Problems visualizing results with Wavescan - MMSIM 6.2 - IC 5.1.14

Hi Pedro,

This option works with PSF format only. Please make sure your format is PSF and not SST2 which is now the default. If you are using PSF, Did you restart your Cadence session when you have setup this variable ? I really doubt it ! You really need to close down your Cadence session and restart it again in order for your environment to enherit the chunk mode.

You can still view your 70G file though. Simvision can open this file if it is in SST2 format. Otherwise, the only available option is using VIVA which is the new viewer in IC6.

Cheers,

Riad.

Hi Pedro

Since the filename is "tran.tran", it seems that the output format has been correctly set to psf. For sst2, it would have been "tran.tran.trn". Just like what Riad has suggested, first we have to be sure that the variable is in effect. What do you get if you execute the following command in ciw?

getShellEnvVar("PSF_WRITE_CHUNK_MODE_ON")


Best regards
Quek

 Hi, thanks for your replies! it looks like I had a mistake on one my files and I was using sst2 format instead of psf. Looks like everything works fine now :)

By the way, do you know how could I invoke  Simvision from the command line?...I know this is a very newby question, but...

Thanks,

Pedro

Hi Pedro

You can use either of the following cmds:

unix>simvision &

OR

unix>simvision -64bit &

Best regards
Quek

Hello.

I encountered a different problem, but it came down to the same thing discussed here, so I thought it might be better to just keep everything in one place.

Namely, I  am trying to do an S-parameter analysis, but somewhere about a half of the simulation it stops and I get:

Fatal error found by spectre at freq=... during SP analysis 'sp'.

write error on PSF file: Error detected in psf library while writing to file 'sp.noise.sp'

Browsing the Internet I came to this problem - and realized that my sp.noise.sp is actually exactly 2 GB of size. So I started looking for solution. First I set the variable:

PSF_WRITE_CHUNK_MODE_ON=true

export PSF_WRITE_CHUNK_MODE_ON

And when I ask for getShellEnvVar("PSF...ON") in CIW, I do get "true".

Then I found that this only works when you work with psf mode. Looking into my spectre.out file I realized that it holds a line "-format sst2". Then I found that in order to change that I am supposed to go to ADE -> Setup -> Environment... and set "userCmdLineOption" to "-format psfbin", without the quotes. Saved the state, of course.

Which I did. Then I closed down all the Cadence stuff, even the terminal, started everything all over again, loaded the state holding the change I just did, and run the simulation.

Again I have the same thing, and again when I look into the spectre.out file I see "-format sst2" in there.

The versions I am using are: IC5 1.41.151 and MMSIM 11.10.509.

What am I supposed to do and how do I actually change the format? 

Thank you very much.

p.s. If you have some additional time, what exactly are these formats and what is the actual difference? Why do we even have two?

First of all, the forum guidelines tell you not to post on the end of an old post - this one is 4 years old. But forgiving that for now, here's some answers:

1. The "chunk mode" is only supported for PSF, and even then only for transient (not for SP or AC type data). 
2. Changing the format via the usrCmdLineOption isn't really the right way to do it. For a start, ADE will output -format sst2 and then later you'll have -format psfbin too on the command line. Spectre will actually use whichever format is specified last on its command line - so psfbin will win. However, certain mapping functions in ADE will still think that sst2 was used, and hence it might have issues plotting currents in particular. As a result, the right thing to do would be to use:
   envSetVal("spectre.envOpts" "simOutputFormat" 'string "psfbin")
   before starting ADE.
3. It's somewhat irrelevant picking psfbin or sst2 anyway, since if sst2 is used as the format, it's only used for transient data. For sp and ac data etc it used psfbin anyway since sst2 doesn't support representing the kind of data needed.
4. The reason why there are multiple formats is historical. There are some very old formats supported by spectre (e.g. nutmeg), and then there is psf (in various flavours) and sst2. sst2 originated from the digital tools, and was made the default in ADE because it handled certain large transient databases more quickly than the older psfbin. In more recent IC version (IC615, IC616) the default is now "psfxl" which is a much newer format which scales much better when dealing with huge amounts of transient data (it's still using psfbin for some of the small signal analyses though even if you pick psfxl). In IC61X there's a field on a form in ADE now to choose the output format as well as a cdsenv setting.
5. I'm pretty amazed that an sp analysis could produce that amount of data. You'd have to have a very large number of ports and also a very large number of points in your frequency sweep, I'd have thought. Even with noise... so I'd check your setup.

Regards,

Andrew.

First of all, sorry about the breaking the rules - I really do apologize. I'll keep that in mind from now on (except for this post :) ).

Thank you very much for your time and a very thorough and clear answer - I really do appreciate it.

It did hit my mind to reduce the number of points in the simulation, but once I got hooked on this format issue, I really wanted to go through it to the end.

From your answer I understood that it already is at the end. Namely,

1. in order to do SP analysis, psfbin is the format used anyway, there is no choice (at least for my version of software).

2. psfbin allows chunks only for transient analysis, so I am limited to a 2 GB file anyway.

Therefore, in order to do my analysis, the only way out is to reduce the accuracy of the simulation, be it reducing the number of points or the number of parasitics or any other way.

Once again, I am very greatful - thank you very much.

For an sp analysis, the parasitics shouldn't make a difference. It's only going to depend on the number of ports and the number of points in the sweeps.

Regards,

Andrew.