How can I automate noise margin calculation ?

Can you explain how you are computing the noise margin (maybe with pictures?) and also precisely what you mean by V_IL, V_IH etc?

BTW, Sorry to all folks for the spam in the forums today - we've been fighting a bit of a battle today against a major attack against some idiots who clearly find this amusing.

Regards,

Andrew.

Hi Andrew,

thanks for your reply and forget those stupid spammers :-)

I am talking about DC transfer characteristics of cmos circuits, maybe this Image here explains my intention http://vlsisystemdesign.com/images/images/nm2.jpg

V_IH = minimum HIGH input voltage
V_IL = maximum LOW input voltage
V_OH = minimum HIGH output voltage
V_OL = maximum LOW output voltage

Given these voltages HIGH and LOW noise margin can be calculated as follows:

NM_H = V_OH - V_IH, NM_L = V_IL - V_OL

1. Calculate derivative of transferfunction (output slope of the equivalent cmos inverter)
2. Look up the input voltage (V_IL, V_IH) for which the derivative is closely to -1
3. Look up the output voltage (V_OH, V_OL) for the in step 2 determined input voltages
4. Calculate NM_L, NM_H

Step 1.)  can be easily solved using the "deriv()" function of the calculator, but I am stuck with point 2. and 3..
Is the calculator best suited for this problem or are there better ways to do this?

Cheers,

Marten

Hi Martin,

I was faced with a similar calculation on a few occasions. What I resorted to was to an ocean script following a DC simulation. Specifically, within the script, I ran a DC transfer function of the circuit using relatively small input voltage increments (0.1 mV if I recall). Following its completion, I could determine the derivative of the transfer function and then the input voltage at the point of interest where the derivative matched -1.  I iterated on the output voltage to locate the value closest to the value corresponding to the input voltage of interest using a search. This provided the data for the noisemargin computations. The script ran over all corners to complete the picture. This may not be the most efficient method, but it worked for me.

I will let you be the judge!

Good luck!

Shawn

Hi Shawn,

nevermind, you are neither the first nor the last one mispelling my name ;-)

Your solution sounds good but maybe you can provide a small code snippet ?

I tried "x=getData("/out")" and x=v("/out") for accessing the values, but I guess those functions do not return arrays through which I can iterate.

Thanks in advance,

Marten

Now I did the following:

;Dertermine input voltage, for which the derivative of the transferfunction crosses -1
V_il=cross(deriv(getData("/out")), -1 1 'either )
V_ih=cross(deriv(getData("/out")), -1 2 'either )

;Get the output voltage with respect to the calculated V_il, V_ih
V_oh=value(getData("/out"), V_il)
V_ol=value(getData("/out"), V_ih)

;Calculate Noise Margin
NM_h=V_il-V_ol;
NM_l=V_oh-V_ih;

I am not sure about the first step, will there in any case only be two solutions?
Maybe someone with more experience about cross() function can comment on this solution.

Maybe this one is even more "secure":

V_il=cross(deriv(getData("/out")), -1 1 'falling )

V_ih=cross(deriv(getData("/out")), -1 1 'rising )

Since this is working fine within the SKILL IDE, I tried to add this script as an ocean measuremet into my ADE XL setup.

But when running Monte Carlo, nothing happens. No outfile gets written and also I am not able to check the "save"-checkbox for saving the calculated datasets. Plotting can be activated, but is useless since my script returns single integers.

Do I have to convert my script to OCEANXL ?

Cheers,

Marten