[time-nuts] Once again about counter calibration
Tom Van Baak
tvb at leapsecond.com
Sun Apr 26 12:59:38 EDT 2015
The trick is to closely synchronize your 1PPS generator, whether you use the 1970's method of a string of seven '7490 decade divider chips (common reset) or a 1900's method of a sync'able MCU divider such as a picDIV (http://www.leapsecond.com/pic/picdiv.htm).
It's so simple it's irresistible. The PIC s/w is there for free. The PIC h/w is less than 1$.
> On Apr 26, 2015, at 8:17 AM, Dave Martindale <dave.martindale at gmail.com> wrote:
> Yeah, I considered saying that. But if you don't have a TI counter, you need some way of resetting the divide-by-1e7 chain so the two 1 Hz pulses are close enough in time that you can see them on the scope at some reasonably fast sweep rate. Yes, you can used delayed sweep, but how stable is the delay? If you do have a TI counter, then the accuracy of the counter's time base also factors into the reading (though you don't really care about absolute timebase frequency, just drift).
> A compromise method might be to divide the 10 MHz down to 10 kHz or 1 kHz. Then the nearest adjacent "wrong" integer multiple of 1 Hz where the drift would be zero is 1 part in 10,000 or 1 part in 1000 off the nominal frequency. Any decent crystal is unlikely to start out 50 PPM or more off frequency, and really unlikely to be 500 PPM off frequency, so this mostly eliminates the wrong ratio problem. Yet you get one cycle of the scope input signal every 0.1 or 1 ms, giving a reasonable chance for one of those edges to drift close enough to the 1 PPS reference to measure the drift at a fast sweep rate.
> - Dave
> On Sun, Apr 26, 2015 at 5:58 AM, Tom Van Baak <tvb at leapsecond.com> wrote:
>> > The problem with using a 1 Hz reference when looking at a nominal 10 MHz
>> > signal is that you will get a stable scope display with no drift when the
>> > input is *any* integer number of cycles/sec. So 10,000,000 Hz will give a
>> > stable display, but so will 9,999,999 Hz and 10,000,001 Hz. Unless you
>> > know that your 10 MHz signal is already within 0.5 Hz of the correct
>> > frequency, the drift method is likely to cause you to adjust to the nearest
>> > integer number of Hz, not exactly 10 MHz as you want.
>> One solution to this problem is to divide the 10 MHz to 1PPS and then compare the two 1PPS signals, using a 'scope or a TI counter.
>> The horizontal sweep of your 'scope and your patience will determine the resolution of the measurement. For example, at 1 ns/div you can easily resolve a 1e-11 frequency difference within a minute.
>> 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