This is exactly why I should not have started that thread. But since I started this, I have to finish it.
I confused you with specification requirement and acceptance requirement. Here is where this came from: In our business (custom military hardware), we receive a specification from the customer, we design the equipment, we then write a test procedure to verify compliance to the spec. The customer reviews and approve the test procedure. Up to that point, no problem. In an ideal world (and I am sure in many other areas of business, like the one where you operate), the acceptance criteria used in the test procedure would start from the spec requirement, deduct the measurement uncertainty, the accuracy of the test equipment and the calibration uncertainty, some people will even put everything in the RMS blender, then what's left is the allowable worst case measurement. For example, if you have to meet <0.20V of temperature drift on a power supply, and you can measure temperature drift with 0.05V total adjusted error, the test data should show less than 0.15V drift for the equipment to be certified. Sounds easy enough. Even I understand it. In my real world of defence department equipment contracting (mostly microwave, but not exclusively), there is no allowance for measurement or calibration uncertainty except in some very unusual circumstances. The acceptance criteria happens to be the same as the spec level 99% of the time. There is nothing fraudulent about it. The customer reviews and approves the test procedure and acceptance criteria, and either the customer themselves or the on-site government inspector witnesses 100% of the testing and certifies the test data. I know it is tempting to think that we do that because we are either stupid or crooks or both because then it makes everybody else look smarter, but unfortunately that is not the case. We do that because if we applied all the uncertainties, we would end up with a negative acceptable band, i.e. the accuracy of the test equipment is insufficient to guaranty that the spec is met if everything were at worst case. So, in my real world, we collectively (customer and vendor) turn a blind eye, telling ourselves that the probability that the equipment is worse than measured is about the same as the probability that the equipment is better, and everybody goes home happy, most of the time. Now please leave the accusations of fraud at the door. Thank you. I am done with this for now. Didier KO4BB > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of Mike S > Sent: Saturday, May 10, 2008 7:24 PM > To: Discussion of precise time and frequency measurement > Subject: Re: [time-nuts] HP 5370B > > At 08:02 PM 5/10/2008, [EMAIL PROTECTED] wrote... > > This has nothing to do with fraudulently claiming compliance, it > >just started as a well reasoned discussion of tolerances and the > >implications thereof. > > There can be no other conclusion from the statements made. > > "Most microwave network analyzers have amplitude resolution > of 0.01dB, while their accuracy is just around 1dB in most > cases...I have had to argue too many times that a piece of > equipment with a 2dB p-p requirement on flatness was just > fine when it measured 2.01dB on the HP network analyzer. I > would not have gotten in that argument if the data had been 1.99dB." > > "I have never had a piece of equipment rejected because a reading was > 1.99 for a spec of 2 max" > > The statements were made with regard to instrument > resolution/accuracy/precision. Clearly, measuring 2.01 (or > 1.99) on an instrument with an accuracy of 1 does not allow > compliance with a specification of 2 to be met. > > It was only later that the red herring of significant digits > was brought up. A specification of "2" is ambiguous in that > regard anyway. > No virus found in this outgoing message. Checked by AVG. Version: 7.5.519 / Virus Database: 269.23.11/1422 - Release Date: 5/8/2008 5:24 PM _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
