Professional Documents
Culture Documents
Chapter 1,2,4
From
Howard Johnson & Martin Graham; High speed Digital Design A hand book of Black Magic, Prentice Hall PTR, 1993.
For every electrical parameter, we must consider the frequency range over which
that parameter is valid.
The bandwidth of an electronic filter is the part of the filter's frequency response
that lies within 3 dB of the response at the center frequency of its peak.
Fknee = 0.5/Tr Fknee = frequency below which most energy in digital pulses concentrated
Tr = pulse rise time
Fknee is used as the practical upper bound on the spectral content of digital signal
0
- 2 3
1+2+3+4+5
1+2+3+4
Square wave: Y = 0 for - < x < 0 and Y=1 for 0 < x <
Tr
0.35
20dB/decade
Tr
Pw
T
40dB/decade
1
T
Harmonic Number
1 3 5 7 9 …...
RNS, MS, MIT
Estimating the Frequency Content
0.35
• Where does that famous equation F come from?
Tr
• It can be derived from the response of a step function into a filter
with time constant tau
V Vinput (1 et / )
• Setting V=0.1Vinput and V=0.9Vinput allows the calculation of the
10-90% rise time in terms of the time constant
Propagation delay of conducting wire increases in proportion to the square root of the dielectric
constant of surrounding medium
Delay per inch for PCB traces depends on both the dielectric constant of the PCB material and the trace
geometry.)
Trace geometry determines whether the electric field stays in the board(signal propagates slow) or goes into the
air (effective dielectric is half).
Outer layer PCB traces are always faster than inner traces
Lumped System : The system small enough for all points to react together with
uniform potential .
Step response measurement shows us just what we need to see: what happens when a
pulse hits a circuit element.
W e can characterize DUT by observing the step response and using these thumb rules.
Resistor display a flat step response. At time zero, the output rises to a fixed value and holds steady.
Capacitor display a rising step response. At time zero, the step respose starts at zero but then later rises to a
full valued output.
Inductor display
RNS, aMS,
sinking
MIT step response. At time zero, the output rises instantly to full size and then later
decays back towards zero.
Ordinary capacitance
The fact that the induced field always opposes the change is
an example of Lenz' law.
electromotive force (emf) energy per unit charge
Disadvantages
It requires more pins and wires.
Signal swing:
Twice the noise margin
Return current
Is a constant DC value.
For bipolar V1 = -V0 , IR = 0.
This technique equalizes low pass nature of RC line by preemphasizing the high frequency
components of the signal.
Solution
1. care full matching of components.
2. digitally trimming the termination resistance.
3. by splitting signal return into clean and dirty paths.
Kerry Bernstein & et. al., High Speed CMOS Design Styles,
Kluwer, 1999.
Data integrity requires zero slope reversal through the valid signal levels.
• In many applications , both valid signal levels and the entire interface protocol are standardized.
• DRAM (both Synchronous Link DRAM & Direct Rambus DRAM)
▫ Use terminated , Rambus signaling logic (RSL), small signal swing interface levels.
▫ Define clocking and control signal strategies with objective of maximizing memory bandwidth.
The characteristics impedance for a transmission line having and inductance L per unit
length, a capacitance C per unit length, and a low resistance may be approximated as:
In systems where the distance between the driver and receiver may be kept short and
source and load impedances may be carefully controlled, these networks may be series-
terminated. Series termination is preferable because there is no DC power consumption
in the signal network.
Rule of thumb: if the distance from driver to receiver is short enough that the round trip
transmission line delay is less than the pulse width of the transmitted data, point-to-point
termination may be used.
The load capacitance (receiver, ESD, Packaging) introduces some inherent impedance
mismatch is the system.
Signal reflections in series terminated network result in un acceptable switching delays under
these conditions: load termination is used instead
Two ideal load terminated configurations common to HSTL, SSTL and other small swing
interface protocols.
• HBM pulses represent a discharge of a 1kv pulse over nearly 200ns, an approximation of a
human body contacting a grounded device.
• MM pulse described as damped, oscillatory current sources with peak current of 7 to 16A at
frequencies of 7 to 16 Mhz. these events which can occur during machine handling.
• CDM pulses are lower in voltage, but longer in duration than HBM and MM pulses,
representing the flow of charge accumulated on-chip to an external ground.
RNS, MS, MIT
ESD protection
ESD pulses can cause irreversible damage such as
Fused metal wires
Thermal breakdown of PFET and NFET devices (occurs when the high drain current localized silicon
heating and irreversible damage).
The protection circuit requires significant area to dissipate large currents, adding 1 to 3 pf of
capacitance to the chip pad.
This circuit merits careful optimization to minimize capacitive loading of high frequency I/O
signals.
In integrated circuits, electrical power is distributed to the components of the chip over a
network of conductors on the chip. Power network design (IC) includes the analysis and
design of such networks.
Networks tradeoffs
Performance
Reliable
Effective
The power distribution network distributes power and ground voltages from pad locations to
all devices in a design.
The package supplies currents to the pads of the power grid either by means of package leads
in wire-bond chips or through C4 bump arrays in flip chip technology.
Although the resistance of package is quite small, the inductance of package leads is significant
which causes a voltage drop at the pad locations due to the time varying current drawn by the
devices on die.This voltage drop is referred to as the di/dt-drop.
Therefore the voltage seen at the devices is the supply voltage minus the IR-drop and di/dt-
drop.
Excessive voltage drops in the power grid reduce switching speeds and noise margins of
circuits, and inject noise which might lead to functional failures. High average current
densities lead to undesirable wearing out of metal wires due to electromigration {EM).
RNS, MS, MIT
Challenges in design
the challenge in the design of a power distribution network is in achieving excellent voltage
regulation at the consumption points notwithstanding the wide fluctuations in power demand across
the chip, and to build such a network using minimum area of the metal layers.
These issues are prominent in high performance chips such as microprocessors, since large amounts
of power have to be distributed through a hierarchy of many metal layers. A robust power
distribution network is vital in meeting performance guarantees and ensuring reliable operation.
explicitly added decoupling capacitances are not free and increase the area and leakage power
consumption of the chip.
most commercial tools focus on post-layout verification of the power grid when the entire chip design is
complete and detailed information about the parasitic of the power and ground lines and the currents drawn by
the transistors are known. Power grid problems revealed at this stage are usually very difficult or expensive to fix
decisions about the structure, size and layout of the power grid have to be made at very early
stages when a large part of the chip design has not even begun.
Due to the growth in power consumption and switching speeds of modern high performance
microprocessors, the di/dt effects are becoming a growing concern in high speed designs.
Clock gating, which is a preferred scheme for power management of high performance
designs, can cause rapid surges in current demands of macro-blocks and increase di/dt effects.
Designers rely on the on-chip parasitic capacitances and intentionally added decoupling
capacitors to counteract the di/dt variations in the voltage. But it is necessary to model
accurately the inductance and capacitance of the package and chip and analyze the grid with
such models, as otherwise the amount of decoupling to be added might be underestimated or
overestimated. Also it is necessary to maintain the efficiency of the analysis even when
including these detailed models.
Power grid analysis can be classified into input vector dependent methods and vectorless methods.
The input vector pattern dependent methods employ search techniques to find a set of input
patterns which cause the worst drop in the grid. A number of methods have been proposed in
literature which use genetic algorithms or other search techniques to find vectors or a pattern
of vectors that maximize the total current drawn from the supply network. Input vector-
pattern dependent approaches are computationally intensive and are limited to circuit blocks
rather than full-chip analysis. Furthermore, these approaches are inherently optimistic,
underestimating the voltage drop and thus letting some of the supply noise problems go
unnoticed. The vectorless approaches, on the other hand, aim to compute an upper bound on
the worst-case drop in an efficient manner. These approaches have the advantage of being fast
and conservative, but are sometimes too conservative, leading to overdesign.
Most of the literature on power network analysis deals with the issue of computing the worst
voltage drops in the power network. Electromigration is an equally serious concern, but is
attacked with almost identical methods. Instead of the voltage at each node, EM analysis solves
for current in each branch, and instead of a voltage limit, there is a current limit per wire,
depending on its layer and width.
Other IC applications may use only a portions of the flows mentioned here. A gate array or
field programmable gate array (FPGA) designer, for example, will only do only the design
stages, since the detailed usage of these parts is not known when the power supply must be
designed. Likewise, a user of FPGAs or gate arrays will only use the analysis portion, as the
design is already fixed.
Noise source
Crosstalk
Sagging and recovery of the power supplies
Ringing on both Vdd and GND power supplies (power supply network).
C is capacitance
Width and thickness of wire
Linear spacing between the lines
Separation distance to any other conductor such as wiring levels above and below
Type of dielectric
Distance of common run length shared by the two wires.
Wires are longer and overlap more, noise pulse are taller or wider due to pulse travelling down a
resistive line (short pulse wider).
One can make this noise pulse larger by putting the driver of the victim wire farther away(shield
driver)
Rule of thumb: one only needs to account 3-4 longest aggressor wires for any particular victim wire.
Design perspective
One can space the wires farther apart,
Keep the wires as short as possible
Increase the driving strength of the victim lines
Make sure direction of transitions of the neighboring wires are in a direction that do not cause logic
noise.
It is synchronous in nature since both the victim and aggressor net are switching at
the same time (race condition at victim).
Timing convention governs when a transmitter drivers symbols onto the signal line and when
they are sampled by the receiver (periodic, aperiodic).
A good timing convention is one that manages timing noise, skew and jitter in an efficient
manner and allows the system to operate over a wide range of data rates.
Timing nomenclature
Delay and transition times
Relative timing b/w two
transitions is measured from
50% point of one transition to
50% transition of the other.
Computer Systems
Testability
Maintaining current state of its storage node to controlling the behavior of the logic being tested
Design of remainder of the latch can be greatly affected.
Cycle boundaries: in almost every design, latches are used at the cycle boundaries to pass valid
logic states from one pipeline stage to the next.
Mid-cycle boundaries: the insertion, or movement, of latches to some intermediate point
between cycle boundaries is a key element of cycle stealing.
Execute logic function: one method of reducing latch delay is to merge it with logic function
from either its input or output side.
Transition from one circuit style to another.
Dynamic to static
Single rail to differential
Pulse to static and static to pulse.
Initiate test of a portion of the logic.
Latch
Is defined as a single bit storage element that can store only a single value in response to a
control signal (D-Latch).
Level-sensitive storage element is one that passes data from the input to the output
when the control signal is active.
A single level-sensitive latch is the basic building block for all latches and flip-flops.
Edge-triggered latch
Is one that samples the input data on either a rising or falling edge, of a control signal.
Sampling: data does not enter the latch until the activating edge of the control signal
Capture: data that propagates into the storage element before the activating clock edge the latch is
said to be captured.