This action might not be possible to undo. Are you sure you want to continue?

# APPLICATION

NOTE 95-1

FliOW

~~

HEWLETT

PACKARD

S-Parameter Techniques for Faster, More Accurate Network Design

ABSTRACT. Richard W, Anderson describes s·parameters and jlowgmphs and then relates them to more familiar concepts slIch as transducer power gain and voltage gain. He takes swept·frequency data obtained with a network analyzer and uses it to design amplifiers, He shows how to calculate the error caused by assuming the transistor is unilateral. Both narrow band and broad band amplifier designs are discussed, Stability criteria are also considered, This article originally appeared in the February 1967 issue of the Hewlett-Packard [ournal.

represent the excitation of the network (independent variables), and the remaining two represent the response of the network to the excitation (dependent variables). If the network of Fig. 1 is excited by voltage sources VIand V 2, the network currents II and I~ will be related by the following equations (assuming the network behaves linearly): 11 I~ = Y21VI + Y~2V2

= YIlVI + YI2V

Z

(1) (2)

**T IN EAR NETWORKS, OR NONLINEAR NETWORKS operating .I.....J with signals sufficiently small to cause the networks to
**

respond in a linear manner, can be completely characterized by parameters measured at the network terminals (ports) without regard to the contents of the networks. Once the parameters of a network have been determined, its behavior in any external environment can be predicted, again without regard to the specific contents of the network. S-parameters are being used more and more in microwave design because they are easier to measure and work with at high frequencies than other kinds of parameters, They are conceptually simple, analytically convenient, and capable of providing a surprising degree of insight into a measurement or design problem. For these reasons, manufacturers of high-frequency transistors and other solid-state devices are finding it more meaningful to specify their products in terms of s-pararneters than in any other way. How s-parameters can simplify microwave design problems, and how a designer can best take advantage of their abilities, are described in this article,

Two-Port Network Theory

In this case, with port voltages selected as independent variables and port currents taken as dependent variables, the relating parameters are called short-circuit admittance parameters, or y-parameters, In the absence of additional information, four measurements are required to determine the four parameters Ylll Y21, Y12, and Yn. Each measurement is made with one port of the network excited by a voltage source while the other port is short circuited, For example, y,w the forward transadrnittance, is the ratio of the current at port 2 to the voltage at port 1 with port 2 short circuited as shown in equation 3, = VI V2 = 0 (output short circuited)

Y21

~I

m

If other independent and dependent variables had been chosen, the network would have been described, as before, by two linear equations similar to equations 1 and 2, except that the variables and the parameters describing their relationships would be different. However, all parameter sets contain the same information about a network, and it is always possible to calculate any set in terms of any other set.

Although a network may have any number of ports, network parameters can be explained most easily by considering a network with only two ports, an input port and an output port, like the network shown in Fig, 1. To characterize the performance of such a network, any of several parameter sets can be used, each of which has certain advantages, ach parameter set is related to a. set of four variables associated with the two-port model. Two of these variables

I~ --0

I

Port 1

"2

Port 2

Fl.!:! .. 1. a,'lIffr!11 twO·p0l'/fletwork,

and a2 are normalized incident voltages. At higher frequencies these measurements typically require tuning stubs. on the other hand. (13) • v. lnctdent ~I a.' These parameters describe the interrelationships of a new set of variables (a. 2 . Input reflection coefficient with the output port terminated by a matched load (Zr. separately adjusted at each measurement frequency. == IIZo 2v'Z.) alld parameter definitions. They are defined in terms of the terminal voltage Vi' the terminal current L. MTT·13. 2...f/rlllV/lJ1! 'III" . (6) voltage wave incident on port 2 Y'4.f (a" a. == 0 == S22 == 111111 I~ II III 'II I. /1 III" r~ . and S12 are: a.. S-parameters. are usually measured with the device imbedded between a 50il load and source. This is difficult to do even at RF frequencies where lead inductance and capacitance make short and open circuits difficult to obtain. This means that scattering parameters can be measured on a device located at some distance from the measurement transducers. Kurokawa. == Vi + Z. (7) Dependent voltages: VI - variables b... then. For most measurements and calculations it is convenient to assume that the reference impedance Z. Generalized scattering parameters have been defined by K. but a tuning stub shunting the input or output may cause a transistor to oscillate. K. IzZo vz: Vz" network (9) The linear then: equations describing b1 the two-port are (10) (11) == sllal S21 l bz The s-parameters == a + +8 812a2 22 2 a z. The independent variables a.. 'Power Waves and the Scattering Matrix. and b. MarCh. Vz. on Micro· a. are normalized voltage wave reflected (or emanating) from port 1 reflected • b. unlike terminal voltages and currents. 822. For the remainder of this article.' IHE Transactions wave Theory and Techniques.. making the measurement difficult and invalid.. 1965. Kurokawa.. do not vary in magnitude at points along a lossless transmission line. to reflect short or open circuit conditions to the device terminals.. as follows: a1 VI + 11Zo == --'---=~"-- 2¥Z: _ voltage wave incident on port 1 VZ. Not only is this inconvenient and tedious. and == 0). ( ln . and an arbitrary reference impedance Zi' as follows . sets a.. voltage wave reflected (or emanating) from port 2 Vz. No.S-Parameters The ease with which scattering parameters can be measured makes them especially well suited for describing transistors and other active devices. b. all variables and parameters will be referenced to a single positive real impedance 4. == Z.2. are normalized complex voltage waves incident on and reflected from the i lh port of the network. provided that the measuring device and the transducers are connected by low-loss transmission lines. Another important advantage of s-pararneters stems from the fact that traveling waves. is positive and real. 821. Measuring most other parameters calls for the input and output of the device to be successively opened and short circuited. (12) al == 0 == Output reflection coefficient with the input terminated by a matched load (Zs == Z. and there is very little chance for oscillations to occur. The wave functions used to define s-pararneters for a twoport network are shown in Fig.II 2vlReZii (4) (5) where the asterisk denotes the complex conjugate. == 0). and b. The variables a. TWO·PORT NETWORK S11 == ~I a1 Sl1.. v.). Vol. Vr2 (8) Vbz == z 2 vz.

(1 (1 - + S[l) Sl1) (17) }=~ z. source minus the power delivered to the input of the network. a. = 0 = Reverse transmission (insertion) gain with the input port terminated in a matched load. and easily manipulated to determine matching networks for optimizing a circuit design. This relationship between reflection coefficient and impedance is the basis of the Smith Chart transmission-line calculator. Consequently. Flow graph oj network oj Fig. reflected or emanating from the output of the network. load. Power that would be delivered to a Z. + Zo Vs no (16) and where ZI ~ ~ ZI = z.• 821 = 812 = ~I ~I a. quantities which are often of more interest than the corresponding voltage functions: 12 = Power reflected from the network input Sll 1 Power incident on the network input reflected from the network output Power incident on the network output 1 12 = Power S22 Power delivered to a Z. a. load and source = Power = incident on the input of the network. source = Transducer 18[212 power gain with Z. some of the y-parameters described earlier in this article are not so familiar. For example. b1. from a Z. Power incident on the load. The above equations show one of the important advantages of s-pararneters. = Power available from a source of impedance Zoo = Reverse transducer power gain with Z. load Is·'[12== ----------Power available from Z. a. + Zo VI 11 is the input impedance at port I. the reflection coefficients 811 and S22 can be plotted on Smith charts. (15) Notice that bs = z. Power reflected from the load. insertion gain gives by far the greater insight into the operation of the network. Clearly. By comparison. and la212 = Power incident on the output of the network. Ib112 Network Calculations with Scattering Pafam'LrI = Power available = Power = Power reflected from the input port of the network. and various power waves: lallZ Hence s-parameters are simply related to power gain and mismatch loss. Ib212 = = 3 . and bz. 2. the y-pararneter corresponding to insertion gain S21 is the 'forward transadmittance' Y21 given by equation 3. both familiar quantities to engineers. Fig. 3. Another advantage of s-pararneters springs from the simple relationships between the variables at> a2. converted directly to impedance. namely that they are simply gains and reflection coefficients. load source. = 0 = Forward transmission (insertion) gain with the output port (14) terminated in a matched load.

No. Vol. A fJ3th is a continuous succession of branches. ) N. In the following equations the load and source arc described by their reflection coefficients rr. March.. lliMC arc two paths from b. in fig._ S21 S...MHI no combinations of three or more nontouching loops: ~hr!d(.. driving point characteristics. respectively.'lim around the loop.l__._". 'Simplified Signal Flow Graph Analysis: Microwave Journal. ifi. Hllllo arc three individual toops. 3 there is only one forward path from bs to b. 2.thnldon~ On p_ 11 is a table of formulas for calculating many often-used network functions (power gains. Fig. = P snv _~fi[_ (1 - l1'sl2) _ G'l' = l--(). . _ -'-"-"--T-=-S~~-i. and its ~". 3 is the flow graph representation of the system of Fig. not touching the k" forwMd path. "" path gain of the k'~ forward path A "" 1 ~ (sum of all individual loop gains) (sum of the loop gain products of all possible combinations of two nontouching loops) . This confirms the definition of Sll: Using scattering parameter flow-graphs and the non-touching loop rule..'s + S12 1') + (s" S12 I'L 1\).(f. and z-parameters. to a.n il !". K.t 1M v~hJl~ a to~ this network ~s of . 1960. referenced to the real characteristic impedance Z... 3 shows that if the load reflection coefficient rI. (19) where T.) II" i'll' 'lis" l'. is zero (ZI. 11.. The simplification of network analysis by flow graphs results from the application of the "non-touching loop rule:' This rule applies a generalized formula to determine the transfer function between any two nodes within a complex system. Nov. Node an represents the wave coming into the device from another device at port n and node b" represents the wave leaving the device at port n. .. y-. Loop gain is the product of the branch IIlIJlii.__ s~. "" The value of /::. 'J.performance but also form a natural set of parameters for use with signal flow graphs". No.1'') Using the non-touching "". = (a. • Transducer power gain deliv~red to the load Power available from the source on load) II'LIZ) Ptrcflccted from load) G" 1 = ..__. to b. Hunton. no 111l((~ boing encountered more than once.'"IJ'".__ S22 1'[. but flow graphs make s-pararneter calculations extremely simple. Flow graphs will be used in the examples that follow. it is easy to calculate the transducer power gain with arbitrary load and source.c. The non-touching loop rule is explained in footnote 4. Also included in the table are conversion formulas between s-pararneters and h-._ S'2 1S"2 1'1. a. The solution T (the ratio of the output variable to the input variable) is . and l's.. 6.). The complex scattering coefficients are then represented as multipliers on branches connecting the nodes within the network and in adjacent networks. Fig.= + _~~. A loop is a path which originates and terminates on the same node."X?~~ = P(incident = Ibzlz (1- PI. l) + . only one combination of two ""it'I<i'''hlil~ 1011(11.-L-l'~ (1 b" P'sI2) (1 loop rule. MTT·S. 'Analysis of Microwave Measurement Techniques by Means of Signal Flow Graphs: IRE Transactions on Microwave Theory and Techniques. + v. it is not necessary to use signal flow graphs in order to use s-pararneters.: their path gains are s"s"l'L aM '" '(".(sum of the loop gain products of all possible combinations of three nontouching loops) 2) Voltage gain with arbitrary source and load impedances + v..3. 1963.djvoly. + b fio = v. ". etc.j". = '-b._.. Zo) there is only one path connecting b.. In a signal flow graph each port is represented by two nodes. Kuhn. which are other parameter sets used very often for specifying transistors at • 4 . Path gain is the product of all the branch multipliers along the (lolh. and a forward path is a path connecting the input node t~ the output node. and I recommend them very strongly. Vol. where no node is encountered rnuru than once. Two other parameters of interest are: with the output termination • + S'" S'2 1\ I) Input reflection coefficient arbitrary and Zs = Zo' S 11 .. ---Sll Sll (1 - 822 1'r) The nontouching loop rule provides a simple method for writing the solution of any flow graph by inspection. I- Sll 1'g - 1\ r. H~~: hM!~~iq~Vilni(_~f~ fn_w'! h~ 1\) ll. (flow graph rules prohibit signal flow against the forward direction of a branch arrow).2. Of course.) in terms of scattering parameters.. - 11 '. (20) lor ""illillo.

3 dB point of the magnitude curve. They represent the actual s-parameters of this transistor between 100 MHz and 1700 MHz.--'I j. 4(d) to that of Fig.-. along with additional data on s21 below 100 MHz. or linear-phase.35 em of line. (21) w" where To = 112 ps w Wo = 2r. passing through +135 0 at 125 MHz. 4(e) required 3. and then rolls off at a slope of 6 dB! octave. with line stretcher adjllSled to remove linear phase shift above 500 MHz. By adjusting a calibrated line stretcher in the network analyzer. 4(d).. The magnitude of S~l is essentially constant to 125 MHz. Amplifier Design Using Scattering Parameters The remainder of this article will show by several examples how s-parameters are used in the design of transistor amplifiers and oscillators.. and mathematical manipulations will be omitted wherever possible. This expression is 0 s _ ~.• Fig. and their relationship to s-parameters is indicated. measured with a vector voltmeter. ~s] ~ 2 10 dB/em -30dB 1700 MH~ 100 MHz 100 MHz 1700 MHz (a) (b) 10 dB/em S~l 10 (m/em S-. Outermost circle Oil Smith Chart overlay corresponds to [s. 5 . 5. 4(e) resulted. a compensating linear phase shift was introduced. S2' behaves like a single pole in the frequency domain. the emphasis in these examples will be on s-parameter design methods. (c) s". These oscillograms are the results of swept-frequency measurements made with the new microwave network analyzer described elsewhere in this issue.Model 8410A Network Analyzer. After removal of the constant-delay. the phase angle of $". (e) s. Measurement of S·Parameters Most design problems will begin with a tentative selection of a device and the measurement of its s-parameters.. Fig.1 = 1. varies linearly with frequency above about 500 MHz.C) + .. equivalent to a pure time delay of 112 picoseconds. (b) s". (a) SIl. The phase angle of S~l' as seen in Fig.. and the phase curve of Fig.T.S?l 1700 IYlHz (c) lower frequencies. Scale [actor same as (a).- --------. In other words. To keep the discussion from becoming bogged down in extraneous details. f land f"mx>are also given. In Fig.. . S parameters of 2N3478 transistor ill CO//1I1J()II-emiller configuration.2 125 MHz S~l" = 21 dB The time delay T" = 112 ps is due primarily to the transit time of minority carriers (electrons) across the base of this npn transistor.:. 4. 4(d) is replotted on a logarithmic frequency scale. for this transistor [Fig.f = 2".!] OdB -110" +20" MHz (c) 100 500/eln (d) i. measured by -hp. the . 4(e)] Q varies from 180 at dc to + 90 at high frequencies. X = 11.S'2'l cle· j-u. 4 is a set of oscillograms containing complete s-parameter data for a 2N3478 transistor in the common-emitter configuration. and it is possible to write a closed expression for it. To go from the phase curve of Fig.. Two important figures of merit used for comparing transistors. the magnitude of S2l from Fig. (d) SOl. component.

11. representing 5011. 6.. Simple series-capacitor.. points corresponding to sll> s" ll' 822. The next step in the design is to synthesize matching networks which will transform the 5011 load and source impedances to the impedances corresponding to reflection coefficients of s* 11 and s* 22. 6. p.. 11). this one being just the reverse of the one in Fig. the circle to be used for finding series C is the one passing through the center of the chart. as they are in this case. are shown in Fig..g. as shown by the solid line in Fig.. In determining how much error will be introduced by assuming Sl" 0. A small feedback factor means that the input characteristics of the completed amplifier will be independent of the load.1 from Fig. Since reverse gain Sl2 for this transistor is quite small50 dB smaller than forward gain S"I' according to Fig.e.Voltmeter. because setting Sl2 equal to zero will make the design equations much less complicated. First. The circles to be used are those passing through s" 11 and s" ~2' as shown by the dashed lines in Fig. operating between a 5011 source and a 5011 load.1 11m M 11. appears in Fig. calculated from the measured parameters. and the output wilt be independent of the source impedance. 4 replotted on logarithmic [requency scale. isolation of source and load is an important consideration. respectively. the first step is to calculate the unilateral figure of merit u.1211 - -lrLIZ) S221'JJ2 (24) 10 MH~ 100 MHz IOGHz FREQUENCY Fig. 6. 5. Data below 100 MHz measured with -hp-: 8405A VeclO. Next. the normalized values of Land C needed for the matching networks are calculated from readings taken 'from the reactance and susceptance scales of the Smith Charts. sign.03. Returning now to the amplifier design. shunt-inductor networks will not only do the job. and optimized for power gain at 300 MHz by means of Iossless input and output matching networks. 10 . calculated from s parameters (see text). Now if 01'\! is the transducer power gain with Sl2 = 0 and 01' is the actual transducer power gain. his is small. 11. and s* 22 at 300 MHz are plotted. 5.&. Values of Land C to be used in the matching networks are determined using the Smith Chart of Fig.. I I I111II II I II smull reverse gain. Increasing shunt inductive susceptance moves impedance points clockwise along constant-conductance circles. its length equal to the magnitude of the reflection coefficient being plotted.Ir. 7.162 UJ ED UJ ::J f- ~ to o 0 Z -10 " I'-i I 0 ::J f- Z 3162 :i -20 -30 ~ ::. increasing series capacitive reactance moves an impedance point counter-clockwise along these circles. 3. = (22) A plot of u as a function of frequency. The final network and the element values. maximum OT" occurs for 1's = s * 11 and rL = s * 22 (table. Narrow-Ba. 6. I ~ 6 . Bottom curve: unilateral figure of merit. the maximum error introduced by using OT\! instead of 01" is given by the following relationship: (1 + up 1 < 01' 01'\1 < (1 - I uF (23) 1111111 . a combination of constant-resistance and constant-conductance circles is found.12)(l sl1r. but they are on another Smith Chart. In most amplifiers.25 dB 11. If this is so.and of isolating the de bias from the load and source. 6.nd Amplifier Design Suppose now that this 2N3478 transistor is to be used in a simple amplifier. Top curve: Is. S12' illl'tlllllIl 11'1111 islruble property d for a transistor to I h 1 I il~ IIlh r limn that it simplifies amplifier de- When Is.IIII1W !lr rm :lii... is = 1821 2(1 1 V 1 GHz . 4 . In this case. i. Each point represents the tip of a vector leading away from the center of the chart. the matching networks need not be complicated.rom the center of the chart.. using the formula given in the table on p. and its angle equal to the phase of the coefficient. The constant-conductance circles for shunt L all pass through the leftmost point of the chart rather than the rightmost point.. the maximum value of u is about 0. These circles are like the constant-resistance circles. to s" 11 and s* 22' The circles on the Smith Chart are constant-resistance circles..via the inductor . 5. or feedback factor.there is a possibility that it can be neglected. but will also provide a handy means of biasing the transistor . obtained either by setting Sl2 0 in equation 18 or by looking in the table on p.1 and Isd are both less than one. the design problem will be much simpler. leading f.03162 11 - . Once these circles have been located. Which arc corresponds to which element is indicated in Fig. so the srror in this case turns out to be about ± 0.. normalized and unnormalized. Each element's reactance or susceptance is the difference between the scale readings at the two end points of a circular arc.r-. the unilateral expression for transducer power gain. Since this is to be a single-frequency amplifier. enough to justify the assumption T Iii 111M 0.'~'l 20 10 r-.

. 12. Next. Procedures for designing source-matching networks are identical to those used for designing load-matching networks.or z-parameters (for shunt or series feedback respectively) using the conversion equations given in the table. leave the gain the same at 450 MHz. The amplifier is to be driven from a 50!! source and is to drive a 50!! load. = r2 = g21snl ~_--.Broadband Amplifier Design IS21)2 = 13 dB at 300 MHz = 10 dB at 450 MHz Designing a broadband amplifier. a set of constant-gain circles is drawn. the gain contributions of G1 and G2 are varied to compensate for the variations of Go IS2112 with frequency. 4 is to be used in a broadband amplifier which has a constant gain of 10 dB over a frequency range of 300 MHz to 700 MHz. negative feedback. 5. p. Consequently. (i. Suppose that the 2N3478 transistor whose s-parameters are given in Fig. its center is on a line between the center of the Smith Chart and the point representing s" ~2 at that frequency. 11): = When a broadband amplifier is designed by selective mismatching. is a matter of surrounding a transistor with external elements in order to compensate for the variation of forward gain 18211 with frequency. or second. This can be done in either of two waysfirst.e. The distance from the center of the Smith Chart to the center of the constant gain circle is given by (these equations also appear in the table.g2) where Pig. We will use the second method .-1 .. When feedback is used. Each circle is drawn for a single frequency. Although in the general" case both a source matching network and a load matching network would be designed.. that is. G 1 for r s s * 11) for this transistor is less than 1 dB over the frequencies of interest. According to Fig.c. which means there is little to be gained by matching the source. To realize an amplifier with a constant gain of 10 dB.IS2212(1 .. p. G 1 ilia. Fig. 8. and increase the gain by 4 dB at 700 MHz.-..:::. source and load matching networks must be found which will decrease the gain by 3 dB at 300 MHz. it is usually convenient to convert to y. Equation 24 for the unilateral transducer power gain can be factored into three parts: GTu = GOG1G2 where = 6 dB at 700 MHz.-'"". Smith Chart [or 300-MHz amplifier design example.__--.. and a digital computer. 7 . 6.-. selective mismatching of the input and output circuitry. only a load-matching network will be designed. The first step in the design is to plot s" 22 over the required frequency range on the Smith Chart. one which has nearly constant gain over a prescribed frequency range. for this example.

to some point on the 0 dB circle at 450 MHz. representing 50rl. The three constant-gain circles are indicated in Fig.5) . as shown in Fig. three circles will be drawn. one in shunt and one in series. one for G2 0 dB at 450 MHz. One which is satisfactory is a combination of two inductors. 300-MHz amplifier with matching networks for maximum power gain. to some point on the . (0. The series inductor transforms the combination of the son load and the shunt inductance along circles of constant resistance and varying inductive reactance.6 dB.3 156 x C2 C. and one for G. = +4 dB at 700 MHz.gz 1- 2- (1 IS2212(1 - 152212) g2) 50g CaLculations: X'2= 2Q.. 9. For this example. 8..The radius of the constant-gain circle is p _ VI . 7.3 x 109) (3.. G2 ma x for all three circles is (0.01 L. Shunt and series elements move impedance points on the Smith Chart along constant-conductance and constantresistance circles. Since Isd for this transistor is constant at 0.3 dB circle at 300 MHz. Fig. There are undoubtedly many networks that will do this. and to some point on the + 4 dB circle at 700 MHz.= 0.5) (50) = 3 pF = 30. The required matching network must transform the center of the Smith Chart. One for G2 3 dB at 300 MHz.278)-1. = = 1 2. Smlill hur! for IJI'(mtlb(lIIti amplifIer (liMlnll camp/e.42 pF = 25 pF (3. The shunt inductance transforms the son load along a circle of constant conductance and varying (with frequency) inductive susceptance.85 over the frequency range [see Fig. as I explained in the narrow-band design example which preceded this broadband example. 4(b)]. .2Q_ nH= 26 nH 1. =- = ~I " H. or 5.32 156(1 109) = 83 nH L2 = 2" (0.

Stability Considerations and the Design of Reflection Amplifiers and Oscillators When the real part of the input impedance of a network is negative. Fig. Broadband amplifier with constant gain of 10 dB [rom 300 MHz to 700 MHz.64 ·"OA4)'~ 13.• Optimizing the values of shunt and series L is a cut-andtry process to adjust these elements so that -the transformed load reflection terminates on the right gain circle at each frequency. whether they are reflection amplifiers or some other kind.4 " nli Fig.and a negative-resistance device.-cc 1(3.4 nli 20A "Ii !jO~l . and an imaginary part equal to zero (the imaginary part can be set equal to zero by tuning.3) "20. of course.4 nH 50 t. Then the element values are calculated. whatever the phase angle of 1'smight be. Amplifiers.... the reactances and susceptances of the elements can be read directly from the Smith Chart.i') nl-!::: 36.cr:(_~ hom 700 MHz data. the same as they were for the narrow-band design. Referring again to the flow graph of Fig. If at some frequency (25) = the circuit is unstable and will oscillate at that frequency. 12. 10. 9. capacitors would behave in the opposite way. 10 is a schematic diagram of the completed broadband amplifier. reflection amplifiers and oscillators.:-(O.. the ratio of the reflected voltage wave b. 2N3478 36.) Once appropriate constant-conductance and constant-resistance circles have been found.. the corresponding input reflection coefficient (equation 17) is greater than one. The circulator is used to separate the incident (input) wave from the larger wave reflected by the negative-resistance device. Transistor oscillator is designed by choosing tank circuit such that 1''I"S'" 1.-'" L~('l. 3. if the circulator is perfect and has a positive real characteristic impedance Zo.3) (2. = . if 1s'111 < I'~. Reflection amplifier consists oj circulator and transistor with negative input resistance. ". an amplifier with infinite gain can be built by selecting a negative-resistance device whose input impedance has a real part equal to +Z. There is a convenient criterion based upon scattering parameters for determining whether a device is stable or potentially unstable with given source and load impedances. and the network can be used as the basis for two important types of circuits.-I 9 the device is unconditionally stable and will not oscillate. Theoretically.c" . II. A reflection amplifier (Fig. Fig. (This rule applies to inductors. Zs = Zo) and an arbitrary load impedance Zr" as defined in equation 19. to the input voltage wave bs is Fig. "'0' Q where S'll is the input reflection coefficient with 1'8 0 (that is. Two porl w~th S'i\ 1 (R('~I part of mput uu pedance IS negative) Fig. On the other hand.) (0.2 (32) (50) 2-. if necessary). II) can be realized with a circulator-a nonreciprocal threeport device . are not supposed to oscillate. ~ (l. -Z-. with unnonnalized element values.. and -the susceptance component decreases with frequency and the reactance component increases with frequency. Combination oj shunt and series inductances is suitable matching network for broadband amplifier.1<-' l.

39. Thus this transistor would oscillate from 1. Sept. 1961. 5. -F. -Plot 1/S'l1 on a Smith Chart. To make the transistor oscillate. 14. Load impedance ZL is 2oon. Reflection coefficients I'TI and I'T2 are also plotted as functions of the resonant frequencies of the two tank circuits. 'On Scattering Matrices Normalized to Complex Port Numbers. As an example of how these principles of stability are applied in design problems.- 1- 822 l'L Fig. 1/S'11 can be measured directly by reversing the reference and test channel connections between the reflection test unit and the harmonic frequency converter.7 GHz with a parallel tuned circuit. the footnotes of the article.) • 10 . Youla.5 GHz with a series tuned circuit and from 2. Vol. it may also take an external feedback element which increases SI2 S21 to make S'11 greater than one. 1966. July. 13 shows a Smith Chart plot of 1/S'l1 for a highfrequency transistor in the common-base configuration. Fredricks. 'Transistor~ and Active Circuits. In this case the input reflection coefficient S'll is the reflection coefficient looking into the collector circuit. the following articles and books contain information on s-parameter design procedures and flow graphs. Weinert. 13. C. (No s-parameters. No. F. =-Choose a load reflection coefficient I'L which makes Sill greater than unity. Linvill and 1. McGraw-Hili. (If the new network analyzer is being used to measure the s-parameters of the transistor. s' Il and 1'8 must be adjusted so that they satisfy equation 25.7.. 1966. Anderson Additional Reading on S-Parameters Besides the papers referenced in. nough to satisfy equation 25 (the tank circuit 11 II 11 • III i nt plays the role of 1's in this equa- -Richard W. G. Oscillations occur when the locus of I'TI or rT2 passes through the shaded region. Electronics. Gibbons. Dec.Fig. but good treatment of Smith Chart design metbods. 'Scattering Parameters Speed Design of HighFrequency Transistor Circuits. IRE. which means that I'L referred to son is 0. 'How to Use S-Parameters for Transistor Circuit Design: EEE.0 to 2. and the 'source' reflection coefficient 1's is one of the two tank-circuit reflection coefficients. The polar display with a Smith Chart overlay will then give the desired plot immediately. In general..5 to 2. Smith Chari for transistor oscillator design example.6. Vol. 1961. There are four steps inthe design procedure: -Measure the four scattering parameters of the transistor as functions of frequency. -J. No. -D. S'l1 = S11 + S12 S21 1'1. 49. No.) onnect either the series or the parallel tank circuit to the collector circuit and tune it so that 1'T1 or 1'T2 is lnr . 12. 12. Vol. -G. consider the transistor oscillator design illustrated in Fig. Proc. 1'T1 or 1'T2• From equation 19. 18.

..b P G' Avalla Ie ower am Power = Power available from network available from source ISidz (1 - (1 - IS1112) g. 10 .) where: i = 1.sursl211 ...szzrLi'z Z = SUal + slzaz bz = a + 82ZaZ S21 l = GOG1G Go = IS2112 G1= G2 = Maximum Unilateral 11 - 11- 11'512 S11I'S 1 2 II'lY Input reflection coefficient with arbitrary Z. 11 ..2 and I'L 182112 G. ower Gam = Power Power delivered to load available from source Error Limits on Unilateral Gain Calculation 11 ...IsdFI i = 1.IrLIZ) 11 .szzI'LIZ Transducer Power Gain when ISI1I < 1 and IS221 < 1 u- G_ Output reflection coefficient with arbitrary Z. and Z. = 8*11 = 8*22 Av = Constant Gain Circles (Unilateral case: 5" = 0) Vz VI = 0- S21 SZ2I'L) (1 (1 + I'L) + S'l1) -center of constant gain circle is on line between center of Smith Chart and point representing s* Ii =--distance of center of circle from center of Smith Chart: P ower G' am = Power Power delivered to load input to network -radius of circle: ~ 1. max = -1-.IrslZ) (l .• + VI Useful Scattering Parameter Relationships a1b1-bI TWO-PORT NETWORK ~a2 ---b2 + Vz Unilateral Transducer Power Gain (5 . r.2Sll This maximum attained for Voltage gain with arbitrary Z.ISllIZ) (1 .2 Unilateral Figure of Merit Transducer P . 2 = 0) G Tn - 18211 (1 .1-1~.

. 15" r Where B1 [Bz [ ± VB22 - _ s" ..) (1 .) .y.) .su) (1 • -15" + s.s..) - y"y" Y. ..su = 1 + ISl112 B2 = 1 + IS2212- IS2212ISl112- = (h..s.s.) . Y-..Z12Z" (Z" h·.s •• 15.= Condition that a two-port network can be simultaneously matched with a positive real source and load: 52:: = z..1012 I> I> S11= z. ZIti Y12/=~ y" h. 1501 Page Mill Road.) .. 151.... and c are all satisfied..Sit) (1 + 5. h.. y.) + -25" 5. (1 ..z. IS111< 1.. + y.) 5... and zpararneters (z" . z"z" z"z.. + 1) (Z. Y-.5. + 1) + 1) (z. (1 .) + y.5.) (1 + 5. r mS = M* -N* rnL - [HI [ ± VB/ .1) (h . (1 (1 s!1= (Z" (Zit (z..s2' (1 .. + I) 2h" . write: Hewlett-Packard..s. = (1 (1 (1 5.05" (1 ..) (1 . California 94304.5..= (1 1 1 Sl~ + s.s. Source and Load for Simultaneous s" = (1 + Y (1 . Isd to oscillate if s-parametersin terms of h·.' = z.5. b...) + 5..h"h" h" = = (1 ..1) + 1) (z" + 1) - z..s.s. (1 .s. .y. • sn = (1 (~11 + h._Pal~ Alto.) (1 .5.) + Y"Y. In Japan. If hi.(h" s" + 1) (h".= YII= K> 1 or C <1 S11= C = Linvill C factor (l (1 ...) .s.) + 5... Geneva.. y.5..6 FEB 1967 N = 822 - 0 s* 11 For more Information.. Or. C = Linvill C Factor Zl/ (Use minus sign when B[ is positive.) . elsewhere in this table.l + h"h" + 1) (h" + I) .IN*I IS2212..) (1 + 5. Shibuya·Ku. + 1) (h" + - 2h" 1) ..) (I + 5. h"h" y" = (1 . 18..) =zjJ. a.) + 5. and Zl are the actual parameters.loS .) (1 .Y"Y....) + s"s" (1 + 5..... z.. = + I) 2z" (z" + 1) .S.) (I ...) ..S"5....) + 5. Tokyo...' = hllZ.h" M= Sl1 - 0 S * 22 Reprinted Compliments of Hewlett-Packard Journal Vol.Y12Y" + s.) + 5.. see 'Source and Load for Simultaneous Match. In Europe. + YII) (I + y.05 .so.Z"z" (z.. and z-pararneterslisted above are all normalizedto L. 2z" ZlZ= 2s .. No...Conditions for Absolute Stability No passive source or load will cause network a. Post Office Box.S. h.) + s.' = z.'=h. CH·1217 Meyrin 2.s .4 IMlzl 21MI2 2jNI2 41NI2J 1012 1012 J 5 11 - _ (h" .) .. call your local ~IP Sales Office or East (301) 948·6370 • Midwest (312) 255·9800 • South (404) 955·1500 • West (213) 877·1282.. then r. Y2'l'=~ " ---z..) (1 . 1·59·1. z.' = h" h '- = ~JZQ 1- y" -z.05" K_ - 1 + 101 S!l 2 18111 2 IS128211 2 IS2212 Match 'to (1 (1 + y. is negative.y"y" + 1) + 1) h"h . (1 + s. Maximum Available Power Gain If K > GA 1. Yo~ogawa·Hewlett·Pac~ard.h"h" h" = (l ....) (1 YII) (1 + + y. (h" + 1) (h" h..= (1 + 5".) .05. Switzerland.s..1) (z" + 1) .) (1 .s.) + 5...S...5.Z.) II) -2Y21 (1 + y.l1. For definition of B.J (1 .) (1 + 5.) + 5..5.) + 5.. (1 .5.) (1 + 52.) + 5.IM* I ISl112 .) .Z...s.L ]1 - Zo Z. plus sign when B...z....h •.05" -2s" Linvill C Factor .. (1 ~ s.5..1012 IS12s211 .. jS12Sd .) (1 . + 1) (z" + 1) ....y..C =K-1 su= = (1 + -2y" YIl) (1 + y.) .Zo y 1_.h"h" h. 5952-'130 PRI'Hn:O 1M u. Yoyogi..05" (1 + 5..5.= (1 + 5.s. z. c.. max = 1~21 (K S12 ± K=C·1 The h-. and z-psrameters in terms of s-parameters <1 b..z..s..z.