DS90LV012A: Statistical Inferences for Timing Data?

Kurt Schmidt

Part Number: DS90LV012A
Other Parts Discussed in Thread: DS90LV011A,

We're doing a timing analysis on a signal chain that contains both the DS90LV012A and DS90LV011A differential parts. We are trying not to use the worst case min and max propagation delays but instead do a statistical analysis to give probability spread on a more typical system.

In the datasheet of both these parts, there is a ‘part-to-part skew’ specified –

Should we/can we infer any statistical information of the delays from the above highlighted parameters? Any additional statistical information you can provide on these parts would also be helpful.

over 7 years ago

0 Kurt Schmidt over 7 years ago

TI__Genius 11015 points

Data table failed to get included in original post. We are specifically interested in parameters tSKD3 and tSKD4 from the Switching Characteristics table at the top of page 5.

0 Lee Sledjeski over 7 years ago in reply to Kurt Schmidt

TI__Mastermind 43795 points

Hi Kurt,

These devices are mature - a nice way to say old. I will work to get as much data as I can on those parameters.

So far all I can find is the simulation reports which have slow/typ/fast information across supply voltage and temperature.

What type of statistics are you looking for?

Regards,

Lee

0 Kurt Schmidt over 7 years ago in reply to Lee Sledjeski

TI__Genius 11015 points

A probability distribution would be very good. Can we say the min-to-max specs are six-sigma specs? Or are they even better than that? For example many of TI's OpAmp min/max specs are hard tested limits such that there can never be a part outside of those limits. This makes them even higher than six-sigma.

Basically anything that will give us guidance in estimating the probability distribution of end-to-end performance when all parts in the signal chain are put together.

0 Lee Sledjeski over 7 years ago in reply to Kurt Schmidt

TI__Mastermind 43795 points

Hi Kurt

The datasheet maximums for device to device skew seem to be aligned with the slow-typ-fast model data.

I would say to develop a statistical model which fits these limits at +/- 6 sigma.

Regards,

Lee

0 Kurt Schmidt over 7 years ago in reply to Lee Sledjeski

TI__Genius 11015 points

Thank you.

While I do appreciate that we are somewhat limited by the very mature nature of these devices, just assigning a six sigma distribution to the min/typ/max numbers would imply that TI can ship parts outside of the spec? According to the data sheet this should not be possible over time/temperature/lot.

We'll run with what we have and truncate any outliers in the simulation but I am not sure this is necessarily the correct answer.

Interface

Interface forum

DS90LV012A: Statistical Inferences for Timing Data?