mp3 compression squelch remove

[Andrew Smith]

Does anyone know of any tools that will read in a heavily compressed mp3 file and repair the squelching? Not sure if such a thing exists.

Andrew

[Rick Reineke]

Never heard of "squelching" in terms of MP3 files. Once an MP3 is encoded, it is what it is. I recall other folks asking if converting an MP3 to PCM (aiff or wave) would bring back the original quality.. No, it would just be a copy of the MP3 along with all the data compression artifacts.

[Andrew Smith]

Well it's hard to describe. I guess it could otherwise be described as the "mp3 compressed too hard" sound. :-P

Andrew

[Greg Miller]

I doubt very much that you're referring to actual squelch, which is a special kind of noise gating originally designed for radio transmission, especially FM mode.

Out of curiosity, what are the details of the MP3 file? Music or voice? Mono or stereo? Bitrate? Sample frequency?

[Andrew Smith]

It's a voice recording. The file that prompted my thinking about this is at http://www.culthelp.info/audio/jan_g...sted_years.mp3

It has audio cassette hiss in it, but it also has nasty compression artefacts. It's interesting in that if an mp3 was your last archive of an old recording .... would it be the end of the line in terms of enhancing enough to make a better quality copy?

Andrew

[Steve House]

Originally Posted by Andrew Smith

It's a voice recording. The file that prompted my thinking about this is at http://www.culthelp.info/audio/jan_g...sted_years.mp3

It has audio cassette hiss in it, but it also has nasty compression artefacts. It's interesting in that if an mp3 was your last archive of an old recording .... would it be the end of the line in terms of enhancing enough to make a better quality copy?

Andrew

Once information is gone, it's gone for good. MP3 has its uses as a distribution format but IMHO should NEVER be used as an original recording or archival format. Storage space for bits is cheaper than dirt and using lossy compression schemes to reduce its use is a false economy.

[Andrew Smith]

Yup. Preaching to the converted.

Andrew

[Rick Reineke]

Aside from the tape hiss (which can be significantly attenuated via EQ and/or NR software and/or expansion), I don't hear any really objectionable data compression artifacts. Fortunate it's mono, so has the equivalent quality of a 48kbs stereo MP3, that and the 22kHz sample rate are usable for spoken word content.

[Greg Miller]

I hear some strange qualities to that audio, but I'm not sure they all can be blamed on the MP3 compression.

For one thing, the voices sound rather muffled. Bad mic? Bad master recording? Bad EQ? Bad processing? But here is a problem with using such a low bitrate. From what I can find quickly, an MP3 file encoded at 24kbps, with a 22.05kHz sample rate, can contain a maximum audio bandwidth of ~~ 6.2 kHz. A quick look at the test file confirms this: the HF cutoff is roughly 6.0kHz. That certainly accounts for some of the missing HF information in the voices. And that HF information is gone forever.

Aside from that, when I listen closely to the original file, the tape hiss does not sound continuous. It seems to rapidly fluctuate in level. EDIT: A good example of this is the section from 3:41.5 to 3.42.8. Was this a problem with the tape transport? Or was there some bad processing somewhere along the line? Maybe an attempt at NR, which might have also affected the voice clarity? Whatever the cause, if it affected the HF tape hiss, it may well have affected the HF of the voices as well. Again, there is no way magically to recover that fidelity.

Bottom line: 24kbps, with a HF limit of 6kHz, is adequate for speech intelligibility. (After all, the public telephone network uses a HF limit of 3.0 to 3.5kHz.) But it doesn't sound great.
A mono file at 32kbps and 22.05kHz sampling would have a HF limit ~~ 7.6 kHz.
A mono file at 48kbps / 22.05kHz would have a HF limit of ~~ 10.5kHz.
And with each increase in bitrate, not only does the HF bandwidth improve, overall fidelity improves.

To answer your general question: "sometimes." e.g. if the original MP3 file also had bad AC hum, that could be removed later with NR, resulting in an improvement over the MP3 file. But you cannot restore missing frequencies, and you cannot improve bad fidelity (i.e. "accuracy") of the program material.

And the original should be as clean as possible, before encoding to MP3. If there is AC hum in the original, some of the bandwidth of the MP3 file will be used to encode the hum. That amount of bandwidth then will not be available for encoding of the desired audio. The same is true of any noise (such as tape hiss) in the original. The codec goes along, moment by moment, deciding which audio is "important" (which is then encoded) and which audio is "unimportant" (which is then discarded permanently). So you want to remove as much of the unimportant audio as possible, before encoding. This is especially important at low bitrates (like 24kbps).

As for your specific sample file: it will never sound good. It's too far gone. Nothing you can do to make it sound realistic. Intelligible, yes. Good, no.

[Jon Fairhurst]

In general, one can use an expander to increase dynamic range. Many compressors can perform an expanding function. Unfortunately, expansion can increase noise, so a noise reduction pass might be necessary as a final step.

Would the result be better than the MP3 source? I guess it depends on the content (which I didn't listen to) and the preference of the listener.

If the simple, aggressive compression was done on the final, you might get decent results if you don't push it too hard. If the compression varied by EQ range, that would be tougher. If the compression was done on the original tracks before the mix, you'd be hard pressed to recover that.

However, there are some spectral analysis tools that can separate different instruments. Do that, expand by track, and put it back together, and you will get a result. I don't know that the result would be good, but you never know...

[Greg Miller]

This is all voice (as stated by the OP, and as confirmed by listening). There are no instruments.

I don't think there was any question about dynamic compression. The discussion was more about data compression when the file was converted to MP3.

Downward expansion might somewhat reduce the background hiss, but IMHO that is not the main problem. The main problem is the "grunge" and "mud" in the voice. The intelligibility is acceptable, but the voice sounds "unpleasant" to me. I think some might be the result of low-bitrate MP3 [data] compression, but without knowing the history of this recording and all the previous generations and processing, it's kind of a crap shoot. At any rate, I doubt that it can be made to sound "nicer" than it is now. Take a listen.