Advanced Technologies to Accelerate

Mixed Signal Simulation

© 2005 The MathWorks, Inc.

Pieter J. Mosterman

Senior Research Scientist

The MathWorks, Inc.
Agenda

 Introduction
 What is Link for ModelSim?
 Demo
 Mixed Signal Simulation
 Summary

2
What is the problem?
 “Hardware verification is itself become more challenging.
Verification times have increased with rising gate count and as
overall design complexity grows. According to a survey by Collett
International Research in 2002 only 39% of designs were bug free
at first silicon, while 60% contained logic or functional flaws. More
than 20% required 3 or more silicon spins. A Collett survey also
showed that nearly 50% of total engineering time was spent in
verification.”

 Hardware/Software Co-verification by Dr. Jack Horgan

http://www.edacafe.com/magazine/index.php?newsletter=1&run_date=29-Mar-2004

3
What are the pains?
1. Time and effort to verify a design: As designs get more
complex, the test benches are an order of magnitude more
complex, and consume 40-60% of project resources.
 Test bench HDL code will *not* be synthesized – i.e.,
will not be a part of the shipping product – “throw-
away” code
 HDL test benches need to run in HDL simulators, and
HDL simulators are *extremely* slow
2. Time and effort to construct and maintain test benches:
For each line of HDL design code in a design, a user
typically needs 10 lines of HDL test bench code to
simulate, test, and verify that 1 line of HDL code.
 Constructing a test bench in a textual language is at
least as complex as the original design itself
 Maintaining the test bench from one generation of a
design to the next is very resource-intensive

4
What are the pains? Solutions?
 Engineers need to verify that ASIC/FPGA
implementations correctly match their system
specifications.

y=f(x) Link

 Using the Link for ModelSim, these engineers can

co-simulate their MATLAB and Simulink designs
with equivalent Verilog and VHDL.

5
Agenda

 Introduction
 What is Link for ModelSim?
 Demo
 Mixed Signal Simulation
 Summary

6
Link for ModelSim makes system-level
hardware verification possible.

 Fast and bidirectional cosimulation interface

between MATLAB and Simulink, and
ModelSim

7
Agenda

 Introduction
 What is Link for ModelSim?
 Demo
 Mixed Signal Simulation
 Summary

8
 Live
Demo MATLAB
Demo

 Edge detection demo

– Design space exploration using floating
point
– Conversion to fixed point
– Co-simulate with HDL

9
Link for ModelSim allows engineers to
share models instead of I/O vectors.
 The HDL is verified in the context of an entire
system and not just as a stand-alone
component.

 System performance metrics, e.g. PER, BER,

S/N ratio can be measured.

10
Agenda

 Introduction
 What is Link for ModelSim?
 Demo
 Mixed Signal Simulation
 Summary

11
Hybrid Dynamic Systems
 Two types of behavior
– Continuous
– Discrete
 An embedded controller
– Plant
• Continuous-time
behavior
• Sporadic discrete
events
– Controller
• Discrete-time
behavior
• Frequent periodic
events
12
Executing a Hybrid Dynamic System

 Integrate continuous behavior

x  f ( x, u, t )

 Discrete event behavior

– Time events
• Pre-determined time of
occurrence
tn  nextEventTime()
– State events
• When a model variable
exceeds a threshold
g ( x, u , t )  0
13
How to handle a discrete event
 Time-driven
– Time integrated
• Integrate up till event time
• Inefficient for time events
– Sampled time
• Run scheduler at lowest
rate
• Inefficient for widely
spaced events

 Event-driven
– Jump to event time
immediately
– Does not apply to state
events 14
 Where do these paradigms apply?
 Discrete event intensive models
– Controller area network (CAN) in automobiles
 Many events that do not affect continuous behavior
 Expensive to stop the numerical solver for each of
these window

dc motor

current
measurement

lights controller roof controller window controller

15
Efficiency
 Two separate solvers
– Time-driven solver
• Numerical solver integrates time (step h)
h k
k1  hf ( xn , yn ) k3  hf ( xn  , yn  2 )
2 2
h k
k 2  hf ( xn  , yn  1 ) k 4  hf ( xn  h, yn  k3 )
2 2
k k k k
yn 1  yn  1  2  3  4  O(h5 )
6 3 3 6
• Sampled time event schedule
(time up to which to integrate)
– Event-driven solver
• Event calendar
20 [ms] open_tonneau
2020 [ms] move_roof_up_cmd
2150 [ms] move_down_window
2250 [ms] stop_moving_window
16
 Event-driven leads
time event

state event

Continuous Time:

This sets an This sets a

event in the time-event
event-driven This event in the
solver never occurs, time-driven
but it is solver
already
processed!

Scheduled Events:

Processed Events:       
17
 Time-driven leads
This event
never occurs,
but it is time event
already
processed!
state event

Continuous Time:

This sets a This sets an

time-event event in the
in the event-driven
time-driven solver
solver

Scheduled Events:

Processed Events:   
18
Questions
 Can we restrict the modeling constructs that are allowed?
– For example, no state events allowed to trigger events in
the event-driven solver
 What inaccuracy is acceptable?
– Not do zero-crossing detection
– Run event-driven part at a quick base rate
 What efficiency is necessary?
– Lock step approach
– Use one solver all together?
• Do you lose the ability to handle a batch of discrete events
independently?
– Other techniques?
• Model differently
 What is the preferred configuration?
– Time-driven leads, store continuous-time state
– Event-driven leads, store discrete event state
– Alternatives?
19
Agenda

 Introduction
 What is Link for ModelSim?
 Demo
 Mixed Signal Simulation
 Summary

20
Summary of key features

 Integrate system level design with

implementation
 No duplication of testbench design effort
 Verification of system level properties

 Advanced Simulation Technologies Are

Required

21