The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
General >> Tech Talk >> 125C digital standard cell models non-optimum?
https://designers-guide.org/forum/YaBB.pl?num=1190989814

Message started by jorobins on Sep 28^th, 2007, 7:30am

Title: 125C digital standard cell models non-optimum?
Post by jorobins on Sep 28^th, 2007, 7:30am

I'm an analog RFIC design engineer who is used to the freedom of having device models across a whole slew of temperatures so I know very well how my circuits behave at pretty any temperature I want. But apparently, in the digital world it is not the same. Most fabs seem to offer only models for standard cells and such characterized @ 3 (junction) temperatures -40C, 25C and 125C. Here are some of my opinions about this. Just wanted to know if others have any comments or if I'm missing anything obvious.

(1) 125C seems too high a (junction) temperature under most conditions (exceptions MAY be military & automotive). Assuming a worst-case ambient of 85C (commercial IC's typical maximum rating), based on normal packaging thermal resistances, I would bet a worst case pessimistic junction temperature of about 110C. So models @ 110C would be optimum choice for the max. temperature. But then you say "that's just a 15C difference between 110C & 125C. That should not be significant ". But I think once you get around to those sort of temperatures, things start behaving exponentially (metal current carrying capability, leakage currents etc.) in addition to much worse delays. So the digital designer is basically overdesigning his block for a condition that may not exist thereby taking a penalty in area and power consumption. I'm going to take a SWAG (Scientific Wild Ass Guess) and say that maybe you are taking a 10% penalty in area and power. Any of guys have any opinion on this?

(2) 25C is also not the typical junction temperature that the devices are going to see. Assuming a room/ambient temperature of 25C, you are looking @ 45 to 50C junction temperature. Typically, most of us in the analog design group do our nominal sims around 50C to get our typical performance numbers. So even the nominal temp. models for the digital models do not really reflect reality.

My sense is that these temperature values that these models are given are due to a "historical anomaly" as well as "fab modeling convenience" and no one has bothered to rethink about the relevance of these today. It is of course easier for the fabs to just give a 125C model to cover industrial and military applications as well, but because of that the 99% of ICs that are not going to be used at that temp, are probably going to be overdesigned. 25C was always considered "room temp", but no one really typically bothers to differentiate between room temp and junction temp.

If you think that I'm just spewing out something and missing something critical, I'm all ears.

Jose