Derating factors and k-factors

[pdude]

Hi,

I would like to know if there is any difference between the K-factors seen in the liberty vs the derating factors we use for AOCVM ??
I have seen it written in many places as K-factor derating but since we have the K-factors in liberty and in case of multi-voltage it is 0 but we use AOCVM files i wanted to know the difference and why we cannot use k-factors in order to do the OCV.

Thanks:thumbsup:

 

[erikl]

Liberty shouldn't ever get derated!

SCNR! ;-)

 

[pdude]

Liberty shouldn't ever get derated!

SCNR! ;-)

Hi,

Then why do we have K-factors ? It is used for scaling the various components of the liberty according to change in voltage or temperature is what i read.
A lil elaboration of your Answer would be great

Thanks

 

[erikl]

A lil elaboration of your Answer would be great

You didn't get me at all, so sorry! It was a fun answer. Not even the SCNR note led you there?

Seems you don't know the meaning of liberty. I'd suggest to look it up in an English dictionary!

Anyway, I'd think your thread would fit much better in the ASIC Design Methodologies and Tools (Digital) forum, because your question addresses the PrimeTime & STA tools' usage. I'd recommend to open a new thread - with new wording - in that forum, and close this one.

May be this thread or this article could help you a bit. However there's no mention of k-factor derating, unfortunately.

 

[pdude]

Hi,

Thanks for writing in. This in the 1st time would have been more helpful, was happy to see a moderator answer me. And yeah it was naive of me to ignore the SCNR remark :bang:

I do not say i know a lot, just learning, and thanks for the patience to point out the threads and articles. However, i think i know the meaning the liberty quite well. You dont need a dictionary to find a VLSI terminology. Anyways, please refer to this link : https://www.csee.umbc.edu/~cpatel2/links/641/slides/lect05_LIB.pdf

The liberty does have k-factors and it is used for scaling it or 'derating' the values of transition. setup etc..
This part is clear to me.

My question was entirely different and yeah i am not claiming i always ask the right thing but i would prefer a SCNR with a proper answer.
Thanks for your time and the links.

 

[erikl]

... Anyways, please refer to this link :
https://obrazki.elektroda.pl/3149671300_1420790942.png

Yeah, here I can see where your error comes from: this US Maryland University paper made the same error, at least in its title. In sight of this lapse, your error may be forgiveable, I'd say!

Liberty means freedom - you probably now may understand my ironic note above.

Here, however, a Library Timing File is meant:

My question was entirely different ...
The liberty (!) does have k-factors and it is used for scaling it or 'derating' the values of transition. setup etc..

AFAIK, AOVCM resp. OCV deals with path-lengths dependent delays, whereas k-factor derating concerns PVT delay dependencies.

 

[FvM]

I agree that liberty shouldn't be derated.

Nevertheless it's also the name of a Synopsys created library format.

 

[erikl]

I agree that liberty shouldn't be derated.

Nevertheless it's also the name of a Synopsys created library format.

I have to admit I didn't know this before - shame on me!

So I'm sorry, pdude, to have rendered you confused by my previous answers!

Still it's my opinion, too, that liberty shouldn't get derated!

 

[pdude]

I agree that liberty shouldn't be derated.

Nevertheless it's also the name of a Synopsys created library format.

Thanks for clarifying.
I understand that derating (AOCVM) concept but i just want to know that how is the k-factor method different from derating , considering both include multiplication of some factors with the delay and/or other factors.

 

[ro9ty]

Hi,

Thanks for writing in. This in the 1st time would have been more helpful, was happy to see a moderator answer me. And yeah it was naive of me to ignore the SCNR remark :bang:

I do not say i know a lot, just learning, and thanks for the patience to point out the threads and articles. However, i think i know the meaning the liberty quite well. You dont need a dictionary to find a VLSI terminology. Anyways, please refer to this link : https://www.csee.umbc.edu/~cpatel2/links/641/slides/lect05_LIB.pdf

The liberty does have k-factors and it is used for scaling it or 'derating' the values of transition. setup etc..
This part is clear to me.

My question was entirely different and yeah i am not claiming i always ask the right thing but i would prefer a SCNR with a proper answer.
Thanks for your time and the links.

Given the fact that derating is mentioned whereever you find reference to k-factors it is quite obvious to get confused but there is quite a lot of difference between k-factors and derating factors used for AOCVM.

If you open your.lib (or LIBERTY) :grin: then you will find something called nominal operating conditions. These are the PVT conditions on which the lib has been characterized, say Pslow, V90, T100 but now suppose inside your dsign you specify an operating condition as Pslow,V95,T110 then this differes from the characterized values. So your values in the .lib have to be scaled to bring it upto the mark of the operating condition, say if some transition time was 20 then for more Voltage it should be more (ideally) so it is multiplied with a k-factor (say 0.25) then it becomes 20+20*0.25 = 25 (just for example) This is k-factor scaling but nowadays since we use MCMM we prefer to supply the tool link libraries for every operating condition. So, the k-factors nowadays is 0, however major tool makers have decided to include interpolation in MCMM too.
.
About derating in OCV -> Say your operating condition is Pslow,V90,T100 and you have the link library for this but to do OCV i.e add extra pessissism you will multiply a factor to increase the delay in launch path and reduce it in capture path (setup) . This is multiplied to al the delays.
Now understand tis that OCV will be added on top of k-factors if different set of operating conditions are used and if the library uses interpolation but because we do not use k-factors now-a-days so only OCV derating factor comes into picture.

Hope it helps.

Thanks,
ro9ty

- - - Updated - - -

Hi,

Thanks for writing in. This in the 1st time would have been more helpful, was happy to see a moderator answer me. And yeah it was naive of me to ignore the SCNR remark :bang:

I do not say i know a lot, just learning, and thanks for the patience to point out the threads and articles. However, i think i know the meaning the liberty quite well. You dont need a dictionary to find a VLSI terminology. Anyways, please refer to this link : https://www.csee.umbc.edu/~cpatel2/links/641/slides/lect05_LIB.pdf

The liberty does have k-factors and it is used for scaling it or 'derating' the values of transition. setup etc..
This part is clear to me.

My question was entirely different and yeah i am not claiming i always ask the right thing but i would prefer a SCNR with a proper answer.
Thanks for your time and the links.

Given the fact that derating is mentioned whereever you find reference to k-factors it is quite obvious to get confused but there is quite a lot of difference between k-factors and derating factors used for AOCVM.

If you open your.lib (or LIBERTY) :grin: then you will find something called nominal operating conditions. These are the PVT conditions on which the lib has been characterized, say Pslow, V90, T100 but now suppose inside your dsign you specify an operating condition as Pslow,V95,T110 then this differes from the characterized values. So your values in the .lib have to be scaled to bring it upto the mark of the operating condition, say if some transition time was 20 then for more Voltage it should be more (ideally) so it is multiplied with a k-factor (say 0.25) then it becomes 20+20*0.25 = 25 (just for example) This is k-factor scaling but nowadays since we use MCMM we prefer to supply the tool link libraries for every operating condition. So, the k-factors nowadays is 0, however major tool makers have decided to include interpolation in MCMM too.
.
About derating in OCV -> Say your operating condition is Pslow,V90,T100 and you have the link library for this but to do OCV i.e add extra pessissism you will multiply a factor to increase the delay in launch path and reduce it in capture path (setup) . This is multiplied to al the delays.
Now understand tis that OCV will be added on top of k-factors if different set of operating conditions are used and if the library uses interpolation but because we do not use k-factors now-a-days so only OCV derating factor comes into picture.

Hope it helps.

Thanks,
ro9ty

- - - Updated - - -

Hi,

Thanks for writing in. This in the 1st time would have been more helpful, was happy to see a moderator answer me. And yeah it was naive of me to ignore the SCNR remark :bang:

I do not say i know a lot, just learning, and thanks for the patience to point out the threads and articles. However, i think i know the meaning the liberty quite well. You dont need a dictionary to find a VLSI terminology. Anyways, please refer to this link : https://www.csee.umbc.edu/~cpatel2/links/641/slides/lect05_LIB.pdf

The liberty does have k-factors and it is used for scaling it or 'derating' the values of transition. setup etc..
This part is clear to me.

My question was entirely different and yeah i am not claiming i always ask the right thing but i would prefer a SCNR with a proper answer.
Thanks for your time and the links.

Given the fact that derating is mentioned whereever you find reference to k-factors it is quite obvious to get confused but there is quite a lot of difference between k-factors and derating factors used for AOCVM.

If you open your.lib (or LIBERTY) :grin: then you will find something called nominal operating conditions. These are the PVT conditions on which the lib has been characterized, say Pslow, V90, T100 but now suppose inside your dsign you specify an operating condition as Pslow,V95,T110 then this differes from the characterized values. So your values in the .lib have to be scaled to bring it upto the mark of the operating condition, say if some transition time was 20 then for more Voltage it should be more (ideally) so it is multiplied with a k-factor (say 0.25) then it becomes 20+20*0.25 = 25 (just for example) This is k-factor scaling but nowadays since we use MCMM we prefer to supply the tool link libraries for every operating condition. So, the k-factors nowadays is 0, however major tool makers have decided to include interpolation in MCMM too.
.
About derating in OCV -> Say your operating condition is Pslow,V90,T100 and you have the link library for this but to do OCV i.e add extra pessissism you will multiply a factor to increase the delay in launch path and reduce it in capture path (setup) . This is multiplied to al the delays.
Now understand tis that OCV will be added on top of k-factors if different set of operating conditions are used and if the library uses interpolation but because we do not use k-factors now-a-days so only OCV derating factor comes into picture.

Hope it helps.

Thanks,
ro9ty

- - - Updated - - -

Hi,

Thanks for writing in. This in the 1st time would have been more helpful, was happy to see a moderator answer me. And yeah it was naive of me to ignore the SCNR remark :bang:

I do not say i know a lot, just learning, and thanks for the patience to point out the threads and articles. However, i think i know the meaning the liberty quite well. You dont need a dictionary to find a VLSI terminology. Anyways, please refer to this link : https://www.csee.umbc.edu/~cpatel2/links/641/slides/lect05_LIB.pdf

The liberty does have k-factors and it is used for scaling it or 'derating' the values of transition. setup etc..
This part is clear to me.

My question was entirely different and yeah i am not claiming i always ask the right thing but i would prefer a SCNR with a proper answer.
Thanks for your time and the links.

Given the fact that derating is mentioned whereever you find reference to k-factors it is quite obvious to get confused but there is quite a lot of difference between k-factors and derating factors used for AOCVM.

If you open your.lib (or LIBERTY) :grin: then you will find something called nominal operating conditions. These are the PVT conditions on which the lib has been characterized, say Pslow, V90, T100 but now suppose inside your dsign you specify an operating condition as Pslow,V95,T110 then this differes from the characterized values. So your values in the .lib have to be scaled to bring it upto the mark of the operating condition, say if some transition time was 20 then for more Voltage it should be more (ideally) so it is multiplied with a k-factor (say 0.25) then it becomes 20+20*0.25 = 25 (just for example) This is k-factor scaling but nowadays since we use MCMM we prefer to supply the tool link libraries for every operating condition. So, the k-factors nowadays is 0, however major tool makers have decided to include interpolation in MCMM too.
.
About derating in OCV -> Say your operating condition is Pslow,V90,T100 and you have the link library for this but to do OCV i.e add extra pessissism you will multiply a factor to increase the delay in launch path and reduce it in capture path (setup) . This is multiplied to al the delays.
Now understand tis that OCV will be added on top of k-factors if different set of operating conditions are used and if the library uses interpolation but because we do not use k-factors now-a-days so only OCV derating factor comes into picture.

Hope it helps.

Thanks,
ro9ty

- - - Updated - - -

Hi,

Thanks for writing in. This in the 1st time would have been more helpful, was happy to see a moderator answer me. And yeah it was naive of me to ignore the SCNR remark :bang:

I do not say i know a lot, just learning, and thanks for the patience to point out the threads and articles. However, i think i know the meaning the liberty quite well. You dont need a dictionary to find a VLSI terminology. Anyways, please refer to this link : https://www.csee.umbc.edu/~cpatel2/links/641/slides/lect05_LIB.pdf

The liberty does have k-factors and it is used for scaling it or 'derating' the values of transition. setup etc..
This part is clear to me.

My question was entirely different and yeah i am not claiming i always ask the right thing but i would prefer a SCNR with a proper answer.
Thanks for your time and the links.

Given the fact that derating is mentioned whereever you find reference to k-factors it is quite obvious to get confused but there is quite a lot of difference between k-factors and derating factors used for AOCVM.

If you open your.lib (or LIBERTY) :grin: then you will find something called nominal operating conditions. These are the PVT conditions on which the lib has been characterized, say Pslow, V90, T100 but now suppose inside your dsign you specify an operating condition as Pslow,V95,T110 then this differes from the characterized values. So your values in the .lib have to be scaled to bring it upto the mark of the operating condition, say if some transition time was 20 then for more Voltage it should be more (ideally) so it is multiplied with a k-factor (say 0.25) then it becomes 20+20*0.25 = 25 (just for example) This is k-factor scaling but nowadays since we use MCMM we prefer to supply the tool link libraries for every operating condition. So, the k-factors nowadays is 0, however major tool makers have decided to include interpolation in MCMM too.
.
About derating in OCV -> Say your operating condition is Pslow,V90,T100 and you have the link library for this but to do OCV i.e add extra pessissism you will multiply a factor to increase the delay in launch path and reduce it in capture path (setup) . This is multiplied to al the delays.
Now understand tis that OCV will be added on top of k-factors if different set of operating conditions are used and if the library uses interpolation but because we do not use k-factors now-a-days so only OCV derating factor comes into picture.

Hope it helps.

Thanks,
ro9ty