Why don't these signals null?

James Price > wrote:
>
>The recordings weren't played by a human twice. A 10 sec. *recording* of a
>guitar DI was played into an amp/cab twice, consecutively, one after the
>other, the output of which was recorded and cut up into separate tracks.
>Both tracks were time-aligned and the phase inverted on one.

Ahh, I was thinking you were comparing the original DI'ed signal with the
reamped signal.

If you're comparing two reamped signals, the difference is likely the
difference between the two reampings, and the differences are likely
more than 30 dB down.

What is different between the two reampings? Well, the noise is different,
because noise always is (that's the definition of noise... it's not correlated
with signal). Likely the microphone position is a little different (maybe
less than a millimeter), the air currents in the room are a little different
(which is a big deal with a source that has an uneven radiation pattern like
a guitar amp), and if that's a tube amp it's likely warmed up a little more
or a little less and so slightly inconsistent.

You'll know for sure when you listen to the difference signal. Is it all
120 Hz hum? Is it all hiss? Is it correlated enough that you can hear a
guitar note somewhere in it? The best thing about the null test is that you
can use your ears (and an FFT) to figure out what the difference is, as well
as just measuring the absolute difference as a scalar number.

Remember that if you're looking at a difference that is 30 dB down from the
signal, you're seeing a difference of 1/1000th of the signal. It does not
take much.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

On Wednesday, October 23, 2019 at 8:01:59 AM UTC-5, wrote:
> the op is trying to null the 2 re-amped re- recordings to each other.
>
> OP, you got a -30 dB to -50 dB result which is very good.
>
> There are bound to be slight differences in room noise between the two takes. Or a window curtain moved slightly.

All recordings are close-miked, so room noise is minimal, and the recordings
are short(ie. approx. 10 secs). More importantly, if room noise were a
factor, the clean guitar recordings would not have nulled.

My best guess is that distortion introduces characteristics that are random
to a degree.

James Price > wrote:
>On Wednesday, October 23, 2019 at 8:01:59 AM UTC-5, wrote:
>> the op is trying to null the 2 re-amped re- recordings to each other.
>>
>> OP, you got a -30 dB to -50 dB result which is very good.
>>
>> There are bound to be slight differences in room noise between the two takes. Or a window curtain moved slightly.
>
>All recordings are close-miked, so room noise is minimal, and the recordings
>are short(ie. approx. 10 secs). More importantly, if room noise were a
>factor, the clean guitar recordings would not have nulled.

What does the residual sound like?

>My best guess is that distortion introduces characteristics that are random
>to a degree.

What do you mean by "distortion?" Because there are a number of different
things you could mean by that.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."