Analog Circuits Cookbook Ian Hickman INEWNES Newnes An imprint of Butterworth-Heinemann Ltd Linacre House Jordan Hill, Oxford OX2 8DP 6<J^ member of the Reed Elsevier group OXFORD LONDON BOSTON MUNICH NEW DELHI SINGAPORE SYDNEY TOKYO TORONTO WELLINGTON First published 1995 © Ian Hickman 1995 All rights reserved. No part of this publication may be reproduced in any material form (including photocopying or storing in any medium by electronic means and whether or not transiently or incidentally to some other use of this publication) without the written permission of the copyright holder except in accordance with the provisions of the Copyright, Designs and Patents Act 1988 or under the terms of a licence issued by the Copyright Licensing Agency Ltd, 90 Tottenham Court Road, London, England W1P 9HE. Applications for the copyright holder's written permission to reproduce any part of this publication should be addressed to the publishers British Library Cataloguing in Publication Data Hickman, Ian Analog Circuits Cookbook I. Title 621.3815 ISBN 0 7506 2002 1 Library of Congress Cataloging in Publication Data Hickman, Ian Analog circuits cookbook/Ian Hickman, p. cm. Includes bibliographical references and index. ISBN 0 7506 2002 1 1. Analolg electronic systems. I. Title. TK7867.H527 621.3815-dc20 94-30919 CIP Printed and bound in Great Britain by Biddies Ltd, Guildford and King's Lynn Foreword Electronics World + Wireless World is undoubtedly the foremost electronics magazine in the UK, being widely read both by professional electronics engineers on the one hand and electronics hobbyists and enthusiasts on the other, both in the UK and abroad. The first article of mine to feature in the magazine, then called simply Wireless World, appeared back in the very early 1970s. Or was it even the late 1960s; I can't remember. Since then I have become a more frequent - and latterly a regular - contributor, with both the 'Design Brief feature, and occasional longer articles and series. With their straightforward non-mathematical approach to explaining mod- ern electronic circuit design, component applications and techniques, these have created some interest and the suggestion that a collection of them might appear in book form found general approval among some of my peers in the profession. This book is the result. Inevitably, in the preparation for publication of a magazine which appears every month, the occasional 'typo' crept into the articles as pub- lished, whilst the editorial exigencies of adjusting an article to fit the space available led to the occasional pruning of the text. The opportunity has been taken here of restoring any excised material, and of correcting all (it is hoped) of any errors in the articles as they appeared in the magazine. The articles have been gathered together in chapters under subject head- ings, enabling readers to home in rapidly on any area in which they are particularly interested. A brief introduction has also been added to each, indicating the contents and the general drift of the article. 1 Advanced circuit techniques, components and concepts Current feedback opamps Many voltage feedback op-amps from the venerable 741 onwards are internally compensated for gains down to unity. As a result, at higher demanded gains, their bandwidth becomes increasingly restricted. Fully or partially uncompensated ('decompensated') types are available, permitting improved performance at a given gain. However, current feedback op-amps offer high-speed performance without special compensation measures, the speed being only slightly reduced as the demanded gain increases. CFBOs: delivering speed at any gain Current feedback opamps have been around for quite a few years now and are winning more ready acceptance with both semiconductor manufacturers and design engineers. Initially, they were introduced by the more specialist analog IC manufacturers such as Comlinear and Elantec, but since then the mainstream semiconductor manufacturers such as Analog Devices, National Semiconductor, Harris Semiconductor, etc. have all come into the market place with offerings. High-speed CFBOs The notable feature of CFBOs (current feedback op-amps) is their high speed and wide bandwidth, qualities that do not deteriorate too markedly as the demanded gain rises from unity to times ten or even more. This is in marked contrast to a conventional VFBO (voltage feedback op-amp) with internal compensation for gains down to unity, such as the 747, as explained in Ogden (1992). Even with a VFBO, the bandwidth reduction with increasing gain is largely avoided when an op-amp with external 2 Analog Circuits Cookbook compensation is used. Figure 1.1 shows the closed loop bandwidth achieved by the earliest popular monolithic op-amp, the 709, when the compensation is correctly selected, according to the set gain. However, although the band Frequency Responca For Output Voltage Swing as R4 Various Closed-Loop Gains a Function of Frequency FREQUENYC (Hz) Tf X(OM)Ui fNC Figure 1.1 The venerable 709 externally compensated op-amp provides a wider (small signal) bandwidth at high gains than even a modern voltage feedback op- amp internally compensated for gains down to unity although the large signal bandwidth suffers at low gains width is maintained remarkably well with increasing gain, it still amounts to only a few per cent of the bandwidth of a CFBO. Furthermore, like a VFBO which is internally compensated for gains down to unity, a CFBO does not need any external compensation components selected according to the required gain (except perhaps when seeking to wring the last ounce of performance from the circuit). This is because there is only one internal node at which there is any significant voltage swing, namely the point immediately preceding the unity gain output buffer, (Figure 1.2). As a non- inverting amplifier, the / terminal is connected to ground via a resistor R s and to the output via a resistor Rj. The input is applied to the high imped ance JVI terminal and the gain is (Rf+ R V^s- Due to the unity gain of the s Figure 1.2 In a CFBO, all gain is raised as current gain, except for the node imme- diately preceding the unity gain output buffer. So there is only one significant lag Rt Ct, rather than two or more as in a VFBO. (Courtesy Analog Devices) Advanced circuit techniques,components and concepts 3 input buffer and the very low value of the amplifier's / terminal internal resistance R{ (tens of ohms), the input voltage at the NI terminal is closely n reproduced at the / terminal, the small difference being equal to I [ times n Ri. Current mirrors cause I[ to flow in R the input impedance (typically n n b several megohms) of the output buffer, giving the device a low-frequency open loop voltage gain of RtlRin- The open loop current gain is simply equal to the current gain of the output buffer. I i is the difference between n the current in R and that in Rj, a CFBO working basically therefore as a s transimpedance amplifier. Testing the LM6181 I recently received a sample of the National Semiconductors' LM6181 CFBO, mounted in an evaluation board and with accompanying data sheet, applications leaflet and SPICE model on floppy disk. The device itself is packaged in an 8 pin DIP outline, with the usual pin-out. The specified typical -3dB bandwidth at a closed loop gain A of +2 and a 1 kQ load is v 100 MHz and it is still well in excess of 50 MHz when driving 100 Q - such as a back-terminated 50 Q cable. The circuit of the evaluation board is as in Figure 1.3, which shows the positions of the links and optional components exactly as found on the board as supplied. I was impatient to experiment with the circuit, and having an oscilloscope with a 300 MHz bandwidth but (at the time) no spectrum analyser, time domain testing was obviously indi- cated. The rise and fall times of the 10 MHz squarewave output of my 10 Hz-lOMHz video oscillator are a leisurely 20 ns, which would scarcely exercise the LM618Ts capabilities. The test circuit of Figure 1.4 was therefore built up, using groundplane construction and a stripline con- nection between pin 11 of the Harris CD74AC00 quad two-input NAND C6 10|i| CllOn (Note 4) rO J1 Vout Vin —ww— R9 Q VAV- R6 • 49R9 (Notel) Q0 2 Figure 1.3 LM6181 evaluation board, as supplied 4 Analog Circuits Cookbook Figure 1.4 Testing the LM6181 's performance, showing the 74AC00 speed-up circuit and extensive use of pads to avoid reflections and standing waves gate and the BNC output socket. This output was connected to a 50 ohm three-way resistive splitter, one output of which was applied to channel 1 of the oscillosope via a coaxial lead with a 10 dB pad at the splitter end and a 20 dB pad at the oscilloscope end. With 30 dB of pads and 6 dB loss in the splitter, the (almost) 5 V pp amplitude out of the NAND gate board might be expected to be 75 mV pp, until one remembers that the scope presents a high-input impedance, leaving the 20 dB pad unterminated. Consequently the channel 1 input is in fact 0.15 V pp (Figure 1.5(a) upper trace) 100 mV/division. As a test signal, it is not as clean as one might wish, but at a little over 2 ns, the rise and fall times are good and fast. The decou- pling included a ceramic chip capacitor connected directly from pin 14 of the NAND gate to groundplane, but the device was socketed, resulting in longer leg lengths and the consequent ringing shown The lower trace of Fig 1.5(a), channel 2 at 200 mV/div, shows the LM6181 output, there being, as in channel 1, a total of 30 dB of pads in circuit, the scope end again being unterminated. The LM6181 drives the output coax via a 49.9 £1 source resistor, which in conjunction with the x2 voltage gain set by R 4 and R 5, should result in unity overall gain. The apparent gain of x2 is due to the fact that the 20 dB pad at the amplifier's input port is also unterminated, the amplifier presenting a high-input impedance due to operating in the non-inverting mode. (Of course the unterminated pads do not result in any significant mismatches, since the input of an unterminated 10 dB pad presents a 20 dB return loss to the source.) The output waveform is a fair copy of the original, though the risetime is clearly degraded somewhat. This turned out to be due to the pole created by Rg Advanced circuit techniques.components and concepts 5 (a) (b) (c) (d; (e) Figure 1.5 (a) Performance of the LM6181 on its evaluation board, at unity gain: upper trace, input waveform (channel 1, 100 mV/div); lower trace, output wave- form (channel 2, 200 mV/div); (b) with R9 shorted; (c) timebase speed increased from 50 ns/div to 10 ns/div; (d) demanded-gain increased by 10 dB; (e) a peaking network of 10 pF in series with 180 Q. connected in parallel with R5 (Rs) 6 Analog Circuits Cookbook and the board and device capacitance at the op-amps's JVI input, pin 3. Figure 1.5(b) lower trace shows the result of shorting /?g; the output risetime is improved, whilst the ringing is slighdy increased compared with the input (upper trace) - a value for R$ somewhere between zero and 200 Q would seem to be optimal. Figure 1.5(c) is the same as Figure 1.5(b) except that the timebase speed has been increased from 50 ns/div to 10 ns/div. The output rise and fall times (lower trace) are about 5 ns. Given that the amplifier's bandwidth when driving a 100 Q load is about 55 MHz, this is a very creditable performance bearing in mind the rule of thumb for a single pole response which says that risetime t (ns) is approximately equal to 350/band- width (MHz). Incidentally, the 16 ns mid-swing delay evident between the two traces indicates the length of the extra coaxial cable in8 the path via the amplifier. Assuming the wave velocity in the cable is 2xl0 m/s (2/3 of that in free space) the length is indicated as 3.2 m, whereas it actually measured only an extra 2.2 m (11 ns). The remainder is due to the tracking on the evaluation board and the 5 ns propagation delay in the device itself. Demanding more gain Obtaining a higher gain with a CFBO means increasing the ratio of Rjio R, just as with a VFBO. But this is achieved by lowering R , not by increasing s s i?f (which would have a deleterious effect upon bandwidth). The increased gain is thus obtained by increasing the drive current to the input buffer portion of the CFBO (see Figure 1.2). Figure 1.5d shows the result of increasing the gain by 10 dB (from x2 to x6 approx), by connecting a 180 £2 resistor in parallel with R; an additional 10dB pad has been inserted s between the splitter and the evaluation board so channel 2 deflection factor is unchanged. Some deterioration in performance is evident, but not nearly as much as would be the case with a VFBO internally compensated for gains down to unity. Would the sort of compensation techniques which have been used for decades by the designers of video amplifiers and oscilloscope deflection stages work with the LM6181? Only a very little experimentation was needed to provide the answer yes, as Figure 1.5(e) shows. Here, a peaking network (consisting of a second 180 Q resistor in series with 10 pF) has been connected in parallel with R . Apart from the s slower rise and fall times, the output trace is an exceedingly faithful copy of the upper input trace. Indeed it is a more faithful copy than either Figures 1.5(a) or (b). As indicated above, when changing the gain of the stage, normally Rjis left at 820 £2 and R adjusted as required. However, there are circumstances s where it may be beneficial to change the value of Rp such as when driving capacitive loads. Load capacitance in conjunction with the CFBO's output resistance adds another pole in the loop, reducing the phase margin and, at Advanced circuit techniques,components and concepts 7 ABaV nd wi-dth- vLs 1 Rp, a*nRd Rs 1*ft Rf & Rs (kA) (a) (b) (c) Figure 1.6 (a) With suitable values of Rf and RS/ the LM6181 is capable of driving a load capacitance as high as 100 pF. (b) Normally, if Rf = Rs = 820 Q,, the LM6181 would oscillate with 100 pF of capacitive load. In this example the feedback Rf and Rs values are scaled to 1.2 kQ, so that the closed loop gain is Av = +2, but the open- loop bandwidth decreases, maintaining adequate phase margin, (c) By scaling both Rf and Rs the closed-loop gain stays constant but the bandwidth changes the extreme causing oscillation. At a gain of +2 and with a 100 pF load capacitance (Figure 1.6(a)), the LM6181 would oscillate if Rj- Rs - 820 £2. Scaling both to 1200 £1 keeps the closed loop gain at +2 but decreases the open loop bandwidth, restoring an adequate phase margin (Figure 1.6(b)). The way that bandwidth changes with /^scaling, for the the case where A v= -1, is illustrated in Figure 1.6(c). As so often in electronics there is more than one way to achieve the desired result. Thus with Rj- R s- 820 Q, even a 48 pF load will result in excessive ringing, Figure 1.7(a), but buffering it from the LM618Ts output with a 47 ohm resistor as in Figure 1.7(b) gives a much more satisfactory result, Figure 1.7(c). Although the illustrations have all related to the non-inverting circuit, the LM6181 and other CFBOs can also be used in the inverting connection. Indeed some (a) (b) (c) Figure 1.7 (a) With Rf = Rs = 820 Q, a 48 pF load will cause excessive ringing, (b) This can be avoided by buffering the capacitance from the op-amp with a 47Q. resistor giving a healthier result (c).
Description: