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: My wave editor adds noise when saving (Read 58267 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

My wave editor adds noise when saving

Reply #100 –
You've made some confusing statements and your pictures don't help me much.

Quote
I have to amplify to flatten the clipped peaks

Flattened peaks are the easiest to see sign of clipping. Proper, undistorted audio peaks are not flat. They will be either sharp peaks or rounded peaks, with no flat spots. different editors display waveforms differently, however. Some show you what an good scope on the output of the DAC will show, some don't.

If you amplify beyond 0dBfs with fixed point files, you will get flatten peaks. That is clipping, That is bad.

If you amplify similarly in floating point, the peaks will disappear off the display. If the display allows you to reduce vertically, doing so will bring the peaks back on screen, showing you they are unchanged.

You first picture looks to me like waveforms with too much vertical zoom to show the peaks. If the vertical zoom was reduced, proper peaks might show up. The second picture shows badly clipped peaks. However, it does not seem to be on the same scale as the first so comparisons between the two are not very meaningful. I don't understand either what you are trying to accomplish or what you believe you are showing us with those graphics.

Jitter is a timing error, sample to sample. The clock is not running at a consistent rate.

If the recording clock had enough jitter to produce an audible impact, the recording is likely to be useless, a loss. Logically there isn't any way to repair the error. The only way the data could produce audio that sounds right would be for the output clock to exactly duplicate the recording clock's inconsistencies, with the data lined up exactly in synch with the timing variations that created each sample.

The jitter isn't in the audio data itself, so operations on the audio data can't effect it. Perhaps you are working on something else but calling it jitter?

My wave editor adds noise when saving

Reply #101 –
I hope this discussion doesn't run for another 100 posts over the OPs misconception about jitter.

Am I the only one annoyed that he still hasn't told us which editing program was adding dither upon saving?

My wave editor adds noise when saving

Reply #102 –
Okay, when I said jitter I was using that word as a generic statement of anything that frequently fluctuates to an annoying level. In this case, the cycle goes like this: 3 seconds of the audio are fine, then one second is nearly-silent then 3 seconds are good then one second is silent again, so I had to amplify and denoise those parts, but that is over and done with.

Onto the clipping: this new source was an MP3 re-encode of whatever the original was from, which was clipped. The high MP3 compression distorted the originally-flat clips, made them look like this: http://i55.tinypic.com/2hdy8ls.png

So I amplified that to flatten them again or the declipper won't work, but when I saved the peaks became distorted again, even with 32-bit. Is there a way I can, you know, keep my digital audio data as it is when I save, bit-for-bit?

My wave editor adds noise when saving

Reply #103 –
How are you saving, as mp3? This is a never ending cycle into madness. Mp3 may be fine as a final format -- after every change that is ever going to be done is done -- but should never be used when working on the data.

Floating point files are not changed by saving or reopening. Something else has to be happening. However, if you are just amplifying in floating point, you are not flatting the peaks (thanks goodness!). The peaks will disappear off the top of the display because they are now above 0dBfs, but their form will not change in the slightest. They will not have been clipped more by the amplification.

If you do “flatten the peaks” you are increasing the clipping distortion. Regardless of whatever  your declipping software does after that, the result will be worse than before.

Name the wave editor, so we can have some idea of what might be going on.

My wave editor adds noise when saving

Reply #104 –
No no no, I said my source is an MP3 file. I edit as wave and always save as wave, I'm not that dumb. Basically the MP3 compression screwed up the peaks, declipped them without actually... declipping. So I need them flat again or the declipper won't work, so yes I will need to clip it 1 or 2 DB higher. This is a very good declipper though and works really well. However, saving as 32-bit FP somehow preserved the distorted clips despite being amplified above 0 DB, and 32-bit INT simply distorted the flattened peaks again. What is the culprit here?

I'm using Izotope.

My wave editor adds noise when saving

Reply #105 –
Amplifying above 0dB in floating point does not clip. The shapes of the peak change not at all, they just raise above the top of the display.

If you set the peaks at 0dBfs, then convert to 16 bit, then amplify by the smallest amount that will clip to the extent you want, then convert back to floating point to continue working, you may be able to get what you want.

If the program unconditionally adds dither when converting to 16 bit, the variations at the peaks will be increased. This will not be helpful. So, if you can save the 0dB floating point step, then open in an editor where you control the options, you might be successful.

Open in this other editor. Convert to 16 bit without dither.  Amplify above 0dB to clip. Save. Make sure no dither is used at any time.

Open again where you want to declip, convert to floating point, do the job.

My wave editor adds noise when saving

Reply #106 –
I tried saving as non-dithered 16-bit but the once flat peaks still turn into http://i56.tinypic.com/21sbdf.png

What editor reliably has dither OFF by default so I can try it?

My wave editor adds noise when saving

Reply #107 –
That's normal and correct and has nothing to do with dither. The blue line is a so-called sinc interpolation of the discrete-time sample values (white blocks) and shows you how the analog waveform of your digital representation will/would look like after digital-to-analog conversion. Nothing to worry about.

By the way, the best thing to do here is to decode the MP3s to floating-point, then apply a compressor/limiter (or simply reduce the amplitude of the entire signal slightly) to get all peaks below 0 dBFS. No need to declip, the MP3 internally is not clipped!

Chris
If I don't reply to your reply, it means I agree with you.

My wave editor adds noise when saving

Reply #108 –
You were right, it was an error of the editor. I opened it in another one and they were perfectly flat. The samples themselves were straight but for some reason izotope displayed them as tiny sine beams.

The declipping was awesome in some areas but sucked in others. Now I know I can't complain about a process which by definition cannot restore something that isn't there, but upon examining the areas where it performed superbly and made the audio much more pleasant to listen to, it realistically extrapolated and added a best-fitting valley on top of those flat, severed hills. But in the areas where it didn't do much, or made the audio sound worse, the flat peaks have been actually deamplified and replaced with valleys pointing downwards. I'm not an expert, but I would think that if I were to guess the missing samples in a clipped peak, they would not have a lower amplitude than the peak. So I think I can realistically say that this declipper fails, not because it didn't want to do the impossible, but because it didn't even attempt to do anything a normal declipper would.

What are some of the highest quality declippers you guys would recommend?

In case a picture might aid my description. Sometimes it turned http://i51.tinypic.com/2vsok85.png into http://i53.tinypic.com/rt2buh.png when its supposed to do something like http://i53.tinypic.com/2d0i645.png Excuse the crappy drawing, but you get the point.

My wave editor adds noise when saving

Reply #109 –
Quote from: Inds on
You were right, it was an error of the editor. I opened it in another one and they were perfectly flat. The samples themselves were straight but for some reason izotope displayed them as tiny sine beams.

It's not the error of the editor, it's a feature. Digital sample values are shown with white squares, while D/A-reconstructed analog waveform is shown with blue line.

Quote from: Inds on
But in the areas where it didn't do much, or made the audio sound worse, the flat peaks have been actually deamplified and replaced with valleys pointing downwards.

Please post a before/after sample, pictures are not enough to judge.

My wave editor adds noise when saving

Reply #110 –
Quote from: Alexey Lukin on
Quote from: Inds on
You were right, it was an error of the editor. I opened it in another one and they were perfectly flat. The samples themselves were straight but for some reason izotope displayed them as tiny sine beams.

It's not the error of the editor, it's a feature. Digital sample values are shown with white squares, while D/A-reconstructed analog waveform is shown with blue line.

Quote from: Inds on
But in the areas where it didn't do much, or made the audio sound worse, the flat peaks have been actually deamplified and replaced with valleys pointing downwards.

Please post a before/after sample, pictures are not enough to judge.


ht#tp:/#/ww#w.me#diaf#ire.co#m/?85r#9brr7#26y2#7ky Remove the number signs

My wave editor adds noise when saving

Reply #111 –
This clipping is too extreme to be repaired by RX.
But maybe iZotope will take a note...

My wave editor adds noise when saving

Reply #112 –
Quote from: Alexey Lukin on
This clipping is too extreme to be repaired by RX.
But maybe iZotope will take a note...


Any declippers you would recommend, or is RX essentially the best out there?

My wave editor adds noise when saving

Reply #113 –
Quote from: Alexey Lukin on
It's not the error of the editor, ...
The topic is about adding (dither) noise when saving. I have to agree with the OP that it's not clear with iZoptope RX (haven't tried the new version 2 yet) which dither setting is used upon "save". The "save-as" window gives details, but IMO it would be better if the dither settings are clearly displayed somewhere (permanently?).
Kees de Visser

My wave editor adds noise when saving

Reply #114 –
Whenever you are saving to a format with less bit depth that the current RX window format (usually 32-bit float), white TPDF dither is applied. If you need to control that, Save As should be used. For example, when you open a 16-bit file, edit it (internal resolution grows to 32 bits at this point), and save — dithering is applied because the file is converted to 16 bits.

It's the same in RX 2.

My wave editor adds noise when saving

Reply #115 –
Quote from: Inds on
Any declippers you would recommend, or is RX essentially the best out there?

I know 2 that are good in different situations: iZotope and CEDAR.

My wave editor adds noise when saving

Reply #116 –
Quote from: Alexey Lukin on
For example, when you open a 16-bit file, edit it (internal resolution grows to 32 bits at this point), and save — dithering is applied because the file is converted to 16 bits.
That is not how my RX works (Mac, v1.2.1). When I use "save as" to disable dither, that setting is also used for subsequent "save" commands. I use this (feature?) often when I make tiny (short) modifications like manual spectral repairs and don't want to dither the whole file.

My wave editor adds noise when saving

Reply #117 –
Kees, you are right, I forgot how it works.
It actually remembers your most recently used dithering setting from the Save As dialog.