In my test I used sox to do ALL the generation and mixing.=A0 And audacity =
to do comparative analysis.
When mixing(RIGHT.wav) the common tone (3rd/C) between +M(MIDDLE.wav) and -=
S(SIDE_M.wav) cancelled out.=A0 Kind of expected, but really plays with the=
gain levels of each channel.=A0 And the derived M/S (and -S) from the L/R =
matrixed file(STEREO.wav).
$ sox -t null /dev/null -s -2 -c 1 -r 48000 null.wav trim 00:00:00 00:00:30
Taking that null file and amplify in audacity (Effect -> Amplify), I get a =
new peak level near -40dB.=A0 (baseline)
Amplify of the Original +S (SIDE_P.wav) and Original -S (SIDE_M.wav), gave =
a new peak level near -34dB.=A0 Track -> Mix and Render && Effect -> Amplif=
y.
Amplify of the Orignal +M (MIDDLE.wav) and Derived -M (MIDDLE3.wav), gave a=
new peak level near -24dB.=A0 Track -> Mix and Render && Effect -> Amplify=
.
Amplify of the derived +S (SIDE_P2.wav) and derived -S (SIDE_M2.wav), gave =
a new peak of 0dB.=A0 AKA audible results.=A0 Track -> Mix and Render && Ef=
fect -> Amplify.
Does it prove anything?=A0 Probably not as you're not likely to have much e=
xactly matching content between M and S in practical application.=A0 And it=
could just be the quality of sox as an editor.=A0 Or audacity.=A0 Or the s=
ample rate / resolution.=A0 Or my math.=A0 And mixing was best guess since =
that plays with the gain.=A0 And that's probably not how one normally handl=
es M/S content.
But I would imagine that in any editor, if you were to mix M and S to L and=
R, then save as.=A0 Then load the new file and derive M and S and derive -=
S, that S and -S wouldn't completely cancel each other out.=A0 Where in a p=
erfect world, they would.=A0 And is an indication that some degredation has=
occurred.=A0 Perhaps not in the audible range since you have to amplify si=
gnificantly to notice.=A0 But enough in my test that the resulting +S and -=
S did have audible results.=A0 Perhaps not the typical editing path for M/S=
either.=A0 And like I said probably doesn't prove much given the synthetic=
/ worst case scenario content.
It should be fairly simple to test with any M/S content.=A0 Does the invert=
ed derived M cancel out the original M?=A0 Exact match, no degradation woul=
d suggest that it does.=A0 But it appears that it doesn't.=A0 If you're eff=
ectively adding a noticeable dB of junk per edit(even if it's not audible),=
enough to motivate someone to do the matrix in post and not in the field.=
=A0 But it depends on what one considers good enough.=A0 And results probab=
ly vary greatly between software packages.
- James
--- On Sun, 8/22/10, Rob Danielson <> wrote:
|