( ESNUG 422 Item 2 ) -------------------------------------------- [02/19/04]
Subject: ( DAC 03 #39 ) A Detailed Review Of NeoLinear NeoCircuit 2.0.6
> I have seen the NeoCircuit demo, and gone to a half-day workshop on
> their "RAD" flow. I have also seen a demo of ADA's tools. Both ADA and
> NeoLinear have intelligent ways for you to set up constraints for device
> sizing of an analog circuit topology. Since I haven't run the two tools
> on a test design, I cannot say whether one or the other does a better
> job converging on an optimized solution. This would be an important
> thing to do if you were choosing one of the two tools. If one tool
> consistently found a good solution in fewer iterations (and therefore a
> shorter time) that would be strongly in its favor.
>
> - from http://www.deepchip.com/items/dac03-39.html
From: Jean-Philippe Dieusaert <jean-philippe.dieusaert=user domain=st spot calm>
Hi, John,
I did an eval of NeoLinear's NeoCircuit 2.0.6 in June. Since then,
NeoLinear provided a new release that is quite improved compared with
2.0.6. We are only starting to use release 2.2.7 now so I cannot make
any comment on it yet. I can comment release 2.0.6 only.
NeoCircuit 2.0.6 Strengths:
- Easy to use. Don't need a long training to use it.
- A junior designer can use the tool to size the circuit (with the
help of an expert designer)
- Sizing can be done on multiple CPUs during the night (when the
- simulation licenses are normally not used)
- Robustness: does not crash. The tool is stable
- Gives good results in sizing with a reasonable time ... if your
constraints are correctly defined (see 1st weak point below).
- Really useful for resizing after migration from 1 technology to
another one.
- Also useful in case of simulation models update
- Automatic HTML datasheet generation of the sized circuit is very
useful
- Automatic back-annotation of the solution into the schematic
- Integrated inside Opus
- If everything is correctly defined, you are sure the tool checked
every simulation. As opposed to doing it manually, where you always
could forget to check one simulation after an update of your
schematic
- It allows you to see the bottleneck of your design : where
NeoCircuit is stuck
- It could be nice to recheck the topology of your design to improve
it during sizing
- It allows matching between devices
- It can be used to minimize of maximize some goals : you have an
already good manual solution, but it can be that the tool finds a
better solution than you.
- Allows you to measure only some parameters (no sizing, just
measuring)
- Can be used with several different simulators
- Expression toolbox is very useful or reuse of .extract in Eldo
possible
NeoCircuit 2.0.6 Weaknesses:
- The more you use the tool, the more you get experienced with it ...
I mean that you start to learn some tricks that are never mentioned
in the training. And these tricks are quite important to obtaining
good results with the tool.
- If you start to know how the tool reacts, you can define your
constraints more easily to achieve the results you want to have.
- In 2.0.6, you can sometimes have some strange results (that will
not happen anymore once you gain expertise). E.g.
1) If you make 2 same runs, you can sometimes get a different
solution.
2) If you make 1 run with certain variable range, you get a
solution ... and if you increase your variable range, the
previous solution you found may not be found anymore (while it
should as you relax your constraints)
3) If you found a solution that was much better than the goal
you wanted to have, if you make a 2nd run with a goal value to
reach just a little bit smaller that the solution found, the
tool may not found the solution anymore, while the solution
exists as it found it just in the previous run.
- No corner integration in 2.0.6, nor any good methodology to use
corner analysis with the tool. You have to recheck the results out
of the tool, yourself, to check the corners. This seems to be
improved in 2.2.7 ... but need to be checked. I'm worried about the
simulation time
- Error messages are not understandable ... seems improved in 2.2.7.
- The setup of the constraints takes some time.
- Sometimes you need to make several runs to tune your constraints in
order to find a solution.
- No visibility of what the tool is doing during its sizing (seems to
be improved in 2.2.7) except with the "compare" function.
- Doesn't work with a version of Eldo below 5.4 (which is often used
for old DesignKit at ST)
- If you want to make your sizing in parallel: it uses a lot of CPU
and simulation licenses.
- After having defined your goals and variable ranges, afterwards it
is not possible to classify them. The only order for displaying
them is the entry order.
- Can sometimes give different results if you run a stand alone
simulation on the backannotated schematic with the found solution.
Mainly with Eldo and .extract using a formula (GBW, PM, ...)
- Need to know up front for your technology the allowed values for
your variable ranges. Because your CDF callbacks in Cadence are not
triggered until you back-annotate, you can throw your solution away
if you gave a too big variable range for 1 parameter that is now
out of the allowed technology rate.
But the latest release still need to be checked because I've seen the
improvements and they seem very good.
Conclusion:
NeoCircuit (2.0.6 !!) can be a valuable tool in optimizing, migration of
technology, and simulation model update.
There is added value in term of time gain, a better solution, a junior
designer can do the sizing, compared with a manual sizing, and the tool
was quite good and gave us better results in a reasonable time.
- Jean-Philippe Dieusaert
ST Microelectronics Zaventem, Belgium
|
|
