I have a bootleg recording of a concert I went to a couple of weeks ago (Toad the Wet Sprocket) that I want to clean up and convert to mp3. I am looking for any suggestions on how to improve the sound quality as much as possible. As I understand it this was recorded with pretty good equipment and I have the WAV files. There are three things that I want to do to the wavs before converting to mp3.

1. Normalize them, they are only about half as loud as they should be. No problems here, I have several pieces of software that will do this.
2. Decrease the bass/boost the mids and highs. I have a shareware program called Sound Studio that looks like it will do this OK.
3. This is the one I am not sure how to do. The music seems to be mostly mids. If you think of the waveform that would represent the audible range all the lows and highs are smashed into the mid range. Is there any way to flatten that wave form to make the lows more low and the highs more high? I am thinking of the way Photoshop will let you set white and black points for a photo and stretch the available data to fit that new curve. Does this even make sence to do with audio? Can anybody recommend a program to do this (that runs on Mac OS or linux/freebsd) or at least give me a name for the process that I am trying to describe?

So the plan would be to concatinate alll the files into a single wav run the three operations above on the resulting file (so modifications are consistent accross all of the concert), chop back into individual files and convert to mp3. Does this sound like a reasonable way to do this? Any other suggestions on getting the most out of a less that perfect audio source?

Thanks,
-Mike

If you gave me a snippet of the audio, say, a verse and a chorus, I would listen to it and give you my opinion on how I would alter it.

If you gave me a snippet of the audio, say, a verse and a chorus, I would listen to it and give you my opinion on how I would alter it.

Cool, thanks for the offer. Here is about a minute out of one of the songs converted to 320kbps mp3 (so the file isn't too huge), but otherwise unaltered.

http://macgeek.dyndns.org/sample.mp3

Thanks,
-Mike

This is the one I am not sure how to do....


What you described sounds to me like most of the bass and treble frequencies got lost in the recording. This would be typical of bootleg recordings; bass gets absorbed by a mass of bodies, and treble can get absorbed by eg, the material that a pocket is made of, the microphone's frequency response etc.

But you also mentioned needing to reduce the bass. This bass that needs reducing, is it sub-bass range? (ie 30-80Hz, wouldn't get absorbed so much by a mass of people), or is it higher? If it's higher, it could be room resonance - that gives a boomy appearance to the sound despite there not being any real bass in the music. (Assuming that the room is rectangular, there could be up to 3 prominent resonances that should be fairly easy to reduce).

(In fact, thinking about it more, it's probable that the microphone wouldn't pick up much sub-bass anyway..)

Whatever,
A decent EQ, HW or SW should be able to make it much more listenable (31 band graphic, or a sweep EQ (more commonly, but incorrectly called a parametric EQ, such as the empeg's EQ))

Got a sample?

What you described sounds to me like most of the bass and treble frequencies got lost in the recording.

Well, I don't think so. I know a lot of the recorded songs pretty well and what should be the bass seems to be there, it is just higher in the range so that in my jeep (scary that it is the best listening environment I have) the bass is heard through my mids rather than felt through the subs.

tt could be room resonance - that gives a boomy appearance to the sound despite there not being any real bass in the music

That sounds like a pretty accurate description.

A decent EQ, HW or SW should be able to make it much more listenable

Yep, just running it through a software eq and starting to cut requencies below about 640hz and then normalizing the result seems to help a lot. I haven't listened to the result through real speakers yet though.

Got a sample?

See my response to Tony above.

Thanks,
-Mike

Well, I don't think so. I know a lot of the recorded songs pretty well and what should be the bass seems to be there, it is just higher in the range so that in my jeep (scary that it is the best listening environment I have) the bass is heard through my mids rather than felt through the subs.

A microphone and a tape recorder (used normally at normal speed) can't alter the pitch of an instrument. It can just color that pitch by introducing a different frequency response.

Remember that an instrument like a bass guitar covers a wide frequency range. What's happening is that certain frequencies are getting lost while others are being accentuated. The pitch of the bass notes has not been changed, which is what you're claiming it's doing.

What you're talking about here is a recording that needs plain old EQ and compression. The recording engineer's first two tools in his toolbox. Nothing magical about it, just the subtle art of audio mixing.

What's happening is that certain frequencies are getting lost while others are being accentuated. The pitch of the bass notes has not been changed, which is what you're claiming it's doing.

OK, just trying to describe what it sounds like to me. So basically what it boils down to is that the bass has been lost for whatever reason and I need to EQ down the lower frequency mids which will hopefully remove enough of the muddiness so that you can at least clearly hear the singer and then I need to compress the result to make it sound louder.

So my next question would be on compression. The app I am using has a compression filter that looks like this...



I am at least quasy familiar with an EQ, but I don't really know anything about compression. Can somebody explain what all those parameters do and what I might want to tweak for this recording?

Thanks for your help,
-Mike

It isn't totally obvious that you need compression per se, but it's as good a method of normalising the volume as any...

Compression is usually used to reduce the dynamic range of a track, ie the variation between loud and soft passages. It does this without changing the overall 'shape' of the waveform though.

The first stage of compression is the bit that makes loud passages quieter.

The threshold level defines the signal level at which the compression will kick in, in this case shown at -20dB; now in a hardware compressor, that dB level would usually be with respect to line level, in SW, I'm guessing that 0dB refers to the maximum signal magnitude that can be represented.
The peak/RMS toggle defines the method used for calculating the current signal magnitude. For reducing clicks, thumps or anything else that is transient, peak might be a good choice. I'd probably try RMS first for your needs though. (Although since we've already suggested that you don't need to compress the music, it shouldn't make much difference.)

Ratio defines the amount by which the magnitude of the signal will be reduced once it goes over the threshold. Compression is dynamic - the more the signal goes above the threshold, the more it will be compressed.

Attack defines how quickly the compressor will 'kick in' when the magnitude raises above the threshold. Similarly, release defines how long the effects of the compression will hang around after the magnitude drops below the threshold.

The post gain is what makes this useful to you - we know that the threshold and ratio settings have given us headroom, and the post gain will expand the signal into it, without clipping.

But enough of the technicalities - leave attack, release and ratio at those settings as they look reasonable. Work on the threshold/post gain first...Actually the settings there look reasonable as a starting point too.

Cool, thanks for the detailed description. Running it through the compressor with the default settings gets me something that is more pleasent (I can at least understand what the singer is singing throughout the passage). I am going to create a few samples to play through real speakers tomorrow. Is there a correct order to run the filters in (should I adjust EQ before compressing or vice-versa) to avoid discarding bits?

-Mike

Compression before EQ.

If you were a mastering facility, you'd run it through and expensive multiband compressor, which compresses various frequencies different amounts.