Audio: Theory Vs. Practice

Back to post

Comments

wolfbass wrote on 9/21/2004, 7:28 PM
Bob: You're pretty close.

The situation:

My son and his class (6 & 7 y.o.) did a "Circus' performance, and the ringmaster was talking over the top of some quite crappy background music. I got a copy of the background music and was going to attemp to get the voice stand out more by reversing the phase of the cd music.

After reading the posts here, and realizing that the whole list of things included on the recording was conspiring against me, I'll probably try something different.

Sorry about the lack of detail in the first post, and follow up after, it was around 1 AM last night when I posted, and I've been asleep since.

Cheers,

Andy
Erk wrote on 9/21/2004, 9:49 PM
Red,

I know you hang out mostly in the audio forum, but with creative suggestions like that, you ought to come over here more often!

Greg
farss wrote on 9/22/2004, 1:07 AM
If all that you need is to make the voice stand out more above the music then Eq would be the simplest thing to try. Certainly no miracle is going to occur but anything will help. It still might also be worth trying the invert and subtract trick on the music. You may get a useable reduction in the music level without making it sound wierd which could make things even worse as the listener will be distracted by the odd sound.

Good luck and let us know how it pans out.

Bob.
Rednroll wrote on 9/22/2004, 6:07 AM
"Red,
I know you hang out mostly in the audio forum, but....."

Oh you'll be seeing me a liitle more often...I plan on getting into video editing more, once I build my knew PC and I know I'll be coming to this forum with lots of stupid video questions , that have already been asked a gazillion times. If you guys promise to help me out with my video questions, I promise to throw some audio tips your way.

Thanks,
Red
Chienworks wrote on 9/22/2004, 9:09 AM
Another issue that seems to have been ignored so far is sampling phase alignment. Forget the "inverse phase" phrase for a moment, especially since it's a misnomer. What i'm talking about is making sure that the samples from the recorded track line up exactly with the samples from the CD. Even if they are both recorded at the same rate, there is no guarantee that the samples are recorded at the same time! The music that came out of the PA system is for all practical purposes a continuous analog stream. The recording made of this may sample the wave form at the same rate as the original, but may get each sample a tiny bit before or after the original. This doesn't make much difference in lower frequencies, but in higher frequencies you can get a completely different waveform based on where the sample is taken.

For an extreme example, suppose the original is a 11025Hz sine wave that just happened to be sampled at 0°, 90°, 180°, & 270°. These four samples will be 0, +1, 0, & -1 respectively (with 1 being full scale of whatever volume level was recorded). When played back through an analog system and rerecorded digitally there is no guarantee that the sound will be resampled at exactly the same points. The new recording might have sampled the original sound at 45°, 135°, 225°, & 315°. These values will be +.7, +.7, -.7, -.7. This is nothing like the original waveform. When the original is subtracted from the new recording the resulting samples will be -.7, +.3, +.7, -.3. This certainly isn't 0, 0, 0, 0. Of course, real world samples will vary all over the lot from this extreme example, but the chances of getting it exactly spot on are just about nil.

One more wrench to throw into the works is that the original music was probably 44.1KHz from a CD and the recording is 48KHz from DV tape. Now try to match those samples together and the problem is compounded much more!

In this situation i would say we are approaching impossibility about as closely as possible. Or something like that. ;)