[time-nuts] our favorite topics

[David]

That is always the danger when using parts for characteristics not
guaranteed in the specifications.

Sometimes a process just becomes obsolete necessitation new parts to
be fabricated on a new process.  Or a process may have enough
variation that some lots or parts meet unguaranteed specifications and
others do not.  Occasionally a minor update is made to correct a
problem or improve yield that significantly changes unguaranteed
specifications.

And of course the company could be bought resulting in the process or
parts you are relying on being discontinued.  I am currently worrying
about this with Linear Technology being bought by Analog Devices and
NXP being bought by Qualcomm.  In the case of Qualcomm, I cannot see
them being in the discrete parts business.

As far as testing, nobody likes to test for noise or low leakage for
that matter.  Test time costs money and low frequency noise testing
especially takes a lot of time.  The example I like to use for this is
the LMC6081 ($0.83) and LMC6001 ($5.76) operational amplifiers; the
later is identical to the former except it spends a lot more time on
the tester to guarantee its lower input bias current.  Common small
signal transistors are usually specified with 50 or 100 nanoamps of
leakage even though it is often 1000s of times lower because that is
as good as the automatic testers can do quickly.

On Sun, 30 Oct 2016 17:06:52 -0400, you wrote:

>It has to do with the manufacturing process and a reduction in cost. I can  
>not speak for other companies, Infinion "killed" the good phase noise   
>performance but the large signal noise is not specified in the data sheet so  
>they are legally  "clean"
> 
>>In a message dated 10/30/2016 4:56:38 P.M. Eastern Daylight Time,  
>>davidwhess at gmail.com writes:
>>
>>You  mentioned suitable transistor availablity being an increasing
>>problem and I  have run across that myself.  Do you expect Qualcomm's
>>aquisition of  NXP to have an impact?
>>
>>NXP is currently the best source I have for fast  complementary pairs
>>or even just fast PNPs.