matlab interface and PSF data format

 IMHO makes no sense.

 1. PSF is a proprietary waveform format of Cadence, you will  not 
     find any specification or documentation about.
 2. The Analog Waveform Display (AWD) which evaluated the PSF will be replaced
     finally in DFII 6.1. by WaveScan which has the wave form format SST2.
     So you would develop something for a dead end.

The best would be if Cadence would support an open waveform standard.
There are change requests for this, but as far as I know there has not much
happened in this direction.

 Bernd

Is SST2 a cadence proprietary waveform format also?

How about the calculator function flow? Has anyone try to use it to interface with matlab?

 A couple of comments here.  

 1) PSF supports an ASCII mode which is pretty easy for creative coders to write to and the "psf" executable can converts ascii to binary and back if the ascii performance is not acceptable.   Just take any psf file and run the "psf" executable on it and you will see the ascii output (% psf tran.tran)

2) PSF and SST2 are both proprietary formats also Cadence does as part of its OASIS simualtion integration to ADE provide access to writing PSF but you must purchase this integration kit to gain access.   For newer formats like SST2, access is not provided.

3) I've shared in my feedback to the subject "How to plot a graph" (Feb09) the process of using "csv" files to import external data into WaveScan & ViVA XL as this is pretty simple as well as the option to use the calculator function "getAsciiWave".

4) If you have VCD data then you can easily convert to SST2 using the simvisdbutil exectuable that comes in the IUS (AMS) hierarchy.

 Hope these options help,

 jons

I agree!. Although a object oriented should be nicer even. It would be easy to look to my houndres of Gbytes of data with something like python. It is time overcome this 30 years type of analog tools. Please Cadence! 

This thread is pretty old so your response was against the forum guidelines which ask you not to post on old threads.

Anyway, we've had a Spectre Toolbox for Matlab which allows spectre results to be read directly into Matlab since something like 2005 or 2006 (I forget exactly when it was introduced). We also introduced a new, more powerful interface from ADE Explorer and Assembler to Matlab which allows Matlab to be used for big data processing on simulation results in a much simpler fashion - including querying across all of your large numbers of sweeps and corners without necessarily needing to access all the individual data if you don't want to.

This was launched at this year's CDNLIve EMEA (mentioned in the Breakfast Bytes blog) - for more details contact your account team. There's also a Rapid Adoption Kit on this.

We did consider a Python interface too - but whilst Python is undoubtedly popular for data processing (using things like Numpy, Matplotlib etc), in the Engineering community Matlab is much more commonly used - so it was better to implement something that had wider appeal.

Regards,

Andrew.

Hello Andrew, 

Thanks for the reply, and sorry for positing a out-of-topic message. I was looking for alternatives to skill to run my simulations and posprocess the results in the context of analog/RF circuits. Python or matlab are both good.  I know the video and know the integration with matlab. I use it.  I am not sure if the new version would  allow me to control the simulations from matlab. One of the strengths that make spectre/ocean the default standard 20 years ago was that all of it was easily controlled and scripted using lisp at that time a state of the art tool(Skill). Unfortunately with the years this is really outdated. Skill lacks way behind compared with matlab/python in numeric or plotting capabilities, not without mention easy to use and number of existing libraries. As the complexity of the verification grows the solution by CAD vendors seems to go to more complicated GUIs, but that would not be sustainable as the number of corners, complex optimizations is growing much rapidly that anyone can make a GUI. In my opinion a good API in any of the two language that can also run the simulations is required so a single script can control variables, run simulations and pos-process the results.

In Python or Matlab  we can implement complex optimization algorithms for our circuits or using similar pos-processing algorithms that we will use during measuring. In some way the solution now is what cadence did 20 years ago.  The best simulator, as still spectre-RF is, with a state-of the art scripting language.