From: "geoffrey (g.g.) skanes" <[email protected]> Sender: "geoffrey (g.g.) skanes" <[email protected]> Message-Id: <"10075 Thu Oct 10 10:48:12 1996"@bnr.ca> To: [email protected] Subject: re:dBmargin
Cynthia: With the absence of a stated tolerance on your test results, you owe it to yourself and your customer to include a margin on top of the limit. At the risk of overdesigning a product, anything in excess of 10 dB margin is too much--I've found 6 dB is more manageable for radiated EMI and 3 dB for conducted. The reasons for margins: - there's as much as an 8 dB variation from site to site on RE - the FCC accepts this as part of the NSA filing you do on your site without having to correct your measurement data - unless you have the luxury of sample testing a large volume / low cost item (unlike myself), you don't have an opportunity to determine the statistical variation characteristics from one sample to the next. (The 80/80 rule in EN 55022 is somewhat impractical for large, distributed telecom systems.) - Measurement uncertainty - this is receiving alot of debate within many standards fora and committees - an entire workshop was dedicated to this topic in the last EMC symposium and for a very good reason: as a customer or a test house, you need to know what level of confidence exists in the measurement results, especially in view of the former discussion. The former notwithstanding, you have no legal (IMHO, I'm not a lawyer) obligation to meet the Part 15 limits with margin. So, how do you insist your customers include a margin? Start with the arguments above--there are so many reasons why a 0 dB pass in one test could be a 3 dB pass or a 3 dB fail somehwere else. Is the customer willing to accept the risk of failure on a subsequent test (stop shipment or even worse, recall/retrofit in the field)? To ensure there's no bickering over 0.2 dB, establish the test criteria prior to testing in a written contract/agreement. It is your responsibility to convince the customer why you think the margin is necessary before you switch on the spectrum analyser. If they still don't buy the arguments you have two options: refuse to test, or, test and make very specific recommendations for improvements to attain the desired margin with a clear risk statement if the improvements are not incorporated. Clear as mud! Regards, Geoff Skanes Nortel Technology
