[time-nuts] Prescalers ?

[Richard (Rick) Karlquist]

On 1/24/2019 5:17 AM, Bob kb8tq wrote:
> Hi
> 
> If you have a one time need to get to 1.3 GHz then some sort of cobbled together dead bug chip solution
> would be my pick of how to get it done. Cost wise it wold be hard to beat. There are also cooled up approaches

It may seem like prescalers are a simple fool proof technology,
but actually they are far from it.  I got educated 40 years ago
when I designed the HP 5334B counter C channel.  The 5334A used
an HP made divide by 10 prescaler that cost $100.  I replaced
it with a $2 COTS divide by 16 prescaler.  There was a production
test for the 34A that used an HP8660 synthesized sig gen.  I
wanted to leverage this test for the 34B.  The test IIRC involved
driving the C channel with 1.3 GHz at the minimum specified
sensitivity.  But it called out using a mini-circuits 1 GHz
high pass filter between the 8660 and the 5334A.  What?  Turns
out that the wideband noise floor of the 8660 corrupted
the measurement on the 34A, unless this filter was used.
Also turned out that the 5334B
with a completely different C channel had the same problem.
OTOH, the 5316 did NOT have this problem.  It used a different
HP made prescaler than the 34A.  What was different is that
the 5316 prescaler had STATIC flip flops.  Unfortunately, I
don't know of any COTS static flip flops that are available.
If the signal you are trying to measure is very clean, you
can get good results just about any prescaler.  You will need
to arrange for the drive level to be in the "sweet spot" for
that prescaler.  Otherwise, all bets are off.  Getting a prescaler
that works over a wide dynamic range is whole 'nother discussion.

Rick N6RK