Problem with ocean script

You didn't say which IC version you were using, but I was able to (I think) reproduce this using the obsolete monteCarlo/monteRun functions in IC617. What I get is this:

iteration (rea abs(IDC("/V1/PLUS")) (A)

1000000 0.0010374721630195
1000000 0.0010428013956719
1000000 0.0010327899393591
1000000 0.001046730050344
1000000 0.001054421433944
1000010 0.0010291098681963
1000010 0.001051662346496
1000010 0.0010464787884481
1000010 0.0010567360862235
1000010 0.0010333544648214

Looking more closely, the issue is how the logFile is written by spectre in this obsolete mode - the raw data has a rounding error in the iteration field, and so it's not ocnPrint that is doing anything wrong. Given that use of these old OCEAN functions is no longer supported:

*WARNING* IC5141 based Monte Carlo scripts will not be supported in future release of IC61 OCEAN.
Users should migrate their Monte Carlo scripts to OCEANXL

I just tried running from ADE Explorer (which would be the same as doing it from OCEAN XL) and then used openResults() on the resulting directory and ocnPrint as above, and I get:

mcparamset abs(IDC("/V1/PLUS")) (A)

1000001 0.0010421889
1000002 0.001057634932
1000003 0.00104455059
1000004 0.001039197119
1000005 0.001046654913
1000006 0.001049822413
1000007 0.001051257347
1000008 0.001037424219
1000009 0.001063850924
1000010 0.001046753547

Note that the index is called mcparamset in the new flow - however, as you can see, it's not truncated. Given that the old monteCarlo/monteRun flow is obsolete, I think there is no chance of getting it fixed - you should migrate to the new ocnxl-based way of running Monte Carlo (or perhaps using maeRun if using the new Maestro views...)

Regards,

Andrew.

One other thing I meant to ask - why are you setting the starting iteration to be greater than 1 million?

Regards,

Andrew

Hi Sir, 

That is because my experiment needs many circuit instances. I have already collected 1 million Monte Carlo simulation result of my circuit, but I still need  one more million for my experiment. 

Ohh, can I ask what is the difference between "Process" and "Mismatch" when setting the Monte Carlo simulation?

My understanding is that: "Process" has larger process variations than "mismatch". For example, if my circuit is just a memory pixel, I try to run monte carlo simulation 64 times. If "Process" is selected, does that mean the 64 monte carlo simulation results are produced by 64 circuit instances (1 instance 1 pixel)?  If "mismatch" is selected, does that mean the 64 monte carlo simulation results are give by 64 pixels in one circuit instance (1 instance 64 pixels)? 

Thanks,

Zheng Yue

Process means that one set of random numbers are generated across all the devices in the design. This corresponds to the die-to-die variation you might get from different samples. The mismatch variation models the local variation of each device (transistor) - so the remaining random variation between devices on the same die. You can't say that process has larger variation than mismatch - they are modelling different effects.

The best approach for simulating variation on large arrays is to use the high sigma variation techniques provided with the Virtuoso Variation Option with ADE Explorer/Assembler in IC617 and later - also available with ADE GXL too.

In your other reply you said you are using IC5141 (which was released in 2004, and stopped being supported in 2011), so that doesn't support OCEAN XL functionality, since that was introduced with the advent of ADE XL in the IC61X release streams. You probably should move to something more recent than a 15 year old release stream that stopped being supported over 7 years ago.

Regards,

Andrew.

That is very helpful. Thank you very much!

That is very helpful. Thank you very much!