Tom Bao wrote:
> 
> Can anyone tell me if the new CISPR 16-2 is published? I'm studying
> section 9 CISPR 16-1987. Not sure if I understand the difference
> between acceptance probability and confidence. Do you mind explaining
> it to me in a little detail? Or any other publications or papers discuss
> this on-going compliance?
> 
> Now the target, dB margin:
> 
> If you have been bored by the topic, just stop here.
> 
> If your products go to Europe, you have to comply with clause 35 in
> section 9, CISPR 16-1987. FCC does not define a statistic formula
> but they do expect to see production variations in EMI. Assume
> we all follow this 80/80 rule. This is the first assumption. Then
> 
> /X + k Sn <= L or L - /X >= k Sn
> 
> where L - /X is the mean value of margin, Sn is the standard deviation
> of emission level with sample size n, and k is the contant which makes
> the formula satisfies the 80/80 rule. There are assumptions here as
> well.
> 
> If we choose n = 5, then k is 1.51. This can be calculated. The formula
> can be rewritten as,
> 
> margin >= 1.51 x standard emission level deviation
> 
> There are several components in the standard emission level deviation
> 
> 1. EUT difference, components, shielding, etc.
> 2. Test setup difference, cable position, accessories, etc.
> 3. Site imperfection
> 4. Attenuator, temperature, time
> 5. Measurement errors
> 6. Atenna factor error
> 
> Give 1dB on each item, we have 6 dB. By careful design and the type of
> product, it's possible to reduce it to 2 to 4 dB. So we need a margin
> of 3 to 6 dB in order to comply with 80/80 requirement. There are many
> variables in the calculation. By setting different values, you may have
> many different results. It eventually becomes your business decision
> how much risk you want to take. There is one thing for sure. "Limit Is
> Limit" does not comply.
> 
> Why can "Limit Is Limit" exist? Because some authorities does not take
> limit as limit, and chances are that you won't be picked. Just like
> you won't be stopped when you drive at 60 mph in NJ (55 limit here).
> It would be nice if someone can analyze the risk a little bit further.
> By better understanding acceptance probability and confidence, we may
> be in a better position.
> 
> More than two years ago, I put a message in TREG saying it's not fair
> requesting 6 dB margin and agreed with "Limit is Limit".
> 
> Anyway, if any of the assumptions I made is not realistic, please do
> point out. Thanks.
> 
> Regards,
> Tom
> http://www.rcic.com

Recently, we had a discussion on this and I think the 
consensus was 'most' agree or realize that "the limit is 
the limit" is bad business especially when confronted 
with a possible hostile competitor overseas in a test 
of which you have very little control. 

I emailed a few statisticans on the 80/80 rule 
and got more than royally ripped apart by them.  
Trying to explain the 80/80 rule to a bunch of 
statisticians was similar to putting my head 
into a very large fan, if you get my drift.  

The 80/80 rule it turns out is an extremely loose "rule".  
Besides alot of technicalities which I could at best 
keep up with, the discussion started out with the observation 
that 20% of a SAMPLE (not necesarily near the population) 
is allowed to fail - very bad constraint. 

Next, hands down, there's no substitute for plain hard data 
from a variety of devices (a LARGE sample) and their results.  
You just can't sit back with a company producing 1000 products 
where 1 or 2 devices have been tested maybe 2 or 3 times 
in a lab and make any sort of reasonable prediction that 
a stats person would sign onto.  I'm sure you're aware of this. 

Confidence levels are based on large numbers of samples being used.
And at the bottom of this all is the standard deviation of your 
results for a large number of products.  It really starts getting 
complicated for any further disussion.  

Bottom line is - the standard is a worst case scenerio, 
or at least supposed to be.  NOT something we should 
shoot for as a target - something we should be well under. 

If you are unlucky enough to be pulled into a lab, 
how confident are you that the samples ARE NOT one 
of the 20% failure allowed?  It could ruin your whole day.

*******************************************************
-------------------------------------------------------
The comments and opinions stated herein are mine alone,
and do not reflect those of my employer.
-------------------------------------------------------
*******************************************************

Reply via email to