Compression/Expansion Tips (Battling Distortion)

GoldWave general discussions and community help
Post Reply
jdeligiannis
Posts: 34
Joined: Sat Mar 05, 2005 1:50 am

Compression/Expansion Tips (Battling Distortion)

Post by jdeligiannis »

I usually use Goldwave 4.26, so I will be referring to that version's terms.

In audio class, I was told that radio stations use compression/expansion to even out the levels between various songs, especially when playing songs from different eras. In other words, all songs played will seem to have the same loudness, or dB levels.

I am playing with compression/expansion, but I don't see a way to do what the radio stations do. If I am going to create a custom CD, I want to make all songs on the CD have the same levels, just like the radio stations, where quality remains virtually the same, but only different in the loudness sense. How can I compress/expand all my songs to the same levels, without actually going in on each one and adjusting the settings.

Also, it seems that when using compression/expansion, putting the smoothness to 0 would be ideal, except for the distortion it causes. Putting this to 0 targets the threshold right on. If compressiong, it's not so bad to go with a smoothness around 0, but with expansion, a terrible static is created. The only way I can find to prevent the static, other than not using expansion, is by increasing the smoothness up a lot. However, this will defeat the entire purpose. If I increase the smoothness, now I won't be targeting the low levels, but rather the higher levels, which I just compressed! Is there any good advice to be given? I want to boost the low levels (expand) but without getting the distortion. Is there a setting that will prevent the distortion I get when the smoothness is set to 0 on expansion?

Thanks.
GoldWave Inc.
Site Admin
Posts: 4372
Joined: Wed Mar 10, 2004 6:43 pm
Location: St. John's, NL
Contact:

Re: Compression/Expansion Tips (Battling Distortion)

Post by GoldWave Inc. »

In GoldWave v4.26, you can use the Volume | Maximize effect to set the rms level to bring the volume of each song to a similar level. In GoldWave v5, you'd use the Match Volume effect and Batch Processing to apply the same average level to a group of files.

Chris
DougDbug
Posts: 2172
Joined: Wed Feb 16, 2005 3:33 pm
Location: Silicon Valley

Post by DougDbug »

This is a difficult thing for a computer to do. It really takes a human to judge the "loudness" of a song. A song with strong high-frequency content is going to sound louder than a song with the same level of bass. You'll have to use your own judgement for songs that have loud parts and quiet parts... And, some songs are supposed to be quieter than others.

It's probably best to Maximize all of the songs first. Now, all of the songs will have 0dB peaks. If they are "ripped" from CDs, most of them will already be maximized...but you won't know that 'till you run Maximize.

Then, listen to them. If some sound too loud or too quiet, you can then compare the average levels with Volume Match. Use the average as a starting-point, but trust your ears.

Most of the songs won't need any more adjustment, unless there's one quiet-sounding song, and all of the others need to be reduced... Note that once you've Maximized, you can't increase the volume without clipping (unless you compress), so you'll have to decrease the loud song(s) rather than increase the quiet-sounding ones.


-------------------
Years ago, in the analog days, I was a DJ on my college radio station. The DJ had to watch the VU meter and adjust the volume manually. We would set it so that the peaks hit about +3dB. The absolute peaks would have been a bit higher, because a VU meter is not a peak-level meter and it doesn't respond instantly. (We didn't "ride" the volume control... we would adjust it for each song as-necessary.) Then, there was a peak-limiter in-line before the transmitter to prevent over-modulation. (If you over-modulate, you get in trouble with the FCC.)

I'm not sure if the limiter (legal limit) was calibrated to 0dB... If it was, then the combination of boosting the level to +3dB and then limiting the peaks back-down to 0dB would provide some compression.

If the DJ turns-up the volume, the average level increases and it sounds louder. But, the peaks can't increase 'cause they are limited... That's one form of compression. A similar technique is used to a greater extent with radio commercials.

The problem with compression is that it "squeezes" the life out of the music. One of the things that makes live music sound better than recorded music is the greater dynamic contrast... the difference between the loud sounds and the quiet sounds.

I assume that DJs no longer need to control of the music level... CDs are mostly normalized (maximized) so their levels are more constant than records were, and it's much easier to do this stuff automatically with digital electronics than it was in the old analog days.
piano nick
Posts: 423
Joined: Wed Jun 16, 2004 8:33 pm

Post by piano nick »

Good post Doug; some excellent advice.

I'll bore you with the process I do (which follows all the advice Doug has given):


I'm in the process of "matching" about twenty songs for a cd, and what I do is this (bear in mind this is mostly piano music - surprise!!)

1. If the dynamic range is large (subjective, but starts to make sense after a few years), I apply the default REDUCE PEAKS (Effect/Compressor/Expander), but with ZERO attack and release. I've found that even the small default attack/release tends to "clip" piano sounds (as though I didn't hold the notes long enough).

2. This slight compression is only applied to the piano track - pianos tend to have a very large dynamic range, particularly the way I play.

3. If the difference in the peaks and valleys is large (a good example is the classical piece Claire de Lune), I will apply a small amount of compression twice, but don't usually use the same THRESHOLD or MULTIPLIER (but alway zero attack/release). Note carefully what Doug said - don't squeeze the life out of the dynamics or pretty soon, it will sound as exciting as the drone from a table saw.

4. If there are other tracks, they are mixed and then in Goldwave, I MAXIMIZE to -0.20 dB. This sounds close to clipping but is not. Note that this is the final wave editing operation. Everything is done before MAXIMIZING.

5. Then I compile all the songs for one album into a PLAYLIST in Windows Media Player and play them while I'm doing something else (not watching the wave file on the screen).

6. Any overly "loud" songs will immediately stand out, and they are then "backed off" in overall volume with the CHANGE VOLUME COMMAND, or if something is obviously wrong, do the mix over. The point of this is to underline what Doug said - only your musically experienced human ear/brain can make the final decision as to how it sounds. In addition to the frequency effect that Doug points out, some songs just do not sound correct if they are too loud (a slow sad blues DOES NOT want to be LOUD).

7. Again, as Doug said, TRUST YOUR EARS.


PN
donrandall
Posts: 550
Joined: Wed Dec 01, 2004 11:06 pm
Location: Denver, Colorado

Post by donrandall »

The advice to "trust your ears" is sound advice.....ooooh.....sorry, heh, heh, heh......yah sure I am.

I occasionsally run into occurences where a sound that looks like it should be quite loud, judging by the wave size, just isn't as loud as you would expect. I have also noticed that certain sounds that seem to be very loud, really do not produce a visible wave that is nearly as loud as you might expect. I don't know how to explain it, but I have a layman's theory that it has something to do with the "density" of the sound.

I work with voiceovers, so my observations are in that regard. I have noticed that a smaller wave produced by speaking in closer proximity to the mic will sometimes seem louder than another wave, produced by speaking back, off mic just a little bit, which will actually be perceived as not being as loud. It seems counter intuitive - the big wave generally corresponds to greater perceived volume and the smaller wave usually corresponding to less perceived volume.
DougDbug
Posts: 2172
Joined: Wed Feb 16, 2005 3:33 pm
Location: Silicon Valley

Post by DougDbug »

I just ran-across a program called Volume Balancer ($15). I haven't tried it out, but I have another very good program (Wave Repair) written by the same programmer.
Post Reply