david.garner@[EMAIL PROTECTED]
wrote, on 09/03/08 17:18:
> On Jul 7, 9:35 pm, Andrew Beckett <andr...@[EMAIL PROTECTED]
>
> wrote:
>> bageduke wrote, on 07/05/08 02:44:
>>
>>> I am doinglongtransientsimulation for 100us. The simulation is
>>> still running, but everytime when I tried to plot signal, it only
>>> plotted up to 36us, but actually output log shows the simulation
>>> already run more than 50us without any error, and thetransientresult
>>> file size still keep increasing.
>>> Have you ever seen this problem before? Does this mean the result is
>>> corrupted? Or is there anyway I can fix the problem?
>>> Thanks a lot.
>> Are you using IC5141 with spectre from an MMSIM release? If so, you
probably
>> need to re-enable the "chunk" mode for writing PSF. In older spectre
versions,
>> big files used to get split into 2Gbyte chunks, because large file
sup****t was
>> not particularly common in many OS versions. However, that's no longer
true - so
>> spectre from MMSIM61 changed to write largetransientPSF into a single
big
>> file. However, sup****t for reading such big files wasn't added into
IC61 (I
>> think), and so if running a new spectre with IC5141, you need to do:
>>
>> For Spectre writing PSF data:
>>
>> setenv PSF_WRITE_CHUNK_MODE_ON true
>>
>> For UltraSim writing PSF data:
>>
>> setenv PSF_LARGE_FILE_ON false
>>
>> Regards,
>>
>> Andrew.
>
> Hi Andrew,
>
> I have a similar problem. I am simulating a sigma-delta converter and
> only the first 19 clock cycles would be displayed, despite simulating
> for 1024 or more, and the simulation clearly running through to
> completion.
>
> I enabled the above environment variables and found that I could see
> the first 177 cycles in AWD - a vast improvement - and the entire
> waveforms in Wavescan. But any calculator functions I do on those
> waveforms are only using the first 177 cycles.
>
> This does not appear to be related to the tran.tran file size - I have
> tried using various errprst settings and only saving every 10 points,
> etc.
>
> Any ideas?
>
> David.
I've just found some issues related to changing the stop time when using
SST2.
The problem seems to be that the wrong precision gets used if you lengthen
the
stop time, and what can happen is that the x-values hit the 64 bit limit
with
the original precision (summarizing it slightly). This is covered in CCRs
574546
and 517273 - for reference if you contact customer sup****t.
This has just been fixed in IC 6.1.2.500.17 - but I don't believe it is
due to
be fixed in IC5141 at the moment.
Probably the simplest thing would be to change the output format to PSF.
Use
the following .cdsenv:
spectre.envOpts simOutputFormat string "psfbin"
or the following SKILL in your .cdsinit:
envSetVal("spectre.envOpts" "simOutputFormat" 'string "psfbin")
(note, you can't change this if you're already running ADE). You'll most
likely
need to force a new netlist (Simulation->Netlist->Recreate) to ensure the
mappings are updated correctly.
Best Regards,
Andrew.
|
