The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> Berkeley Design Automation
https://designers-guide.org/forum/YaBB.pl?num=1183646486

Message started by ACWWong on Jul 5^th, 2007, 7:41am

Title: Berkeley Design Automation
Post by ACWWong on Jul 5^th, 2007, 7:41am

Hi folks,

Any one used Berkeley Design Automation's analog or RF fastSpice simulators and would like to comment on the performance/issues they have faced when using the tools ?

http://berkeley-da.com/

cheers
aw

Title: Re: Berkeley Design Automation
Post by paulm on Jul 17^th, 2007, 1:17pm

I use their analog fastspice simulator and it works very well. As they claim, I do get about a 5x faster simulation with the same results that I get from Spectre. In some cases the simulation will run up to 10x or 12x faster, it depends on the circuit.
It is a fairly new tool and there have been some small bugs with the tool, but no show stoppers. It seems that all of the big bugs have already been worked through.

Hope this helps,

PM

Title: Re: Berkeley Design Automation
Post by Jess Chen on Jul 17^th, 2007, 1:29pm

Does any one know if it supports VerilogA or VerilogAMS or any other behavioral modeling language?
And if so, is it full support or just enough to support device models?

Title: Re: Berkeley Design Automation
Post by ACWWong on Jul 17^th, 2007, 2:19pm

verilogA is supported.

The computational efficency savings over Spectre/Hspice looks very impressive (especially for larger circuits) as does the growing band of adopters...

Waiting for the verilog-D (and verilogAMS i guess) interface which is promised in 6 months... until then its seems good for top level dc & transient sims where spectre/hspice grinds to a halt...

Title: Re: Berkeley Design Automation
Post by ACWWong on Sep 30^th, 2007, 3:22am

So far I am impressed with the speed of BDA's spice solver for transient simulations, and satisfied with the accuracy... and there haven't been any real issues, which given the variety of netlists and veriloga behaviourial code we have thrown at it, is comforting.

So my next question is: would anyone like to comment on their experience with the BDA's RF tool offering, especially in reference to accuracy and comparison to spectreRF ?

Title: Re: Berkeley Design Automation
Post by pancho_hideboo on Dec 1^st, 2007, 6:04am

Hi.

I also have large interest in Analog FastSPICE and RF FastSPICE.
RF FastSPICE seems to have both HB and shooting PSS/QPSS.

I think if I get BDA's solution, both ADS(GoldenGate) and MMSIM(Spectre) will not be needed.
So I've contacted with Berkeley Design Automation.
We were very surprised to hear the prices, very very expensive, while our budget is very cold.
But I can't understand why their SPICE is very fast without any accuracy loss.

There seem to be many many good reputations about BDA's SPICE.
Why their SPICE could be fast ?

Title: Re: Berkeley Design Automation
Post by Jaw on Dec 1^st, 2007, 4:52pm

I guess that optimization based on today's computer architecture has been made. Traditional SPICE is over 30 years old, and the computer architecture has been changing a lot since then.

Title: Re: Berkeley Design Automation
Post by James Bond on Dec 4^th, 2007, 4:56am

Sounds very interesting, does anybody know if it support circuit optimization?

james

Title: Re: Berkeley Design Automation
Post by pancho_hideboo on Dec 4^th, 2007, 6:31am

http://www.berkeley-da.com/news/news_pr/news01_pr_2007_11_06.html

I expect a starter of this topics knows this tool very well.

Title: Re: Berkeley Design Automation
Post by ACWWong on Dec 4^th, 2007, 11:54am

James, i don't believe any optimization tool is offered.

Pancho, we evaluated the analogFastspice tool and decided to take it to speed up our brute force large netlist transient simulations. Although there a few niggles, its generally very easy to use.

Title: Re: Berkeley Design Automation
Post by achim.graupner on Dec 4^th, 2007, 9:57pm

Hi Folks,

are there any experience regarding mixed-mode simulation. We have started to use AMS-Designer with ultrasim solver. We observed a nice speedup compared to the spectre solver. Pure ultrasim simulations of the full chip are still to lengthy,

Regards, Achim

Title: Re: Berkeley Design Automation
Post by pancho_hideboo on May 2^nd, 2008, 1:13am

In Cadence MMSIM7.0, cadence has announced speedup of Spectre as Turbo technology.
http://www.cadence.com/company/newsroom/press_releases/pr.aspx?xml=042908_analog_design&lid=cdn_pr

On the other hand, BDA announced transient Noise analysis improvement.
http://berkeley-da.com/news/news_pr/news01_pr_2008_04_30.html

Is there anyone who compared speed of BDA's FastSpice and Cadence's Spectre-Turbo ?

Title: Re: Berkeley Design Automation
Post by byang on May 2^nd, 2008, 9:22am

I have seen many public simulation benchmark results that are almost meaningless. The benchmark results only give the size of the circuits, e.g. # of MOSFETS and RC's, and simulation time. Why this is meaningless? It is because for the same circuit, one can simualte 1ns or 1ms transient time. The simulation time will differ by orders of magnitude. Then there is also the computer difference. However, now this difference is probably not that large.

I would like to give some GSIM benchmark data that is much more meagingful. Hope that others will release their benchmark result the same way in the future. Then users will know exactly how fast the simulator, rather than reading meaningless marketing stuff.

Circuit Size: 17477 bsim3v3's, 53632 capacitors, 600 resistors, ...
Circuit Type: Sigma Delta ADC
Matrix Size: 37336
Transient Simulation Stop Time: 63us
Total # of Simulation Time Steps: 1037144

Simulation Elapse Time (Wallclock Time) : 28 hours

byang
http://www.gemini-da.com

Title: Re: Berkeley Design Automation
Post by pancho_hideboo on May 2^nd, 2008, 9:33am

I don't think so.

Using same computer and same netlist with same accuracy setting,
I can feel obvious turn around time difference.

This is not numeric time results but body feeling.

If I can feel speed without any special tuning for simulation, it is a good simulator.

Cadence tool often needs special tuning.
Even though I do tweak many laborious simulation settings for tuning, there are many cases I can not feel remarkable speed up at all.
Rather it results in common speed regardless of its very very high price of Cadence Tools.

Body feeling speed is important.

Title: Re: Berkeley Design Automation
Post by byang on May 4^th, 2008, 12:53pm

Hi, pancho_hideboo,

I agree with you actually. I think I didn't make myself clear enough in my previous post. I wanted to say that whether it is Intel CPU or AMD CPU doesn't make a big difference.

If the simulator is different, then there is a big difference certainly.

In fact, I believe very few Spice simulator can get close to the benchmark result in gave in my previous post. Many Spice simulators should be 10x or more slower than that benchmark. The purpose of my posting the benchmark data is to give readers a concrete reference data for measuring whether a simulator is fast or slow.

byang
http://www.gemini-da.com