CIRCUIT THEORY A N D APPLICATIONS, VOL. 2, 13-22 (1974) ANALYSIS OF NONLINEAR MODEL FOR OPERATIONAL AMPLIFIERS IN ACTIVE RC NETWORKS ANDERS FORSEN AND LARS KRISTIANSSON Telecommunication Theory, Royal Institute of Technology, Stockholm, Sweden SUMMARY In active RC networks certain nonlinear phenomena may occur due to the nonlinear properties of the amplifiers. It is shown that one of these, the slew rate limitation, may give rise to the jump phenomenon. The class of networks to be studied are described by a third order nonlinear differential equation. The theoretical and experimental results are found to correspond well. INTRODUCTION While studying active filters in the higher part of the LF spectrum and with moderate Q ( ‘Y 5 ) certain phenomena arose which led to a study of methods of calculation for nonlinear analysis of active filters. A mathematical model for slew rate limited amplifiers in RC networks is derived and analyzed. The model is sufficiently accurate for study of one of the phenomena, the sudden jumps in the output amplitude which may occur when a sinusoidal input slowly varies in frequency or amplitude. This so-called jump phenomenon is known in other contexts. For instance a second order filter with nonlinear inductance permits an amplitude jump in the voltage across the inductance. The phenomenon is called in this context ferroresonance and is described by a second order nonlinear differential equation, Duffing’s equation. In our case the nonlinearity consists of the limitation of slew rate in the operational amplifier, which, with the other components of the active filter, leads to a higher order nonlinear differential equation. The following notation will be used throughout the analysis: Lower case letters : time functions Upper case letters : Laplace transforms In the figures upper case letters will be used when Laplace expressions are present. OPERATIONAL AMPLIFIER MODEL By using the Miller effect one can compensate an operational amplifier with only one capacitor. Using this method the first differential stage exhibits a dominant pole, so that the gain vs. frequency characteristic will have a slope which does not exceed 6 dB/octave down to 0 dB. In Figure 1 is depicted a simplified model of a typical Miller-compensated amplifier.2 We now seek a relation between u , and i which permits us to replace the first differential stage by a voltage controlled current generator. From Figure 2 we find by means of well known current-voltage relations for the transistor . - I. N i,(exp (nuBl) - 1) to i uB1- N i,(exp (nuBz)1) uBz + R,(io - 2i) Received 9 April 1973 Revised 12 June 1973 0 1974 by John Wiley & Sons, Ltd. 13 = u1 (3) 14 ANDERS FORSEN AND LARS KRISTIANSSON t t OR C Figure 1. Miller-compensatedamplifier (simplifiedmodel) t io-i ti Figure 2. Emitter-degenerated differential stage where R 4 kT = - z 40 volt-' is N at room temperature. 0.2. A (4) (5) Neglecting is equation (6) may easily be derived I = 10 1 +exp(Ro,C,) (6) If we let Re + 0, i.e. C1-, 1, equation (6) changes into the known expression for the simple differential stage : I = i0 1 +exp (nu,) We now insert the nonlinear current generator in our model and obtain the diagram in Figure 3. We use Thevenin's theorem and obtain for the equivalent voltage source ueq = Ri = ioR 1 1 +exp(Ru,C,) (7) ANALYSIS OF NONLINEAR MODEL FOR OPERATIONAL AMPLIFIERS IN ACTIVE RC NETWORKS 15 t Figure 3. Amplifier model with nonlinear current generator This means that, for the input signal ul = 0, ues = (i0R/2),which leads to u2 # 0. In actual fact this DC voltage is eliminated from the circuitry aspect. As equivalent voltage source u (Figure 4), accordingly, we have Figure 4. Voltage model for nonlinear amplifier The relation between V,(s) and V ( s )is obtained from the Laplace relation or Now it is simple to divide the model into a linear and a nonlinear block. Let Au be the output voltage of the memoryless nonlinearity. This dummy variable is named u ; . The nonlinearity is followed by a one pole low-pass filter. See Figure 5, where the following relations prevail : A= 1 ( A 1)RC + 16 ANDERS FORSEN AND LARS KRISTIANSSON Figure 5. Block diagram for nonlinear amplifier PARAMETER IDENTIFICATION Usually, the parameters used in the preceding analysis are not available. We shall now identify and express these parameters in well known quantities-pen-loop gain, bandwidth and slew rate. By defining Fo as the small-signal open loop gain we may easily derive the relation Fo=-- ioRA QC, 2 2 or 2F0 -ioRA =- nc, 2 From the small-signal behaviour of the model we deduce I =0 (16) 1 where o1equals the small-signal 3 dB bandwidth. Owing to interior time constants in the amplifier the rate of change of the output voltage is limited. When the amplifier is largely overdriven at high frequencies this phenomenon causes a high distortion. In extreme cases the output signal will be a triangular wave. This maximum rate of change of the output voltage is often called the slew rate ( p ) of the amplifier. If we examine our model under this high frequency overdriven conditions we may easily derive i.e. the relation linking u1 and u; in the model will be Thus we can identify and express the parameters i. and (i0RA/2) in terms of the 'black box' parameters: Fo = small-signal open-loop gain o1= small-signal 3 dB bandwidth p = slew rate As we shall later use the describing function technique, we must calculate the describing function for the n~nlinearity.~ We assume u,(t) = B cos ot (19) u;(t) = BF(B) cos at (20) This procedure implies that we disregard the generation of harmonics in the nonlinearity. u; can now be easily calculated by Fourier series expansion : BF(B) = - A o1 n/20 exp [(2F00,/p)B cos or]- 1 cos at dt exp [(2F001/p)B cos ot] 1 + ANALYSIS OF NONLINEAR MODEL FOR OPERATIONAL AMPLIFIERS IN ACTIVE RC NETWORKS 17 or *i2 exp [(2F00,/p)Bcos ot]- 1 cos wt dt exp [(2F0w,/p)Bcos or] 1 + This integral, together with the expression for the linear portion of the amplifier, now permits us to make simple nonlinear analyses of active filters with slew-rate-limited amplifiers. EXAMPLE OF NONLINEAR ANALYSIS To check the validity of the amplifier model, a study was made of the behaviour of a simple active lowpass filter. The filter in Figure 6 is recognized as the well known UG c~nfiguration.~ -'invout- - - e = 1 29 f5L s2+sl+l 29 a Figure 6. UG low pass network We shall use in the calculation a normalized cut-off frequency of 1 rad/s. Proceeding from Figure 6 we can draw a corresponding block diagram, where N is a memoryless nonlinearity. Figure 7. Block schematic for UG network The two transfer functions H , ( s )and H 2 ( s )are easily calculated from Figure 6 by means of the superposition theorem 1 I H,(s) = s2 + s(2Q+ 1/Q)+ 1 = s2 + s(2Q + 1/Q) + 1 2Qs H2(S) We now make the simplified assumption that w 1 -w1 s+w, within the frequency range of interest. s 18 ANDERS FORSEN AND LARS KRISTIANSSON From Figure 7 we can calculate the relation between the three voltages &(s), V2(s)and V3(s). w1 K,,(s)Hl(s)= V2(s)+ V3(s)(l -Hi+))S The corresponding differential equation is We now assume according to the describing function technique u,,(t) = fincos(wt + 4) u2(t) = B C O S W ~ u3(t) = BF(B) cos mt Insertion of these time functions in the differential equation (27) gives us - fi,w sin ( w t + 4 ) = Bw3 sin wt - B W1 - Bw sin wt - w 1BF(B)w2cos wt -BF(B)w sin a t + wlBF(B)cos wt Q (31) By splitting of the equation into sine and cosine parts we obtain the two corresponding equations (32) and (33). 0 1 - finw cos 4 = B o 3 - BO - -BF(B)w (32) Q - P,,,wsin 4 = -B w2-o,BF(B)w2+wlBF(B) (33) Squaring of the equations and summation gives { pin= (34) This equation relates the amplitude pinof the input signal to the input amplitude B of the nonlinearity N. To complete the analysis we must find a relation between the amplitude of the output signal, and the output amplitude of the nonlinearity and an expression for the phase shift in the filter. From the block diagram in Figure 7 with the simplification (25) we obtain directly the relation . tu,, As the signal after the nonlinearity N passes an integrator, the phase shift between the output signal and the signal after N will be - 4 2 rad. From equation (32) we obtain We can now summarize the preceding results : where Po,,,is calculated from equation (34) and (35). arg H(jw) = II ---4 2 ANALYSIS OF NONLINEAR MODEL FOR OPERATIONAL AMPLIFIERS IN ACTIVE RC NETWORKS 19 As it is difficult to form a notion of the significance of these relations, a computer programme was written for graphic presentation of the equations. In Figures 8(a) and 8(b) we see the result of a run with amplifier parameters according to the nominal values for the operational amplifier 741. A vOUT 10 5- 0 0 5 10 15 4 N Figure 8(a). Relation between input and output amplitudes for UG network with Q-value = 5 and resonant frequency = 15.9 kHz PHASE -;1 Figure 8(b). Dependence of the phase shift on the input amplitude for the filter in Figure 8(a) 20 ANDERS FORSEN AND LARS KRISTIANSSON These parameters are : DC gain: 100,OOO Gain-bandwidth product : 1,OOOkHz Slew rate: 0-5 V / p s The input frequency in this simulation is 0.8 when normalized to the resonant frequency. We can draw an interesting conclusion from the curves. For certain values of the input signal amplitude there is more than one value of the output signal amplitude and of the phase shift in the filter. The possibility of more than one output signal for a given input signal is known in the literature. The transfer from one type of output signal to another takes place in jumps and this phenomenon is called 'the jump phenomenon'. A good description of this phenomenon can be found in Reference 1 pp. 130-132. MODEL VERSUS REALITY A comparison was made between a series of measurements of an active UG network and corresponding theoretical calculations. The parameters of the amplifier in question were measured as accurately as possible and used as input data for the programme. Figures 9-12 show the results of this comparison both for amplitude and phase displacement. These parameters have been common for all the figures : Amplifier : DC gain: 100,OOO Gain-bandwidth product : 1,350 kHz Slew rate: 0-54VIps Filter : Resonant frequency : 15.9 kHz Q-value: 5 The input frequency normalized to the resonant frequency is given in the figure legends. In all of these figures the calculated results are represented by fully drawn and the measured results by dashed lines. !5' 1 Figure 9. Input frequency (normalized):0.6 ANALYSIS OF NONLINEAR MODEL FOR OPERATIONAL AMPLIFIERS IN ACTIVE RC NETWORKS A -270 -180- Figure 10. Input frequency (normalized):0.6 Figure 1 1 . Input frequency (normalized):1.2 21 22 ANDERS FORSEN AND LARS KRISTIANSSON A - -270 -180' c - 90 I 1 % 00 5 10 15 Figure 12. Input frequency (normalized): 1.2 As will be seen, there is close correspondence and we note that we cannot verify the parts of the calculated curves which have a negative slope. We see from the curves that a jump can take place only when the input signal frequency is lower than the resonance frequency for the filter. This property, that a jump can occur only at frequencies either lower or higher than the resonance frequency, is typical of the jump phenomenon. In the literature the phenomenon is customarily ascribed to soft or hard spring force. CONCLUSIONS A simple but useful model of a slew rate limited operational amplifier has been achieved by considering the saturation of the input differential stage. The model gives a good description of the jump phenomenon in a simple active network. More complicated networks should also be possible to study with the applied technique. REFERENCES 1 . Stoker, Nonlinear Vibrations, Interscience, New York, 1970. 2. W. E. Hearn, 'Fast slewing monolithic operational amplifiers', IEEE J . of Solid-Slate Circuifs,SC-6, 20-24 (1971). 3. J. C. West. 'Analytical techniques for nonlinear control systems, 101-1 32, The English Universities Press Ltd., London, 1960. 4. R. P. Sallen and E. L. Key, 'A practical method of designing RC active filters', IRE Trans. Circuit Theory, CT-2, 7 4 8 5 (1955).