[time-nuts] And, for my next trick, 50Hz
bruce.griffiths at xtra.co.nz
Sat Oct 4 10:38:07 UTC 2008
Steve Rooke wrote:
> 2008/10/4 Bruce Griffiths <bruce.griffiths at xtra.co.nz>:
>> With interpolation you dont need to slow the signal transitions too much.
>> A transition suficient to allow 3 or 4 samples to be taken during the
>> transition is adequate.
>> If the signal slew rate is too slow sound card input noise and the
>> finite ADC resolution will increase the measuremnet noise.
> OK, I understand what you mean. Guess it will have to slewed somewhat
> to try and get the 3 to 4 samples on a rising/falling edge though. If
> we are looking at a 1KHz signal and sampling at 44KHz, that means we
> have to slew the transition to something like 100us to guarantee
> getting enough samples during the edge. So that is a 1/10 of the input
> signal or a roll-off of 10khz. That happens to fall nicely inside the
> bandwidth of the sound card.
> So how do we time-stamp these samples considering they will be
> buffered by the card and not read independently. Perhaps we don't have
> to as we can read each buffer when it it is full and time-stamp at
> that point. We know that each sample is taken at 44KHz and can easily
> count the sample number to calculate the time.
> I'll have a look at this and see what I get. I really need to check it
> against a known standard so will solve that problem first but this
> would be an interesting way of using a PC to check frequency.
Such interpolation is more effective when one channel has a reference
signal input (eg PPS or a known frequency) and the other channel has the
frequency to be measured.
More information about the time-nuts