Forwarded thoughts on the dB margin . . . 

<---- Begin Forwarded Message ---->
Return-Path: [email protected]
List-Post: [email protected]
Date: Fri, 11 Oct 1996 10:12:05 -0700
From: [email protected]
Subject: dBmargin
To: [email protected]

The issue of margins are always difficult.  The limits are just that -- limits 
and they must not be exceeded.  Because one is only sampling one product and 
basing all future results on that one product some margin is strongly 
recommended.  The CISPR recommendation is based on the statistical 80/80 rule 
namely 80% of all products tested must pass 80% of the time.  To meet this 
rule TUV and VDE insist for their private label mark that all readings must be 
a least 2 dB below the limit.  AUSTEL insist that the test report state the 
amount of uncertainty in the measurement readings.  This implies that the 
measurements taken should be below the "measurement uncertainly" to insure a 
true passing  condition.  The "radiated emission measurement uncertainty they 
accept without detail measurements is 6 dB.  CKC believes that each company 
product manager must try to determine the typical variance normally expected 
from their particular product.  We suggest for small products that up to six 
identical units be measured and the variance determined on a product by 
product basis and use the CISPR 11 statistical test to derive the "margin" 
needed to insure the 80/80 rule is complied with.  Most experts feel that 
without actual product data, a margin of 3-6 dB should be set.  However, we 
have customers accept 1 dB margin even without margin data but this the 
exception not the rule.  Cost factors always cloud the margin issues.    

Chris Kendall
CKC Laboratories, Inc.



<----  End Forwarded Message  ---->


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~                                         
^       Todd Robinson           ^
^       Marketing Manager       ^
^       CKC Laboratories, Inc.  ^
^       WorldWide EMC           ^
^       [email protected]            ^
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~                                               
                          

Reply via email to