Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Dynamics Compressor for FB2K (Read 32564 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Dynamics Compressor for FB2K

Reply #25
Quote
Compression is a one way process, sorry. No way to recover the original stuff.

dbx?

<edit> not that dbx will be useful as an expander towards the goal desired here, but rather that it is an example of compression-expansion. And Dolby's similar methods for that matter.

Dynamics Compressor for FB2K

Reply #26
Quote
Quote
Compression is a one way process, sorry. No way to recover the original stuff.

dbx?

What dbx? They have different maths than the rest of the world?

Dynamics Compressor for FB2K

Reply #27
well, mind explaining yourself when Dolby and dbx both contradict what you're saying?

Dynamics Compressor for FB2K

Reply #28
Please post the exact statements here, I'd love to see them explain how they work around the laws of information theory.

Dynamics Compressor for FB2K

Reply #29
you're unfamiliar with their tape noise reduction systems?  Both companies employ fixed-slope compression and expansion schemes to reduce tape hiss and increase dynamic range.

Dynamics Compressor for FB2K

Reply #30
Quote
fixed-slope

When's the last time you bought a CD compressed with a standard fixed slope?

Dynamics Compressor for FB2K

Reply #31
Quote
Quote
fixed-slope

When's the last time you bought a CD compressed with a standard fixed slope?

that's not the point, I don't think that what's being asked for here is something that will bring every single compressed CD back to the exact state it was in before it was compressed.  Obviously such a thing would be impossible, because you have no way of knowing the exact way it was compressed.  The point I'm trying to make is that it is possible to expand something, since the principle's been in use in the industry for years and years and years, when Garf seems to be pretty sure that it's impossible.  Yes, it would be impossible to do what you're talking about, and Garf's right in stating that getting the original signal back when you don't know how it was compressed in the first place isn't possible.  But it also sounds like he's saying it's impossible to expand anything at all, actually.  When clearly it isn't impossible, or else dbx noise reduction would never have worked.  And it did.

Dynamics Compressor for FB2K

Reply #32
they are 2 various typologies:

1) dbx or dolby ( a, b, c, s ) are compressor-expander systems to reduce hiss on analog tape.
original signal is compressed when recording and expanded when play and restore original dynamic.

2) expander is a signal processor what increases dynamic when play.
obviously it  can't restore original dynamic but it can improved some recordings.
not to use all the time.

for exemple:

expander
Music is my first love.

Dynamics Compressor for FB2K

Reply #33
GARF , do you think add preset support, same for equaliser?
Music is my first love.

Dynamics Compressor for FB2K

Reply #34
Quote
But it also sounds like he's saying it's impossible to expand anything at all, actually.

I _NEVER_ said that. I said it was impossible to get the original back. Note that even if you know the compressor, it will be impossible to get the exact original back, if there's a multiple-to-one mapping. And this happens in all compressors.

So you can do some expansion and get something reasonable, but never what was the original. Simple maths. You can't make anything out of nothing.

Expansion is a lot harder to make work well than a compressor, and I don't know of any ones that I consider good enough in general use to make it worth the hassle.

Dynamics Compressor for FB2K

Reply #35
Quote
GARF , do you think add preset support, same for equaliser?

Do you think it would be usefull?

Dynamics Compressor for FB2K

Reply #36
Yes, i think it would be usefull, same as preset support is usefull for equaliser.

But I think that what I think is less important than you think on this subject, because it is you who made work...
if you think that it is useful and not too complicated to make I would think that it is well.
if not, the wise would say: compressor without preset support is better that no compressor! 

EDIT: please , do you want explained how parameters act on signal, to educate ignorant people, me the first.
Music is my first love.

Dynamics Compressor for FB2K

Reply #37
Someone posted a link to an explanation about compressors before in this thread, explains it better than I could probably.

I'll consider presets.

Dynamics Compressor for FB2K

Reply #38
Quote
I _NEVER_ said that. I said it was impossible to get the original back. Note that even if you know the compressor, it will be impossible to get the exact original back, if there's a multiple-to-one mapping. And this happens in all compressors.

So you can do some expansion and get something reasonable, but never what was the original. Simple maths. You can't make anything out of nothing.

Expansion is a lot harder to make work well than a compressor, and I don't know of any ones that I consider good enough in general use to make it worth the hassle.

OK, explain this then:

Compress
1  amount over 5 is -4 so output 1
2  amount over 5 is -3 so output 2
3  amount over 5 is -2 so output 3
4  amount over 5 is -1 so output 4
5  amount over 5 is 0 so output 5
6  amount over 5 is 1 so output 5 + 1/2(1) is 5.5
7  amount over 5 is 2 so output 5 + 1/2(2) is 6
8  amount over 5 is 3 so output 5 + 1/2(3) is 6.5
9  amount over 5 is 4 so output 5 + 1/2(4) is 7
10 amount over 5 is 5 so output 5 + 1/2(5) is 7.5

Expand
1  amount over 5 is -4 so output 1
2  amount over 5 is -3 so output 2
3  amount over 5 is -2 so output 3
4  amount over 5 is -1 so output 4
5  amount over 5 is 0 so output 5
5.5 amount over 5 is 0.5 so output 5 + 2(0.5) is 6
6  amount over 5 is 1 so output 5 + 2(1) is 7
6.5 amount over 5 is 1.5 so output 5 + 2(1.5) is 8
7  amount over 5 is 2 so output 5 + 2(2) is 9
7.5 amount over 5 is 2.5 so output 5 + 2(2.5) is 10

Looks like the original was recreated to me. Explain how it wasn't, please.

Dynamics Compressor for FB2K

Reply #39
Quote
Looks like the original was recreated to me. Explain how it wasn't, please.

Most audio formats (including CDDA) don't store floats, they store integer values.
A riddle is a short sword attached to the next 2000 years.

Dynamics Compressor for FB2K

Reply #40
Quote
Most audio formats (including CDDA) don't store floats, they store integer values.

so? Got absolutely nothing to do with what I just said.

<edit> Happy?
512
1024
2048
4096-2048->2048(0.5)->3072
8192-2048->6144(0.5)->5120
16384-2048->14336(0.5)->9216

512
1024
2048
3072 1024(2)+2048=4096
5120 3072(2)+2048=8192
9216 7168(2)+2048=16384

Dynamics Compressor for FB2K

Reply #41
4096 4097

Dynamics Compressor for FB2K

Reply #42
ABX the rounding errors then.

Dynamics Compressor for FB2K

Reply #43
I can't ABX MPC --insane either, that means I can get the original back via some procedure you will no doubt explain us?

Dynamics Compressor for FB2K

Reply #44
you're using integer storage as an excuse to not admit you're wrong.

<edit> in other words, the only reason you say it's impossible to get the original back is because only integers are stored in the one format you were thinking of. Even though it would more than likely be impossible to tell any difference if you actually did try compressing and then expanding some 16bit audio.  Yes, I understood from the beginning that if you were dealing with 16 bit integers you would not be able to get exactly the same data back because of the margin of error when dealing with odd value samples.  But with the actual results, I sincerely doubt you'd hear anything different, and I'm sure you do to.

Dynamics Compressor for FB2K

Reply #45
Yes, I assume the music is not stored in floats after compression. This is true for any practical format I know of. Lossy audio can be floats but suffers from the same problem that the mapping is not invertible. Moreover, compressors have variable gain to attain a constant average loudness. You cannot figure out the original loudness levels either even if you use floats.

No, it does not only have to do with odd/even problems, the sharper the compression, the more values are mapped to the same output value.

It gets even more fun when the compressor causes hard limiting, nothing will save you in that case.

Dynamics Compressor for FB2K

Reply #46
now you're mixing issues again. I'm not talking about guessing the compression method of any random CD in order to obtain exactly what was there before the compression occured, as my example clearly was a given scenario. In one instance we're talking about a compressor that does a straight 2:1 compression after a given point, and then you start talking about variable gain compressors as if they were even related to the process I'm talking about. Fine. You have problems admitting when you're wrong.  Citing extreme and/or unrelated scenarios to try to prove you're right.  Yes, you're right in your scenarios, I could not guess a random compression method from a random CD.  I'll drop it. Fact remains, if you couldn't get out what was put in, dbx's product never would have worked.  And it did.  You're arguing that an expander would be worthless because I can't guess those values or have enough accuracy to get the original back even if I could.  Well, it wouldn't be worthless because I don't need guess how it was compressed or have enough accuracy to get the original back after guessing, because I only need to make it sound better than its current state.

<edit> much like clipping restoration isn't useless just because it has no way of knowing whether or not the signal during the clipped portion would have been a smooth curve matching the slopes of the waveform before and after the clipped portion.  Yes, there could have been more than a simple curve in that portion of the signal, but that doesn't make the clipping restoration process worthless.  No, it doesn't necessarily give you what was originally there before it was clipped off, but it does give you an unclipped signal.

Dynamics Compressor for FB2K

Reply #47
Quote
I'm not talking about guessing the compression method of any random CD in order to obtain exactly what was there before the compression occured, as my example clearly was a given scenario.


True. In such a case, you can do a reasonable reconstruction that would probably be audibly the same, though not identical to the original. Is this case typical? Not at all.

Quote
You have problems admitting when you're wrong. 


My original point still stands as strongly. You cannot invert a compressor for all practical purposes, you can't invert a one-way function, and you cannot invert something that maps multiple values into one.

You haven't come an inch closer to refuting this.

Quote
Fine. You have problems admitting when you're wrong. 


Is this another way of saying you have run out of arguments?

Quote
Citing extreme and/or unrelated scenarios to try to prove you're right. 


There is nothing extreme to the scenario I quoted, it's the typical one on most CD's. It's not unrelated either, since it's the one you actually have to deal with in practise.

I refuted your theorethical example, and you say my refutation doesn't matter in practise. Right. But your example wasn't practical, so that's not an argument.

Quote
Well, it wouldn't be worthless because I don't need guess how it was compressed or have enough accuracy to get the original back after guessing, because I only need to make it sound better than its current state.


This I wont argue, it might be possible to make something that reasonably guesses the used method and makes some additional dynamics out of it. Would it work well? So-so, depending on what the original is.

But you can't get it back.

Dynamics Compressor for FB2K

Reply #48
Quote
<edit> much like clipping restoration isn't useless just because it has no way of knowing whether or not the signal during the clipped portion would have been a smooth curve matching the slopes of the waveform before and after the clipped portion.  Yes, there could have been more than a simple curve in that portion of the signal, but that doesn't make the clipping restoration process worthless.  No, it doesn't necessarily give you what was originally there before it was clipped off, but it does give you an unclipped signal.

Yes, you are apparrently agreeing exactly with what I already said:

Quote
...I _NEVER_ said that. I said it was impossible to get the original back...

...So you can do some expansion and get something reasonable, but never what was the original. ..


So, if you agree with me, why are you wasting my time?

Dynamics Compressor for FB2K

Reply #49
Quote
My original point still stands as strongly. You cannot invert a compressor for all practical purposes, you can't invert a one-way function, and you cannot invert something that maps multiple values into one.

You haven't come an inch closer to refuting this.

Quote

Fine. You have problems admitting when you're wrong. 


Is this another way of saying you have run out of arguments?

Fine, I'll continue if you want, and I'll ask again, will you explain how dbx didn't work then? If you can't explain how dbx's noise reduction technique did not work then there is no point in you saying one more word about it, because you are stating that it could not work, yet, it does.