[MarkK]

I presume you mean high pass, and yes you are dead right. It sounds
like his high pass isn't entirely so. I have seen this in the past
where the number of bins treated in the FFT is wrong. It tends to
leave a small residue of DC rather than all of it though.

d

lets say the system full scale is + 255 and - 255 for sake of discussion...

if you have a waveform like applause that is highly asymmetric,....
lets say it is +255 for 4 samples and 0 for 1000 samples, that waveform,
even though it is 0 for many samples DOES have a DC component.

If you remove the DC component then the result will be +253 for 4 samples
and -3 for 1000 samples. (those may not be the exact numbers, I'm just
trying to demonstrate a point)

Now if you looked at that waveform and saw the long string of -3's you
might incorrectly conclude that there was a DC offset.

The point I'm trying to get across is that the measure of the DC value
depends strongly on the time duration over which you make the measurement.

For example if you put the waveform through a true high pass filter with a
1 Hz cutoff, there may be an apparent short term DC value over the duration
of say 0.1 second but not over the duration of several seconds.

Mark