"danielsan" <
[email protected]> wrote in message
news:eo2dnaxwmLaV_2_eRVn-oA
[email protected]..
> I need some help. I am sampling a signal proporational to our sample in
I'm having trouble understanding your question.
> frequency. The output range is 5-7.5KHZ. I can sample up to 384 ksps with
> 10 bit accuracy. Do I need 10 bits or is 8 enough? I can sample as low as
> 15 khz, is that enough for accuracy? I would like accuracy to 4 hz. I
You are mixing up several terms here and that might be why I have trouble
interpreting this.
You have to be clear when you talk about 'accuracy'. State whether you mean
frequency accuracy, amplitude accuracy, etc. It seems like the 4 Hz number
is your frequency accuracy requirement. The bits in your ADC have no
relevance to this. To keep things simple, I'd say you should sample your
signal at a rate that is approx 4 times your highest input frequency in
order to not lose any information (you can go a little lower if you are
careful and I'm avoiding the topic of undersampling to keep it simple).
Your frequency accuracy requirement will determine the length of the FFT
you'll need.
Since you need 4 Hz accuracy, you'll need to collect data for 250ms. If you
are sampling at 30ksps, this will result in a 7500 point FFT (or 8192 points
to be power of 2 friendly).
> don't care about computational efficiency, Ijust want to use a fourier
> transform to extract peak frequency. what min sampling freq, min accuracy
> do I need? All I am looking for is a simple to for loop. Thanks
What is a 'to for' loop?
Your 10 bit vs 8 bit decision depends on what kind of dynamic range you need
on your measurements. You did not indicate if you are measuring
sinusoids/tones, modulated signals...are you perhaps looking for spurs or
harmonics in the presence of other signals. Say you have a full scale signal
and you want to look at another signal that is 90dB below this full scale
signal, your 10 or 8 bit A/D isn't going to be enough. A good rule of thumb
is about 6dB per bit of SNR (and any extra you get with processing gain).
Cheers
Bhaskar