[time-nuts] Low cost 6+ GHz Prescaler

Rick Karlquist richard at karlquist.com
Tue Mar 30 18:30:34 UTC 2010


It has been interesting reading about the prescaler project.
I designed the prescaler for the HP 5334B, and am familiar with
the prescaler for the HP 5386.  You really want a static
flip flop like the 5386 used.  The commonly available prescalers
are "dynamic".  They are designed to be driven by a clean signal
from a VCO in a synthesizer.  If you try to measure a signal
with a wide band noise floor (such as an HP8660), most prescalers
will miscount in the lower order digits.  Only the 5386
will work correctly, in my experience.  You won't notice
this if you count some free running oscillator, because you
don't know its correct frequency.  Another issue is that some
prescalers lose sensitivity at low and high frequencies.
Trying to fix this by adding wideband amplifiers exacerbates
the wide band noise floor issue.  Signal conditioning with
ECL line receivers is another non solution.  You also need
a power detector to gate the prescaler off in case the power
is too high or too low.  Of course, you get to determine
where to set those limits, which tend to be frequency dependent.
You can't rely on the idea that you can tell if the prescaler
is miscounting by looking at the stability of the display.
A stable display can be wrong, and a variable display can
be correct and simply reflect the stability of the DUT.
There is still the problem of the noisy source.

If you are trying to design a prescaler board, it is important
to understand the intended application and properly test the
board for that application.  It is a lot more difficult than
most people think it is.

Rick Karlquist N6RK




More information about the time-nuts mailing list