[time-nuts] First results with the ACAM GP22 chip
th.allgeier at gmail.com
Sun Nov 29 17:33:36 EST 2015
As threatened in earlier postings I can now show some results gathered with the ACAM GP22 TDC chip and evaluation board. Admittedly this is only from its "range 1" at the moment.
The setup was suitably simple to match my almost non-existent level of expertise and available kit: I have a very old analogue Thandar function generator. This I set to frequencies of 1, 10, 100, 1000 and 10000 Hz square-wave. Its adjustable 50 Ohm output was fed into a short bit of coax cable which at its other end was connected to the start pin of the GP22. The same output also fed into a 5 metre length of coax cable which was connected to the stop pin. Both cables were terminated with 75 Ohms (the cable was 75 Ohms as well), and 2 scope channels were hooked up to these pins so I could see what I was doing.
As the theory predicts I got a delay of around 24 ns and the GP22 measured this with a "double" resolution of around 50 ps. Recordings were made of 1000 samples each at these 5 frequencies and dumped into Excel spreadsheets. No averaging was selected on the GP22, so these are 1 shot samples.
Not to overload everybody's inbox I have uploaded the files to my homepage: www.stanton-instruments.co.uk/page26.html
(Those not interested in old balances ignore the rest of the website. Those not interested in the GP22 performance ignore the whole link.)
At 10 Hz I let the whole setup run for 30 minutes which gives around 64000 samples, the limit of the eval software supplied with the kit. Again a spreadsheet was compiled with those values.
It certainly looks pretty good to me - if the chip does the same in range 2 I'd be quite happy. Short term nearly all readings are within the 100 ps band. Over the half hour that band is wider but there appears to be no measurable drift. There was no temperature control, other than the stuff just sitting on my desk with the heating on in the house. That said I don't think there was more than one degree C change over the whole duration of the experiment.
There seems to be some frequency dependency as the absolute values of the measured delay are not identical. But then there is a good chance that the rise time of the function generator also varies a little with the selected frequency, and I suppose this would be enough to explain the small differences - surely my delay line will react to that.
Incidentally a very large reel of RG6 coax is supposed to arrive some time next week. The intention is to cut a length to give somewhat over 500 ns delay and repeat the entire palaver in range 2. If this proves successful I will try it with the full reel (250 m). I don't intend to push it any further, I would expect the performance of the chip to be similar whether the period to be measured in 1 µs or 70.
Thanks again for all the suggestions so far.
More information about the time-nuts