INTERNATIONAL JOURNAL OF CIRCUIT THEORY AND APPLICATIONS, VOL. 25, 483—502 (1997) A MACROMODEL OF SAMPLE-AND-HOLD CIRCUITS U. JO®RGES*, G. JUMMEL, G. MU®LLER AND M. REINHOLD Institute for Fundamentals of Electrical Engineering and Electronics, Dresden University of Technology, Dresden, Germany SUMMARY A concise macromodel of sample-and-hold circuits for the simulator SPICE is proposed. This generalized model is independent from actual technical realizations and is based upon compromises regarding the representation of exact circuit structures in the model. Different types of feedback and hold capacitor connections corresponding to the standard external circuit application are accounted for in the model internally. The macromodel allows one to simulate arbitrary user circuits with respect to the behaviour in both the time and frequency domains including error parameters and the temperature dependence of several parameters. Model parameters are extracted for the integrated sample-and-hold circuit AD585 from Analog Devices as an example. Simulation results and selected diagrams are compared with the manufacturer’s data. ( 1997 by John Wiley & Sons, Ltd. Int. J. Circ. ¹heor. Appl., 25, 483—502 (1997) No. of Figures: 25 No. of Tables: 2 No. of References: 8 1. INTRODUCTION Sample-and-hold circuits are an important link between analogue and digital signal processing. As memories for analogue values they have an analogue signal input, an analogue signal output and a digital control input (Figure 1). Substantial circuit parts are the analogue switch controlled by a digital signal and the hold capacitor. In addition, sample-and-hold circuits contain input and output buffer amplifiers. Sample-andhold circuits for high accuracy use an overall feedback.1—4 Without any doubt, macromodels of sample-and-hold circuits are necessary for simulating complex mixed analogue—digital systems. However, powerful simulation models have not been available yet. Largely independent from various technically realized circuit structures known so far, a generalized macromodel5,6,7 for this circuit family has been developed which is based upon data sheet parameters.3,4 For creating the model, techniques known from modelling operational amplifiers have been adapted. Thus compromises have to be made regarding the exact representation of the circuit structures to be modelled. Various types of feedback and hold capacitor connections are represented by a single model structure according to the standard application circuit. The advantage of this approach is that parameter extraction can be done only from data sheet parameters, even for circuits whose internal structure is unknown. The model principally contains linear passive elements and controlled voltage and current sources. The SPICE diode model is used as a non-linear element for voltage limitation and approximation of characteristics. The sample-and-hold model has its own reference node which is different from the global reference node ‘0’ in the SPICE simulator for the purpose of implementing the model as a subcircuit into any arbitrary application circuit. Noise is mainly caused by aperture jitter and random phase jitter of the sampling clock. The effects of aperture jitter can be evaluated by modulating the aperture delay with a control voltage signal. * Correspondence to: U. Jörges, Institute for Fundamentals of Electrical Engineering and Electronics, Dresden University of Technology, Dresden, Germany. CCC 0098—9886/97/060483—20$17.50 ( 1997 by John Wiley & Sons, Ltd. Received 19 April 1996 484 U. JO®RGES E¹ A¸. Figure 1. Principle of sample-and-hold circuit for high accuracy Figure 2. Hold-to-sample transition 2. THEORY OF OPERATION AND PARAMETERS OF SAMPLE-AND-HOLD CIRCUITS The definitions of the dynamic parameters of the hold-to-sample transition are illustrated in Figure 2. The acquisition time t is defined as the period from the hold-to-sample transition of the control signal to the A#2 time instant where the output signal enters a specified tolerance band around the steady state value following an input signal step of half the full scale range (FSR). The acquisition time consists of the switching delay t , SW the slewing interval t and the settling interval t . The definitions of the dynamic parameters of the S-%8 SS sample-to-hold transition are illustrated in Figure 3. The sample-to-hold settling time t is the period from SH the corresponding clock transition to the instant where the output signal enters a specified tolerance band. The aperture delay t is the time interval from the clock transition to the instant where the switch actually A opens. Table 1 gives all circuit parameters which are covered by the model. 3. FUNCTIONAL BLOCKS OF MODEL The entire model is divided into functional blocks, each representing a subset of circuit parameters. The connection of the model blocks described in this section is shown in Figure 17. For the clock input stage and Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) ( 1997 by John Wiley & Sons, Ltd. MACROMODEL OF SAMPLE-AND-HOLD CIRCUITS 485 Figure 3. Sample-to-hold transition Table I. Characteristics of sample-and-hold circuits Supply characteristics Supply current Supply voltage dependence Power supply rejection Input characteristics Input resistance Input capacitance Bias current Temperature dependence Offset voltage Temperature coefficient Output characteristics Maximum output voltage swing Maximum output current Output resistance Control input characteristics Min. HIGH voltage level Max. LOW voltage level Input currents Input capacitance I S1,/ dI /d» S S PSR 1,/ r * C * I IB I (¹) IB » IO d» /d¹ IO » O.!91,/ I O.!91,/ r 0 » H.*/ » L.!9 I H,L C C-, ¹ransfer characteristics, sampling (tracking) mode Closed-loop voltage gain A V Gain error *A V Temperature coefficient of gain dA /d¹ V Gain—bandwidth product GBP Slew rate SR 1,/ Rise time t 3 Peak overshoot p OS Acquisition time t A#2 ¹ransfer characteristics, hold mode Aperture time t A Effective aperture time t A%&& Aperture jitter *t A Sample-to-hold settling time t SH Peak overshoot voltage » OH Hold step » HS Feedthrough attenuation A FT Cut-off frequency f 'FT Leakage current! I D Droop rate! d» /dt O ! In most data sheets either the leakage current or the droop rate is given. They can be calculated from each other for a known hold capacitance value. for modelling the temperature dependence of the bias and leakage currents several interchangeable model blocks are available. To associate the model block diagrams with the entire model diagram, the same node numbers as in Figure 17 are used throughout this section. The elements in the model block diagrams are designated by the same symbols which are used in the equations whereas in Figure 17 their equivalent SPICE notation is used. ( 1997 by John Wiley & Sons, Ltd. Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) 486 U. JO®RGES E¹ A¸. 3.1. Input stage The input stage (Figure 4) represents the parameters input resistance r , input capacitance C and bias * * current I . The bias current is generated by a current source I which is controlled by the temperature stage IB IB (see Section 3.8). The offset voltage is considered in the output stage (see Section 3.7). For sample-and-hold circuits with external offset compensation terminals the offset voltage is adjusted such that it is compensated in the hold mode by the hold step. The subsequent stages are controlled by the voltage across the input terminals using controlled sources. Node 10 is the reference node of the macromodel. 3.2. Frequency response and slew rate stage In the frequency response and slew rate stage (Figure 5) the closed-loop gain A , the gain—bandwidth V product GBP, the positive and negative slew rates SR and SR , the peak overshoot p and the acquisition 1 / OS time t are modelled. A#2 The analogue switch is modelled in SPICE by a two-dimensional polynomial controlled current source representing a voltage controlled transconductance with an on-resistance r "1/g . This source is control0/ 0/ led by the voltage at node 30, which is limited to 1 V (see clock-shaping stage, Section 3.5). G 0 for v "0 C-, i "g v v " SW 0/ C-, 45 g v for v "1 V 0/ 45 C-, (1) The current source i with its transconductance g is controlled by the sum of the input voltage and the 2 .2 feedback voltage across the hold capacitor C . The small-signal amplification of this stage is H A "g R 0 .2 2 (2) Figure 4. Input stage Figure 5. Frequency response and slew rate stage Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) ( 1997 by John Wiley & Sons, Ltd. MACROMODEL OF SAMPLE-AND-HOLD CIRCUITS 487 In principle the value of A can be chosen arbitrarily, because the stage is operating as an ideal integrator. 0 The amplification is fixed at A "1000 to minimize the influence of the voltage-limiting diodes D and 0 SR1 D on the slew rate. SR/ For calculating the large-signal behaviour, the step response is divided into the slewing time t and the S-%8 sampling-mode settling time t intervals (Figure 2).8 The switch delay t is modelled in the clock-shaping SS SW stage. During the slewing time one of the diodes, D or D , is conducting. Therefore the voltage across the SR1 SR/ capacitor C is constant, which results in a constant current charging the capacitor C . The limiting voltages 2 H » and » together with the time constant q determine the positive and negative slew rates SR and SR . SR1 SR/ 1 1 / During the subsequent sampling-mode settling time interval the circuit operates in the small-signal region. C q "R C " H (3) 1 0/ H g 0/ » "q SR (4) SR1 1 1 » "q SR (5) SR/ 1 / The lower absolute value of the slew rate values SR, and thereby the corresponding value of the limiting voltage » , is used to determine the parameters of the sampling-mode settling characteristic. The acquisition SR time of a sample-and-hold circuit is defined in such a way that the sampling process starts when v steps C-, from 0 to 1 V, with the input voltage having changed by *º "FSR/2 and the initial voltage being v "0. I CH The sampling-mode settling corresponds to the large-signal behaviour following an input step with an amplitude *» and the switch being on, neglecting differences during a short period where the circuit is I operating in the small-signal region before entering the slewing interval. The limiting phase is characterized by a linear rise SR in the voltage across the hold capacitor C . At the H time instant t"t the voltage v is no longer limited. The amplitude *» of the input voltage step is S-%8 C2 I divided into two regions according to the two regions of the circuit in sampling mode. *» "*» #*» (6) I I,S-%8 I,-*/ q SR *» " 1 (7) I,-*/ A 0 The transition from the slewing region to the linear region occurs at the time instant t"t with the voltage S-%8 across the hold capacitor and its derivative being the initial conditions. v (t )"*» "*» !q SR/A CH,S-%8 S-%8 I,S-%8 I 1 0 dv (t) CH,S-%8 "SR dt t/tS-%8 From equations (8) and (9) the slewing time t is obtained. S-%8 q v (t ) *» t " CH,S-%8 S-%8 " I! 1 S-%8 SR A SR 0 The frequency response is described by the second-order transfer function K C D 1 A p p (s) » 1 2 " G(s)" CH,-*/ " 0 q q s2#(1/q )s#A /q q (s!p ) (s!p ) » (s) I,-*/ 1 2 1 2 2 0 1 2 with the time constants q "R C and q "R C @q 1 0/ H 2 2 2 1 ( 1997 by John Wiley & Sons, Ltd. (8) (9) (10) (11) (12) Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) 488 U. JO®RGES E¹ A¸. Thus, the gain—bandwidth product GBP is determined. GBP"A /2n q (13) 0 1 Because of the overshoot behaviour in the time domain, the poles p of the transfer function (11) have to be 1,2 complex conjugate. SC A BD A 1 2 1 0! p "a$jb, with a"! and b" 1,2 q q 2q 2q 1 2 2 2 For zero initial conditions the step response v (t) of the system has the form CH,-*/ a v (t)"*» 1!eat cos (b t)! sin (b t) CH,-*/ I,-*/ b C A BD (14) (15) The first extreme of this function provides the relative overshoot factor p . OS p "[email protected] (16) OS This yields the ratio k of the time constants q and q . 1 2 n 2 q 1 #1 k" 2" (17) q ln p 4A 1 OS 0 The linear interval follows the slewing interval. The step response (15) is adjusted with the factors k and 1 k to the values of the function and its deriviative at the end of the slewing interval. In order to simplify the 2 following equation, the time is set to t"0 at the end of the slewing interval. CA B D v C A BD a (t)"*» 1!eat k cos (b t)!k sin (b t) CH,-*/ I,-*/ 1 2b (18) v (0)"0 (19) CH,-*/ dv (t) CH,-*/ "SR (20) dt t/0 Comparing the coefficients in equations (18)—(20) the constants k and k are determined. 1 2 k "1 (21) 1 SR #k "1!2A k (22) k " 1 0 2 a*» I,-*/ The time-decaying term in (18) describes the settling behaviour. The sampling-mode settling time t as a part SS of the acquisition time is approximated by the time instant where the enveloping curve reaches the relative deviation e from the steady state value *» . I a 2 v (t)"*» k2#k2 eat (23) CH,%/7 I,-*/ 1 2 b K SC ABD 2q SR 2 e*» " [email protected], t "t !t !t (24) I I(4A k!1) SS A#2 S-%8 SW 0 The slewing time t is given by (10), the delay t of the switch is dealt with in Section 3.5. There are three S-%8 SW equations, (13), (17) and (24), for determining the two time constants q and q . The parameters slew rate, one 1 2 or more acquisition times, relative overshoot and gain—bandwidth product have to be adapted. For example, Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) ( 1997 by John Wiley & Sons, Ltd. MACROMODEL OF SAMPLE-AND-HOLD CIRCUITS 489 if the slew rate, the gain-bandwidth product and one acquisition time are selected, the overshoot is fixed. Contradictions in parameter consistency are eliminated by optimization. 3.3. Stage for modelling feedthrough attenuation Feedthrough attenuation is modelled frequency dependently (Figure 6). From the cut-off frequency f it 'FT is falling by 20 dB per decade. The output signal of this stage is coupled to node 42 of the hold capacitor by the controlled voltage source v via the capacitance C . The effective capacitance at node 42 of the FT FT hold capacitor is set to C*"1 nF. The capacitance C is determined according to the feedthrough H FT attenuation A . FT C "C*][email protected] (25) H FT The inductance ¸ is calculated from the cut-off frequency f . FT 'FT ¸ "1/2n f R (26) FT 'FT FT The feedthrough attenuation not being able to fall below 0 dB, the following equation also has to be valid. C FTg R "1 C* FT FT2 H The voltage-controlled current source i has a transconductance g . FT FT (27) 3.4. Clock input stage The clock input stage models the clock input currents I and I as well as the input capacitance C . The H L C-, logic levels are transformed into normalized currents and fed to the clock shaper (see Section 3.5). There are various functional block models for the clock input stage. The coupling of the input functional block model and the clock shaper is accomplished by the current source i at node 63. This source provides V a current of 1 mA in the sampling mode and !1 mA for a minimum high-level » or a maximum H.*/ low-level » at the inputs respectively. Figure 7 shows the functional block model for a single asymmetriL.!9 cal input. Figure 6. Stage for modelling feedthrough attenuation Figure 7. Functional block model for asymmetrical clock inputs ( 1997 by John Wiley & Sons, Ltd. Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) 490 U. JO®RGES E¹ A¸. Figure 8 illustrates the functional block model of the clock input stage for a symmetric input, which is also intended for modelling an asymmetrical input with a reference input. 3.5. Clock shaper The clock shaper (Figure 9) consists of a pre-limiter, a delay stage and a limiter. The delay stage models the aperture time. Furthermore, the aperture jitter can be evaluated using an additional controlling voltage » . M0$ The delay stage between nodes 66 and 67 is represented by a second-order lowpass filter with the transfer function (28). With the grading factor set to Q"1·172 the rising and falling edges of the output pulses have their inflection point at their 50% value. The current source i has a transconductance g . 77 77 » 1/¸ C VV VV G (s)" 67"g R V VV V s2#(R /¸ )s#1/¸ C » 63 V VV VV VV u2 0 "g R VV V s2#(u /Q)s#u2 0 0 (28) Figure 8. Functional block model for symmetrical clock inputs Figure 9. Clock shaper Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) ( 1997 by John Wiley & Sons, Ltd. MACROMODEL OF SAMPLE-AND-HOLD CIRCUITS The group delay t ',TP 491 of the second-order lowpass filter is determined by t "1/Qu (29) ',TP 0 The group delay t is set to the effective aperture time t to yield the centre frequency u . ',TP A%&& 0 u "1/Qt (30) 0 A%&& The group delay t of the signal path is calculated from ' q #q 2 t" 1 (31) g A 0 The switch delay t is the sum of the effective aperture time and the group delay of the signal path. SW t "t #t (32) SW A%&& ' The driving voltage v of the switch is limited to a range between 0 and 1 V using a tanh function, which is C-, implemented by the bottom circuit in Figure 9. v "» #v 61 61,0 61,1 The constant part » of this voltage adjusts the operating point of the diode D at I "10 lA for the clock 61,0 F 0 input voltage v "0. The value of the resistor R is assigned such that the resulting voltage drop at the IC-, F operating point is negligible (R I "1 mV) compared with the forward voltage drop of the diode. The diode F 0 current is approximately an exponential function in a defined clock input voltage range. The resistor R limits the input current to reasonable numerical values during an input overload and converts the diode F current I into a voltage. Therefore equation (33) is valid for the voltage v across the resistor R with the DF 62 F conditions I @I and u AR I , where I is the saturation current of the diode. S 0 61 F 0 S v !v v 62 !1 +R I exp 61,1 v "R I exp 61 (34) 62 F S F 0 » » T T The controlled voltage source v obeys a tanh function. C-, 1 2 1 2 (35) v2 ]1 V! v2 v v " 62 62 C-, C-, I R I R 0 F 0 F exp(2v /» ) v 1 61,1 T ]1 V" tanh 61,1 #1 ]1 V v " (36) C-, exp(2v /» )#1 » 2 61,1 T T The voltage » can get arbitrarily close to the limits 0 and 1 V but never exceed them at overload. This is C-, one important property of the circuit. The controlled current source I compensates the influence of the F minimum conductance G , which is introduced in parallel with the diode by SPICE. The temperature .*/ dependence of the transfer characteristic caused by » is compensated by the additional control of the voltage T source v by » . A modulation of the aperture time is possible by shifting the threshold voltage by » . F 5%.1 M0$ This facilitates the simulation of the aperture jitter. C A A B B D A B A B C A B D 3.6. Stage for modelling hold-mode settling time and hold step The hold-mode settling time is modelled using a damped resonant circuit (Figure 10). Its output signal together with the hold step is transferred onto the hold capacitor (Figure 11). The voltage at node 75 is calculated from (37) provided that a voltage step of !1 V has been applied to node 30, where g is the T transconductance of the controlled current source i . T g 1V » "! T (37) 75 C s2#(1/R C )s#1/¸ C T T1 T T T ( 1997 by John Wiley & Sons, Ltd. Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) 492 U. JO®RGES E¹ A¸. Figure 10. Stage for modelling holdmode settling time Figure 11. Coupling of hold step and hold-mode settling wave-form onto hold capacitor The inverse Laplace transform provides the voltage v in the time domain. 75 g 1 v (t)"! T exp(!a t)sin(b t)]1 V (38) 75 C b T with 1 1 1 2 a" and b" ! ¸ C 2R C 2R C T T T1 T T1 T With the resistor R set to 1 k), the three free parameters g , ¸ and C can be determined if the resonant T1 T T T frequency, the damping factor and the absolute amplitude of the hold-mode settling wave-form are given. The hold-step voltage » from the controlled voltage source v is capacitively coupled onto the hold HS HS capacitor C via the capacitance C (Figure 11). H HS D» D (39) C " HS C* HS 1V H SC A BD The value of the hold capacitor C is determined such that the effective capacitance C* at node 42 is exactly 1 nF. H H (40) C "C*!C !C H HS FT H 3.7. Output Stage The output stage models the output resistance r , the maximum output voltages » , the maximum 0 O.!91,/ output currents I and the offset voltage » . The polynomial controlled current source i of the O.!91,/ IO AV output stage (Figure 12) adds the offset voltage to the output voltage v of the frequency response stage. 42 Additionally, the non-linear distortion, the gain and its temperature dependence are modelled. The current source i models the power supply feedthrough. The output voltage source v is two-dimensionally PSR 54 controlled by both the voltage v and the reference potential of the sample-and-hold circuit. Thus the output 53 voltage is referred to node 10 of the model. The controlled current source i realizes the output resistance O using the voltage drop across the current source itself as the controlling voltage. Limiting the output current is done by restricting the controlling voltage. The resistor R provides a DC path to node 21 as required in 21 SPICE and is set to 1 G). 3.8. Temperature stage The temperature stage (Figure 13) provides a Kelvin-proportional voltage » for modelling linear 5%.1 temperature dependences using the SPICE model (41) of the linear temperature-dependent resistor (b"0). R(¹)"R(TNOM) [1#a(¹!TNOM)#b(¹!TNOM)2] (41) The parameters of the current I and the resistor R are extracted such that the temperature stage provides T T a voltage » "1 V at the temperature ¹ (SPICE-Option TNOM). The temperature coefficient is set to 5%.1 0 a"1/TNOM. The voltage » works as a linear temperature-controlled input value for any other 5%.1 Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) ( 1997 by John Wiley & Sons, Ltd. MACROMODEL OF SAMPLE-AND-HOLD CIRCUITS 493 Figure 12. Output stage Figure 13. Stage for generating temperature-dependent voltage Figure 14. Functional block model for temperature dependence of drift current controlled sources in the model. The approximation to a non-linear behaviour is realized using appropriate polynomial coefficients of the controlled sources. Another temperature stage with an exponential behaviour consists of the voltage source » , the diode TD D and the current source I . The parameter extraction is done for the desired behaviour. The DT TGD exponential dependence of the diode saturation current on temperature is used. A B A B ¹ [email protected] E (1!¹ /¹) 0 I(» )"I exp ' T SO ¹ n» 0 T0 (42) The bias current and drift current sources are interchangeable functional block models, because the linear temperature dependence requires voltage control whereas the exponential temperature dependence requires ( 1997 by John Wiley & Sons, Ltd. Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) 494 U. JO®RGES E¹ A¸. Figure 15. Stage for modelling supply current current control of the current source. Figure 14 shows the functional block model for the drift current. The functional block model for the bias current has the same structure. 3.9. Stage for modelling supply current In the model the supply current depends on the output current (Figure 15). The output current is assigned to the positive or negative power supply respectively according to its sign by the diodes D and D . The IO1 IO/ sources i and i generate the supply currents. The resistors R and R model the quiescent currents. S1 S/ S1 S/ 4. MODEL OF SAMPLE-AND-HOLD CIRCUIT AD585 The sample-and-hold circuit AD585 is used as an example for a complete model in this section. 4.1. Block diagram and complete model See Figures 16 and 17. 4.2. Subcircuit listing Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) ( 1997 by John Wiley & Sons, Ltd. MACROMODEL OF SAMPLE-AND-HOLD CIRCUITS ( 1997 by John Wiley & Sons, Ltd. 495 Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) 496 U. JO®RGES E¹ A¸. Figure 16. Block diagram. The ports marked with an asterisk are not available in the model Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) ( 1997 by John Wiley & Sons, Ltd. ( 1997 by John Wiley & Sons, Ltd. Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) Figure 17. Complete model 498 U. JO®RGES E¹ A¸. 4.3. Comparison of simulation results and data sheet parameters See Figures 18—25 and Table II. Figure 18. Test circuit for simulation Figure 19. Transfer characteristic Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) ( 1997 by John Wiley & Sons, Ltd. Figure 20. Frequency response of open- and closed-loop gain Figure 21. Small-signal step response, sample-to-hold overshoot and hold step Figure 22. Slew rate ( 1997 by John Wiley & Sons, Ltd. Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) Figure 23. Acquisition time Figure 24. Aperture time Figure 25. Temperature dependence of bias current Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) ( 1997 by John Wiley & Sons, Ltd. 501 MACROMODEL OF SAMPLE-AND-HOLD CIRCUITS Table II. Comparison between simulation results and data sheet Parameter Symbol Conditions Data sheet Min. Supply characteristics Supply current Power supply rejection Input characteristics Input resistance Input capacitance Offset voltage Offset voltage Bias current Bias current Output characteristics Output resistance max. output voltage max. output current Clock characteristics min. HIGH-voltage max. LOW-voltage HIGH input current LOW input current Sampling-mode characteristics Gain Slew rate Acquisition time 0.01% error range 0.01% error range Rise time Peak overshoot Gain—bandwidth product Full power bandwidth Hold-mode characteristics Aperture time Sample-to-hold settling time Peak overshoot voltage Hold step Feedthrough attenuation Cutoff frequency Droop rate Droop rate I S1,/ PSR R I C I » IO » IO I IB I IB ¹ "25 °C A T —T .*/ .!9 ¹ "25 °C A ¹ —T .*/ .!9 2 3 2 5 » H.*/ » L.!9 I H I L (Hold) (Sample) » "4 V, fixed H » "4 V, fixed !50 L t 3 p OS GBP FPB t A t SH » OH » HS A FT f 'FT d» /dt 0 d» /dt 0 $10 1012 10 I "$10 mA 0 » "$15 V S1,/ » 50 $12.5 $50 2 0.8 50 1 $10 "20 V OSS *» "10 V O *» "20 V O » "50 mV OSS » "50 mV OSS » 3 5 — — 2 160 "20 V OSS º "0 V I º "0 V I fixed ¹ "25 °C A ¹ —T .*/ .!9 Unit $10 70 mA dB 0.5]1012 10 3! 4.5! 2 0.5—5 ) pF mV mV nA nA 50 $12.7 $54 m) V mA 0.8 2 50 !50 V V lA lA 1 $10 V ls~1 Max. $6 70 r 0 » O.!91,/ I O.!91,/ A V SR 1,/ t A#2 Typ. Model 35 500 !15 3 !3 86 — 1 Dbl. ev. 10 °C 3 5 150 28 2 160 35 500 !15 !3! 86 100 1 Dbl. ev. 10 °C ls ls ns % MHz kHz ns ns mV mV dB kHz mV ms~1 mV ms~1 ! The hold-step is compensated by offset voltage trimming. 5. CONCLUSIONS In this paper a generalized macromodel of sample-and-hold circuits for the simulator SPICE based on the data sheet characteristics has been presented. The model is independent from actual technical realizations and is based upon compromises regarding the representation of exact circuit structures in the model. Different types of feedback and hold capacitor connections corresponding to the standard external circuit ( 1997 by John Wiley & Sons, Ltd. Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) 502 U. JO®RGES E¹ A¸. application were accounted for in the model internally. The macromodel allows us to simulate arbitrary user circuits with respect to the behaviour in both the time and frequency domains, including error parameters and the temperature dependence of several parameters. The efficiency of the model was proved by comparison of simulation results and data sheet parameters of the integrated sample-and-hold circuit AD585 from Analog Devices. In addition to this paper, the model was used successfully for the considerably different circuits SHC605, HA5351 and HA2420. ACKNOWLEDGEMENT This paper is part of the project ‘Model Library of Complex Analog Components’ sponsored by the German Ministry of Research and Technology. REFERENCES 1. F. Moraveij, ‘A high-speed current-multiplexed sample-and-hold amplifier with low hold step’, IEEE J. Solid-State Circuits, SC-26, 1800—1808 (1991). 2. N. Greitschus, ‘Sample-and-hold circuits and their implementation in microelectronics’, Dissertation, Institute for Fundamentals of Electrical Engeneering and Electronics, Dresden University of Technology, 1992 (in German). 3. ¸inear and ¹elecom ICs for Analog Signal Processing Applications, Harris Semiconductor, Melbourne, FL, USA, 1993—1994. 4. Design-In Reference Manual, Analog Devices, Norwood, MA, USA, 1994. 5. V. Pose, ‘Switch models for sample-and-hold circuits’, Diploma ¹hesis, Institute for Fundamentals of Electrical Engineering and Electronics, Dresden University of Technology, 1994 (in German). 6. M. Reinhold, ‘Macromodeling for sample-and-hold circuits’, Diploma ¹hesis, Institute for Fundamentals of Electrical Engineering and Electronics, Dresden University of Technology, 1994 (in German). 7. U. Jörges, G. Jummel, G. Müller and M. Reinhold, ¹he ¹º-Dresden Sample-and-Hold Circuit Model, Model Library of Complex Analog Components, VDI/VDE Technologiezentrum Informationstechnik G.m.b.H, Teltow, 1995 (in German). 8. J.-Ch. Lin and J. H. Nevin, ‘A modified time-domain model for nonlinear analysis of an operational amplifier’, IEEE J. Solid-State Circuits, SC-21, 478—483 (1986). . Int. J. Circ. Theor. Appl., Vol. 25, 483—502 (1997) ( 1997 by John Wiley & Sons, Ltd.