Counting (interneal timebase)events with only one counter

Question

Hello,
i am using a PXI-6115 card. This card is external clocked with 8 MHz. But I want to measure only about every second. So I divided the external clock with the internal counter 0. This works fine.
Now my problem. Now I want to know the external clock rate. Is it possible to build a pulse period measurement with only one counter??
I thinking of counting internal timebase- pulses with counter 1 and to start and stop the counter 1 by the output of counter 0.
Is this possible and how??
Thanks for help, Ruediger

Answer 1

Yes, there's a way to measure pulse periods with a single counter. The trick is to perform a buffered period measurement and sum the periods. The size of the buffer will determine the quantization/roundoff error.
The source of this error is that you have no control over the initial phase relationship of your external clock and your internal timebase. There's a nominal ratio between the two frequencies, and you'd start by expecting ratio * (buffer size) total cycles counted.
However, your actual count must be an integer, so you'll get either the next-highest or next-lowest integer, but you can't predict or control which one.
Example: If you collect 1 period at 8 MHz using the internal 20 MHz timebase, you'll capture either 2 or 3 e
dges from your timebase, implying a measured frequency of either 10 MHz or 6.6667 MHz.
If you collect 1001 periods, you'll capture a total of 2502 or 2503 timebase edges, implying a measured frequency of either 8.0016 MHz or 7.9984 MHz.
Notes:
1. In many cases, you can set up for buffered event counting, using the internal 20 MHz clock as a source and your external 8 MHz as a gate.
The advantage is that the buffered values already represent cumulative time so you don't need to sum them.
A disadvantage would be that the internal 24-bit counter value will roll-over in less than 1 second at 20 MHz. I typically capture periods and then sum in software because I can convert to floating point between capture and sum if necessary to produce both high-resolution and long duration measurements.
2. You should typically ignore the very first value in the buffer and work only with the others. If you want to capture 1001 legitimate periods, size the buffer for at least 1002
-Kevin P.