BootlegZone

Made for and by bootleg lovers
It is currently Fri Jul 21, 2017 12:39 pm

All times are UTC + 1 hour [ DST ]




Post new topic Reply to topic  [ 165 posts ]  Go to page 1, 2, 3, 4, 5 ... 7  Next
Author Message
Remake
 Post subject: Downconverting & dithering discussions/comparisons
PostPosted: Fri Feb 20, 2009 4:40 am 
Offline

Joined: Tue Feb 03, 2009 5:21 pm
Posts: 437
NOTE: This thread was split from this topic in the "Audiophile Factory" area: viewtopic.php?f=19&t=22613&st=0&sk=t&sd=a
-DLedin

Audioarchivist wrote:
Anyway, I'm looking into SoX tonight - how easy (or hard) is it to learn and use?
It's quite easy.

I was going to open a thread with the following instructions, but could not find the way to do it so here they go:

Installing SoX om PC and using it is a breeze provided you have installed Foobar2000 (http://www.foobar2000.org) before. In other words, we are going to use Foobar as the GUI.

Just download this file: http://www.hydrogenaudio.org/forums/ind ... opic=67373 unzip Its a plugin of the SRC programmed for use with Foobar. Unzip the file and you will find a file called 'foo_dsp_resampler.dll' and two txt files. Take the .dll file copy it and go to 'Program Files' > 'foobar2000' and paste it in the 'components' folder you'll find there.

Next, run Foobar, click on any music file you have in the play list (if you don't have any yet, just drag an audio file into the list) to select it and then right click over it. This will open a menu. At the bottom of the menu you'll find an option that says 'Convert,' run the mouse pointer to it and another menu will open to the right. Click on '...' This will open the 'Converter Setup' window. In it you will program the use of the SoX resampler.

Let's say you want to have an option where you can resample down to 44.1 kHz anytime you want. In order to set an option for this choose your target format on the Output tab. I suggest WAV, which is usually set by default. Then click on the 'Processing' tab and look for the text Resampler (SoX) in the right hand window. Select it and click on the <= arrow. This will move it to the left window. Select it there and you will notice that the 'Configure Selected' button will highlight.

Click on the 'Configure Selected' button and choose the following:
Samplerate: 44100
Quality: Very High
unclick 'Allow aliasing'
mark 'Steep filter'

and, finally, under 'Phase response' choose 'linear'

Now, hit 'OK', then save the preset by giving it a name on the top of the next window. Then, click on 'OK' and you are done!

Now, if you want to create a resampling preset for, say, 96000, so that you can upsample from a lower sampling frequency or downsample from - say - 192, go through all the previous steps and just choose 96000 on the Samplerate window.

Now, if you want Foobar to use the same resampler for playback, go to File > Preferences > Playback > and under the latter click on DSP Manager. This will open a window. Follow the same steps as you did on the similar window before and choose, for example, a 96000 Samplerate this time. Once you have configured that, all music you play on Foobar will play back at 96 kHz.

It's all as easy as this. :)

_________________
"Don't take yourself so seriously. No one else does."


Top
 Profile  
 
dodog1
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Fri Feb 20, 2009 5:04 am 
Offline
User avatar

Joined: Mon Dec 29, 2008 10:10 pm
Posts: 1005
Well, geez, you make it all easy with a GUI. Damn you.

But, you'll feel a sense of accomplishment if you do it via line-command. Right? (hears crickets chirping). Yeah, you may enjoy the GUI more, plus you'll be interfacing with the best Audio Player/converter EVER created (totally my biased opinion).

_________________
Keep your mouth shut, you're squawking like a pink monkey bird. And I'm busting up my brains for the words.


Top
 Profile  
 
pbthal
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Fri Feb 20, 2009 5:29 am 
Offline
User avatar

Joined: Sat Nov 15, 2008 9:48 pm
Posts: 1100
dodog1 wrote:
Well, geez, you make it all easy with a GUI. Damn you.

But, you'll feel a sense of accomplishment if you do it via line-command. Right? (hears crickets chirping). Yeah, you may enjoy the GUI more, plus you'll be interfacing with the best Audio Player/converter EVER created (totally my biased opinion).


Oh please you can't compare Audio Player/converters, this thread is out of control


Top
 Profile  
 
BobNET
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Fri Feb 20, 2009 5:31 am 
Offline

Joined: Wed Mar 28, 2007 5:55 pm
Posts: 71
dodog1 wrote:
sox -3 -r 96000 -c 2 input.wav -2 -r 44100 -c2 output.wav dither -fmodified-e-weighted
That's about half of my processing chain right there...


Top
 Profile  
 
dodog1
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Fri Feb 20, 2009 6:09 am 
Offline
User avatar

Joined: Mon Dec 29, 2008 10:10 pm
Posts: 1005
pbthal wrote:
dodog1 wrote:
Well, geez, you make it all easy with a GUI. Damn you.

But, you'll feel a sense of accomplishment if you do it via line-command. Right? (hears crickets chirping). Yeah, you may enjoy the GUI more, plus you'll be interfacing with the best Audio Player/converter EVER created (totally my biased opinion).


Oh please you can't compare Audio Player/converters, this thread is out of control


Once you 'drop your box of '45's, I'll ABX them on different players. Of course, that's after I properly dither down those 24/96 files you're known to give out. :good:

_________________
Keep your mouth shut, you're squawking like a pink monkey bird. And I'm busting up my brains for the words.


Top
 Profile  
 
Remake
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Fri Feb 20, 2009 6:22 am 
Offline

Joined: Tue Feb 03, 2009 5:21 pm
Posts: 437
pbthal wrote:
dodog1 wrote:
Well, geez, you make it all easy with a GUI. Damn you.

But, you'll feel a sense of accomplishment if you do it via line-command. Right? (hears crickets chirping). Yeah, you may enjoy the GUI more, plus you'll be interfacing with the best Audio Player/converter EVER created (totally my biased opinion).


Oh please you can't compare Audio Player/converters, this thread is out of control
Not quite, the Foobar audio player's resampler is based on the same source code as the SoX resampler. I have compared downsampled files done with SoX with ones done with iZotope's sample rate converter and the latter's results are smoother and keep more of the soundstage of the source file. But, hey, SoX is free and gives 'cleaner' results than quite a few others.

Try it and see. :)

Thus, the only case I can think of where I would not use the SoX one is if I had the iZotope sample rate converter.

_________________
"Don't take yourself so seriously. No one else does."


Top
 Profile  
 
pbthal
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Fri Feb 20, 2009 6:53 am 
Offline
User avatar

Joined: Sat Nov 15, 2008 9:48 pm
Posts: 1100
dodog1 wrote:
pbthal wrote:
dodog1 wrote:
Well, geez, you make it all easy with a GUI. Damn you.

But, you'll feel a sense of accomplishment if you do it via line-command. Right? (hears crickets chirping). Yeah, you may enjoy the GUI more, plus you'll be interfacing with the best Audio Player/converter EVER created (totally my biased opinion).


Oh please you can't compare Audio Player/converters, this thread is out of control


Once you 'drop your box of '45's, I'll ABX them on different players. Of course, that's after I properly dither down those 24/96 files you're known to give out. :good:



Tsk tsk..we cannot compare apples to oranges...citrus wars


Top
 Profile  
 
Audioarchivist
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Sat Feb 21, 2009 12:59 am 
Offline
User avatar

Joined: Mon Dec 18, 2006 8:17 pm
Posts: 92
Location: Denman Island, B.C.
pbthal wrote:

Tsk tsk..we cannot compare apples to oranges...citrus wars

Comparing apples to what? We were talking Led Zeppelin, not The Beatles!!! haha
Compare Capitol's to EMI/Parlophone's - what?


Top
 Profile  
 
Remake
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Sun Feb 22, 2009 5:19 am 
Offline

Joined: Tue Feb 03, 2009 5:21 pm
Posts: 437
Have any of you here experimented with the software we have commented above? If so, what is your take on it?

_________________
"Don't take yourself so seriously. No one else does."


Top
 Profile  
 
DLedin
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Sun Feb 22, 2009 5:46 am 
Offline
User avatar

Joined: Sat Apr 06, 2002 11:36 pm
Posts: 2503
Location: Not here
Remake wrote:
Have any of you here experimented with the software we have commented above? If so, what is your take on it?
As you know, I'm just getting started to working with this. What I'd like to do is something like what you suggested where I take a track from something I've already ripped and downconvert it to 16-bit via the foobar/SoX application. Then I'll compare it to the same track that had been downconverted via my normal way which was through Audition. I'd also like to post the results here for others to compare. There have been times when I just don't hear a subtle difference but others do. If that happens here, I'll begin using this other method for downconverting with my next project. I may even go back and re-convert some of my other projects if this proves out and the difference is worth it. It's easy enough for me to do since I work all in 32-bit (float) space up to the point of preparing the 24 & 16-bit audio for uploading. I've saved all of my final 32-bit files so they're ready to downconvert all over again if I want.

Just to clarify, using the foobar/SoX will downconvert and dither for me? (or do I dither somewhere else, like Audition?)

Finally, any suggestions for a test track from my already existing projects?

-DLedin


Top
 Profile  
 
TaterBones
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Sun Feb 22, 2009 5:56 am 
Offline
User avatar

Joined: Thu Jan 17, 2008 11:07 pm
Posts: 88
Location: East of Austin
foobar/SoX will downconvert & dither.

Perhaps something from ATMP or A Space in Time?

I like the settings Remake suggested.
Remake wrote:
Click on the 'Configure Selected' button and choose the following:
Samplerate: 44100
Quality: Very High
unclick 'Allow aliasing'
mark 'Steep filter'


P.S. foobar's ABX utility is one way to compare the different versions (foobar/SoX vs. Audition).


Top
 Profile  
 
markshan
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Sun Feb 22, 2009 6:05 am 
Offline
User avatar

Joined: Tue May 15, 2007 3:30 am
Posts: 4422
TaterBones wrote:
foobar/SoX will downconvert & dither.



Well, this is quite helpful. I misunderstood and thought I would have to incorporate two new steps. Maybe checking this out won't be as much work as I thought.

_________________
Did you use a compass to get out of this place?


Top
 Profile  
 
Remake
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Sun Feb 22, 2009 6:27 am 
Offline

Joined: Tue Feb 03, 2009 5:21 pm
Posts: 437
DLedin wrote:
Just to clarify, using the foobar/SoX will downconvert and dither for me? (or do I dither somewhere else, like Audition?)

Although I think I saw some dithering setting on the full command-line-managed version of SoX, there is no such thing in the Foobar SoX plugin and, even if there were, I would no use it, especially because I imagine it would be of the lowly white noise type.

As I said before, the best dithering algorithm out there is iZotope's MBIT+.

Other than that, I would go with the triangular one in Audition.

_________________
"Don't take yourself so seriously. No one else does."


Top
 Profile  
 
DLedin
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Sun Feb 22, 2009 6:54 am 
Offline
User avatar

Joined: Sat Apr 06, 2002 11:36 pm
Posts: 2503
Location: Not here
Remake wrote:
DLedin wrote:
Just to clarify, using the foobar/SoX will downconvert and dither for me? (or do I dither somewhere else, like Audition?)

Although I think I saw some dithering setting on the full command-line-managed version of SoX, there is no such thing in the Foobar SoX plugin and, even if there were, I would no use it, especially because I imagine it would be of the lowly white noise type.

As I said before, the best dithering algorithm out there is iZotope's MBIT+.

Other than that, I would go with the triangular one in Audition.


OK, so does this mean that after running the track(s) through the Foobar/SoX setup, that I need to then run the track(s) through Audition's triangular dithering (since I'm not going to fork out the $$$ at this time for something that has MBIT+)?

Also, for others trying to work with this, when I ran my first test track, starting with 32-bit (float) audio, I ended up with 44.1/32-bit. I found that by going back into the "Converter Setup" at Foobar and selecting the "Other" tab, I could select the bit-rate for the outputted file to 16-bit. OR....should I be ending up with a 44.1/32-bit file after running through Foobar/SoX and I then take that 44.1/32 file to Audition to convert to 44.1/16-bit with triangular dithering?

-DLedin


Top
 Profile  
 
dodog1
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Sun Feb 22, 2009 4:07 pm 
Offline
User avatar

Joined: Mon Dec 29, 2008 10:10 pm
Posts: 1005
DLedin wrote:
Also, for others trying to work with this, when I ran my first test track, starting with 32-bit (float) audio, I ended up with 44.1/32-bit. I found that by going back into the "Converter Setup" at Foobar and selecting the "Other" tab, I could select the bit-rate for the outputted file to 16-bit. OR....should I be ending up with a 44.1/32-bit file after running through Foobar/SoX and I then take that 44.1/32 file to Audition to convert to 44.1/16-bit with triangular dithering?


That's funny. I tried this with my 24/96 version of ATMP (DLedin's I believe (actually, a question on that, where I got this "Magoocus" was the UL'er. Did he just post your drop?)), and I got a 24 bit, 44.1 file. I thought I did something wrong. I've been a bit busy and hadn't had time to putz with this since then.

Oh, and as far as Sox is concerned, regarding dithering and filtering, from the PDF manual:

Quote:
Dithering deliberately adds a small amount of noise to the signal in
order to mask audible quantization effects that can occur if the output sample size is less than 24
bits. The default (or with the −t option) is Triangular (TPDF) white noise. The −r option can be
used to select Rectangular Probability Density Function (RPDF) white noise. Noise-shaping (only
for certain sample rates) can be selected with −s. With the −f option, it is possible to select a particular
noise-shaping filter from the following list: lipshitz, f-weighted, modified-e-weighted,
improved-e-weighted, gesemann, shibata, low-shibata, high-shibata.


So, it appears that the Triangular dither in SoX is of the "lowly white noise" type. How is this different than a triangular dither in Audition? Also, what's the difference between a dither like "triangular" and a noise-shaping filter?

Maybe DLedin or Pbthal would be nice enough to provide a 24-bit sample file that we could all work with, that way we could actually make comparisions with each other (SoX vs. Izotope, vs. UV22, etc.)? Or, maybe a 32-bit sample might work best. Once all these different tests were compiled, maybe DLedin could start a new thread posting them all and having people post what they feel the differences are (similar to the solid-state vs. tube-cut thread). Even better, maybe all those 24-bit lovers out there who always claim that we're missing out on 16-bit (who also have access to $$$ equipment) could ABX the files (like Tater suggested) and post what it IS we're actually missing.

_________________
Keep your mouth shut, you're squawking like a pink monkey bird. And I'm busting up my brains for the words.


Top
 Profile  
 
DLedin
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Sun Feb 22, 2009 5:33 pm 
Offline
User avatar

Joined: Sat Apr 06, 2002 11:36 pm
Posts: 2503
Location: Not here
dodog1 wrote:
...I tried this with my 24/96 version of ATMP (DLedin's I believe (actually, a question on that, where I got this "Magoocus" was the UL'er. Did he just post your drop?)...


Actually, if you used the Magoocus sourced audio, yes that was from me but there's a flaw in the audio from which Magoocus worked with. Last year at about this time I had someone smarter than me (and that's a lot of folks) determine that any of the 24-bit audio I'd been putting up was not true 24-bit audio. It was actually 16-bit/96kHz audio that had been "padded" to 32-bit audio to start with. This was due to a setting within Audition that I was able to subsequently change. So Magoocus' great work on his DVD-A was done with the "padded" audio. I have since gone back & re-done a number of my transfers, including ATMP, so that the hi-rez is true 24-bit audio. If you want to start with true 24-bit audio for that title, you need to grab the current 24-bit version that's out there.


dodog1 wrote:
...Maybe DLedin or Pbthal would be nice enough to provide a 24-bit sample file that we could all work with, that way we could actually make comparisions with each other (SoX vs. Izotope, vs. UV22, etc.)? Or, maybe a 32-bit sample might work best. Once all these different tests were compiled, maybe DLedin could start a new thread posting them all and having people post what they feel the differences are (similar to the solid-state vs. tube-cut thread). Even better, maybe all those 24-bit lovers out there who always claim that we're missing out on 16-bit (who also have access to $$$ equipment) could ABX the files (like Tater suggested) and post what it IS we're actually missing.


I can certainly provide a 32-bit track that's ready for anyone to play with. Here's a list of completed projects to date: viewtopic.php?f=19&t=21937. Any of them that indicate 24-bit audio I can provide the 32-bit (float) for. Just give me a suggested track and I can make it available. I also agree that, at some point, we should split this thread. A good spot to start the new thread would be the post where Remake originally wanted to start a new topic, when he posted the Foobar/SoX instructions. We probably should put it in the techical section of the forum.

-DLedin


Top
 Profile  
 
Remake
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Sun Feb 22, 2009 6:20 pm 
Offline

Joined: Tue Feb 03, 2009 5:21 pm
Posts: 437
DLedin wrote:
OK, so does this mean that after running the track(s) through the Foobar/SoX setup, that I need to then run the track(s) through Audition's triangular dithering (since I'm not going to fork out the $$$ at this time for something that has MBIT+)?

Also, for others trying to work with this, when I ran my first test track, starting with 32-bit (float) audio, I ended up with 44.1/32-bit. I found that by going back into the "Converter Setup" at Foobar and selecting the "Other" tab, I could select the bit-rate for the outputted file to 16-bit. OR....should I be ending up with a 44.1/32-bit file after running through Foobar/SoX and I then take that 44.1/32 file to Audition to convert to 44.1/16-bit with triangular dithering?

-DLedin
It is normal to end up with a downsampled file at the target frequency resolution, but at the bit depth of the original. Downsampling is for changing the sampliing frequency, dithering is for when you want to change the bit depth to 16 bits. If you are going from 32 bits to 24 you can truncate, although some people also dither that, which I would not normally recommend.
dodog1 wrote:
as far as Sox is concerned, regarding dithering and filtering, from the PDF manual:

Quote:
Dithering deliberately adds a small amount of noise to the signal in
order to mask audible quantization effects that can occur if the output sample size is less than 24
bits. The default (or with the −t option) is Triangular (TPDF) white noise. The −r option can be
used to select Rectangular Probability Density Function (RPDF) white noise. Noise-shaping (only
for certain sample rates) can be selected with −s. With the −f option, it is possible to select a particular
noise-shaping filter from the following list: lipshitz, f-weighted, modified-e-weighted,
improved-e-weighted, gesemann, shibata, low-shibata, high-shibata.


So, it appears that the Triangular dither in SoX is of the "lowly white noise" type. How is this different than a triangular dither in Audition? Also, what's the difference between a dither like "triangular" and a noise-shaping filter?
Yes and not exactly. Here's what the iZotope dithering PDF I suggested reading says about this:

"In Ozone, the noise source is selected by the Type button. Type 1 is a rectangular probability density function, Type 2 is triangular probability density function and MBIT+ is an iZotope proprietary dither technology.

Without getting into too much detail, MBIT+ is (if you have Ozone) the preferred dither type. Beyond our own technologies, a triangular probability density function (TPDF, or Type 2 in Ozone) is a commonly used simple dither function (over other shapes such as rectangular, Gaussian, etc.). The statistical/objective level of a TPDF source is slightly higher than a rectangular function (+4.77 dB), but the perceived "annoyance" or loudness of a TPDF function is lower. We mention this fact because this is one of several cases in dealing with extremely low level noise where the mathematics go against the human perception. And since we're making music for people to listen to and not to measure, we go with the perception.

So, in general, our recommendation is to use the MBIT+ algorithm if you have Ozone, or a TPDF or triangular Type 2 if you don’t have Ozone (but have that option in another dither plug-in)."

_________________
"Don't take yourself so seriously. No one else does."


Top
 Profile  
 
DLedin
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Sun Feb 22, 2009 7:07 pm 
Offline
User avatar

Joined: Sat Apr 06, 2002 11:36 pm
Posts: 2503
Location: Not here
Remake wrote:
DLedin wrote:
OK, so does this mean that after running the track(s) through the Foobar/SoX setup, that I need to then run the track(s) through Audition's triangular dithering (since I'm not going to fork out the $$$ at this time for something that has MBIT+)?

Also, for others trying to work with this, when I ran my first test track, starting with 32-bit (float) audio, I ended up with 44.1/32-bit. I found that by going back into the "Converter Setup" at Foobar and selecting the "Other" tab, I could select the bit-rate for the outputted file to 16-bit. OR....should I be ending up with a 44.1/32-bit file after running through Foobar/SoX and I then take that 44.1/32 file to Audition to convert to 44.1/16-bit with triangular dithering?

-DLedin
It is normal to end up with a downsampled file at the target frequency resolution, but at the bit depth of the original. Downsampling is for changing the sampliing frequency, dithering is for when you want to change the bit depth to 16 bits. If you are going from 32 bits to 24 you can truncate, although some people also dither that, which I would not normally recommend...


OK, so what are my steps? Run the initial 96/32-bit (float) audio through Foobar/SoX at the suggested settings and end up with 44.1/32-bit (float) audio. Then go to Audtion and convert to 16-bit using triangular dithering?

Regarding getting from 32-bit (float) to 24-bit I typically just use the "save as" command in Audition and select "24-bit packed int (type 1, 24-bit)". Am I correct in assuming this is simply truncating from 32 to 24-bit audio with no dithering? When saving/coverting this way, I do not notice any additional processes going on and it's actually quite fast.

-DLedin


Top
 Profile  
 
Remake
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Sun Feb 22, 2009 7:24 pm 
Offline

Joined: Tue Feb 03, 2009 5:21 pm
Posts: 437
DLedin wrote:
OK, so what are my steps? Run the initial 96/32-bit (float) audio through Foobar/SoX at the suggested settings and end up with 44.1/32-bit (float) audio. Then go to Audtion and convert to 16-bit using triangular dithering?
Yes.

DLedin wrote:
Regarding getting from 32-bit (float) to 24-bit I typically just use the "save as" command in Audition and select "24-bit packed int (type 1, 24-bit)". Am I correct in assuming this is simply truncating from 32 to 24-bit audio with no dithering? When saving/coverting this way, I do not notice any additional processes going on and it's actually quite fast.

-DLedin
This (type 1, 24-bit) is the normal way of saving a WAV file in Audition. Just make sure that dithering is unchecked in the 'Options' menu before saving from the 'Save as' menu.

_________________
"Don't take yourself so seriously. No one else does."


Top
 Profile  
 
DLedin
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Sun Feb 22, 2009 7:46 pm 
Offline
User avatar

Joined: Sat Apr 06, 2002 11:36 pm
Posts: 2503
Location: Not here
Remake wrote:
...This (type 1, 24-bit) is the normal way of saving a WAV file in Audition. Just make sure that dithering is unchecked in the 'Options' menu before saving from the 'Save as' menu.
Very good. Thank you.

-DLedin


Top
 Profile  
 
DLedin
 Post subject: Re: Led Zep IV Japan / technical discussions on drops + bitrates
PostPosted: Sun Feb 22, 2009 8:27 pm 
Offline
User avatar

Joined: Sat Apr 06, 2002 11:36 pm
Posts: 2503
Location: Not here
Remake:

OK, one last detail and then I think I'll be ready to post examples of what I've usually been doing and then of going through the Foobar/SoX and then Audition.

Once I've got my 44.1/32-bit file from Foobar/SoX, I then import it into Audition. Here's the settings I typically use for dither:
Image
Are there any setting changes I should be making to this before converting from 32-bit to 16-bit?

-DLedin


Top
 Profile  
 
Remake
 Post subject: Re: Downconverting & dithering discussions/comparisons
PostPosted: Sun Feb 22, 2009 10:18 pm 
Offline

Joined: Tue Feb 03, 2009 5:21 pm
Posts: 437
Here are some quotes about dithering from different parts of the Audition Instruction Manual to answer your doubts:

Dither Transform Results Enables dithering when processing effects such as FFT Filter or Amplify. Adobe Audition does most processing using arithmetic greater than 16-bit, with the results converted back to 16-bit when complete.

During this conversion, dithering provides a higher dynamic range and cleaner results, with fewer distortions and negative artifacts.

With this option enabled, you can approximate 24-bit performance with 16-bit data, because dithering increases dynamic range by about 10 dB. If this option is disabled, audio data is truncated to 16-bit during reconversion, and more subtle information is lost.

The drawback of dithering is that each operation adds a small amount of noise at the quietest volume levels.

However, the trade-off between dithering (adding noise) and truncating the data (creating artifacts and limiting dynamic range) generally favors dithering, so it’s best to select this option.

Use Symmetric Dithering Enables symmetric dithering. In most cases, it’s best to select this option. If it is not selected, a DC offset of one-half sample is added each time data is dithered. Symmetric dithering has just as many samples added above zero as below zero. By contrast, non symmetric dithering just toggles between 0 and 1.

Sometimes in a final dither, this may be desired to reduce the bit range of the dither.However, both methods produce identical audible results in every respect.

Dither Amount For Saving 32-Bit Data To 16-Bit Files Enables dithering when pasting 32-bit audio to 16-bit audio.

The default value of 1 (bit) enables dithering, while a value of 0 disables dithering. For semi dithering, choose a value of 0.5.

With dithering, you can approximate 24-bit performance with 16-bit data, because dynamic range increases by about 10 dB. This allows signals as quiet as –105 dB.

With this option enabled, you can approximate 24-bit performance with 16-bit data, because dithering increases dynamic range by about 10 dB.

Change the bit depth of a file The bit depth of a file determines the dynamic range of the audio. For example, 8-bit resolution provides 256 possible unique volumes, while 16-bit resolution provides 65,536 possible unique volumes. Adobe Audition supports up to 32-bit resolution. You can raise the bit depth of a file to gain a greater dynamic range, or you can lower the bit depth to reduce the file size.

Work at the 32-bit level when processing audio, even if you plan to downsample to 16- or 8-bit for output. You’ll achieve better results than at the 16- or 8-bit level. The only time you may want to work at the 16- or 8-bit level is when processing a very large file on a slow computer.
1 Choose Edit > Convert Sample Type.
2 Select a bit depth from the Resolution list, or enter a custom bit depth in the text box.
3 When you select a lower bit depth, Adobe Audition provides dithering options to help reduce noise and distortion.

Set the following options as desired, and click OK:
Although dithering introduces a small amount of white noise, the result is far preferable to the increased distortion that you would otherwise hear at low signal levels. Dithering also lets you hear sounds that would otherwise be masked by the noise and distortion limits of 8-bit audio.

Enable Dithering Enables or disables dithering. If dithering is disabled, Adobe Audition truncates the audio,
meaning that unused bits are simply chopped off and discarded. The result gives a crackly effect that fades in and out on very quiet audio passages.

Dither Depth (Bits) Sets the bit amount of dithering to be applied. In general, values of 0.2 to 0.7 give the best results without adding too much noise. Note, however, that as this value is lowered, other unwanted harmonic distortion noise appears. (Lower values are usually OKay if you also apply Noise Shaping.)

p.d.f. (probability distribution function)Controls howthe dithered noise is distributed away fromthe original audio sample value.

Usually, Triangular p.d.f. is a wise choice because it gives the best trade-off among SNR (Signal-to-Noise ratio), distortion, and noise modulation. Triangular p.d.f. chooses random numbers that are generally closer to 0 than to the edges –1 or +1 (that is, the chance of 0 being chosen is twice as great as the chance of 0.5 or –0.5).

Noise Shaping Determines the placement when you move noise to different frequencies. The same amount of overall noise is present, but you can place less noise at one frequency at the expense of placingmore noise at another.

You may also specify that no noise shaping is used.

Different curves result in different types of background noise. The type of curve to use depends on the source audio, final sample rate, and bit depth. By introducing noise shaping, youmay be able to get away with lower dither depths to reduce the overall background noise level, without introducing a lot of unwanted harmonic noise.

Note: In general, there are no really good noise shaping curves for audio at 32 kHz or lower.With audio at those sampling frequencies, try the different curves to see if they help, and just choose the one that sounds the best.

dithering (definition) Adds small amounts of noise to a digital signal so that very quiet audio remains audible when you convert from a high bit resolution to a lower one (for example, when converting from 32-bit to 16-bit). Without dithering, quiet audio passages such as long reverb tails may be abruptly truncated.

_________________
"Don't take yourself so seriously. No one else does."


Top
 Profile  
 
DLedin
 Post subject: Re: Downconverting & dithering discussions/comparisons
PostPosted: Mon Feb 23, 2009 12:19 am 
Offline
User avatar

Joined: Sat Apr 06, 2002 11:36 pm
Posts: 2503
Location: Not here
OK.

That's a lot of good info but my question for you, Remake, is would you change any of the settings that I'm showing in the image above? About the only area I'd be in doubt on is the dither depth, which is currently set at 1. Should that be backed off to, say, .7 or even lower?

The reason I'm asking this now is that I don't want to put up a comparison and have someone come back & say, well, you should have used such & such a setting. Let's set this for optimum, based upon what I'm using.

Thanks,
-DLedin


Top
 Profile  
 
Remake
 Post subject: Re: Downconverting & dithering discussions/comparisons
PostPosted: Mon Feb 23, 2009 1:20 am 
Offline

Joined: Tue Feb 03, 2009 5:21 pm
Posts: 437
Hi DLedin,

I understand your interest to get it right the first time, but the truth is that you may have to end up using your ears. Why, you might ask? Basically because it has been some time since I last used Audition for dithering (although my settings for dithering are exactly like the ones you show here) because I've been using MBIT+ dithering for a while now.

This means that I have never dithered an SoX-downsampled file with Audition. So, I don't know if these dithering settings need some tweaking in order to get the best out of an SoX file. But, I do believe that these settings you show now are a good starting point.

_________________
"Don't take yourself so seriously. No one else does."


Top
 Profile  
 
DLedin
 Post subject: Re: Downconverting & dithering discussions/comparisons
PostPosted: Mon Feb 23, 2009 6:12 am 
Offline
User avatar

Joined: Sat Apr 06, 2002 11:36 pm
Posts: 2503
Location: Not here
OK folks, here we go. This is where the rubber meets the road.

I've uploaded three 16-bit tracks, all sourced from the same 32-bit audio file. All of the tracks have some difference regarding the conversion to 16-bit. This will be a blind test and the tracks are simply labelled A,B & C. If you hear some differences or have a preference, please post your observations and, if at all possible, be specific if there's a particular area that you can hear the difference(s).

As none of these tracks have been created using MBIT+ dithering, I would not consider this comparison complete unless some kind individual who has MBIT+ would grab the 32-bit source file, which I'm also making available here, and convert it to 16-bit, 44.1 audio and then post the results in this thread.

So here's the three 16-bit tracks for comparison:
http://www.sendspace.com/file/w041im

...and here's the 32-bit source file:
http://www.sendspace.com/file/jgxn2z

The track used for this comparison is "I'd Love To save The World" from the "A Space In Time" UK LP that I dropped recently.

EDIT: ...And here is the track from the same source with MBIT+ dithering:

http://www.sendspace.com/file/71imwk

-DLedin


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 165 posts ]  Go to page 1, 2, 3, 4, 5 ... 7  Next

All times are UTC + 1 hour [ DST ]


Who is online

Users browsing this forum: No registered users and 2 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Jump to:  
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group