Designer’s Guide Community

M.K.

New Member

Offline

Posts: 3

Mathematical calculations in the offset simulation of the comparator
May 11^th, 2010, 10:45pm

Hi,

I have a question regarding offset simulation of comparators method reported by Achim Graupner.
I am wondering if it is possible to simply measure the difference between the input ramp (at output transition point) and reference (threshold) voltage and consider this variable as offset voltage. Then by using Monte Carlo results for this variable, mean and variance of the offset can be obtained.
If so why we need those mathematical post processing?

Thanks in advance.

Back to top

IP Logged

AnalogDE Senior Member Offline Posts: 137	Re: Mathematical calculations in the offset simulation of the comparator Reply #1 - May 17^th, 2010, 4:31pm What you describe sounds more like a DC offset measurement -- that's fine for DC. but if you want to include transient effects (hysteresis, for example) then the method you describe will be optimistic.
Back to top	IP Logged

M.K.

New Member

Offline

Posts: 3

Re: Mathematical calculations in the offset simulation of the comparator
Reply #2 - May 17^th, 2010, 7:37pm

Hi,

Thanks for your reply.
I agree that DC simulation could not be a correct way and a better way is doing transient analysis. What I described is exatly the same method as in paper (Doing transient analysis by applying an increasing step and running Monte Carlo on this transient analysis). My question is on the last step where mean and variance is calculated using mathematical equations.
Why these parameters cannot directly obtained from the results of Monte Carlo? The only reason I can think about is cases where we have more than one transition in either rising or falling directions. In that case the threshold voltage is undefined.

Back to top

IP Logged

AnalogDE

Senior Member

Offline

Posts: 137

Re: Mathematical calculations in the offset simulation of the comparator
Reply #3 - May 18^th, 2010, 9:29am

There is post-processing because you need to run a monte carlo for each input.

For example if you have your threshold at 0mV differential you would run a bunch of MC simulations around 0mV Vin (-2mV, -1mV, 0mV, +1mV, +2mV), etc. Each MC simulation gives you the number of times your comparator outputs a 1 (or 0). From that, you can fit a line to extract the stddev of the offset.

Unless there's a way you can loop the MC run in your simulator you will have to do some post-processing. (If you know how to do this in eldo, I'd be interested. I had to hack together my own perl script to do it.)

Back to top

IP Logged

M.K.

New Member

Offline

Posts: 3

Re: Mathematical calculations in the offset simulation of the comparator
Reply #4 - May 20^th, 2010, 11:52pm

Hi,
Thanks for reply.
Actually I am doing it in another way. In each MC run, during transient analysis, I am reseting my comparator and applying first voltage (e.g. 0mv) and reading the output (e.g. 1) and then reseting again and applying the next voltage (e.g. 1mV) and waiting for output (e.g. 1), .... during this transient analysis I can find the switching threshold of the comparator for this MC run. After completing the whole MC runs, I have a distribution of the switching threshold of the comparator.
(Note: My comparator should be reset between two consecutive inputs)

Back to top

IP Logged