Parameter distributions of Monte Carlo analysis

Hi, I found a guide on monte carlo analysis on **broken link removed**. According to the guide, there are 3 kinds of different parameter distributions. My questions are:
1) Which distribution function do we usually use and the reason.
2) How many runs of monte carlo is sufficient?
3) What parameters do we usually vary in order to get an idea how good the circuit works? eg. an opamp

Thanks.

Re: Monte Carlo Analysis

1. Normal distribution is a good approximation for most random deviations. Note that the variation in a certain parameter (for example the resistance of a resistor) results from the variations in a significant number of physical quantities (in our example, thickness of the conducting polysilicon layer and the concentration of implanted ions, etc, etc. etc.).
Even if the variations on those physical quantities do not have a normal distribution, statistics tell us the normal distribution is a good approximation for resulting resistance variations (central limit theorem).

2. In practice 100-500 give acceptable results.

3. All electrical devices have random deviations: the resistance in the resistors, the capacitance in the capacitors, etc. In the particular case of MOS transistors usually the mismatches in the Vt and in the β (current gain factor). For more info search for "Pelgrom Law" on google.

Re: Monte Carlo Analysis

2) How many runs of monte carlo is sufficient?
This depends on how many parameters you like to vary statistically.

Re: Monte Carlo Analysis

As a rule of thumb, the number of MC runs is 10 times the number of circuit nodes/transistors in the circuit. I go for the number of transistors in my MC mismatch analysis.