[time-nuts] comparing two clocks
saidjack at aol.com
Sat Feb 22 12:52:43 EST 2014
we just had the pleasure of doing exactly this aligned-1PPS measurement two days ago. I had to measure the difference (noise) of two units that were locked to the same source. To jump ahead, the difference was 0ns +/- about 500ps noise range.
We used an HP 5335, no problem, it jumped back and forth by +/-1ns steps. If I had done very long averages, it may be useful.
Next came an HP 5370A. A bit tedious to set up, but the noise floor of about 40ps was helpful, but the unit had about 200ps offset when in COM-A test mode so needs some adjustment.
Then moved to a DTS-2070, once we found space for it and the correct attenuators to not damage the inputs it was quit funny to see single femtosecond resolution on a ~500ps pulse to pulse noise.
Lastly we used the HP 53132A. This was the easiest to set up. It works fine as long as you stay within about -6ns, if you go earlier then the counter will measure an entire second, adding one second of error from its internal time base, and showing numbers like 0.999,999,997s. Since we were within a 1ns window, the numbers looked almost identical to the DTS-2070 so we know we have a good measurement.
I took the output of the 53132A and ran it through Excel and got a standard deviation of 220ps. Not bad considering some of that was probably counter noise and the counter has 'only' 150ps resolution if I remember correctly.
The 53132A it will be for future 1PPS to 1PPS measurements for me.
Sent From iPhone
On Feb 22, 2014, at 7:25, Bob Camp <lists at rtty.us> wrote:
> On some counters, if both inputs arrive at exactly the same time, they get very confused. The normal approach is to offset one by a few hundred ns or so. The exact offset is fairly non-critical. It’s real value depends entirely on the amount of drift you expect to see over the time period you are checking.
> If your oscillators are off by 1 ppm, they will slip by 1 us per second. If you want to check them for 12 days or more you will need an offset of more than one second. If they are off by 1 ppb, then your offset could be a bit over one millisecond to handle a 12 day run. (12 days is roughly 1 million seconds).
> On Feb 22, 2014, at 8:17 AM, Jimmy Burrell <jimmydburr at gmail.com> wrote:
>> I need some help with a 'noob' question regarding some practical examples in some of the NIST literature. When attempting to compare two clocks, I'm a bit confused on the subject of exactly how to use my counter to compare a delayed clock relative to another. Or perhaps I should just say 'comparing two clocks'. Let's take some concrete examples.
>> Let's say I want to characterize my Morion MV89 ocxo using my HP5335a. Obviously, I can tune the MV89's 10MHz by +/- 1Hz and feed it to the counter's input 'A'. Obviously, I can feed in a second, external reference clock at 10MHz into input 'B'. Suppose, however, I didn't have an external reference clock. Can I compare against the counter's internal time base by hooking a line from the rear jack time base output to channel 'B' input? Or am I making it too complicated? Do I simply plug into input 'A' and go?
>> In a somewhat related question, in this article (http://www.wriley.com/Examples%20of%201%20PPS%20Clock%20Measuring%20Systems.pdf) where two clocks, both divided to 1PPS, were compared, W.Riley makes the following statement, "The two 1 PPS outputs were connected to a Racal Dana 1992 time internal counter having 1 nanosecond resolution, and the start and stop signals were separated sufficiently in time for the counter to function properly". I wonder what exactly is meant by "separated sufficiently in time for the counter to function properly" and how one would go about doing this? For example, is inverting one of the signals sufficient separation? If not, how is this typically done? Delay line?
>> Thank you,
>> 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.
More information about the time-nuts