Kurt Patzik

For acceptance sampling plans you should start with defining the distribution. E.g. if we assume a Binomial distr. we need to define the alpha risk, the success probability for the consumer risk, the the beta risk, and the success prob. for the beta risk. Once you have these values it is strait forward to calc the sample size. If you instead use a hypergeometric distr. you should also define the batch size. Note that you will need many samples, if you like to verify that your failure rate is only a few ppm.

As the others already stated, we are unable to calc a sigma level. Nevertheless, you could use the Bernoulli distr. to quantify your quality. I suppose you have a well defined quality goal like "first pass yield (FPY) must be at least XYZ%". If this is true, you can setup a hypothesis test and check whether or not or FPY is at least XYZ% -- just assume that your FPY is exactly XYZ% and calculate the probability that you obtain no defect.