Re : distance and depth based aocvm variation

[jaya sree]

Hai everyone ,

I am unable to understand why derates values decrease as depth increase and why derated values increase when distance increase. It would be a great help if some one explains.

Thank you
jaya sree

 

[adunsmoor]

The differences are due to what the depth and distances derates are attempting to model.

Distance based AOCVM is used to model global variation effects. It's assumed that different parts of the chip will vary more so that's why as distance increases you see more variation.

Depth based AOCVM is used to model local variation effects. Local variations tend to cancel out over the course of a long path - some delays are a little worse, some are a little better, but as the depth increases the derate approaches 1 (or some other value depending on your methodology).

In my experience, distance based is used much *less* frequently than depth based AOCVM, today.

Does that answer your question? If not I'd be happy to post a more thorough explanation.

 

[jaya sree]

Thank you adunsmoor

I understood distance based analysis.I am unable to figure out why local variations cancel out as depth increases.It would a great help if u explain it

 

[research235]

hello jayasree

Please go throgh the following link. can clear some of your doughts

https://www.synopsys.com/Tools/Implementation/SignOff/CapsuleModule/PrimeTime_AdvancedOCV_WP.pdf

 

[jaya sree]

Yes.Now I have understood the concept .Thank you very much.

 

[jaya sree]

hai , I am bit confused with (distance / depth ) concept and ( graph/path) concept.
In my project , for timing analysis using primetime, first graph based analysis is used and then path based analysis.My understanding is that now aocv (distance /depth) is applied on it.Please correct me if I am wrong.How these two concepts are related.

 

[research235]

Hi Jaya
can you giv a little back grnd on wat stages of P&R are thse two kind of analysis used ?

 

[jaya sree]

Hi ,

i think after final routing is over, i.e during final timing analysis

 

[research235]

hi jaya.

Using depth based analysis at final routed(signoff) is common practice (This gives more accurate results and takes more time),but for CTS, and othr stages graph based can be used which is less accurate(takes less time for analysis).
I hope this answers ur question.

 

[jaya sree]

o.k. Since in my projects , depth based aocv is used , can i make a staement that " at sign off path based analysis (with depth based aocv ) is used and at previous stages i.e CTS etc graph based analysis (with depth based aocv is used)" . Please correct me if my assumption is wrong.

 

[research235]

You can say that, and normally tat is how it shld be done. but check in your timing analysis scripts to confirm teh sme.

 

[jaya sree]

yes. It is like that. I have checked . thank you

 

[nikhil.madhekar]

hi research235,
I still didnt get the clarity on how local variations cancel out as depth increases. please help me on this. i have gone through the synopsis pdf .

 

[research235]

Hi Nikhil

The local variation does not cancel out as the depth increases. But OCV values are calculated as %, ie. STA would consider de rate factors for signoff.
so for the the first cell we de rate the data or clock path, as we go deeper into the logic for every cell data and clock path are de-rated which is over pessimism. so we can say as we go deeper into the path these derate factors get smaller.

Hope this helps u.

 

[jaya sree]

Hai nikhil,

Lets assume that there are 6 buffers in a data path. Generally , in OCV , the cell delays are derated with same number , ( ex . 20ps ).So 20ps is worst delay for each cell. Therefore total worst delay is 20x6 = 120ps.this value is taken and slack is calculated. But there is no rule that all cells will produce 20 ps delay.some cells may be fast.so total delay will be less than 120ps. This effect will be more with more depth.Hence path based analysis is more accurate one.

I hope my explanation is correct.Please anyone correct me if i am wrong

 

[matter]

I agree with most of the comments in this thread. But I also think folks have confused GBA/PBA with OCV/AOCV. Both are entirely different concepts.
PBA is always more accurate than GBA regardless of whether flat OCV or AOCV is used.
Likewise AOCV will be more accurate than flat OCV regardless of whether PBA or GBA is used.