# [time-nuts] Homebrew frequency counter, need help

Bob Camp kb8tq at n1k.org
Fri Nov 28 09:36:43 EST 2014

```Hi

> On Nov 28, 2014, at 6:26 AM, Li Ang <lllaaa at gmail.com> wrote:
>
> I did a little calculation and it's a 10 digits counter.
> log(10,000,000,005) = 10.
> There is still a big gap between this one and 53132A :(

Not if you can hit the advertised time resolution of your chip. Keep on working on this.

>
> For 53132A, the time resolution is 150ps, which I think is 10digits/s with
> interpolator. According to the schematics the only difference between
> 53132A & 53131A is the ADC of interpolator. It is the reason why 53131A
> only has time resolution of 500ps (also 10digits/s). However, 53132A is a
> 12 digits/s counter. I guess the 2 more digits come from software. Linear
> regression maybe ?

More like confusing spec’s on the two counters. The timing resolution is what you want to look at in this case. They do some math on multiple internal readings to (hopefully) improve the frequency resolution. Sometimes that magic works, often on real signals it does not make much difference.

Bob

>
> 2014-11-28 11:15 GMT+08:00 Bob Camp <kb8tq at n1k.org>:
>
>> Hi
>>
>> One way of looking at resolution is at the one standard deviation point.
>> Another way of looking at it is as a +/- 1 digit accuracy point. Each
>> approach has it’s advantages. It’s more common to see single shot timing
>> specified as one sigma and frequency specified as +/- 1 count. Often you
>> need to read the fine print to see just what is being spec’d.
>>
>>
>>> On Nov 27, 2014, at 5:30 PM, LiAng <lllaaa at gmail.com> wrote:
>>>
>>> Thanks for the info.
>>>
>>> What's the standard to be 11 digits/s? For real 11 digits/s, the ADEV
>>> needs to reach the 1e-11 level? I'm not sure if my GPSDO & Rb is stable
>>> enough. Maybe 2 MV89A as the refclk and signal?
>>
>> For the “easy” approach, first feed the counter’s reference back into the
>> input. That will usually give you a “best result no matter what” sort of
>> reading. It also will suppress a variety of problems coming from the
>> reference signal.
>>
>> A source with a <1x10^-12 ADEV at 1 second should be good enough for
>> testing a 10 to 11 digit counter. It’s not going to do the trick for a 12
>> digit device. In the case of a 12 digit device, use a second copy of what
>> ever you are using for the reference for the counter ….
>>
>> ——
>>
>> Another approach, don’t measure frequency, measure period / time / phase.
>> Generating a pulse that is 100 ns wide is fairly easy. Doing so with < 1 ps
>> jitter is not impossible. If your signal source is good to a few ppm, your
>> pulse generation accuracy will be “plenty good enough”. Things like rise
>> and fall times through buffers will be a much bigger deal in the delivered
>> result than the absolute accuracy of the clock feeding the circuit. If the
>> counter measures the resulting pulse with a 10 ps one sigma error, you have
>> a 10 ps counter. If it says that 100 ns is 102 ns, that’s to be expected
>> with a simple pulse generation technique. Yes, you eventually do need to
>> verify that 100 ns is 100 ns, but that can be done a different way.
>>
>> Bob
>>
>>>
>>>
>>> TDC-GP22 has it problem, I will post some data/schematic/source code
>>>
>>> Thanks
>>>
>>> _______________________________________________
>>> time-nuts mailing list -- time-nuts at febo.com
>>> To unsubscribe, go to
>> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>>> and follow the instructions there.
>>
>> _______________________________________________
>> time-nuts mailing list -- time-nuts at febo.com
>> To unsubscribe, go to
>> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>> and follow the instructions there.
>>
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.

```