejudicial to the presentation of his case.” Micron, 645 F.3d at 1328 (citations omitted). That requires more than finding that the documents “could have supported” a defense. Moreover, the ALJ’s thin analysis of prejudice is less than that of the deficient analysis provided by the district court in Micron, which prompted remand. Id. Rambus satisfied even the “heaviest” burden it could have properly expected to bear by

81

PUBLIC VERSION

showing that Respondents’ defenses were not materially prejudiced. CIB at 142-59; CRB at 81-94. The demonstrative lack of prejudice is particularly clear as to the only type of defense the ALJ identified as being prejudiced—invalidity. By its very nature, invalidity depends on patent itself and public prior art documents and cannot be “illuminated by evidence of a non-public nature, e.g., by internal Rambus documents.” Micron, 255 F.R.D. at 151. Indeed, the only two invalidity defenses that Respondents actually pled—invalidity based on written description and prior art—were not impacted by any alleged lost emails. While Respondents also contended their derivation defense was prejudiced, they did not plead and had no basis to plead derivation. i. Written Description

Respondents’ defense based on written description could not have been prejudiced by any alleged document destruction because written description is determined based on the “four corners of the specification.” Ariad Pharms., Inc. v. Eli Lilly & Co., 598 F.3d 1336, 1351 (Fed. Cir. 2010). While Respondents hypothesized about what emails “could have discussed” regarding “the scope of the ’914 patent,” written description is determined based on the specification and nothing more. Moreover, Rambus produced emails between the inventors and many testified that they saved all emails; there is no reason to believe that any emails addressing written description support were destroyed. Respondents deposed all of the living Barth I inventors and the prosecuting attorneys and had the opportunity to question them concerning written description and whether “sampling” was within scope of their invention. Thus, Rambus met its burden of establishing that Respondents’ written description defense was not materially impacted by the alleged document destruction. ii. Anticipation and Obviousness

Respondents’ anticipation and obviousness defenses have not been prejudiced by any alleged destruction of prior art because those defenses turn on public documents. As the district court in Micron recognized, these defenses “do not seem likely to depend from evidence internal to Rambus” “since prior art references (by definition) must be publicly available.” Micron, 255 F.R.D. at 151 82

PUBLIC VERSION

n.59. And even if Rambus had destroyed copies of prior art, it would have been redundant of the publicly available material accessible to Respondents. Indeed, Respondents have enormous

resources and have identified at least 100 alleged prior art references to the Barth I patents, so it is hard to imagine how Respondents have been prejudiced regarding prior art. Rambus also showed that there is also no evidence that any of the inventors discarded prior art, so Respondents have not been prejudiced. See CX-10766C.004-006, JX-14.152 at 301-302, HTr 459, JX-14.037 at 126, JX-14.039-40 at 133, 137, JX-14.0060 at 221-22, JX-14C.192-194 (Barth did not discard competitive information or documents related to Rambus’s intellectual property efforts); JX-128.0077 at 106–108 (Stark does not recall ever deleting articles, publications, patents, patent disclosures, draft patent applications, or documents related to JEDEC); CX-10767C.003, JX049C.183 at 97-98, JX-049C.188-89 at 156-57, HTr. 369 (Hampel did not destroy prior art or documents related to patentability); JX-150C.113 at 174 (Ware did not destroy documents concerning his work on patents). While Respondents suggested that Rambus discarded RamLink and SyncLink documents (RPB at 368, 375), Rambus preserved and produced them. See RX-3341C; RX-3342C; RX-4532C; RX-4272C; RX-3053C; RX-4528C; RX-3054C; RX-2052; RX-3411C. Respondents also speculated that email correspondence from Barth and Diepenbrock could have referred to “flaws in Rambus’s patents” and may have “discussed prior art,” and that handwritten notes from prosecution files could have had a “direct bearing” on invalidity. RIB 25051, 254. But there is no evidence—let alone a “concrete, plausible suggestion”—that Barth I inventors or Diepenbrock ever created such emails or that they were destroyed. Both Barth and Diepenbrock testified that they kept prior art and documents relating to patent and patent prosecution. See CX-10766C.004-006; JX-14C.192-194; JX-37C.051-54, 111-12. Likewise, it is just speculation that handwritten notes reflecting “thoughts on the patentability of the inventions and whether the inventions were obvious” were destroyed. RIB 254. No one testified that they created such notes. And the materials Respondents speculate could have existed would have been privileged and 83

PUBLIC VERSION

Respondents would not have been entitled to them for an invalidity defense. See McCook Metals L.L.C. v. Alcoa Inc., 192 F.R.D. 242, 252 (N.D. Ill. 2000) (“drafts of the specifications, claims, and other parts of the patent application prepared by the attorney . . . necessarily reflect[] the communications between a client and his attorney as the attorney attempts to put forth the invention in the best light possible to protect a client’s legal right.”); Micron, 255 F.R.D. at 151 n.59 (“The court does not generally require the production of draft documents, either as they relate to patent prosecution or contract negotiations.”) Respondents further speculated that “[e]vidence regarding the contemporaneous work of others obtained from Rambus’s contacts with JEDEC would also be ‘relevant to obviousness as a secondary consideration.’” RIB 256-57. But Respondents were not prejudiced since many of them said that they were members of JEDEC at that time and thus would have been aware of the alleged contemporaneous work. And Rambus produced JEDEC documents, including a complete set of JEDEC meeting minutes and Richard Crisp’s trip reports. Accordingly, Rambus met its burden of establishing that Respondents’ anticipation and obviousness defenses were not materially impacted by any document destruction. b. The ALJ Grounded His Finding of Prejudicial Destruction on Documents that Demonstrably Still Exist

The ALJ clearly erred in grounding a finding of prejudice on the alleged destruction of Barth’s emails—those emails still exist. The ALJ’s oft-repeated assertion that Barth’s emails were destroyed rests on testimony from Barth that he did not discard any emails and that he preserved them. ID at 313 (citing Tr. 441:11-24, 452:17-23). No basis exists for the ALJ’s conclusion that Barth’s emails were “admittedly destroyed.” ID at 305. Instead, the ALJ assumed that Barth’s emails were destroyed, stating that “Mr. Barth left behind tens of thousands of emails when he abruptly left Rambus, all of which would have been destroyed under the retention policy.” ID at 313. Yet no evidence exists that any of Mr. Barth’s emails were discarded after his departure. And the 84

PUBLIC VERSION

uncontradicted evidence that they were preserved through the first shred day under the document retention policy is powerful evidence undermining the ALJ’s assumption that they “would have been destroyed under the retention policy. The large number of Barth emails retained, preserved, and produced in this case also belies any notion that some unidentified person at some unidentified time destroyed them intentionally and in bad faith after Barth’s departure. The proposed exhibit list alone includes over 100 emails for which Barth is an author or recipient, many of which the ALJ cites as damaging or even dispositive of improper conduct. Complainant’s Proposed Exhibit List (EDIS ID 459718 at Exhibit 2) (9/21/11); Respondents’ Proposed Exhibit List (EDIS ID 459712 at Exhibit B) (9/21/11); ID at 295-305. There is no evidence suggesting that Barth’s emails were selectively purged and no plausible suggestion of the evidentiary value of such emails. No defensible grounds exist to find prejudicial destruction of Barth’s emails worthy of dispositive sanctions. The evidence shows that even if some Barth emails were destroyed they would be redundant of emails retained by others. For example, to the extent that the Barth I inventors discussed Barth I inventions, the evidence shows that the inventors retained emails pursuant to the document retention policy. See JX-150C.101 at 126-27, JX-150C.113 at 174 (Ware retained emails but lost emails on two occasions unrelated to the document retention policy); JX-47C.038 at 146-47 (Griffin does not recall deleting emails and did not have a practice of deleting emails); JX-49C.187 at 141-44 (Hampel’s practice was to delete emails from his inbox when it became “too large” and may have at times deleted emails); CX-10766C.004-006, HTr. 451, JX-14C.146 at 259, JX-14C.044 at 154, JX14C.193-194 (Barth never deleted emails). As a result, Rambus retained and produced an enormous volume of email, including ones regarding drafting a Barth I patent application. See CX-7243C (email regarding an application for the strobe invention in the Barth I patents); JX-49C.055; JX14C.027-28. The ALJ’s cursory reference to destroyed backup tapes also does not support prejudice. ID at 85

PUBLIC VERSION

313. Backup tapes are redundant—they retain duplicate copies of information stored on the server. Indeed, Rambus’s backup tapes are used to recover from system crashes, and the same data would be carried forward on all future backup tapes. HTr. 1945. Destroying backups tapes does not mean that the documents on the server were destroyed. While not identified in support of prejudice, the ALJ erroneously says that JEDEC meeting materials were destroyed and assessed dispositive sanctions based in part on “documents Mr. Barth received regarding JEDEC.” ID at 316; see also id. at 296 (citing question that “[f]rom time to time, Mr. Crisp would drop off copies of JEDEC meeting materials at your desk; isn’t that right?”); id. at 305 (“[a]lso absent from the evidence are the documents that Mr. Crisp shared with Mr. Barth”). But the events regarding JEDEC at that time do not relate to Barth I or show that Mr. Barth used JEDEC materials to draft Barth I claims. Section V.B.3.c, supra. Mr. Barth was no longer employed by Rambus when the Barth I claims were amended to cover the JEDEC DDR standard in 2000. RIB at 223; id. at 21. Rambus preserved the documents reflecting Rambus’s JEDEC activities in the mid 1990’s, including those involving Mr. Barth. For example, Rambus retained and produced a complete set of trip reports concerning JEDEC, internal communications, and correspondence with JEDEC. RX435C; RX-440C. Rambus has also provided a complete set of JEDEC meeting minutes. See, e.g., Complainant’s Proposed Exhibit List (EDIS ID 459718 at Exhibit 2 (see CX-10428 through CX10503)) (9/21/11). Indeed, in examining Rambus’s conduct at JEDEC, the FTC’s chief

administrative law judge observed: [T]he process here has not been prejudiced as there is no indication that any documents, relevant and material to the disposition of the issues in this case, were destroyed. In fact, Complaint Counsel noted that the record shows “an unusual degree of visibility into the precise nature of Rambus’s conduct.” In re Rambus, Inc., 2004 WL 390647, at *175, 2004 FTC LEXIS 17, at *531-32 (F.T.C. Feb. 23, 2004, Initial Decision); see also Hynix Semiconductor, Inc. v. Rambus Inc., 591 F. Supp. 2d 1038, 86

PUBLIC VERSION

1057 (N.D. Cal. 2006) (stating that “the material JEDEC materials are available”). That was the case here, where the ALJ used the mass of supposedly damaging JEDEC documents preserved by Rambus to fashion a detailed but wrong account of Rambus’s conduct. 2. No Prejudice Resulted from the Destruction of Other Documents

The propriety of the sanctions ruling should be assessed solely based on the alleged destruction of documents cited by the ALJ as supporting prejudice or sanctions (i.e., Barth emails and JEDEC materials). Micron, 645 F.3d at 1329 (remanding for further evaluation of prejudice and sanctions, stating that “the seriousness of [a dispositive] sanction warrants an analysis” of prejudice and “the degree of prejudice”). While mentioning other allegedly destroyed documents throughout the ID, the ALJ did not cite or analyze those documents in assessing prejudice or the appropriate sanction. ID at 312-14. Indeed, the evidence shows that to the extent that these documents were destroyed, Respondents were not prejudiced. a. Conceptual Tree

While not identified as prejudicial, the ALJ said that “Mr. Barth’s testimony also made it very clear that Rambus had destroyed some evidence [the conceptual tree] that he had used in drafting the Barth Patents—evidence that Rambus knew it had a clear duty to preserve as evidenced by Mr. Tate and Mr. Karp.” ID at 300. But nothing in Tate or Karp’s testimony shows that Rambus had a clear duty, or knew it had a clear duty, to preserve Barth’s conceptual trees. Indeed, there is no evidence that anyone other than Barth and his lawyer were aware of the conceptual trees. Moreover, many “conceptual trees” indisputably exist. See, e.g., RX-2468C; RX-

2469C.0012-13; RX-2470; RX-2471C; RX-2472; HTr. 480. The conceptual trees are not “patent charts,” as mischaracterized by the ALJ. See, e.g., RX2471C; ID at 300. They are high level documents listing categories of technology. As one explains, “[t]his document merely contains references to various technology and is not an interpretation of the referenced patent, including but not limited to, the scope of the patent.” Id. Nothing in any tree could have materially aided 87

PUBLIC VERSION

Respondents in their invalidity defenses. As a result, even if there had been a tree that included a reference to P048 for the Barth I patents and that tree had been destroyed, Respondents’ invalidity defenses were not substantially impaired. The high-level information on these “trees” is also cumulative of produced information. For instance, an email from Barth regarding preparing patent applications, including the application that led to the Barth I patents, has the same level of detail as the conceptual trees. Compare CX-7243C (referring to “Power Control Signalling” and “Parallel Serial Byte Masks”) with RX-2471C (referring to “power modes and controls” and “bit and byte masking”). The reference to “strobe” and “strobe mechanism” in CX-7243 is a high-level reference to the strobe in the Barth I patents, showing that Respondents have the type of information presented in the trees. Likewise, Rambus produced licensing presentations that identify Rambus’s technology in the same high-level terms and specifically identify the Barth I technology, i.e., “Strobed write DDR.” See, e.g., CX-71C.016. b. Vincent Files

The ALJ also alleged that Rambus’s outside patent counsel, Lester Vincent, “purged” some Barth application files, although he does not conclude that Respondents were prejudiced by the alleged purging. See ID at 285 (“However, Mr. Moniz had only received the files after Mr. Vincent had already purged them and could not provide the ALJ with any information regarding the relevant time frame.”); ID at 289 (“However, that fact is not relevant in this matter since there is conclusive evidence of a shred by Mr. Vincent of information in those same files, at a time and place that is relevant.”). Far from “conclusive” evidence, Mr. Vincent could not say that he had discarded any documents from the Barth I files. CX-10772C.027-29. Moreover, Mr. Vincent testified that he

would have in any event “retained all PTO correspondence, notes reflecting communications with the Examiner, prior art, conception and reduction-to-practice documents, and all maintenance-related documents.” CX-10772C.038. And Rambus showed that Respondents were not prejudiced even if Mr. Vincent had cleaned certain Barth I patent files. 88

PUBLIC VERSION

Blakely Sokoloff, and Lester Vincent in particular, handled the prosecution of the first four Barth I applications—he prosecuted U.S. Patent Nos. 5,748,914, 5,914,046, and 5,966,731 and partially prosecuted U.S. Patent No. 6,122,688—none of which were asserted in this Investigation. Respondents alleged that Mr. Vincent cleaned the application files for the non-asserted ’914, ’046, and ’731 patents (Mr. Vincent identified these files as P048, P048D2 and P048D3) of draft amendments and claims, handwritten notes, letters to and from Rambus and Mr. Vincent’s firm, printouts of emails, and diskettes with electronic files of drafts. RIB at 251-52. But these

documents, even if they existed and were discarded, would have been privileged and not available to Respondents’ for their defenses. See, e.g., McCook, 192 F.R.D at 252. Moreover, Mr. Vincent said that cleaning up a file meant organizing it, not necessarily destroying any documents. JX-143C.04142 at 129-30; JX-82C.019. Mr. Vincent also said that he was not sure whether notes or drafts ever existed in those files. CX-10772C.028. And since prosecution of the three applications proceeded quickly without much substantive action—two were allowed without a substantive action on the merits (P048 and P048D2) and the other was allowed after one action on the merits (P048D3), it is unlikely that anything of substance would have been in the files. Also, even if some drafts or notes existed and were discarded, any prejudice Respondents may have suffered as a result was minimized because Mr. Vincent maintained electronic drafts of documents from the Barth I files. CX-10530C to CX-10550C; RX-2528C; CX-10555C; CX-

10589C. To the extent that any diskettes may also have been destroyed, they were also duplicative, both of the paper and electronic files. HTr. 2254-55. As to correspondence with Rambus allegedly purged from the P048, P048D2, and P048D3 files, Mr. Vincent said that he maintained correspondence with Rambus in at least three other files—in his “G-files,” in his “chron” files, and in a separate “chron” file. CX-10772C.004. As a result, any discarded document would have been duplicative of correspondence in other locations. Indeed, Mr. Vincent produced extensive

correspondence with Rambus concerning the Barth I files. See, e.g., CX-10530C; CX-10555C; RX89

PUBLIC VERSION

2528C; CX-10589C; CX-10651C; CX-10715C; CX-10729C; CX-10709C; CX-10710C; CX10712C; CX-10713C; CX-10719C; CX-10720C; CX-10724C; CX-10745C; CX-10574C. Thus, Rambus showed that Respondents were not prejudiced by any cleaning that Mr. Vincent may have done to the earlier Barth I application files. And neither the Respondents nor the ALJ explained how Respondents’ defenses could have been substantively impaired by any document allegedly discarded by Mr. Vincent. 3. Respondents’ Other Defenses Were Not Prejudiced

Rambus also showed that Respondents’ defenses other than the invalidity defense cited by the ALJ were not prejudiced by any document destruction. CIB at 142-59; CRB at 80-94. For instance, Rambus showed that Respondents’ prosecution laches defense, which the ALJ rejected, had not been prejudiced because prosecution laches is determined based on the conduct of the patentee at the PTO and can be determined from the public file history. See Symbol Techs., Inc. v. Lemelson Med., Educ. & Research Found., 422 F.3d 1378, 1385 (Fed. Cir. 2005). Respondents’ inequitable conduct defense also was not prejudiced since Respondents could not meet the heightened standard established in Therasense, Inc. v. Becton, Dickinson & Co., 649 F.3d 1276 (Fed. Cir. 2011) (en banc), regardless of any document destruction. ID at 246-47. Rambus further showed that Respondents’ defenses based on Rambus’s conduct at JEDEC were not prejudiced because those defenses failed for legal reasons unrelated to any alleged document destruction, as the ALJ found. Indeed, Rambus had no duty to disclose before formal work began and formal work did not begin on the DDR standard until after Rambus left JEDEC. Rambus Inc. v. Infineon Techs. AG, 318 F.3d 1081, 1105 (Fed. Cir. 2003); Hynix, 645 F.3d at 1348. Rambus showed that Respondents’ defense of no domestic industry was not prejudiced by any destruction of draft agreements since domestic industry is determined based on the license agreements themselves. Respondents’ noninfringement defense, which was based on documents in their own control and the testimony of their experts, also was not prejudiced. 90

PUBLIC VERSION

4.

The ALJ Legally Erred by Relieving Respondents of Their Burden and Placing Too High a Burden on Rambus

The ALJ erred by imposing too high a standard for showing no prejudice. While a bad faith spoliator bears a “heavy burden” to show lack of prejudice, the presumption of prejudice is rebuttable. Micron, 645 F.3d at 1328 (quoting Anderson v. Cryovac, Inc., 862 F.2d 910, 925 (1st Cir. 1988) (presumption that missing evidence “substantially interfered with the aggrieved party’s trial preparation” is “a rebuttable one”). Here, the ALJ effectively set an irrebuttable presumption of prejudice by characterizing evidence demonstrating lack of prejudice as irrelevant. The ALJ accused Rambus of “misdirection” and offering “misleading” arguments when Rambus tried to meet its burden by citing the existence of the very documents (e.g., Rambus’s internal prosecution files or employment records) allegedly destroyed. ID at 309-11. But the nature and scope of documents in existence is an important factor in assessing prejudice. Micron, 645 F.3d at 1328 (whether destroyed documents were “redundant” is relevant). At the very least, it is an important factor in assessing “the degree of prejudice suffered.” Id. at 1329. Yet the only reason the ALJ provided for concluding that Rambus did not carry its burden of showing no prejudice is that “Rambus presents no evidence that the documents destroyed were ever cataloged.” ID at 313. While the ALJ said that the “failure to catalog destroyed documents does not necessarily mean that a party cannot show prejudice,” the ID shows he considered it dispositive. Id.; see also id. at 315 (“As Rambus never cataloged its

document destruction it is unclear how Respondents could ever identify any unique piece of evidence.”) That error, combined with the ALJ’s refusal to consider what documents still exist or assess the redundancy of destroyed documents, precluded Rambus from demonstrating a lack of prejudice. The ALJ also legally erred in not requiring Respondents to come forward with plausible, concrete suggestions as to what the destroyed evidence might have been. ID at 313. Micron states that a party alleging prejudice “must only ‘come forward with plausible, concrete suggestions as to 91

PUBLIC VERSION

what [the destroyed] evidence might have been.” 645 F.3d at 1328 (quoting Schmid v. Milwaukee Elec. Tool Corp, 13 F.3d 76, 80 (3d Cir. 1994)). Speculation about what the evidence might have been is not sufficient. Schmid, 13 F.3d at 80. Micron states that the ultimate burden of proving no prejudice shifts to the spoliator when bad faith spoliation is found. Id. It did not, however, hold that a bad faith finding relieves the party alleging prejudice of coming forward with concrete, plausible suggestions. Id. Without any reasonable identification of the documents Respondents would have sought to rely on for their defenses, Rambus’s burden of showing no prejudice would be impossible to meet. E. The Harsh Sanction of Dismissal Is Unwarranted and Contrary to Law

Dismissal is a “harsh sanction” that “should be imposed only in particularly egregious situations where ‘a party has engaged deliberately in deceptive practices that undermine the integrity of judicial proceedings.’” Micron, 645 F.3d at 1328 (citation omitted). “[S]uch sanctions should not be imposed unless there is clear and convincing evidence of both bad-faith spoliation and prejudice to the opposing party.” Id. “[T]he presence of bad faith and prejudice, without more, [does] not justify the imposition of dispositive sanctions.” Id. at 1329. The ALJ “must take into account “(1) the degree of fault of the party who altered or destroyed the evidence; (2) the degree of prejudice suffered by the opposing party; and (3) whether there is a lesser sanction that will avoid substantial unfairness to the opposing party and, where the offending party is seriously at fault, will serve to deter such conduct by others in the future.” Id. (citation omitted). Indeed, “[t]he sanction ultimately imposed must be commensurate with an analysis of these factors.” Id. (emphasis added). The ALJ’s numerous errors of law and fact made in determining anticipation of litigation, bad faith, and prejudice undermined an objective determination of an appropriate sanction. For example, the ALJ’s analysis of degree of fault hinges on the erroneous finding that “the Barth Patents were always part of Rambus’s long term litigation campaign” and the mistaken belief that Mr. Barth copied Barth Patent claims from JEDEC, when the Barth I claims were amended to cover the JEDEC 92

PUBLIC VERSION

DDR standard well after the last shred day. ID at 316; see Section V.B, supra. The ALJ also erred in failing to recognize mitigating circumstances like the fact that no asserted Barth I patent issued until years after the last shred day, no infringing product was marketed until well after the last shred day, Rambus committed to licensing and peaceful business resolution over many years, and actual litigation on the Barth I patents did not begin until 10 years after Rambus adopted its document retention policy and eight years after Rambus instituted a litigation hold. See Section V.B.3.d, supra. The ALJ also erred in dismissing as “misleading” and devoid of evidence Rambus’s reliance on advice of fully informed counsel in conceiving and implementing the document retention policy. See Section V.C.1.a, supra. These relevant factors show that the degree of fault as to destroyed Barth I documents is low and does not warrant imposing a dispositive sanction. This is not a case where documents were destroyed while litigation was pending or soon to be pending. See Gorelick et al., Destruction of Evidence §§ 9.3, 13.5 (2009) (a parties’ prelitigation obligation to suspend document destruction has been a notorious “gray area”). Regarding the degree of prejudice, the ALJ merely restates his erroneous and cursory conclusion that the Respondents were prejudiced by the destruction of Barth emails. Compare ID at 313 (“Email correspondence, from Mr. Barth alone, could have provided support for Respondents’ invalidity contentions.”) with ID at 316 (“The types of documents destroyed by Rambus related to the Barth Patents, such as Mr. Barth’s emails, documents Mr. Barth received regarding JEDEC, etc., would likely be ‘at the heart’ of Respondents’ invalidity defenses.”) Further, having deemed the existence of documents irrelevant, the ALJ never considered that any destroyed documents would be cumulative or redundant of the records retained and produced. Similarly, the ALJ did not consider that Respondents’ invalidity defenses were not dependent on documents unique to Rambus. The ALJ has not shown that the degree of prejudice that the Respondents suffered warrants imposing the drastic sanction of unenforceability. The ALJ’s refusal to consider a lesser sanction rests on “the egregious nature of Rambus’ 93

PUBLIC VERSION

conduct.” ID at 316. That mere restatement of the degree of fault factor is not a meaningful analysis of an appropriate sanction. The only case cited in support is distinguishable: in Leon, the plaintiff in a discrimination case deleted thousands of files, after litigation was filed, that would “likely be at the heart of [the defendant]’s defense.” Leon v. IDX Sys. Corp., 464 F.3d 951, 960 (9th Cir. 2006). Indeed, rather than substantively analyze alternative sanctions, the ALJ falls back on his adverse credibility determinations and concludes that “[s]uch an inference is particularly useless where the Respondents cannot obtain candid or forthright testimony from Rambus’s witnesses.” ID at 317. But those adverse credibility determinations have no merit. See Section V.C.4, supra. Moreover, were that testimony truly unreliable or had Rambus truly “sculpted” evidence (ID at 317), the ALJ could have precluded Rambus from relying on the testimony of its witnesses or the “sculpted” documentary evidence. Further, sanctions like those suggested by the Staff that are issue dispositive, rather than case dispositive, are not susceptible to the concern raised by the ALJ. For instance, “precluding Rambus from antedating prior art (e.g., due to an alleged inability to challenge evidence of conception and reduction to practice)” (SIB at 104) resolves that issue in favor of Respondents, but still permits Rambus to demonstrate validity over that art, which is based purely on the public documents themselves. But the ALJ did not explain why the Staff’s suggestion did not “avoid substantial unfairness to the opposing party.” Micron, 645 F.3d at 1329. The ALJ’s circular reasoning that lesser sanctions are not appropriate because there is bad faith spoliation would require the sanction of dismissal in every case of bad faith spoliation. Micron makes clear that is not the law, and dismissal is unwarranted.27 Id. at 1329 (“bad faith and prejudice, without more, [does] not justify the imposition of dispositive sanctions”). Micron mandates that, in Dismissal is also unwarranted because, as the Federal Circuit has explained, “[b]efore sanctions may be imposed, CIR 210.36(b) and Fed. R. Civ. P. 37 require a party to fail to comply with an ‘order’ concerning discovery.” Genentech, Inc. v. Int’l Trade Comm’n, 122 F.3d 1409, 1422 (Fed. Cir. 1997). Rambus has not violated any discovery order. 94
27

PUBLIC VERSION

determining the appropriate sanction, the court “ must ‘select the least onerous sanction corresponding to the willfulness of the destructive act and the prejudice suffered by the victim.’” Micron, 645 F.3d at 1329 (citation omitted). The ALJ erred in not following that mandate. VI. THERE IS NO PATENT EXHAUSTION The ALJ’s finding that certain products are subject to exhaustion should be reviewed

because it is based on fundamental errors of fact and law. A. The ALJ Erred in Finding that Authorized Sales from Samsung to Place in the U.S. Take

The ALJ erred in finding that “authorized sales” from Samsung to

occurred in the

U.S., a requirement for finding exhaustion. Fujifilm Corp. v. Benun, 605 F.3d 1366, 1371-72 (Fed. Cir. 2010). had the burden of establishing that the allegedly exhausting sales took place that

within the U.S., but offered only speculation based on contradicts all other evidence, as well as common sense. 1. The SamsungSales Take Place in

The ALJ erred in concluding that a U.S. sale takes place between Samsung, a Korean company with no DRAM memory manufacturing facilities in the U.S., and that delivered to It is undisputed

Samsung memory devices are

Thus,

own witnesses admit the “transaction as a whole” takes

place outside of the U.S. TransCore, L.P. v. Elec. Transaction Consultants Corp., 563 F.3d 1271, 1278 (Fed. Cir. 2009). The only “evidence” the ALJ relied on to establish that the sale takes place in the U.S. is a special notation of

95

PUBLIC VERSION

conditions that apply to clearly state that

purchase orders—the same purchase orders

relies on—

Thus, the ALJ clearly erred in failing to hold that the transaction between Samsung and 2. when viewed as a whole, is not a U.S. sale.

The Samsung/Elpida Sales Are Not Authorized Sales

The ALJ also fundamentally erred in relying on the sale of DRAMs, i.e., memory devices, instead of identifying any authorized sale of a memory controller. “The longstanding doctrine of patent exhaustion provides that the initial authorized sale of a patented item terminates all patent rights to that item.” Quanta Computer, Inc. v. LG Elecs., Inc., 553 U.S. 617, 625 (2008)(emphasis added); see also United States v. Univis Lens Co., 316 U.S. 241, 250-51 (1942) (stating “where one has sold an uncompleted article which . . . is within the protection of his patent . . . he has sold his invention so far as it is or may be embodied in that particular article.”) (emphasis added) Because there has been no finding of an authorized sale of a memory controller or product that incorporates such memory controllers, Rambus’s patent rights in the memory controller cannot have been exhausted. The ALJ erroneously concluded that Samsung and Elpida are authorized to sell JEDECcompliant memory devices “under the ’353 and ’109 patents,” because “the Samsung/Elpida agreements grant a license to

CX-4339C.0008, 11; RX-4075C.0013, 20. The Agreements expressly 97

PUBLIC VERSION

which is exactly the case here.

CX-4339C.0007; RX-

4075C.0012. Thus, with respect to the DRAMs of Samsung and Elpida, the agreements license only

B.

The ALJ Erred in Finding that the Samsung/Elpida Memory Devices Substantially Embody the Barth Patents

The ALJ also made both legal and factual errors in concluding that “the Samsung/Elpida memory devices substantially embody the Barth Patents.” ID at 335. For patent exhaustion to apply, inter alia, the article sold must substantially embody the claimed invention if the only reasonable and intended use of the article sold is to practice the claimed invention. Quanta, 553 U.S. at 630-34, 638. The ALJ failed to apply this standard. Rather than determining whether the DRAMs purchased from Samsung and Elpida substantially embody the asserted claims of the Barth patents, the ALJ inexplicably assessed whether the claims “exclude memory devices.” ID at 335. But that does not establish that the DRAMs “substantially embody” the claimed invention. Under a proper analysis, the DRAMs do not substantially embody the asserted claims of the Barth I Patents. The asserted Barth I claims are directed to memory controllers and steps performed by the memory controller for controlling a memory device. For example, claim 11 of the ’405 patent is directed to a “method of controlling a semiconductor memory device,” including “providing a plurality of control codes” to a semiconductor memory device. While the claims reference a memory device, the asserted claims are directed at methods of controlling memory device, not a memory device itself. But a DRAM, i.e., the semiconductor memory device, is controlled by a memory controller, not itself; and a memory controller provides the plurality of control codes to the DRAM, not the DRAM itself. Because the licensed DRAMs do not control themselves, they cannot

substantially embody or “all but completely practice” the claimed methods of controlling a memory device, as Quanta requires. Id. at 633, 638. 98

PUBLIC VERSION

The ALJ’s conclusion that “a significant portion” of the infringement evidence is “from the perspective of a memory device” also does not establish that the DRAMs substantially embody the asserted claims directed to memory controllers. Rambus’s infringement proof includes evidence showing that memory devices receive the claimed signals issued by a memory controller, but that does not mean the memory device substantially embodies the claims. Rather, this evidence shows that the memory controllers infringe the asserted claims because they practice the claimed methods by controlling the memory devices in the manner claimed. That the memory devices can be

controlled in the claimed manner because the DRAMs’ operation permits such control does not establish that the DRAMs practice the claimed method. See, e.g., JX-3.0046. The ALJ’s conclusion that memory devices substantially embody the asserted claims of the Barth Patents is also inconsistent with his findings regarding the scope of the same claims in the 661 Investigation. There, the ALJ found that “the asserted claims of the ’353 Patent are method claims that lack a specific recitation of memory device sampling circuitry and are directed at the controller’s functionality.” 661 ID at 27 (emphasis added). It is the controller that embodies, and infringes, the asserted Barth claims, and the standard memory components cannot embody those claims as the ALJ himself has construed them. VII. CONCLUSION For the reasons set forth above, Complainant Rambus Inc. submits the Commission should review the ID due to errors of fact and law, as well as a matter of Commission policy. In particular, the Commission should review the ID as to the invalidity over alleged prior art of all asserted claims, unclean hands as to the Barth I patents, and exhaustion as to certain products.

99

PUBLIC VERSION

Sign up to vote on this title
UsefulNot useful