[time-nuts] Phase noise confusion II

[Bob kb8tq]

Hi

Back quite a while ago, I went back and visited my buddies / ex-coworkers  
at Motorola Franklin Park ( = the Motorola crystal / oscillator group). Based on 
my past experience with the yields on a “quartz bar to oscillator” fab process,  
I asked:

	How do you build a “6 sigma” oscillator? …..

The answer:

You divide the process up into about 6,000 independent “points of failure”. You
then postulate that those failures can only be “detected” at the final test end of things.

I’m not suggesting that is a *wrong* answer, only that it was a somewhat surprising
approach to solving the “problem”. 

Bob

> On Apr 10, 2020, at 12:27 PM, Richard (Rick) Karlquist <richard at karlquist.com> wrote:
> 
> 
> 
> On 4/10/2020 5:47 AM, Bob kb8tq wrote:
>> Hi
>> It is not at all uncommon to …. errr …. make that decision, regardless
>> of what the customer might think about it :)
>> Even with that sort of decision, the whole process of measuring a one
>> sigma and multiplying by 6 depends very much on the underlying
>> processes (noise or maybe something else ….) being well behaved /
>> Gaussian sort of things. One can at least on paper construct situations
>> that do not meat that “well behaved” constraint …..
>> 
>> Bob
>> 
> 
> 
> A few decades ago, Motorola had a "six sigma" program
> which made everyone think they were a statistician just
> because they knew that buzzword.
> 
> Rick N6RK