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.

[SOLVED] Monte Carlo Simulation in HSPICE 2008.03

Status
Not open for further replies.

rmanalo

Advanced Member level 4
Joined
Feb 8, 2017
Messages
107
Helped
16
Reputation
32
Reaction score
16
Trophy points
18
Location
Philippines
Activity points
979
Hello again everyone,

I'm doing simulations now with my circuit biased in subthreshold and I have problems in MC Simulations. In general I want to do an MC Simulation on my temperature sweep on the circuit from 0degC-125degC. So I listed my questions in the following.

[1] Is MC Simulation the same with Process corner Analysis?
[1.1] A colleague said it's basically the same except Process corner Analysis simulates the worst case for MC Simulation. Is he correct?
[2] I read the HSPICE Manual "hspice_sa" on how to do it. I basically get the premise. But manual requires me to input the amount of variation on each parameter e.g.

Code:
.param Length=agauss(5u, 0.5u, 3)
M1 node1 node2 node3 node4 model W=10u L=Length
.temp 0 125 1 monte=30

the problem is, the "agauss" command is just me playing with the code. I don't know the right values for a realistic variation for each parameter. What should I do?
[3] Here is my main problem, since I'm biasing my circuit in subthreshold, the main contributor to variation is the threshold voltage. This parameter cannot be specified in the netlist, i.e. .param vth=agauss(nominal, deviation, sigma), (unlike length and width).

Thanks in Advance guys.
 

UPDATE:

[2] I read the HSPICE Manual "hspice_sa" on how to do it. I basically get the premise. But manual requires me to input the amount of variation on each parameter e.g.
So I scanned through "hspice_sa" again and realized that I need a variation block to get what I need. So I got this question answered.

[3] Here is my main problem, since I'm biasing my circuit in subthreshold, the main contributor to variation is the threshold voltage. This parameter cannot be specified in the netlist, i.e. .param vth=agauss(nominal, deviation, sigma), (unlike length and width).
part of this is answered through the variation block. But I still don't know how much should each parameter vary. I am using TSMC 0.18um Technology. Does anyone know the typical values for this technology or should I use relative values from books such as from Baker. (e.g. vth varies by 20%)

Also question [1] still stands.
 

But I still don't know how much should each parameter vary.
I am using TSMC 0.18um Technology.
TSMC provides model files with statistical data.
See model files.

[1] Is MC Simulation the same with Process corner Analysis?
No.

[1.1] A colleague said it's basically the same
except Process corner Analysis simulates the worst case for MC Simulation. Is he correct?
No.

What is the worst case for MC Simulation ?
Is is 3*sigma ?, 8*sigma ?

We can give boundary codition as MC option in current almost all comercial simulator.
It is close to corner analysis results.
 
Here's an (old) C@dence tutorial on MC simulation which might be helpful: View attachment Cadence_Monte_Carlo_simulation_tutorial.ppt

You'll see that the Gauss distribution of the parameters to be varied is defined differently: by their standard deviation. Don't expect that the given std values are realistic - they are only examples.

Corner simulations are one single simulation per corner: for a w.c. (worst case) simulation you define w.c. parameters, e.g. "slowest" process (slowest transistor reaction, stated by the fab), lowest operation voltage, and highest operation temperature (on chip!).

MC simulations use randomly selected values of the parameters to be varied within the stated sigma values of the distributions for the chosen number of runs. If the model file has the necessary variation parameters, you can run mismatch or process variation runs, or runs which use both variations at the same time.

You can even combine corner and MC simulations by additionally stating w.c. (or b.c.) environment parameters (op. voltage, temperature, any combinations) and run mismatch or process variation MC simulations for these (possibly extreme) environment conditions.

Sigma values are also needed for the evaluation of the histogram - to calculate the expected yield (s. p. 27).
 
Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top