When using a FFT should I convert the AtoD samples into a Voltage?

[btb4198]

when using the FFT for DSP should I convert the A to D samples back into a Voltages ?

so

A to D sample = sample voltage * ( bits / max Voltages)

so

sample Voltage = (A to D sample ) *(max Voltages/ bits)
that is right, right?

anyhow should I keep the samples as digital samples or convert them to analog samples
before I run them into the FFT?
why or why not?

I kind of think it should matter... but I do not know . I am new to this.

I do think that the A to D / mic on a computer might remove all the negative numbers, but i do not know is you would get negative numbers from a microphone...
do you ?

 

[UltrafastPED]

The units are only a scaling factor ... they don't change the relative values of the amplitudes.

They would be important if you were comparing one spectrum with another: you need to be consistent.

 

[meBigGuy]

Do all your computations in the digital domain, and afterwards scale your results to the analog domain if you need to. For example you may need to know the actual energy level of a modulated signal within a bandwidth.

I get in trouble with the 1/N normalization thing all the time. Should I use 1/N for forward, inverse, or 1/sqrt(N) on both.

 

[berkeman]

when using the FFT for DSP should I convert the A to D samples back into a Voltages ?

so

A to D sample = sample voltage * ( bits / max Voltages)

so

sample Voltage = (A to D sample ) *(max Voltages/ bits)
that is right, right?

anyhow should I keep the samples as digital samples or convert them to analog samples
before I run them into the FFT?
why or why not?

I kind of think it should matter... but I do not know . I am new to this.

I do think that the A to D / mic on a computer might remove all the negative numbers, but i do not know is you would get negative numbers from a microphone...
do you ?

A microphone puts out an AC signal. How are you handling that at your ADC? Are you offsetting the zero crossing voltage of the input signal to bias it up to half of the input voltage range of the ADC? If so, you will want to subtract that offset out of your digital data before doing the FFT, or else just subtract the DC component out of the final FFT data.

If your ADC circuit outputs 2's complement data to represent the AC waveform, you will probably need to do something before the FFT, I would think...

 

[meBigGuy]

If you get A/D numbers from a sound card or wave file they are 2's complement positive and negative numbers. Is that what you are doing? Just crunch on them to your hearts content.

 

[btb4198]

yes, I am getting them from a sound card on my computer ...