0% Interest for 24 Months! Learn more »
(800) 222-4700
  • Español: (800) 222-4701
Cart
Microphone Month

Sweetwater Forums [Archived]

After 15 years of great discussions, the Sweetwater Forums are now closed and preserved as a "read-only" resource. For discussions about current gear, check us out on Facebook, YouTube, inSync, and our Knowledge Base.

Best Software to Remove Distortion

orangutang

I've searched everywhere for the best solution to remove distortion from a low encoded MP3. Everyone always says that there is no way to fix it, not even the equalizer. Is there some kind of software that uses "artificial intelligence" to remove distortion? Sorry, I don't know much about music editing.
February 19, 2006 @05:56am
AndyH

Have you seen the drawings of a sound wave, a wiggly line going up and down about a central axis? Do you have an idea of the way a sine wave looks on a oscilloscope? This is just to give us some common starting point to explain to you why what people say is true.
The music can be graphed or displayed as such a line, going up and down in the vertical or Y axis as time progress along the horizontal or X axis. It is always and unconditionally the case that if the volume level isn't zero the line is moving either up or down as the music is played.
While you can't visually tell very much about how the music sounds from looking at a graphical display of it, it is the case that every different instant of the music produces its own display, different from any other. If you play the same music file over and over it will always produce the same trace on the display for any given part of the song. Its wiggly line will always be the same each time, but it will always be different than the display for any other piece of music.
Distortion is some change in that wiggly line. Something has altered the music which can be seen as a difference in the line. To remove the distortion you must return the line to exactly as it was before the distortion got into it. What should it look like? You can't say unless you have an undistorted copy to compare it with, and neither can any program. Since all music is different than all other music, there is no way to look at the line and say "this part is wrong," or "that part should go like so instead of like that." There is nothing that can appear on the display that might not be right for some piece of music even if it is terrible for this particular piece of music.
February 19, 2006 @08:29am
eyeo dbp

There are several things that CANNOT be removed from audio and my friend distortion or audio clipping is at the top of that list.
February 24, 2006 @05:47pm
BagHun

What they say is true. There are apps and plugins that attempt to reduce noise, however.
February 24, 2006 @07:48pm
orangutang

I don't know anything about audio editing but isn't there some kind of frequency where the audio starts to get distorted? If there is, why can't you just lower the frequency?
March 2, 2006 @06:53am
robertruetz

The reason there is "distortion" in the first place, is because too much information has been lost in the conversion to MP3 (at least I'm guessing this is what you're describing). Therefore cutting anything else out, would actually do more damage. There is no way to put the frequencies back that were lost during encoding. This is an unfortunate consequence of encoding MP3s at low bit-rates. Your best bet is to do the encoding process over again, this time with a higher bit-rate. Sorry, but there is no magic bullet for this one.
Le_Singe
March 2, 2006 @03:15pm
sabianq

The reason there is "distortion" in the first place, is because too much information has been lost in the conversion to MP3 (at least I'm guessing this is what you're describing). Therefore cutting anything else out, would actually do more damage. There is no way to put the frequencies back that were lost during encoding. This is an unfortunate consequence of encoding MP3s at low bit-rates. Your best bet is to do the encoding process over again, this time with a higher bit-rate. Sorry, but there is no magic bullet for this one.
Le_Singe

yea, i guess it is an incorrect assesment to say you want to "remove" distortion as distortion is an absence of information.
lets say you have a picture of a wave printed on a piece of paper.
now take a pair of shears and cut off the top of the wave and thow away the removed section.
look at the picture, notice how the wave is "clipped" and the wave ends where you made the cut? that, my boy is distortion, an absence of information.
how can you remove "nothing"?
but i do understand what you are asking.
is there a program out there that can rebuild the wave using an algorithm that might be able to extrapolate information based on the information preceeding wave termination?
well, the straight forward answer is,
not yet.
id say, that given the way computer power if increasing, if you wait 15 or so years, im sure there will be a program out there that can do exactly what you are asking easley and accurately.
so either build a time machine, write an algorithm, re do the take, or just wait for a while.
your answer will happen.
March 2, 2006 @04:00pm
sabianq

I don't know anything about audio editing but isn't there some kind of frequency where the audio starts to get distorted? If there is, why can't you just lower the frequency?

audio gets distorted when there is to much gain for the equipment to handle.
frequency has nothing to do with distortion.
frequency is pitch. when a horn or piano or guitar player plays scales, he/she is going up and down the frequency range, the higher the frequency, the higher the note.
if you lower frequency, you will change the pitch of the music and/or sample and that is totally undesirable.
March 2, 2006 @04:07pm
softwizz

I've read the postings on this thread, and I think none of them is completely correct.
The CEDAR audio restoration system developed in Cambridge, England is capable of achieving at least some of the manipulations which are said on here to be impossible. For example, it can remove a click or pop and replace it with appropriate audio which it has synthesized intelligently from the surrounding clean audio - the equivalent of an 'invisible mend' on a tailored suit. This is done by complex fourier analysis functions which calculate the frequency content of the material.
For another example, Steinberg WaveLab is capable of changing the duration of a recording while preserving the pitch of the sounds, and also of changing the pitch of a recording while preserving its duration.
Both these systems achieve their results by creating completely new waveforms which reflect accurately the program material. It's not much of a stretch from that, to an algorithm which - for example - detects unpleasant odd-order harmonics and removes them from the recording.
The only issue, for me, is who has developed such algorithms, and where can I buy an implementation?
October 24, 2009 @01:53pm
AndyH

There are many programs that do what Cedar and Wavelab do, some better, some worse. Clicks & pops and broadband LP noise are additions to the recording that can generally be removed, leaving the music near enough to the original to be (usually) quite acceptable. These process do not produce “completely new waveforms” but attempt to effect as little of the original as possible to achieve their results. Time stretching (or compressing) and pitch changing are, relatively speaking, child’s play.
There are many kinds of audio distortion. Distortion of several kinds is often intentionally induced in many genres of music, electric guitar being a very common example. Basically, distortion exists where the output of any process does not equal the input, not including a change in scale (i.e. aside from changing the volume level (which will also be distorting if not completely linear)). Distortion is almost always a one way mapping, it cannot be reversed for the reason that the output does not carry the information of what was done to produce the change, nor what the input was like before the change. This will never change, regardless of computer power.
Additionally changing the output so it sounds better than the first result is sometime possible. Clipping can sometimes be “fixed” to sound better by making the transition from before the clip to after the clip smoother and more gradual than the abrupt change of the clip itself. This can sometimes be quite acceptable to the listener, but it is not a reversion to the original waveform.
A side effect of most distortion is the production of new harmonics and overtones. These may be desirable, such as in the cited guitar playing, or undesirable such as LP in-groove damage from a bad stylus or poor alignment. Filtering out higher frequencies, and doing a kind of click/pop removal on them, can make something that sounds closer to the original simply because it now does not have all the distortion generated additions. The basic distortion, the difference from the original, will still be there, however.
MP3 encoding, the original topic of this thread, reduces the resolution of parts of the music that are less audible. It then uses a compression process (loosely comparable to zip encoding a data file) to reduce file size. When decoded for playback, the file becomes the same size as the original recording, but the parts whose resolution was reduced cannot be returned to their original form -- there is no information in the encoded data to tell what that original was. This perceptual encoding can work extremely well, so that the output, although quite different mathematically from the input, sounds indistinguishable from the original to almost everyone.
However, the reductions can be carried to extremes in order to produce very small files. The lower the resolution, the smaller the data can be compressed. Complete silence, the ultimate resolution reduction, can be fit into a very small file. Uncompressed, it is exactly as large as any completely full audio file of the same duration.
Also, some encoders work less well than others. The encoding is based on empirically obtained models of human hearing. Translating these into the encoding program algorithms is involved and can be done under more than one interpretation. Some encoders, LAME in particular, have been fine tuned through extensive listening /comparison tests.
The lower bitrate encodings can take the resolution reduction to extremes beyond sounding much like the original, Their legitimate goal is usually to produce something intelligible, without regard to what the original sounded like. For example, an audio comment on any presentation, such as a slide show. When applied to music, it can be easy to tell their sound is different than the original, but only through comparison, sample by sample, with the original data can one know what the difference is.
October 24, 2009 @08:46pm
ecc83

What we might call "linear" distortion, due to a curved I/O characteristic certainly can be improved by various techniques, pre-distortion (Nagra tape machines) feedforward...
Clipping cannot be rescued by analogue means AFAIK but there are several digital app's that claim to be able to. One "freebie" you could try is Sony Soundforge.
Dave.
October 25, 2009 @05:57am
mantronix

I've searched everywhere for the best solution to remove distortion from a low encoded MP3. Everyone always says that there is no way to fix it, not even the equalizer. Is there some kind of software that uses "artificial intelligence" to remove distortion? Sorry, I don't know much about music editing.

There is nothing that will correct the compression created by an MP3 file unfortunately.
November 3, 2009 @03:22pm
sabianq

bottom line,
distortion due to lost information can not ever ever be restored.
but it may be covered up if it is done correctly.
cheers!
November 4, 2009 @03:15pm
AndyH

While I think we have pretty well covered the fact that the problems cannot be undone, a significant mis-impression crept in near the end. MP3 encoding reduces resolution, adds noise, and almost always involves low pass filtering to remove higher frequencies, but there is no compression of the audio. True, the resulting mp3ed audio data is compressed for storage, conceptually the same as non-audio data being compressed with zip, but that is undone by decoding, just as a zip file is returned to normal mode by unzipping. “Compression” as the term is used in audio processing, is not part of the mp3 process.
November 5, 2009 @09:26am