Hans T. Mellberg wrote: > Not one of you respondents have mentioned the fact that if you perform the > 80/80 > rule on more than 2 units, you will discover that in order to pass the 80/80 > test you must have a certain margin below 0 dB! (assuming that the readings > from > both units are not identical, an unlikely event. That margin delta erodes > when > the sample size increases. That is because the arithmetic mean is added to a > confidence factor which includes a standard deviation and a sample size > factor. > When added, the sum of those must be below or equal to the limit. So, for > production sampling, 0 dB margin will work out to be a failure using the 80/80 > rule! Your best hope then is that when using a small statistical lot size, the > arithmetic mean is small so that the added margin is small. > Hans T. Mellberg > EMC Consultant.
I don't doubt what you say, I have to ask for more help with an explaination, Hans. Say I have a total of 100 units that I require 80 to 'pass' or 'fail' some test. Also, I take a sample of 80 units to test and I require 64 to pass testing (80/80 rule). The probability that I will get 64 to pass from the sample of 80 is 24.37%. (hypergeometric distribution here) If I have a total of 10 units that I require 8 to pass. I sample 6 units for 4 to pass. The probabilty that 4 samples will pass = 33.33% Slim pickings if you ask me. The 80/80 rule is extremely libral. The issue at hand that Cynthia is going to run into very quickly is this: Cynthia manages to get her hardware manager to change policy (magically) with the new 6dB margin. But, later, during routine sampling, more samples begin to fail at the -6dB margin. Again, the margin doesn't matter now with the math if you impose a pass/fail with the 80/80 rule. Now, she will be faced with correcting this deviation with an ECR that costs the company money. She will be hard pressed to get that ECR through even though company policy deems it. The ECR may be torpedoed by declaring it a insignificant change (not madatory) and there it will sit, or worse, disappear. I've been there and the frustration doesn't stop. Trying to push this through was difficult and the engineering staff dug their heels in and called it a "non-problem". Maybe difficult to understand but true. My support goes out to Cynthia. ******************************************************* Doug McKean [email protected] ------------------------------------------------------- The comments and opinions stated herein are mine alone, and do not reflect those of my employer. ------------------------------------------------------- *******************************************************
