The MOS Tetrode - the the Triodes Adjust Themselves for the Different Cases Feedback Capacitance

134 PHILlPS TECHNICAL REVIEW VOLUME 30

T.Okumura

In the various designs of MOS transistor that have so far been produced the feedback capacitance is relatively high. This means that they are less suitable for amplifying signals 'of very high frequency, The article below describes a related circuit element which can be usedforfrequencies up to 200 MHz and higher. This new device has been developed at the Takatsuki-Osaka laboratory of the M atsushita Electronics Corporation (MEC) a company ownedjointly by the Japanese company Matsushita Electronic Industries (MEI) and the Philips Group of Companies, and active in various fields of common interest.

Introduetion The most advanced of the various field-effect tran- the anode current in both valves is about the same. If the grid G2 sistors known at present is the metal-oxide-semicon- is decoupled to prevent the appearance of an a.c. voltage, then the circuit behaves like a triode of very high internal impedance ductor (MOS) transistor. MOS transistors are relatively (anode impedance) Re and very high amplification factor ,."t; simpleto manufacture and they combine the advantages the transconductance gm t is about the same as that of one ofthe of high input impedance and characteristics that give triodes. In this case, which corresponds to the situation in the very little cross-modulation. However, they are not MOS tetrode, we have [3]: very suitable for amplification at frequencies above (1) about 100 MHz because of their relatively high feed- (2) back ("Miller-effect") capacitance. ""t = /-111-'2 + 1-'1 R; ""1""2 The same problem is encountered in thermionic and devices in the triode valve. The usual solution adopted , . . (3) there has been to replace the triode either by a pentode or by a pair of riodes connected in a "cascode" circuit. from which these statements can readily be verified. In solid-state electronics an analogous arrangement to the cascode has been made up from two MOS transistors, and this gives a feedback capacitance much less than that of a single MOS transistor. At Matsushita Electronics Corporation we have been looking into Fig. 1. A valve cascode circuit the possibility of producing a practical circuit element using triodes. The circuit can of even better performance by making a casco de of be adjusted in such a way that the characteristic curves are two MOS transistors as a solid circuit [1]. Our invest- analogous to those of a single igations have been successful and their practical end- triode. The internal impedance and the current amplification result is a new circuit element, the MOS tetrode [2]. This factor are however much higher, new device does indeed have the desired very small and the feedback from anode to control grid is much less. The feedback capacitance, and it also surpasses the ordi- transconductance is about the nary MOS transistor on two other counts: in circuits same. with automatic gain control (AGC)-readily produced with the MOS tetrode, as we shall see below - there is much less cross-modulation with the tetrode, and there Fig. 2 shows a diagram of the cross-section of one is much less variation of the input capacitance with of our designs of a MOS tetrode, next to a section of an input voltage. The characteristics are also very stable. ordinary MOS (triode) transistor. The diagram shows that the tetrode has two control electrodes (gates) GI In a valve cascode circuit (jig. 1) the operating conditions of two triodes are such that' the grid current is very small, and and G2. Between the two gates there is an N-type region which' is surrounded by the P-type silicon of the sub- N-type Dr. T. Okumura, is with the- Reseárch Laboratory of Matsushita strate, This region is the "island", which serves Electronles Corporation, Takatsuki, Osaka, Japan, as the drain electrode for the triode formed by S, GI 1969, No. 5 MOS TETRODE 135

on Vg, and the same is therefore true for the current Id which flows through it from S to D. GdD~ ~ If curves of the variation of this current with the ~ ." voltage Vd - Vs are plotted for constant Vg, it is found a that the first part of each curve is parabolic and the remainder is very nearly a horizontal straight line (fig. 3; Vs is assumed to be zero, as usual). The maximum value Id sat of the current is the ordinate value of the straight-line region and occurs consequently in a large interval of Vd values (saturation). The value of Id sat is dependent on Vg (fig. 3b). In the "saturated" region Id and Vg are related at constant Vd by b lel sat = 1- fJ(Vg - Vth)2 (4) Fig. 2. a) Diagrammatic cross-section of a MOS transistor (MOS triode). S source. D drain. C metal gate electrode, insulat- to a very good approximation. Here fJ is equal to ed from the other parts of the device by a thin oxide layer O. Cfl-JI2, where C is the capacity between the gate Sand D are of the same semiconductor material as the rest of the substrate (shaded) but are of the opposite conduction type. electrode and the channel, fl- the mobility of the major- b) Diagram of the cross-section of a MOS tetrode on a P-type ity carriers (the electrons in this case) and I is the length silicon substrate. There is a third N-type region, the island J, between Sand D. There are now two gate electrodes, Cl and G2. of the channel. Vth is the threshold voltage, i.e. the The device can be considered to be an integrated cascode circuit value of Vg for which Id becomes zero. The variation of two MOS triodes. Each of the two circuit elements has a substrate contact which of the current with Vg is thus quadratic (fig. 3c). It is usually connected to the source. The standard symbols can be shown directly from (4) that gm sat, the value for use in circuit diagrams are shown on the left. The arrow represents the source. of the transconductance under saturation conditions,

°o~--~~------vth -~ Q b fig. 3. a) The Id- Vd characteristic for a MOS triode for a fixed value of Vg. The curve can be fairly accurately represented by a combination of a parabolic region and a nearly horizontal straight line. b) Set of J,,- Vd characteristics for various values of Vg. c) Ie- Vg characteristic for constant Vd (see the dashed curve in b). At Vg < Vth the current fd is virtually zero. The parabolic and the horizontal part of the Ie- Vd characteristics meet at Vd = Vg - Vlil.

and I and as the source electrode for the triode formed IS given by: by T, G2 and the drain electrode D. The island has no gmsat = ('Md satJàVg) = j3(Vg - Vth). (5) contacts. In most applications the signal to be ampli- fied is applied to Gl and a capacitor is connected be- The transconductance gm sat is thus not independent tween the electrode G2 and earth. The substrate has a of Vg. contact (not shown), which is connected to the source. It is the very nearly quadratic relation between Id sat Before describing the special electrical and techno- and Vg which accounts for the very small cross-modu- logical features of our new tetrodes, we should perhaps lation obtained from the MOS transistor [41. If a recapitulate some ofthe more important features ofthe ru Characteristics and production methods for solid circuits are MOS triode. The fundamental mechanism in the oper- discussed in: A. Schrnitz, Philips tech. Rev. 27, 192-199, 1966. ation of a MOS triode (fig. 2a) which has a P-type [2J During the development of our tetrode some details of similar work were given by N. H. Ditriek and M. M. Mitchell at a meet- substrate, is the formation at the substrate surface of a ing ofthe IEEE Electron Devices Group in October 1965. thin N-type layer when a positive bias Vg is applied to [3J See G. Klein and J. J. Zaal berg van Zeist, Precision electronics, Centrex Pub!. Co. Eindhoven 1968, Chap. 18. the gate (inversion). The charge density of the mobile [4J This is discussed by P. E. Kolk and I. A. Maloff, Electronics majority carrier in this layer, the "channel", depends 37, Dec. 14, 1964, p. 71. 136 PHILIPS TECHNICAL REVIEW VOLUME 30

radio receiver is tuned to a, particular frequency, the modulation of the signal received is affected to some extent by the modulation' of other transmit- From this relation and the similar one for Id2 it fol- ters, particularly those at neighbouring frequencies. lows that The magnitude of this effect is mainly determined by the third-power term --:- and to .a lesser extent by the ViSI = -!-[Vgl+ Vg2 :- 2 Vth - higher-power odd terms - in the expansion of Id sat as - {(Vg! + Vg2 - 2Vth)2 - 2(Vg2 - Vth)2}t], (9) a power series in Vg. For the MOS transistor these and terms are very small and the cross-modulation is there- (3 fore correspondingly small. . !ct = '8 [Vg2-- Vgl + {(Vg! + Vg2 - 2Vth)2-:- MOS transistors for use in computers usually have - 2(Vg2 - Vth)2}t]2: (10) a substrate of N-type silicon, arid hence a P-type channel. In these transistors Vth usually has such a value Let u~ now look at a number of special cases. When that the transistor does not conduct when there is Vg! is equal to Vth, the first triode is cut off and the no bias on the gate, corresponding to the usual require- potential ViSIOfthe island becomes equal to Vg2 - Vth; ments in digital applications. MOS transistors for this means that the second triode is just driven into amplifying high-frequency signals, on the other hand, cut-off. If Vg! is allowed to increase indefinitely, then, are preferabl~ made on a P-type substrate. The channel as (9) shows, VISI goes to zero (i.e, Visi becomes equal • then shows N-type conductivity, which is more suitable to Vs) and Id goes to the limiting value -!-(3(Vg2 - Vth)2. hete since the electrons in silicon have a mobility about three times that of the holes. (It is true that the mobility in the layer obtained by inversion is smaller than it is in the bulk materialof similar type, but the relative I, \ I decrease. isabout.the same for both kinds of charge carrier [51.) TDI-..:.J.I.!A--t+--I---l The operation of a MOS tetrode mA 8~~~r4-~-~ Let us now calculate how the current in a MOS Id tetrode depends on the various potentials and charac- 6 teristic quantities. We shall proceed as if the tetrode i was a cascode circuit oftwo triodes : in fact this assump- tion is not quit~jüstified since a tetrode has one substrate contact, while the substrate of each triode of a cascode circuit is connected to its own source elec- TDV trode. We shall also assume that the two triodes are identical, so that they each have the same (3, Vth, etc. If an equation analogous to (4) is set down for the Fig. 4. Operating curves for a pair of MOS triades in a cas- cade circuit, those for the first triode on the left and those two triodes, then since both transistors carry the for the second on the right. The drain potentialof the first tri- same current it can be shown that ode, which is also the source potential for the second, is labelled Visl, since it corresponds to the potentialof the "island" of a tetra de. The curves for ViSl and Vd- ViSl apply for constant . . . . . (6a) Vg2. At a lower value of Vg2 this curve takes up the position indicated by the dashed line, and at higher Vg2 it takes up the or position shown by the dotted line. . (6b)

and This limiting of !ct has advantages on the score of (7) reliability, since a high voltage applied at the gate cannot set up a current large enough to damage the This equation for the saturation current of the tetrode. tetrode is identical with the one that holds for the Finally, we should say a word or two about the most first triode (see eq. 4). The diagrams oî fig: 4 show how important quantity of all in the MOS tetrode - the the triodes adjust themselves for the different cases feedback capacitance. In the MOS triode the feedback (èf. eq. 6b). action arises because the gate electrode overlaps the When Vgl is greater than Visl + Vih, the current drain. electrode a little (fig. 5a). In the MOS tetrode in the fi~st t~i~cie is no longer' saturated, and is given by: the main cause of feedback is that any a.c. voltage 1969, No. 5 MOS TETRODE 137

whose characteristics were like those of fig. 3, and for which D Vtl! = - 2.7 V, we found a value of 6.2 V for Vis! at Vg2 = 5 V, Vd .... ~D as compared with a calculated value of 7.7 V. A much smaller " discrepancy was found when we used triodes made on a much Cfb rl less strongly doped silicon substrate (20 ncm). This agrees with /-- I the calculations of Van Nielen and Memelink [6], from which it G C,••o{/ Vist ",' can be shown that there will be no discrepancy at all in transis- I tors made on an intrinsic subtrate. 1~" ~ 1 Q 5 5 12. Some suitable configurations

Fig. 6 shows a diagram of the cross-section of two c"O}~[ of our experimental MOS tetrodes, which fulfilled r- expectations in every respect. The configuration shown in fig. 6a is very simple and does not really need 1 0':61-_ , __ ~ further explanation. In the configuration of fig. 6b it 1 o ~--~W~--~70~O~--~7~OO~OkQ can be seen that the oxide layer does not have the ---- Rg2 same thickness everywhere. Because ofthe step in thick-

Fig. 5. a) The small stray capacitances in a MOS tetrode between ness the two triodes have different values of C (the G and D and between G and S arise mainly because G slightly capacitance is of course inversely proportional to the overlaps the two other electrodes. The stray capacitance Crb between G and D is responsible for the feedback action. b) Dia- thickness of the oxide layer). In the tetrode of fig. 6b C gram referring to the calculation of the feedback capacitance is greater for the second triode, and it can be shown of a MOS tetrode. c) The feedback capacitance Crb as a function of the resistance in the lead to the second gate electrode. 1f this resistance is greater than a few tens of kn, the advantages that can be gained with the tetrode are partly lost.

at D sets up a corresponding (but much smaller) a.c. voltage at J, and this in turn is coupled back to Gl (fig. Sb). The direct effect of the potentialof Gl on that of D is very small, since G2 functions as a screen- ing electrode, provided that the resistance in the lead of G2 is not too high (fig. Se). The feedback via the island can be calculated as follows (fig. Sb). If the feedback capacitance of the tetrode is represented by Crb and the a.c. current in- duced in Gl by an a.c. voltage VI superimposed on Fig. 6. Diagram of the cross-section of two types of experimental MOS tetrode. In the lower diagram the oxide layer of the Vd is represented by is, then: second tetrode is thinner than that of the first one, which gives a smaller feedback capacitance and also makes it easier to apply (11) automatic gain control (AGC) via G~. or

Crb = CIVI/Vd. that this has the effect of making its current ampli- Since the same current flows in both parts of the fication factor fl greater. It follows that for this confi- tetrode, VI/Vd is equal to R; divided by the total guration Crb will be even smaller than it is for that resistance of the cascode circuit, which is equal to of fig. 6a. This arrangement also has the advantage Rl(l + fl2) + R2. Consequently: that a smaller voltage is needed at G2 when the tetrode is used for AGe. As in other tetrodes the cross-mod- Rl Crb= Cl---·----- R,j CI/fl2. (13) ulation in reverse AGC with Vg2 (see below) is less Rl(l + fl2) + R2 than in reverse AGe with Vgl, so that this configuration This shows that in the MOS tetrode the feedback combines a number of attractive features. capacitance is about fl2 times smaller than that of the A variation of the tetrode drawn in fig. 6a is now triode. [5] O. Leistiko, A.S. Grove and C. T. Sah, IEEE Trans. ED-12, 248, 1965. If one attempts to verify experimentally that ViS! becomes [6] J. A. van Nielen and O. W. Memelink, The influence of the equal to Vg2 - Vtl! when Vgl is made equal to Vth, as the relation substrate upon the DC characteristics of silicon MOS tran- (6) predicts, a fairly large discrepancy is found. With triodes sistors, Philips Res. Repts. 22, 55-71, 1967. 138 PHILJPS TECHNICAL REVIEW VOLUME 30

in production (3 SK 32 MOS tetrode). Its characteristic features will be described below; a diagrammatic cross-section and the method of making it are shown Q in fig. 7. The cross-section has been made different from the one of fig. 6a to reduce as far as possible the effect of the overlap of the two gate electrodes ~:~. nëillnllilliTIfB:9 on S, I and D. Fig. 8 is a microphotograph of the surface of a silicon wafer on which a tetrode like that of b fig. 7fhas been produced.

Experiments have also been carried out with two other configurations besides the ones shown in fig. 6. The first is like the configuration of fig. 6a, but with the island left out. We found that tetrodes of this type were not very successful; the potentialof the oxide between Gl and G2 changes when Vgl

or Vg2 changes, but with a very high time constant. The conduc- Fig. 7. Stages in the production of the Matsushita 3 SK 32 tivity of the channel between Gl and G2 is therefore not uniquely MOS tetrode. a) Oxidation of the silicon substrate. b) Oxide is etched away in a particular pattern and phosphorus is diffused determined by Vgl and Vg2. into the substrate to obtain source, island and drain. c) Removal The fourth configuration that we have investigated [7] is drawn of the oxide layer. d) Second oxidation; the oxide between S, I in fig. 9. Here the oxide layer of the second triode is thicker than and D is etched away. e) Third oxidation. I) Oxide is removed that of the first one. This has the result that Vth is higher for the from parts of Sand D and aluminium is deposited. second triode than for the first, which means that the island potential is fairly high even at Vg2 = 0; consequently at Vg2 =0 the current lel is so large that a fairly high transconductance is obtained.

The Matsushita very high frequency MOS tetrode type 3 SK 32 The 3 SK 32 MOS tetrode, whose configuration has just been described (fig. 7 and 8), is intended for use at VHF (30 to 300 MHz), and is herrnetically sealed in a standard TO-72 package. The Id- Vd characteristics of a sample of this type are reproduced in

fig. ID. The value of Vth is the same for both triades

20 mA

Vgr=2V /' Fig. 8. Microphotograph of a silicon wafer on which a 3 SK 32 I1 MOS tetrode has been produced. Starting at the outside, the 7.5 yellow lines are the source, the first gate, the second gate and the J/' drain. The four dots are the connections. The four lead wires 70 are blurred but can just be made out. 1/ 7 V 1 I 0.5_ V 0 V 70 20V -t{j

Fig. 9. Experimental MOS tetrode in which the second triode Fig. 10. Set of Id- Vd characteristics for the 3 SK 32 MOS is the one with the thicker oxide layer. tetrode for Vg2 = 5 V. At Vgl = 0 the current Id is about 1.5 mA. 1969, No. 5 _MOS TETRODE 139 and is in the region of - 0.6 V.O At Vgl' 0 the satura- T6r------, tion value of Id is about 1 mA, and this means that a mA/V Vci=lQV bias on Gl is desirable in most applications (enhance- T4 ment mode "operation). gm When Vd is set equal to zero and Vgl and Vg2 are made so strongly negative that there is no current f 12 in the tetrode (Vgl = Vg2 = -10 V), the small-signal input capacitance and output capacitance are 4.5 pF 10 and 3 pF respectively at a frequency of 455 kHz. 8 The variation of the feedback capacitance Crb as a 8 function of Id is shown in fig. 11. The value of Crb is about 0.02 pF, except at very low Id. In general Crb is never greater than about 0.035 pF. 6 6 4

a03 pF 2 a~~ ~ ~

Cfb j0.G1 Fig. 12. The transconductance gm of the 3 SK 32 tetrode as a function of Vgl for nine different values of Vg2, with Vd = 10.V!. °0~--~--~~--~6~--~8~--~ro~m-A~ (: . -Id

Fig. 11. The feedback capacitance Crb for the 3 SK 32 tetrode as a function of Id, at a frequency of 445 kHz, with Vd = 10 V and Vg2 = 5 V.

20 dB / / / I Fig. 12 shows how the transconductance varies I G---. / F-~ with Vgl, with Vg2 as a parameter. It can be seen that I I all the curves have a maximum, which means that gm ITV / can either increase or decrease with increasing Vgl. \\ / »: ~ / .. A reduction in Vg2 always gives a reduction of gm. 5 ~------3V _- _--. _~ This means that there are in principle three ways of - - - --=--=--=--=- - ~2=5V - obtaining automatic gain control: two using Vgl and one using Vg2. The most attractive method of the three %~------~5~------~T~0------+'$~mA is the one using Vgl - which is chosen to give an -Id =, operating point in the downward-going part of the curve Fig. 13. The power gain G (solid lines) and the noise figure F - since the magnitude of the variation of gm with Vgl (dashed lines) of the 3 SK 32 tetrode, measured at 200 MHz. Both quantities were measured under optimum conditions and can be anywhere within certain limits; all that has to with a bandwidth of 4 MHz. be done here is to set Vg2 to a suitable value. The fact ;." that a fairly high Vgl has to be used in this method of control so that Id also has a rather high value, is a disadvantage. Fig. 13 shows how the power gain G and the noise The cross-modulation characteristics for. circuits figure F of the 3 SK 32 tetrode depend upon Id at using two methods for obtaining A9C_ are shown the frequency 200 MHz and for three values of Vg2. .in fig. 14. The curves were obtained with an untuned For still higher values of Vg2 the measured curves al- circuit which had two signals applied to it, one with most coincide with those for Vg2 = 5 V. It can be a carrier frequency of 200 MHz and the other with a seen that for normal conditions and optimum matching . . ~ the power gain at 200 MHz is 20 dB or more. _ < _ [7) See also Electronics 39, May 16,)~66, p. 212-213._ 140 PHILlPS TECHNICAL REVIEW VOLUME 30

1V'r------~------~--~ ture at Vd = 10 V and Vg2 = 5 V. At -Vg! = 0 V (or Id = 1.5 mA) the effect ofthe temperature is almost zero, for Vg! < 0 the temperature coefficient is positive but very small, and for Vg! > 0 it is negative. This negative temperature coefficient is another feature which is useful for reliability. (Bipolar transistors do not have this attractive feature, and the current in these is also more strongly temperature-dependent than the current in MOS transistors.) As in other MOS transistors the Vth of our tetrode

Vct=10V 0.01

6pF

Ci o 10 L 20mA ----__ 11rconst.(1.4 V) -d -_-_ Fig. 14. The interfering signal Vlnt (frequency = 212 MHz), t 5 --_ which gives a cross-modulation of 1% in a signal at a frequency of 200 MHz in the 3 SK 32 tetrode, shown as a function of Id for Vd = IOV. The solid line applies for constant Vg2 - i.e. the normal situation with the tetrode driven at Gl - and the dashed line refers to constant Vgl.

\{{=10V carrier frequency of 212 MHz. It can be seen that for 200MHz normal values of Id a cross-modulation of 1 % occurs when the interfering signal is 0.1-1 V, and that the case with fixed Vg! is the most favourable one.

In MOS triodes and tetrodes the input capacitance Fig. IS. The input capacitance Cl of the 3 SK 32 tetrode at a Ci is not entirely independent of the applied voltages. frequency of200 MHz as a function ofthe current Id at constant Vg2 (5 V) and constant Vgl (1.4 V), with Vd = 10 V. The varia- Since this capacitance can contribute to the tuning of tion of Cl with Id is smallest at constant Vgl. a circuit, it is desirable that the variation of Ci should be as small as possible. Fig. 15 shows how Ci varies with Id; in one case Id is controlled by Vg! and in the other it is controlled by Vg2. It can be seen that the most 25 favourable situation is the one in which Vg! is held mA constant. In this respect AGC via G2 is therefore preferable. 20 Since the mobility of the charge carriers and the Id potential Vth both depend on temperature, the current Id and the transconductance gm are also to some extent t T5 temperature-dependent at constant Vd, Vg! and Vg2. 2.0 Conditions can be found for which the two effects TO cancel one another out. If rJ.p, is substituted for (J, it 1.5 can be readily shown from (4) and (5) that the two temperature coefficients are given by: T.O 5 0.5 bh Id bp, V·----b Vth ------= --- 2 rJ.p,Id-- (14) 0- er P, sr sr ' -0.5 ~20 0 20 40 60 80 TOO T200C

bgm gm bp, ö Vth -T - =---rJ.p,--. . . . (15) sr P, bT »r Fig. 16. Variation of Id with channel temperature at constant Vd and Vg2 for a number of values of Vgl. When Vgl is small, Id Fig. 16 shows how Id varies with the channel tempera- is almost independent ofthe temperature. 1969, No. 5 MOS TETRODE 141 is not absolutely constant. Fig. 17 shows values of tion, although it was not very large (about 0.05. V in Vth recorded for four samples over a period of 150 hours), and after that Vth remained virtually 1000 hours. At first there was a relatively rapid varia- constant for all four tetrodes.

0.5~ ~=+10V _~ 0.4 __ _ ,,_ _ _ f~;------~------~~:--10V------

at °0~--~--~2~OO~---L--~40~0~--~--~6~070--~----8~0~0~--~~t==OOOh -t

Fig. 17. A record ofthe variation ofthe values of Vtll measured over a period of 1000 hours for four samples ofthe 3 SK 32 tetrode at a temperature of 150°C. During the experiments S was connected to D and Gl to G2. The voltage between the two pairs of electrodes was +10 V for two of the tetrodes, and -10 V for the other two.

Summary. The MOS tetrode, an integrated cascode circuit of -0.6 V; at Vgl = 0 the current Id is about 1.5 mA, so that a two MOS triodes, has been developed to obtain a MOS device bias is needed for certain applications. The power gain is 20 dB with such a low feedback capacitance that operation in the VHF or more at 200 MHz. The transconductance (about 10 mA/V) band would be possible. The slope is about the same as for a MOS has a maximum for a particular value of Vgl and also depends triode, and the current amplification factor and the internal im- on Vg2, so that there are three possible ways of obtaining AGC. pedance are very high. In normal use the gate G2 of the second The feedback capacitance is about 0.02 pF. The cross modula- triode is kept at a fixed potential (say zero) and the tetrode is tion is very low. At low Vgl the current Id is very nearly inde- driven at Gl. In the 3 SK 32 VHF tetrode developed by Matsu- pendent of the temperature. During the first 150 hours of life shita Electronics Corporation the threshold voltage Vtll is about Vtll changes a little, but after this changes hardly at all. .