Hi Again, I received quite a few responses to my query regarding EMI limits and margins. The results indicate that there are no actual written margins in current standards. There were some old standards, i.e. VDE, from long ago that specified either 2.0 or 2.2 dB margin. But these are no longer mandatory standards, being over-ridden by the EMC Directive. Most test houses still use that amount for GP's (general principals). Most manufacturers use at least 2 dB margin while some are going farther and using up to 6 dB. Some are even including the uncertainty in their requirements. There is an under-current of trying to get uncertainty included in the official requirements but that is meeting resistance and not happening very fast.
I have included below snippets from some of the comments I received. I found them interesting, enlightening and confirming. Thanks for your replies. Scott [email protected] the CB I've worked with insisted on 6 dB. There is no 2 dB figure in law that I'm aware of, you apply a level of margin baised on your measurement uncertainty. The limit is the limit. Period. However, our own internal policy says that in order to meet the requirement of 'due diligence,' we should apply a 6 dB safety margin for both immunity and emissions. There is, however, a written requirement on the TEST LAB in ISO Guide 25 (or equivalent QA spec.) that requires that any statements of compliance take into accout the measurement uncertainties, i.e. the labs cannot state that "...the product complied" if the margin is less than the measurement uncertainty. They can make a statement explaining that, while the measurements were below the limit, the margin was not sufficient to state with 95 % confidence that the product complied. Labs usually have their own way of expressing this. No - There is NO LEGAL requirement. The NATA in Australia has an unpopular approach, the measurement uncertainty of the lab must be applied as a margin to the limit. Since typical uncertainties can easily reach 4 dB in a good laboratory, this can create difficulties for the manufacturers. Unless all laboratories adopt this approach, it will never work, because manufacturers will seek the labs that have the least uncertainty or that ignore the requirement for margin. I believe this is what you can call a "prevailing practice". I dealt with at least one competent body that expected 2 dB margin. However, I did (in a previous employ) submit at least one set of test data that, while overall it was very low in emissions, one radiated emission had only 0.9 dB of margin. Tabular data, sorted by margin, was submitted in the report and no issue was raised. The same data was also submitted under the pre-DoC FCC certification scheme for Class B. It was also accepted without comment. Minimum passing margin of 2 dB was mentioned in old, now obsolete VDE EMC specifications. There is 2 dB rule in CISPR procedures to determine "representative configuration" in a case of multiple units, ports, cables, etc. Accredited test laboratory must have uncertainty calculations for each test method, but it should be ignored when they compare measured value with limit. The "new" thinking today is to take measurement uncertainty into consideration when determining a pass/fail criteria. For example, assume that a lab has 95% confidence that their measurement uncertainty is no more than 3 dB. Also assume that the measured margin or your device is 2.5 dB. The measured value indicates that you pass; however, based upon the measurement uncertainty, the lab cannot say that you really do pass. For them to be 95% certain that you do pass, your margin would have to be no less than their measurement uncertainty - 3 dB in this example. The closest to a margin requirement that I know of is the 80/80 stipulation of CISPR. That still doesn't allow you to exceed the limits for any single emission but by testing multiple units and applying the statistical process to it they have gained some assurance that while the five units might be very near the limits it is slightly more likely that your production units will also meet the limits in the long run. I've had labs issue a "passing" report with .1 dB margin, because that's wha the EUT did. It passed the limits. But I have seen a final report from a very reputable lab and written for a very reputable company that only had 0.7dB of margin under the legal limit. The emission was generated by the EUT. Furthermore, the report was a formal FCC Class B Grant Certification submittal and the FCC issued the Grant. I have worked at companies that imposed an internal requirement of as little as 3dB below the limit on a one off sample. The final decision on margin required is usually a blend of business and technical factors, each case being different. It is not written into EU law. The old VDE "f" mark required 2dB margin. My former employer, (name deleted by S_Douglas), _would_ issue a passing report on any result zero dB or greater margin. Of course there was also a prominent statement of the measurement uncertainty of plus or minus 5.4 dB. From a practical standpoint, I wouldn't expect my customer to have much faith in my product if it had a one dB margin in that uncertainty margin. I'll give you a passing report with zero margin. I'll also advise you to strive for that margin, but if you want to ship while you work on it, that's within you regulatory rights. Any lab which will not write you a passing report with passing data (even if it's at the limit at a billion points) is overstepping their bounds. Last year, CENELEC attempted to formalize a common practice for measurement uncertainty (prEN50222). It failed. In November 1997, during the CISPR/A meeting a decision was made to issue a new international Committee Draft (CD) on this issue. I have not seen an actual draft number, but from what I have read, the new CD will describe a "reasonable" measurement of uncertainty of +/- 3dB for conducted and +/- 5.5dB for radiated. --------- This message is coming from the emc-pstc discussion list. To cancel your subscription, send mail to [email protected] with the single line: "unsubscribe emc-pstc" (without the quotes). For help, send mail to [email protected], [email protected], [email protected], or [email protected] (the list administrators).
