I would measure the counter both ways to characterize it, and you may learn a little more about how it reacts. Use its internal oscillator for the counter, apply the same 1PPS signal to both inputs. Use an external time base for the counter, apply the same 1 PPS to both inputs - make sure the timebase is not related to the 1 PPS. Then repeat, and use a common oscillator for all. You may also introduce delay into the stop channel and see what that causes and any differences, etc. Log your results in your time-nuts log book, you may have a reason to look at it again, down the road. Brian KD4FM

On Tue, 06 Nov 2012 02:14:04 +0100, Magnus Danielson <magnus@rubidium.dyndns.org> wrote: >Hi Bob, > >On 11/05/2012 01:30 PM, Bob Camp wrote: >> Hi >> >> As a practical example - a SR620 will look much better reading it's own reference than it will looking at almost anything else. That said, it's still a good idea to make sure the counter looks good reading it's own reference. If it doesn't look good, then you need to fix something. > >I agree that this is a good self-test strategy, it usually gives a good >clue if something is really bad or not, but as I pointed out, it doesn't >give a fair idea of the measurement noise floor. > >Cheers, >Magnus A lot of counters support a self check mode which counts the internal reference. I tracked down an interesting problem in a Tektronix DC505 by wiring the internal channel A input to measure the internal reference. Noise from the display multiplexing was getting into the level shifter located before the last 4 integrated counter stages and causing spurious counts. Based on the documentation and how it did not match the actual circuit, I suspect Tektronix fiddled with the values in production until it seemed to work but never figured out the real problem or back annotated the documentation.