[time-nuts] Frequency counters

[Hal Murray]

> I want to find a frequency counter that will read to 1 hz precision up to a
> minimum of 30 Mhz that I can clock using the output from my Efratom 101 Rb
> frequency standard. I am quite happy to make one if there is a suitable
> design out there. Any suggestions?

Do you want to read the answer with your eyes, or get it into a computer?

Most of the older HP counters have HP-IB/GP-IB.


I think you could build everything you need in a small FPGA.  I'm thinking of 
something like this:
  Assuming a reference clock, generate a pulse every second.
  Count cycles on your unknown clock.
  On each second pulse, copy the cycle-counter to another register and reset 
the counter.
  That gives you the number of cycles in the previous second.
  If you count in decimal rather than binary, it's simple to send out the 
digits as ASCII.  With a bit more work, you could feed them to a small LCD 
display.

  One complication is avoiding metastability.  My straw man would be to run 
the internal logic at 100 MHz (derived from the 10 MHz reference) and just 
run the input signal through a classic pair of FFs.

  I pulled 100 MHz out of the air.  It's a nice round number.  It might be 
too fast for a simple design with 8 digits in the counter chain.  This whole 
problem is easily pipelined so you can run it at close to the max toggle rate 
in your chip, but that adds complexity.

  In the simplest version of the synchronizer, the receiving clock has to run 
at over 2X the sending clock.  Then the receiving logic just looks for rising 
edges in the input data stream.  You can gain a factor of 2 if you divide the 
sending clock by 2 and look for either rising or falling edges.  That would 
let you run the main clock at 40 MHz which might be slow enough to avoid 
having to think about pipelines.


You can probably to do the whole thing in software if you pick a chip with 
the right counter/timer blocks.  I'd have to check the fine print on the data 
sheet carefully to see how fast the counter/timers will go.  Some of them 
have prescalers so the counter can watch a fast signal, but if you divide the 
input signal too much you won't be able to get your 1 Hz resolution.  But 
maybe you can count for several seconds...




-- 
These are my opinions, not necessarily my employer's.  I hate spam.