Philips Research Reports EDITED BY THE RESEARCH LABORATORY OF N.V. PHILIPS' GLOEILAMPENFABRIEKEN, EINDHOVEN, NETHERLANDS

Size: px

Start display at page:

Download "Philips Research Reports EDITED BY THE RESEARCH LABORATORY OF N.V. PHILIPS' GLOEILAMPENFABRIEKEN, EINDHOVEN, NETHERLANDS"

Iris Mathews
6 years ago
Views:

1 , : VOL. 11 No. 3 JUNE 1956 Philips Research Reports EDITED BY THE RESEARCH LABORATORY OF N.V. PHILIPS' GLOEILAMPENFABRIEKEN, EINDHOVEN, NETHERLANDS R 292 Philips Res. Rep. 11, , 1956 A GENERAL-PURPOSE DRIFT-FREE D.e. AMPLIFIER hy S. LANDSBERG *) Summary. A method for drift rednetion in -D.e. amplifiers is described. The advantage of the method is that the circuit which stabilizes the output voltage of the D.e. amplifier does not interfere with its characteristics. This is accomplished by comparing a fraction of the output voltage with the input voltage by means of a mechanical chopper. The chopped difference is amplified by an A.e. amplifier and its rectified output voltage is fed to the D.e. amplifier, so as to reduce the drift. Thus, the zero-point control circuit acts as a voltage stabilizer, with the input serving as a reference voltage. An experimental amplifier constructed on this principle has a gain of 8000, flat from D.e. to 300 kc/so Input impedance at D.e. is of the order of 100 M.n. The stability is of the order of 10 I-lV (with respect to the input terminals) over a period of more than five hours. Résumé Une méthode est dëcrite, qui permet la réduction de la dérive dans les amplificateurs à courant continu. Cette méthode présente les avantages suivants: le circuit stahilisateur de la tension de sortie de l'amplificateur à courant continu n'agit nullement sur les caractëristiques de ce dernier. Cecis'obtient par la comparaison d'une fraction de la tension de sortie avec la tension d'entrée à l'aide d'un rupteur mécanique; la tension différentielle dëcoupée est amplifiëe par un amplificateur à courant alternatif, et sa tension de sortie redressée est appliquée à l'amplificateur à courant continu de façon à réduire la dérive. Le circuit de commande du zéro agit done comme stabilisateur de tension, la tension d'entrée servant de tension de rëfërence. Un amplificateur expérimental construit sur ce principe a un gain de 8000, la courbe d'amplitude étant plate jusqu'à 300 kc/sol'impédance d'entrëe en courant continu est de l'ordre de 100 M.n. La stahilité par rapport aux bornes d'entrée est de I'ordre de 10 I-lV pendant une période supérieure à cinq heures, Zusammenfassung Eine Methode zur Verminderung des Nullpunktverlaufs in Gleichstromverstiirkern wird beschrieben. Der Vorzug dieser Methode besteht darin, dab die Schaltung die die Ausgangsspannung des Gleichstromverstärkers stahilisiert, dessen charakteristischen Eigenschaften nicht beeinfluût. Dies wird dadurch erreicht, daf ein Teil der Ausgangsspannung mit Hilfe eines mechanischen Zerhackers mit der Eingangsspannung verglichen wird, Die auf diese Weise aus der f *) Present address: SI Bar-Giora Street, Haifa, Israel.

2 162 S. LANDSBERG Spannungsdifferenz erzeugte Blockspannung wird durch einen Wechselstromverstärker verstärkt; dessen- gleichgerichtete Ausgangsspannung derart zum Gleichstromverstärker zurückgeführt wird, dab sein Verlauf erheblich verringert wird. Auf diesc Weise wirkt die Nulpunktregelschaltung als Spannungsstabiüsator mit der Eingangsspannung als BezugsgröBe. Ein nach diesen Prinzip entworfener experimenteller Verstärker verstärkt 8000 X gleichmäbig von Null bis 300 khz. Die Gleichstromeingangsimpedanz hat eine GröBenordnung von 100 Mil; die auf die Eingangsklemmen bezogene Stabilität ist etwa 10 flv in einem Zeitraum von mehr als fünf Stunden.. 1. Introduetion The zero-point drift in D.e. amplifiers has caused considerable concern to designers in recent years. Various solutions to the problem have heen found, but in each of the methods involved some properties (such as bandwidth, input impedance or uniformity of frequency characteristic) are sacrificed. In section 1 of this paper the best known methods for drift elimination are reviewed. In section 2 a novel method for automatic zeropoint control is described that makes possible the construction of a drift-free wide-band flat-response D.e. 'amplifier with a very high input impedance. Section 3 contains the description of an experimental amplifier and some of the results obtained. 2. Review of various drift-free D.C. amplifiers a. Chopper-rectifier amplifier 1) In fig. 1 the block diagram of a conventional chopper amplifier is shown.,the signal el is fed over a large resistance RI' chopped by a mechanical chopper and amplified by an R-e-coupled amplifier. The second contact of the chopper serves as a phase-sensitive rectifying element. The output is obtained from the low-pass filter, consisting of Rand C, by means of which the chopping-frequency component is removed. This kind of amplifier is in principle driftless, but its frequency response is practically restricted to a few cycles/second. A chopper amplifier is therefore applicable only in apparatuses that do not require a fast response.. - Fig.!. Block diagram of a conventional chopper D.e. amplifier.

3 GENERAL.PURPOSE DRIFI'-FREE D.C. AMPLIFIER 163 b. Galvanometer-photoeiectric-tube feedback amplifier 2) The underlying principle in this method is to insert ahead of the D.C. amplifier a device that can amplify the signal without drift. Such a scheme is shown in fig. 2. The combination of a galvanometer with a photoelectric tube in a feedback system as shown operates as a servomechanism that counteracts the drift in the amplifier: Apart from the expense and awkwardness involved in the application of a galvanometer, the frequency response is also limited by the inertia of the moving parts and by the stabilizing network required for preventing oscillation in the system. K*' Fig. 2. Block diagram of a galvanometerphotoelectric-tube feedback amplifier. c. Zero-point control with negative feedback and auxiliary amplifier 3) 4) An all-electric circuit for automatic zero-point control, which enables a fairly good high-frequency performance, is shown in fig. 3. The circuit is a combination of an ordinary D.C. amplifier with a chopper amplifier such as shown in fig. 1. Part of the output signal eo is fed back to the input over the feedback impedance Zf. The resulting voltage e is applied to the chopper amplifier and, after amplification, rectification and :filtering, to the second Zf Fig. 3. Basic diagram of a D.e. amplifier stabilized with the aid of negative feedback and an auxiliary chopper ampli fier.

4 164 S. LANDSBERG grid of the double-triode input stage in the proper phase, so as to counteract the drift originating in the D.e. amplifier. At high frequencies, the A.C. amplifier is practically short-circuited by the R-e filter and the frequency characteristic of the D.e. amplifier is not interfered with. At low frequencies, however, the two amplifiers are actually operating in cascade. The result is therefore, at low frequencies, a high loop gain, which is the product of the gains of the two amplifiers, and a much lower gain at high frequencies, when the D.e. amplifier alone is operating. A sketch of the combined loopgain-versus-frequency characteristic is given in fig. 4. The disadvantage of this method is that the input impedance and the gain of the amplifier are dependent on the relation of Z! to Zi. As Z; is determined by the source impedance, the gain (and frequency) characteristic may differ for various source impedances. Furthermore, stabilizing networks must he inserted in many practical amplifiers to prevent oscillation of the system. This restricts the gain-bandwidth product of the amplifier. For these reasons an amplifier operating on this principle is not suitable as a general-purpose D.e. amplifier; the principle may be succesfully applied only in certain circumstances. Loop-Gain "db Fig. 4. Typical overallcharacteristic of loop gain versus log! of amplifier shown in fig. 3.The gains of the D. C. amplifier and the chopper amplifier are G I and G 2 respectively..~ L-~ l~f 88.Sn 3. Reduction of drift by comparing output,with input voltage a. Principle of operatien. The operation of the amplifier described below can best he understdod with the aid of the block diagram shown in fig. 5. The input signal ei is fed directly to the D.e. amplifier. At the output terminal P we have the voltage eo = A(ei + ed); ed is the drift with respect to the input grid. At P, an attenuator is set to have an attenuation A. At the point Q we therefore obtain the voltage ei + ed. Point Q is connected to the vibrating contact of a mechanical chopper. To another contact the input voltage ei is connected through a large resistance Rl' In this manner the drift voltage ed is chopped and fed to the A.e. amplifier. After sufficient amplification this signal is rectified by the third contact,

5 GENERAL-PURPOSE DRIFT-FREE D.C. AMPLIFIER 165 filtered and fed to the control terminal e in the D.e_ amplifier with such a polarity that it counteracts the original drift. The factor by which the origin':' al drift is reduced in this manner is roughly proportional, and may be made equal, to the overall gain G of the A.e. amplifier. Quantitative relations are obtained by the following considerations. Let Ao = gain from e to P; A = total gain of D.C. amplifier = attenuation from P to Q; G = overall gain of A.C. amplifier; ej = residual drift voltage with respect to thè input terminal; e~ = Aed' Fig. 5. Block diagram ofd. C.amplifier with automatic zero-point control. Chopper compares voltage at Q with input voltage at S. The chopped. difference voltage ed is amplified by the A.C. amplifier, rectifiëd and fèîi. to con: trolïërffiillacc.-- } Assuming ei = 0, we may write (1) Thus I eo eo = --G-:-A-' 1+_0 A (2) and ed ed= ---- GAo l+- - A (3) If the control terminal C is situated at a low signal level in the amplifier, we may have the situation Ao = A. In this case (3) reduces to eá = ed/(l + G). b. Elimination of interference from A.C. amplifier Several kinds of interference with the normal operation of the D.C. amplifier may originate in the A.C. amplifier and its associated input circuit. These, and the steps taken to eliminate them, will now be described. bol. Interference of A.C. signal at chopping frequency This is the most obvious interference originating directly in the A.C.

6 166 S. LANDSBERG amplifier. Without the R-C filter (see fig. 5), it would amount to an interfering signal at the chopper frequency of V~ = ea GAD= Aed at the output terminal. If io is the angular frequency of the chopper and T = RC is the time constant of the filter, the residual interference at the output terminal will he approximately (4) With respect to the input terminals this amounts to ed/wt. The factor rot can easily he made of the order of 10 4 Because ed is of the order of 10 mv, the residual interference ed/rot will he of the order of one microvolt. h-2. Common-mode signals from A.C. amplifier Ideally the A.C. amplifier should he sensitive to. the difference (ei + ed) - ei = ed only. However, this will not he the case unless special steps are taken. The input signal ei is actually applied to the A.C. amplifier as well. At low frequencies, a certain signal will therefore reach point C, thus distorting the frequency characteristic of the D.C. amplifier in this region. Another un7 desirable effect due to the same source is the detection of signals at the chopper frequency which may produce adirect voltage at C, thus shifting the zero point. In addition, signals with a frequency close to the chopper freque~cy will force the zero point to oscillate at the difference frequency. h-3. Unequal delay at the chopper terminals A step-function voltage applied at the input appears at Q with a delay determined hy the handwidth of the D.e. amplifier. At point S, however, the signal is partly de1ayed, in accordance with fig. 6. The time constant that determines this transient is given hy T = Cl(Rl + Rg). The result is that during a certain amount of time determined *) hy T there exists a voltage difference across the chopper terminals Sand Q, thus causing a disturhing interference. h-4. Input circuit of A.C. amplifier and attenuator circuit All the undesired effects mentioned previously can he practically eliminated hy a suitahle design of the A.C.-amplifier input and attenuator circuits. This will he explained with the aid of fig. 7. The A.C. amplifier is provided with a low-pass filter (consisting of Rl' R 2 and Cl) at its input circuit, This filter serves a double purpose: (I) it attenuates spurious signals at or *) We assume here that the delay caused by the D.e. amplifier is very small compared with the time constant T..

7 GENERAL-PURPOSE DRIFl'-FREE D.C. AMPLIFIER 167 ~--~ e. Rg I~ Fig. 6. Response of e. when a step voltage ei is applied at the input. near the chopper frequency, and prevents their interference described in b-2; (2) currents at chopper frequency flowing from the chopper through Rz and RI and through the source impedance are also attenuated, thus preventing these signals from reaching the D.C. amplifier. The attenuator circuit is arranged with the aid of two cathode followers in cascade. The first cathode follower provides the output terminal of the D.C. amplifier. To this terminal a network, identical with the input network, is connected. Its output voltage is applied to the ~econd cathode follower, to which the attenuator consisting of R6 and R7 is connected. Thus equal delays are obtained for.the two terminals of the chopper. In order to eliminate the common-mode effect in the A.C. amplifier, the cathode ofthe input tube is returned to earth through a small resistance R4' which is incorporated in the delay network of the attenuator. This resistance R4 is so adjusted that the cathode and grid of the input tube are driven with nearly equal signals *). Fig. 7. Input and attenuator circuits for preventing common-mode signals in the A.C. amplifier and providing equal delays for both chopper terminals. *) In practice R4 is so adjusted that the common-mode signal at die anode of the input tube is just eliminated

8 168 S. LANDSBERG 4. Description of experimental amplifier The construction of the experimental D.e. amplifier is straightforward. It consists of two cathode-coupled stages in cascade with cathode followers at the input and output in order to provide high input impedance and low output impedance respectively. The amplifier stages are coupled by voltage dividers consisting of constant-current devices in order to avoid loss in gain between stages. High-transconductance tubes are employed in order to ensure a large bandwidth with a minimum number of stages. One control grid in the second stage serves as the zero-point control terminal C (see fig. 5). The chopper consists of a standard 70-c/s self-oscillating vibrator. Because of the automatic zero-point control, ordinary resistors of 5 and 10 per cent accuracy could be used throughout the amplifier. Moreover, all the filaments are fed directly from mains transformers. The anode supply is regulated. The results obtained can be summarized as follows. Fig. 8. Recording of zero-point voltage over a period of one hour. The f1.v scale is given with respect to the input terminals of the amplifier. The same stability was maintained over a ~ period of 5 hours. (a) Drift The drift over a period of five hours is not larger than 10 f.1.v with respect to the input terminals. The settling time of the zero point within these limits is ofthe order of one minute after switching on. However, fluctuations with an amplitude of about 50 f.1.v (peak value) are present. A recording of the zero point over a period of one hour is shown in fig. 8. In comparison, a recording of the zero point of a conventional commercial D.e. amplifier is given in fig. 9. The stability of the amplifier at various signallevels can be

$GENERAL-PURPOSE DRIFT-FREE D.e. k\[plifier 169 seen m fig.$

9 GENERAL-PURPOSE DRIFT-FREE D.e. k\[plifier 169 seen m fig. 10, which shows the input voltage of the amplifier with step variations of approximately 1 mv at the input over a period of 5 minutes and a sudden return to zero input voltage. (b) Hum With short-circuited input circuit the hum IS less than 5 [LV. Fig. 9. Recording of zero-point voltage of a conventional commercial D.e. amplifier. The flv scale is with respect to the input terminals of the amplifier. Upper trace was made after resetting the zero point. / Fig. 10. Recording of output voltage with gradual increments of 1 mv (approximately) over a period of 5 minutes and a sudden return to zero point.

170 s. LANDSBERG 14 m sec. a 1 m sec b 4 [lsec e Fig. ll. Response of amplifier to square waves of constant amplitude and different duration. Outpnt voltage is 80 V.

10 170 s. LANDSBERG 14 m sec. a 1 m sec b 4 [lsec e Fig. ll. Response of amplifier to square waves of constant amplitude and different duration. Outpnt voltage is 80 V. (c) Gain The gain is 8000 (or push-pull) and the frequency characteristic is flat from zero to 300 kc/s (- 3 db). Figures Ha, Hb and He show the response to square waves with constant amplitude at various frequencies. Maximum output is 120 V (or 240 V push-pull).

11 GENERAL-PURPOSE DRIFT FREE D.e. AMPLIFIER 171 (d) Input "impedance The input impedance at very low' frequencies is determined by the grid current of the cathode follower (about 10-8 A). At high frequencies it dr~ps to the value of the input impedance of the filter in the A.C.-amplifier input circuit. The chopper circuit does not draw current from the signal source. (e) Output impedance The output impedance is determined by the design of the output cathode follower. In the amplifier described it amounts to 200 n. (f) Linearity A very high degree of linearity is obtained at low frequencies, since deviations from proportionality between input and output are corrected in the same way as drift. Up to 100 V output the deviation from a linear relation between input and output is less than one per cent. At high frequencies moderately good linearity is maintained because of the push-pull operation throughout the amplifier. : 5. Conclusion The method of automatic zero-point control described in this paper is characterized by the fact that it does not interfêre with the general properties of the D.C. amplifier. Properties such as D.C. mput impedance and shape of the frequency-response characteristic ~re not inhuenc~d by the zero-point control circuit. This is made possible hy the good common-mode rejection of the A.C. amplifier.. The fact that the zero-point control circuit is a voltage regulator whi~}l reacts only to drift variations makes it possible to use even larger gainbandwidth products in this feedback loop, which would reduce the fluctuations seen in fig. 8 still further. This could be attained by employing a chopper with a higher frequency. In the construction of the amplifier no critical adjustments or precision components are required and all heaters are fed directly from mains transformers. The extra cost of the zero-point control circuit is thus considerably compensated. Eindhoven, March 1956 REFERENCES 1) A. J. Williams, Jr., R. E. Tarpley and W. R. Clark, D.C. amplifier stabilized for zero gain, A.LE.E. Transactions 67,47-57, ) G. E. Valley, Jr., H. Wallman, Vacuum Tube Amplifiers, Radiation Laboratory series, Vol. 18, Sec ) Edwin A. Goldberg, Stabilization of wide-hand direct-current amplifiers for zero gain, RCA Review 11, , ') Frank R. Bradley and Rawley McCoy, Driftless D,C. amplifier, Electronics 25, ,

, ' ,,.,EDITED BY THE RESEARCH LABORATORY OF N. V. PHILIPS' CI.OEILAMPENFABRIEKEN, EINDHOVEN, NETHERLANDS

, ' ,,.,EDITED BY THE RESEARCH LABORATORY OF N. V. PHILIPS' CI.OEILAMPENFABRIEKEN, EINDHOVEN, NETHERLANDS J -~-~--:-- --'-- ~~-~.~- -~~... - /. './ - _... ' I~. VOD: '5. NR i. FEBRUARY 1950 Researèh Reports.. EDITED BY THE RESEARCH LABORATORY OF N. V. PHILIPS' CI.OEILAMPENFABRIEKEN EINDHOVEN NETHERLANDS R