415 Tasman Drive Sunnyvale, CA 94089-1706 (T) 408.747.1155 (F) 408.747.1286 (E-mail) sales@aldinc.com (Website) www.aldinc.com
system-on-a-chip using both software and hardwar e tools, thus per-
forming trial and error experimentation on the personal worksta-
tion as well as on the lab bench. Of cour se, the final design will
only be as good as what the designer can measure or predict using
simulation and breadboar d measurements. With simulation added
to the design verification process, convergence upon a design so-
lution is rapid. In addition, this approach allo ws the manufacturer
the possibility to verify his system through actual field tests, lim-
ited production, and beta test sites, before committing to the ex-
pensive of custom IC production.
Actual chip level experimentation and mask iteration are not only
costly, b ut as systems g et increasingl y complex, the probability of
a designer actually developing a w orking ASIC system under the
time and budget constraints diminishes. For complex analog or
mixed analog digital system development intended for ASIC imple-
mentation, the ability to inexpensively and quic kly do design it-
erations and measurements, whether on a computer station, or on a
breadboard, is vital to the ultimate success of the project.
Of course , the a bility to do both software and hardware iteration
produces a desirable combination for engineering analysis. For
certain situations, a quick run on a computer simulator can verify a
circuit conf iguration that performs an intended function. Unfortu-
nately , nice , idealized g raphs presented on a computer terminal do
not always represent actual circuit behavior and in many analog
systems full simulation with regard to all parametric behavior is
impractical if not impossible.
For other situations, such as to verify signal levels and timing of
interface between the ASIC and other board components such as
transducers, and to "feel" the adequac y of an analog signal, a bread-
board as an actual hardware simulator can prove superior as an
engineering tool. When all the pieces of individual subcircuits are
simulated or designed, the time comes to put them together and
verify that they perform as a system. This is a time when a de-
signer needs all the engineering tools at his disposal. As software
simulators and models may assist in spotting problems early on,
hardw are simulators and real hardware help other problems to show
up. This two prong approach to circuit development eliminates
redundant engineering effort when an ASIC needs to be imple-
mented later in a production cycle.
A design initially developed with this approach would not need
additional engineering effort when its time to integrate. More im-
portantly , this approach cut out most of the technical risks of ASIC
implementation because the designer has close feedback and abil-
ity to observe and measure the actual as well as the simulated sys-
tem. Thus the often sta ted goal of a quick implementa tion onto an
ASIC chip can be easily handled as the need and justification de-
velops.
EASY ASIC BREADBOARDING FOR AN ASIC
A vailable from ALD is a design kit that contains a box full of ana-
log IC par ts such as operational amplifiers, comparators, oscilla-
tors and transistors that allow a designer to develop custom cir-
cuitry. Also available from ALD is a proprietary library of
macromodels of these specific analog IC parts de veloped for a stan-
dard SPICE circuit simulator such as PSPICE, in addition a design
manual that contains complete information and useful hints is pro-
vided.
When working with the kit, one can design and sim ulate using the
ASIC library components or breadboard these circuits with Ad-
vanced Linear Devices' standard par ts. The designer can use any
number of analog cells fr om the design kit and in any configura-
tion desired, and later decide on the package or pinout for an ASIC
part. Since ALD's ASIC program is semicustom standard cell ap-
proach, ther e is no rigid limitation on how man y instances of any
of function block, pins, components or even the values used.
Furthermore, digital logic such as HC, 74C and CD4000 family
parts can also be used in the software simulator as well as the hard-
ware prototype . Many software circuit simulator suppliers already
supply dig ital parts by individual par t numbers in their digital li-
brary. These digital parts simulate the logic and performance of,
say, a CD4027 Dual JK Flip-Flop. ALD has developed functional
equiv alent digital cells in its computer libr ary that will perform the
identical logical functions.
A custom prototype system designed around using these generic
digital part types along with analo g parts supplied by ALD can be
readily implemented into an ASIC with virtually no performance
differences. With this complete libr ary of digital MSI functions
such as counters, encoders and decoders, no logic conversion is
necessary from standard part type to implementation of the ASIC.
All digital circuit blocks in ALD's digital library perfor m identical
functions to the generic part types in the simulation library.
Backgound on macromodels
An operational amplifier modeled with detailed transistor level
parameters and cir cuitry may run for five or six hours equivalent
CPU time based on a Pentium class personal computer for each
parametric iteration. W ith several amplifiers and tens or hundr eds
of simulations with perhaps dozens of parameters each requiring
multiple passes to run in order to completely analyze or optimize
an analog system design, analysis of these systems can readily
overwhelm available computer resources.
In addition, many simulation induced problems tend to creep up
using detailed transistor operational models. These problems in-
clude setting the correct internal node conditions, non-converg ence
of the circuit and simulation induced oscillation. In analog simu-
lation, where signal amplif ication, precision v oltages and feedback
are common circuit parameters, circuit behavior of a specific cir-
cuit block may depend not only on the transistor models and the
simulators used, but also on user specified iteration limits and tol-
erances. If these limits are set improperly, circuit may take hours
to converge, generate useless data, or terminate abruptly in the
middle of the simulation.
In the integration of a complex ASIC w here sev eral complex func-
tional blocks are used, a detailed tr ansistor lev el simulation is sim-