In message <09c2d42ff0bfca4b829cdbe89b8f66ffe86...@g3w0637.americas.hpqcorp.net>, dated Tue, 26 Jun 2007, "Conway, Patrick R (Houston)" <[email protected]> writes:
>It seems from Brian's email, and the responses so far, that test margin >is a presumption. > Why? > Why is it presumed that we need any margin at all? Because ALL results of measurements of physical quantities on many nominally-identical samples show a statistical distribution of values. With only a limited number of samples to measure before bulk production starts, the 'shape' and 'width' of this distribution cannot be accurately determined. > If we can answer that question, then the "amount" of margin >needed should follow easily. No, for the reason given above. A small number of samples doesn't explore the whole distribution, especially since they are probably made form one or a few batches of each component, so even those don't show a real-world distribution of values. > > > >Here are some basic questions we compliance engineers should ask of our >companies. > > What problem do we solve by having margin during testing? > (If there is not a well defined problem, then we should >stop right here!) It's an allowance for not knowing the real distribution of values. > > > Does that problem get solved "better" by adding more margin? > (can you show that doubling your margin improves your >problem solution by 2x ? ) If the distribution is Gaussian, and the margin is expressed in linear units, not dB, then doubling the margin usually improves the situation by more than 2x, maybe much more. > > > Are the percentages, probabilities and uncertainties associated >with that margin truly defensible? > (defensible in a mathematical and manufacturing sense, >not a legal sense) Some of them (tolerances on electronic components, for example); maybe others can't reasonably be calculated (e.g. screening effectiveness of an enclosure with large removable parts). > > > If a problem occurs in the field, does the required margin shown >in the test report gain you any advantage in discussions with the >customer or auditing body? > (so, you did one test and passed. Then you manufacture >10 million a month. What does that margin buy you?) It shows that you understand the effects of statistical distribution and took them into account with due diligence. But if you are going to make a million a month, you don't do just one test; you apply the 80-80 rule and test however many the rule requires. At that rate of production, you CAN afford the USD100k it costs. > > -- OOO - Own Opinions Only. Try www.jmwa.demon.co.uk and www.isce.org.uk There are benefits from being irrational - just ask the square root of 2. John Woodgate, J M Woodgate and Associates, Rayleigh, Essex UK - This message is from the IEEE Product Safety Engineering Society emc-pstc discussion list. Website: http://www.ieee-pses.org/ To post a message to the list, send your e-mail to [email protected] Instructions: http://listserv.ieee.org/request/user-guide.html List rules: http://www.ieee-pses.org/listrules.html For help, send mail to the list administrators: Scott Douglas [email protected] Mike Cantwell [email protected] For policy questions, send mail to: Jim Bacher: [email protected] David Heald: [email protected] All emc-pstc postings are archived and searchable on the web at: http://www.ieeecommunities.org/emc-pstc ______________________________________________________________________ This email has been scanned by the MessageLabs Email Security System. For more information please visit http://www.messagelabs.com/email ______________________________________________________________________
