Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

corner simu vs Monte carlo

Status
Not open for further replies.

cretu

Full Member level 2
Joined
Nov 12, 2003
Messages
137
Helped
5
Reputation
10
Reaction score
2
Trophy points
1,298
Activity points
1,224
corner analysis vsmonte carlo

I need to know if anybody can provide some links towards books or artciles which can explain ina formal way the difference between corner and MonteCarlo simulations. As far as i know corner simualtions are 'the real deal" due to the fact they are showing the technology spread. However, monte carlo can show the offset, spread of a certain parameter but is usually included int he corners of the technology.

Can anybody help?
 

corners vs monte carlo

When you fabricate an integrated circuit, depending on the fabrication process, all the devices will be in a different corner. They can be slow, fast, high resistive, low resistive, etc. When you design a circuit, you want it working in any possible corner of the fabrication, that's why you simulate it in all corners.

Now, in whatever corner you are, there is mismatch between devices and that causes errors. You need to simulate your circuit including these mismatch errors.

Both type of simulations are equally important
 

corner vs monte carlo

corner simulations: usually 5 process corners are deifned (nominal-nominal, slow nmos-slow pmos, slow nmos-fast pmos, fast nmos -slow pmos, fast nmos-fast pmos) in the transistor models, but you should consider also the temperature corners (low, nominal, high) and the supply voltage corners.
It works fine for digital circuits, but usually they are not the real deal for analogue circuits.

For Monte Carlo simulations you need statistical values in your process technology package, and then due to a probability distribution the values will vary from run to run.

That is why, Monte Carlo is better to show matching related problems.

If you need books, articles: every classic analogue design book covers that topic (also the free Camenzind: Designing Analog Chips), also in the simulator manuals you can find informations.

To check in simulator I would suggest two very simple circuits: CMOS current mirror and CMOS inverter, just run corner case simulations and Monte Carlos for both and you will see the difference ...
 

In short, the corner sims will let you see the effect of global process variations(a slow nmos is slow nmos throughout the chip) while Monte-carlo can let you see the effect of both global and local mismatches
 

saro_k_82 said:
In short, the corner sims will let you see the effect of global process variations(a slow nmos is slow nmos throughout the chip) while Monte-carlo can let you see the effect of both global and local mismatches

Thanks to all of you for your inputs. I do know all this differences, however I was not able to get something on paper which shows the comparison between the 2 ways of doing things. In my past experience, whenever you managed to do a design which even in the worst corners fitted the specification, then MC was for sure included within. With MC you could check the variations, spreads, but not the entiore technology spread. So, Is anaybody out tehre having any clue where can I get some papers putting this to together?
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top