DISCRETE CONTROLLED PRE-DRIVER FIR MODEL FOR HYBRID IBIS MODEL AMS SIMULATION MAY 09, 2015, TURIN, ITALY

Size: px

Start display at page:

Download "DISCRETE CONTROLLED PRE-DRIVER FIR MODEL FOR HYBRID IBIS MODEL AMS SIMULATION MAY 09, 2015, TURIN, ITALY"

Joleen Preston
6 years ago
Views:

WAEL DGHAIS AND F. H. BELLAMINE waeldghais@ua.pt/wael.dghais@hotmail.co.

1 DISCRETE CONTROLLED PRE-DRIVER FIR MODEL FOR HYBRID IBIS MODEL AMS SIMULATION IEEE Workshop on Signal and Power Integrity (SPI) MAY 09, 2015, TURIN, ITALY WAEL DGHAIS AND F. H. BELLAMINE Institut Supérieur des Sciences Appliquées et de Technologie de Sousse, Tunisia 1

2 PRESENTATION OUTLINES 1-Introductions/Motivations 2- IBIS Model for an Inverter -Input Signal Decomposition -Functions Decomposition 3-IBIS Hybrid Automaton Representation 4-Event-Driven IBIS Implementation 5-Verification/Validation Setup and Results 6- Conclusions 2

3 INTRODUCTION The driver s input signals are encoded as NRZ or PAM signals. Real digital (i.e. discrete) input waveforms have a finite non-zero rise and fall times that describe the switching between the DC high (H) and low (L) states. The buffered output signals that propagate through the PCB trace are analog. V 1 (t) A VDD PU Devices ih(t) VDD 0 dv 1 (t) dt 0 d 2 V 1 (t) dt 2 0 t r A/t r T t f time (s) time (s) -A/t f time (s) Trapezoidal input signal waveform and its derivatives V1 (t) I1 (t) Predriver Stage Control Circuit VSS PD Devices il(t) Driver Last Stage PAD I2 (t) V2(t) The I/O buffer s structure with its main electrical variables 3

4 MOTIVATIONS Methodology: Apply hybrid modelling procedure that integrates both the continuous electrical variables and discrete-controlled signals to describe the driver s model dynamics Previous work: nonlinear-accurate booleanization of continuous dynamics (ABCD-NL) was proposed to capture the AMS component s analog dynamic behavior using purely Boolean models [1]. Objective of this work: extend this approach [1] to analyze the predriver s circuit presented by the equivalent circuit IBIS model [3] and the other parametric approaches [4], [5] and develop a driver s model formulation and implementation. The model is based on a nonlinear hybrid automaton that combines the interaction between discrete-controlled models for capturing the switching behavior (i.e. finite state machine (FSM) model and the local models structure characterized by continuous evolving variables (e.g. nonlinear differential equation (NDE) model). 4

5 INPUT SIGNAL DECOMPOSITION At constant power supplies V DD and V SS, the driver s model NDE representation is : I 2 t = Φ V t, V t, V m t V t = [V 1 t, V 2 (t)] T is the vector of state variables. m describes the order of the system. Φ is a multivariable nonlinear function that defines a NDE relating the observable output current I 2 t with the vector of state variables and its higher derivatives. Model order reduction : The complexity of the NDE formulation can be reduced using a discretization of the driver s continuous nonlinear dynamic trajectory based on input signal or state variables decomposition/partitioning. The choice of an appropriate state quantization technique must be carefully considered in order to balance the modeling accuracy and running complexity 5

6 FUNCTIONS DECOMPOSITION The PU and PD output nonlinear admittances, are described by the F L and F H ( ) functions, while the input is kept at high and low levels and varying the output voltage V 2 t. I L t = F L V 2 t, V 2 t = Φ V 1 = 0, V 2 t, V 2 t I H t = F H V 2 t, V 2 t = Φ V 1 = V DD, V 2 t, V 2 t The function, Φ, can be analytically approximated as a product of two functions, knowing its two samples at two different DC inputs values (e.g. F L and F H ( ) functions), by a linear interpolation with respect to the variable V 1 : I 2 t = w H V 1 t F H V 2 t, V 2 t +w L V 1 t F L V 2 t, V 2 t 6

7 IBIS MODEL FOR AN INVERTER I 2 t = w H V 1 t F H V 2 t, V 2 t +w L V 1 t F L V 2 t, V 2 t w H t = V 1 t V SS and w V DD V L t = V DD V 1 t SS V DD V SS characteristics. (w H t + w L t = 1) capture the input port static switching This formulation has no dynamics reflected in the driver s input port. The transitions between the local models occur instantaneously and it is valid only for predicting signal distortion of one inverter at a low data rate However, commercial drivers require more sophisticated predriver s circuit in order to perform additional functions. 7

8 IBIS HYBRID AUTOMATON REPRESENTATION The switching between the local models is a continuous flow that occurs with respect to time. This switching behavior has to be integrated in the model in order to reflect the real hybrid device trajectories which traverse across this quantized continuous state space. The pre-driver s nonlinear dynamic manifest itself by asymmetry distortion of the input rising, (r), and falling, (f), edges. IBIS specifications introduce the four voltage-time (V-t) tables that are used to compute the scaling timing signals w L n t and w H n t in order to improve the switching accuracy. The event-driven model formulation is : I 2 t = w L n t F L V 2 t, V 2 t + w H n t F H V 2 t, V 2 t ; n = r, f 8

9 HYBRID AUTOMATON REPRESENTATION OF THE OUTPUT CURRENT V 1 (t)=v DD I 2 (t) COMPUTATIONAL PROCESS V 1 (t)=0 F H (V 2 (t),d/dt) F L (V 2 (t),d/dt) High state Low state V 1 (t)<v DD V 1 (t)=v DD V 1 (t)=0 V 1 (t)>0 f w H (t) F H (V 2 (t),d/dt) f c +w L (t) F L (V 2 (t),d/dt) r w H (t) F H (V 2 (t),d/dt) C r +w L (t) F L (V 2 (t),d/dt) Switching state from H to L 0<V 1 (t)<v DD 0<V 1 (t)<v DD Switching state from L to H 9

10 EVENT-DRIVEN IBIS IMPLEMENTATION (1) Instead of using a fixed timing V-t table stored in the EDA model library that will be time-driven controlled, Event-driven linear filter will be developed in order to assure the mapping from the NRZ input signal and the four timing signals, w(t). Assuming that the pre-driver behaves as a linear time invariant system for the rising and falling transitions inputs: An extraction of the impulse response, h(t), from the observable trapezoidal input response was carried out by computing the derivative of the w(t) signals providing the input-output V-t data: h L n t = dw L n t dt h H n t = dw H n t dt ; n = r, f 10

11 EVENT-DRIVEN IBIS IMPLEMENTATION (2) Detection/comparison block : generates the discrete signal controlling/synchronizing the multiplexing of the FIR filters output according to the duration of the high and low levels and the event occurring with rising and falling transitions V 1 t. V 1 (t) Comparision and detection r h H (t) f h H (t) f h L (t) r h L (t) MUX 2:1 MUX 2:1 w H (t) w L (t) The discrete controlled predriver FIR model yields scaling timing signals that controls the driver s trajectory as it crosses the boundaries between the input DC states of the PU and PD driver s stage currents. 11

12 VERIFICATION RESULTS w (t) H w (t) L IBIS IBIS r w (t) L f w (t) L r w (t) H filter filter filter time (ns) f w (t) H filter Predicted timing signals by the filter bank of the predriver s event-driven model implementation, 12

13 3 pf 500Ω VALIDATION SETUP V DD V 1 (t) I 2 (t) V 2 (t) Load V2(t) Pull-UP F H (V 2 (t),d/dt) V 1 (t) Predriver Stage FIR Filters with enable, trigger and control logic w H (t) w L (t) IL(t) Pull-Down F L (V 2 (t),d/dt) + + I2(t) R L - + RL.CL du/dt V2(t) Pre-Driver Model Last stage Model V2(t) I 2 / V 2 Load 13

14 E t (V) E(t) (V) V 2 t (V) V 2 (t) (V) time (ns) Behavioral Model VALIDATION RESULTS Physical TL Model time (ns) time (ns) Behavioral Model Physical TL Model time (ns) Comparison between the predicted output voltage waveform, V 2 t for input data rate of 700Mbps and tr=tf=500 ps. Comparison between the predicted output voltage waveform, V 2 t for input data rate of 400Mbps and tr=tf=500 ps. 14

15 The developed nonlinear hybrid automaton is a well suited modelling approach for capturing highly nonlinear AMS devices by defining the transient evolution of the hybrid AMS DC state under two-level NRZ or PAM multilevel signals excitation. 15 CONCLUSIONS The derivation of the hybrid automaton driver s model that manifests switching continuous behavior since it mixes the event-triggered predriver s switching dynamics with nonlinear continuous steady state dynamics of the PU and PD models of the driver s last stage, The extracted predriver s model is implemented as a discrete/boolean controlled bank of a finite impulse response (FIR) filters approximating its circuit s nonlinear dynamics based on quantized states of the input signal,

16 REFERENCES [1] A. V. Karthik, S. Ray, P. Nuzzo, A. Mishchenko, R. K. Brayton, and J. Roychowdhury. ABCD- NL: Approximating continuous non-linear dynamical systems using purely Boolean models for analog/mixed-signal verification. In ASPDAC 14: Proceedings of the 19th Asia and South Pacific DAC, pp , [2] T.A. Henzinger. The theory of hybrid automata. In Proc. 11th IEEE Symp. Logic in Computer Science, pp , [3] I/O Buffer Information Specification Version 5.1, 2012 [online] Available: ibis.org/pub/ibis/ver5.1/ver5-1.pdf [4] G. Signorini, C. Siviero, S. Grivet-Talocia, I. S. Stievano, Power and Signal Integrity cosimulation via compressed macromodels of highspeed transceivers, Proc. of the 2015 IEEE 18th Workshop on SPI), Berlin, Germany, May 10-13, [5] W. Dghais, T. R. Cunha, and J. C. Pedro A Novel Two-Port Behavioral Model for I/O Buffer Overclocking Simulation IEEE Trans. on Components, Packaging and Manufacturing Technology, pp: October [6] T. Dang, A. Donze, O. Maler, Verification of analog and mixed-signal circuits using hybrid systems techniques, Formal Methods for Computer Aided Design, vol in LNCS, pp Springer, Heidelberg (2004) 16

17 THANK YOU FOR YOUR ATTENTION QUESTIONS COMMENTS?

Multiport I/O Model Computation For Poweraware SI Simulation

Multiport I/O Model Computation For Poweraware SI Simulation EUROPEAN IBIS SUMMIT MAY 11, 216, TURIN, ITALY WAEL DGHAIS AND F. H. BELLAMINE waeldghais@ua.pt/wael.dghais@hotmail.co.uk Institut Supérieur