is mentor-carlo analysis neccessary?

[renwl]

mentor carlo

thanks.

 

[ravet]

mismatch analysis and corner analysis

you mean monto-carlo, in ordinary corners, you assumed that all nmos will be fast or slow at the same time, and this is approx. true if they are close enough to each other, but as they becomes far and far, you may find that one is fast and other is true , monto-carlo lets you define corelation coefiicient between them , when this coefficient 1 then , the 2 transistors follow each other exactly , when this coeff. zero , then one may fast other may slow , this analysis is extremely important in circuits that are very sensitive to mismatches as some ADCs comparators,

 

[rob1012]

monte carlo mentor 2008

If you mean monte carlo analysis... It is necessary to test a circuits performance in worst-case conditions. monte carlo analysis provides random variations to certain devices present in the circuit to verify its efficiency.

 

[nathan80]

corner analysis mentor

if you mean monte carlo analysis : yes it is necessary if you are concern about the cost of the IC.

monte carlos analysis is important to design an analog IC so no chips are lost because of parameter variations

 

[mickey_melomane]

Yes... It is the most convenient method to test the circuit or the system under analysis for worst-case scenarios...

 

[elbadry]

It will be necessary only if you are concerned about the effect of mismatches on your performance. Through monte-carlo, you can for example estimate the input offset voltage of your circuit; something that you can NEVR do without monte-carlo

 

[vcsel_driver]

i am curious. do you still need to use Monte Carlo analysis even when we are still in design stage.

i mean we just tape off the first version of our product for test and validation porposes. should we still use Monte Carlo

 

[renwl]

It's mission impossible to run too much simulation. I wonder to know how many simulation is enough. It will cost too much time and slow down the development. Is it the prerequisite analysis before production?

 

[wallycharlo]

monte carlo will give an accurate picture of how the random variations within your circuit (mismatch etc) will pan out in reality.

'slow' and 'fast' models are not very realistic at all;eg if have,say,10 critical parameters,then 'slow' would tend to push them all into one direction,typically the direction which would make a digital circuit slow down.But in reality these parameteres are nor correlaed in any way,and MC will reflect this by varying them in a random nature.

So if you want to know in advance what your yield migth look like then MC is always a good idea.

@elbadry

well,to be strictly true,you dont need to do a MC to calculate your input offset,if you use the formula

offset = n*Sigma/sqrt(WL) then it will give it pretty good

 

[srieda]

Monte Carlo analysis is very useful to check the working of the circuitry in all situations.
Most of the things have been said by ppl above me.. but they have missed on DEV and LOT variations which is an unique feature of Monte Carlo analysis. It is this feature which tests the working of the chip in possibly the worst conditions..
It is a safety check which can be run if time permits.. But if its a very critical part such as a VCO or something, its always advisable to run Monte Carlo simulations..

 

[suria3]

Monte Carlo analysis is a necessary run to test up your design to pass all the process distribution and the devices mismatches. On top of the 5 process FF,SF,FS,SS,Typ that we run normally, the monte carlo analysis is an additional run to confirm on the successful working of our design through the combination of process and device mismatches.

 

[strennor]

agreed

 

[raduga_in]

I still have a question .. If your fab ensures you
that the design you are goining to fabricate will
surely work if you can get satisfactory results
with corner simulations, Then why do you
need monte-carlo ?

Thanks in advance

Regards

Raduga

 

[suria3]

I still have a question .. If your fab ensures you
that the design you are goining to fabricate will
surely work if you can get satisfactory results
with corner simulations, Then why do you
need monte-carlo ?

Thanks in advance

Regards

Raduga

Corner analysis will ensure your design to work on all the process distribution on the silican fab but monte carlo analysis play more role in the mismatch analysis. In corner analysis, it is more on process paramteres variation and device mismatches between them is not taken into consideration. Whereby in monte carlo we have the chance to mix the analysis of device mismatches together the process to see the result in which really happen during the silicon fabrication.

 

[engrak]

monte carlo is important for analog dsign mainly where mismatch between devices is important , the corner analysis is not good for analog circuits.

 

[filipecbv]

Hi

I'm a begginer in MC anlysis, and I found 3 options to run it: vary process paramenters (lot), where all transistors vary togheter; vary matching parameters (dev), where the transistores doesnt vary togheter, but all they vary; or the both analysis togheter (dev+lot).
But It's very important to apply some layout techniques to minimize the mismatch in the circuit.
So, should I apply the MC analysis to perform the mismatch if I'm appling the layout techniques to minimize the mismatch?
I think that appling the layout techniques, the mismatch reduces satisfactorly. Or not?? So, the MC with mismatch will show the true mismatch??
In summary, what is the relation between MC analysis with mismatch and the layout techniques used to minimize the mismatch??
It's not clear to me.

Thanks
Filipe

 

[little-nemo]

Hi,

the MC parameters given with your models are normally measured with good layout techniques. (for example, to measure the deviation of a Vth, an array of transistor is done (all transistor are well matched in layout) and the Vth is measured for each transistor).

So if you simulate the offset of your opamp with a DEV variation, you will get a realistic offset for a good layout. Of course, if your layout is not well done, your offset will be higher than in simulation.

Hope this help.