Using Montecarlo simulation for finding Rise and fall tile of a SETTLING waveform

It looks as if it is probably working OK. Look in the file mcparam in the monteCarlo directory - this should list (on each row) the name of the parameter (and some additional information) corresponding to each column in the mcdata file. You will be seeing the statistical parameters in this file too - so that's why you're getting additional values. There's a surprisingly large number of statistcal parameters, but maybe that's just what there really are in  the statistics block in the model files.

The mcdata file should have one row per monte carlo iteration.

The message about subckts is just one of those odd things that happens sometimes - I think that's been cured in a later version of spectre (I'm not 100% sure, but I didn't see it).

Regards,

Andrew.

 Dear Andrew,

Thanks a lot for your help.

I just want to run my input.scs file using spectre command like you did. Can you please tell me how I can do that.

In my case after montecarlo run by the ocean, I am not seeing any histogram plot being plotted automatically Can you please tell why it is so and how it can be fixed.

Kind Regards

I just ran spectre input.scs (you might need to give the -I argument to provide any include paths). The only reason I did it this way was because it was quicker for me to put together a simple testcase that I could share without needing any schematics.

The histograms don't get plotted automatically in OCEAN - you'd need to do (after the monteRun()):

monteResults()
histogram("RISE_TIME_27")

You can see the available names from the monteResults() output.

ocnHelp('histogram)

will tell you the other optional arguments for the histogram() function.

Regards,

Andrew.

 Dear Andrew,

I did some modification to the code in .artilinit file.

It is NOT working at all.

Can you please have a look at it and tell whether it should work or not.

 procedure(MyRiseTime(signal CLOCK_DUTY_PERIOD CLIP_START CLIP_END)
  let((clip_I_P MIN_VALUE MAX_VALUE UPPER_THRESHOLD LOWER_THRESHOLD RISE_UPPER RISE_LOWER)
    clip_I_P=clip( signal CLIP_START CLIP_END)
    MIN_VALUE=ymin(clip_I_P) + 0.02
    MAX_VALUE=ymax(clip_I_P)- 0.03
    UPPER_THRESOLD=MIN_VALUE + ( (MAX_VALUE-MIN_VALUE)*0.9 )
    LOWER_THRESHOLD=MIN_VALUE + ( (MAX_VALUE-MIN_VALUE)*0.1 )
    RISE_UPPER= cross( clip_I_P UPPER_THRESOLD 2 'rising )
    RISE_LOWER= cross( clip_I_P LOWER_THRESHOLD 2 'rising )
    RISE_TIME=abs(RISE_UPPER-RISE_LOWER)
    ;;abs(RISE_UPPER-RISE_LOWER)
    if( (RISE_TIME >= CLOCK_DUTY_PERIOD)   then
        CLIP_START = CLIP_START + (CLOCK_DUTY_PERIOD*1.5)
        clip_I_P=clip( v("/MIXER_I_P" ?result 'tran) CLIP_START CLIP_END)
        w = newWindow()
        plot(clip_I_P)
        RISE_UPPER= cross( clip_I_P UPPER_THRESOLD 2 'rising )
        printf("%e \n", RISE_UPPER)
        RISE_LOWER= cross( clip_I_P LOWER_THRESHOLD 2 'rising )
        printf("%e \n", RISE_LOWER)
        abs(RISE_UPPER-RISE_LOWER)
    else
            abs(RISE_UPPER-RISE_LOWER)     
     )
  )
)

My Monte expression is :

 monteExpr("RISE_TIME" sprintf(nil "MyRiseTime(v(\"/MIXER_I_P\" ?result 'tran) %g %g %g)" CLOCK_DUTY_PERIOD CLIP_START CLIP_END))

My monteexpression is abelow:-

monteExpr("RISE_TIME" sprintf(nil "MyRiseTime(v(\"/MIXER_I_P\" ?result 'tran) %g %g %g)" CLOCK_DUTY_PERIOD CLIP_START CLIP_END)) 

No, that should not work. Your function is plotting things - and that makes no sense. This function is called by the simulator during each iteration of the monte carlo as a background process - it invokes the "artil" executable to do this. artil has no ability to plot anything, and so the calls to newWindow(), plot(), etc would fail.

Even if it did work, you'd end up with many windows being opened (if you did a reasonable number of statistical points), which I'm sure you'd find annoying...

The functions called from monteExpr can only compute a scalar result - they cannot do anything more complex like plot anything.

Regards,

Andrew.

 Dear Andrew,

I commented the plot() and printf() commands.

And ran the ocean. But it didn't work.

it is showing the following errors

subckts: writing subcircuits to rawfile.
\o
\o
\o Error found by spectre during Monte Carlo analysis `mc1'.
\o
\o     ERROR (SPECTRE-8003): subckts: Error evaluating ocean expression
\o         `RISE_TIME=MyRiseTime(v("/MIXER_I_P" ?result 'tran)
\o             2.5e-10 8e-07 9e-07)'.
\o
\o
\o
\o Unsuccessfully evaluated export statements (based on return code).
\o
\o Analysis `mc1' was terminated prematurely due to an error

Dear Andrew,

The idea of doing this if-else waveform is that:-

As the wave is settling, one can't predict the time instant where the start of the wave is clipped at the middle of the clock period.

Without the if-else, I sometimes found some wave getting clipped at the middle of the rising edge which gives me erroneous results.

Kind Regards,

Without seeing your data, it's really hard to debug (as I keep saying - customer support is the best avenue for this). The most likely scenario is that the cross() function is returning nil in some cases, and hence the subtraction is failing. Maybe if you do:

        RISE_UPPER= cross( clip_I_P UPPER_THRESOLD 2 'rising ) || 0.0
        RISE_LOWER= cross( clip_I_P LOWER_THRESHOLD 2 'rising ) || 0.0

then it will succeed (albeit giving the wrong answers) but it might highlight where the problem is. If you save the waveform data from the monte carlo by adding ?savefamilyplots "yes" to the monteCarlo() call in your OCEAN script, then you can look at the waveforms for each point and check them after the event.

The code change above says to either return the cross() value, or if cross returned nil, it will set RISE_UPPER or RISE_LOWER to 0.0.

Regards,

Andrew.