Frequency counter

How accurate is the frequency counter in CW Lite?
The frequency counter and an Si5351 oscillator is in disagreement about the frequency output by the oscillator.
(I got a clone of from AliExpress.)


How much of a disagreement are you seeing?
The CW-lite counter works by counting how many times the clock toggles in 2^22 cycles of a 96 MHz clock generated by the SAM3U. In my experience it’s very accurate.


The disagreement between the two seems to be 0.0024 % and I can live with it.
I can assume the frequency counter is accurate and then adjust the offset in Si5351 in software.
Absolute correctness is not important. To have comparable results are.


Depending on the frequency this may be within the expected error, but probably the CW is still fairly close. As J-P mentions the freq counter counts 2^22 = 4,194,304 cycles @ 96 MHz = a 0.043690666… second “gating” period on the measurement. This ends up being up to about 23 Hz off in theory I think:

So in that period it counts N cycles and does (N / 0.04369066 = reported frequency). You can see the error as for example with a 100 MHz input:
100,000,000 * ‭0.043690666… = 4,369,066.666

The counter will of course only count a discrete integer = 4369066 cycles. Then the previous formula reports a 99,999,984.74 Hz frequency, about 16 Hz off in this example. If you end up in a worst-case error you’d be 23 Hz off. I think it can be a little more due to the fact the gating period isn’t synchronized to the input transitions (modern frequency counters use this).

The other caveat - we don’t have a precision/calibrated frequency source on the CW-Lite! So there could be some larger errors due to the XTAL itself drifting on the CW-Lite. Because it’s an XTAL source it should still be more accurate than non-XTAL sources.

Thanks! I really appreciate you taking the time to explain this to me.
No, I am more than 3KHz off on a 16MHz signal. The error is then around 200ppm. I guess the combined error of the two crystals involved gives that result. I assume the Si5351 has the cheapest crystal available. A good crystal can be around 10ppm.
The frequency counter is what it is. The output frequency from the Si5351 can be calibrated by initially setting it to 10MHz and then telling it how far off it is. I will then tell it what the CW-Lite frequency counter says. Then the two devices will be paired and have the same error - whatever that is. I do not need more than that.


Do you happen to know the spec or name of the resonator on the Lite board?

I replaced the crystal on the Si5351 board with one of these:
(It was the wrong physical size but fit the pads well enough)

The error was reduced to 1/3 of what it was. Just sayin’