Quick Dither and Bit Depth Question

[OldTimey]

Hi, i know dither has been discussed quite a bit over the years, but I have a new question regarding it.

Recently I've begun mastering in DP. I take a 24-bit file, put my plugins on it, and then print a 24-bit mix. I have been (up till now) using Adobe Audition to then convert the 24-bit file to 16-bit and applying dither there.

Then I thought, why not just print a 16-bit mix in DP (i don't use bounce to disk for final mixes/masters, just for ref mixes/master since i use UAD DSP, and I prefer to print in realtime by routing to a bus/audio track)

So I have a 24-bit mix in DP. To print a 16-bit mix I switch the audio engine to 16-bit. I then put a dither plugin at the end of the insert chain, (I use sonalksis dither) set to 16-bit etc. Route everything, and then hit record. The resulting file is a 16-bit file with the dither I applied. great.

So the question-should i uncheck DP's internal "Dither" when doing this?

Related question - I have a 24-bit file in DP, but set the audio engine to record to 16-bit, am I losing anything? In other words, is DP using the full 24-bit resolution of the 24-bit file, (within the 32-bit floating point engine) or does it on the fly truncate it to 16-bit before upsampling to 32-bit?

[FMiguelez]

Hi, i know dither has been discussed quite a bit over the years, but I have a new question regarding it.

Recently I've begun mastering in DP. I take a 24-bit file, put my plugins on it, and then print a 24-bit mix. I have been (up till now) using Adobe Audition to then convert the 24-bit file to 16-bit and applying dither there.

Then I thought, why not just print a 16-bit mix in DP (i don't use bounce to disk for final mixes/masters, just for ref mixes/master since i use UAD DSP, and I prefer to print in realtime by routing to a bus/audio track)

So I have a 24-bit mix in DP. To print a 16-bit mix I switch the audio engine to 16-bit. I then put a dither plugin at the end of the insert chain, (I use sonalksis dither) set to 16-bit etc. Route everything, and then hit record. The resulting file is a 16-bit file with the dither I applied. great.

So the question-should i uncheck DP's internal "Dither" when doing this?

No. They are unrelated. If you're referring to the Dither command in one of the menus, that's for other things that do not include BTD.

This is a recent dither thread where I cited and old post by Magic Dave talking about the Dither command:

http://www.motunation.com/forum/viewtop ... 40#p518045

Related question - I have a 24-bit file in DP, but set the audio engine to record to 16-bit, am I losing anything? In other words, is DP using the full 24-bit resolution of the 24-bit file, (within the 32-bit floating point engine) or does it on the fly truncate it to 16-bit before upsampling to 32-bit?

In that same thread, and in an older one, I asked that exact question (and a related one), and nobody ever answered it...

I don't know, and I've always wanted to know. But that's one of the things that only a couple of people can tell us about... They work at MOTU...

What I think, based on my experience, is that, as long as you have auto conversions off, DP should use the full resolution of your 24 bit file, even if you temporarily set the engine to 16 bits.
I switch the engine from 24 to 32 to 16 bits as well, depending on what I want to capture (like you), and I haven't ever noticed anything odd, so I'm operating under the premise that they are totally independent, and the files will use their native resolution, and be "upped" to 32 bFT whenever you do anything to them, regardless of the setting of the engine.

[EMRR]

I almost always work 88.2 or 96, I have to do a post-conversion regardless, so I print mixes 32 bit. I haven't considered the question of bit depth on it's own, wonder if there's real difference printing at end depth immediately versus printing at 32 and then doing a conversion from the soundbites window?

[OldTimey]

No. They are unrelated. If you're referring to the Dither command in one of the menus, that's for other things that do not include BTD.

This is a recent dither thread where I cited and old post by Magic Dave talking about the Dither command:

http://www.motunation.com/forum/viewtop ... 40#p518045

thanks for the reply, i will review the thread, but just to be clear, I'm not using the "bounce to disk" command, i'm routing playback to a new stereo audio track using internal busses and printing a master in realtime. not sure if that makes a difference or is covered in that link you posted but will check.

[FMiguelez]

No. They are unrelated. If you're referring to the Dither command in one of the menus, that's for other things that do not include BTD.

This is a recent dither thread where I cited and old post by Magic Dave talking about the Dither command:

http://www.motunation.com/forum/viewtop ... 40#p518045

thanks for the reply, i will review the thread, but just to be clear, I'm not using the "bounce to disk" command, i'm routing playback to a new stereo audio track using internal busses and printing a master in realtime. not sure if that makes a difference or is covered in that link you posted but will check.

Either way, it won't make a difference. That dither command in the menus is for merging, etc (I think).

I also would like to know how what one records (like you) gets truncated...
I mean, if DP is working everything at 32 bFP, regardless of your audio engine settings, and you print to 24 bits or 16 bits, does DP simply truncate from 32 bFP to 24 or 16 bits when you print your master in real time?
Or does it automatically add dither (with the dither command on in the menu) to mask the quantisation distortion?

MOTU?

[magicd]

When you set the bit depth in DP to 16 that means files you record after that will be recorded at 16 bits. Any 24 bit files in the project will still be 24 bit files. Internally DP does it's processing with 32 bit floating point precision, regardless of the bit depth of the audio files.

If you record in DP at 16 bit resolution DP will quantize the 32 bit float sample to 16 bits. Truncation happens after quantization so you are never truncating bits that contain audio data.

Dither from the DP menu applies to non-real-time processes such as fades. No dithering is happening as you record to 16 bits. So if you want to finish the master in DP at 16 bits, yes you should use a dither plug-in on the output of the mix.

Dave

[Shooshie]

My understanding is that no matter what you select as your soundbite resolution or output resolution, you are working within DP with a 32 bit mix. When you print it to media, you get a converted, dithered version of it in the resolution that you specify.

For any given set of tracks that you are calling final, dither once. For any track that may be converted again, or processed again in any way (other than analog) do NOT dither.

Shooshie
[edit — I didn't see Magic Dave's post. I defer to anything he said. I'd been distracted before finishing the post.]

[Shooshie]

Dither from the DP menu applies to non-real-time processes such as fades. No dithering is happening as you record to 16 bits. So if you want to finish the master in DP at 16 bits, yes you should use a dither plug-in on the output of the mix.

Dave

I did not know this. So, you're saying that the dither in the menu in DP is only applicable to fades and non-real-time processing? So, to dither your final 16 or 24 bit output, you'd have to consciously select "dither" on the dialog, right? And if you don't, it doesn't dither.

Or are you saying that you have to use a plugin to dither on the final output?
This is unclear to me. There are several dithering possibilities, and I'd like to know the do's and don'ts of each of them, including when it is done automatically, and what requires manual intervention to either dither or not dither.

Geez... just a little while ago I thought I fairly well understood dithering in DP. Now, I'm not so sure.

Shooshie

[FMiguelez]

Dither from the DP menu applies to non-real-time processes such as fades. No dithering is happening as you record to 16 bits. So if you want to finish the master in DP at 16 bits, yes you should use a dither plug-in on the output of the mix.

Dave

I did not know this. So, you're saying that the dither in the menu in DP is only applicable to fades and non-real-time processing? So, to dither your final 16 or 24 bit output, you'd have to consciously select "dither" on the dialog, right? And if you don't, it doesn't dither.

Or are you saying that you have to use a plugin to dither on the final output?
This is unclear to me. There are several dithering possibilities, and I'd like to know the do's and don'ts of each of them, including when it is done automatically, and what requires manual intervention to either dither or not dither.

Geez... just a little while ago I thought I fairly well understood dithering in DP. Now, I'm not so sure.

Shooshie

You must add a dither plugin if you want it in your final bounce. It is NOT automatic and there are no "dialogs" anywhere.
The dither option in the menu has nothing to do with bouncing with/without dither.

[OldTimey]

When you set the bit depth in DP to 16 that means files you record after that will be recorded at 16 bits. Any 24 bit files in the project will still be 24 bit files. Internally DP does it's processing with 32 bit floating point precision, regardless of the bit depth of the audio files.

If you record in DP at 16 bit resolution DP will quantize the 32 bit float sample to 16 bits. Truncation happens after quantization so you are never truncating bits that contain audio data.

Dither from the DP menu applies to non-real-time processes such as fades. No dithering is happening as you record to 16 bits. So if you want to finish the master in DP at 16 bits, yes you should use a dither plug-in on the output of the mix.

Dave

thank you dave for your expertise, as always...

what if my final output file is 24-bit...in your opinion should i use a dither set to 24 bits at the end of my plugin chain?

[FMiguelez]

If you record in DP at 16 bit resolution DP will quantize the 32 bit float sample to 16 bits. Truncation happens after quantization so you are never truncating bits that contain audio data.

But isn't truncation the result of quantizing?
If DP's 32 bFP stream is recorded to 16 bits with no dither, then 16 bits are being truncated, thus creating quantizing distortion, correct? Is that what DP does?
Wouldn't adding dither to the recording (printing in real time) process be the ONLY way to avoid quantization (truncation) noise?

Dither from the DP menu applies to non-real-time processes such as fades. No dithering is happening as you record to 16 bits. So if you want to finish the master in DP at 16 bits, yes you should use a dither plug-in on the output of the mix.

Dave

Ok. So to avoid truncation, one must add a dither plugin when recording to a lower bit depth, just like when bouncing.
That makes it clear now. Thanks!

[magicd]

what if my final output file is 24-bit...in your opinion should i use a dither set to 24 bits at the end of my plugin chain?

If your output is set to 24 bits do not add dither.

As mentioned DP uses 32 bit floating point precision in the mix bus. Plug-ins can go up to 64 bit float depending on the design.

Audio interfaces don't deal with floating point numbers. So playback output from DP has to be within the 24 bit sample range. If you don't bring the output level of DP back within the 24 bit sample range you'll get clipping. So, no dither is required for 24 bit output.

Dither in the menu refers to non-MAS functions. This includes crossfades and I'm fairly certain if you convert audio from the Soundbites window mini-menu and choose 16 bit, dither is applied there (if it's checked in the menu). Bounce to disk and record are MAS functions, so dither from the menu has no affect in those processes.

Any time you reduce bit depth in DP it's done by rounding. So for example if you use the MW Limiter and set it's output to 16 bits, there will still be a 24 bit sample coming out of the plug, but any data in the bottom 8 bits of the original 24 bt sample has been rounded into the top 16 bits. If you then print to a 16 bit file the bottom 8 bits are truncated, but because there is no data in those bits, no audio is lost.

Rounding introduces quantization error, which manifests itself as distortion. Dither is the process of randomizing the bottom bits of the sample. The result is the distortion is masked by noise. Noiseshaping is a dither option. Noise shaping biases the frequency response of the dither noise. The idea is that you want the dither noise to be in the same range as the audio content so the audio content does it's best job of masking the noise.

To get to know dither, put the MW Limiter on the Master Fader and start lowering the output bit depth. When you get to six or four bits the dither noise will be obvious. Now switch the noise shaping modes and you'll hear that difference.

I don't always use dither on my final output. I recently did a track that was designed to be a TV theme song. The client wanted it to be at max volume all the way through. Dynamic range was about 10 db. There was no content in the bottom bits. The track doesn't end with a reverb tail or fade. I can hear the dither when it's on and since dither didn't make the over-compressed track sound any better, I turned it off.

One thing for sure, you do not add dither until you print the final master mix. If there is any further processing of the audio downstream you do not want to have dither noise in that signal.

Dave

[Shooshie]

Dither from the DP menu applies to non-real-time processes such as fades. No dithering is happening as you record to 16 bits. So if you want to finish the master in DP at 16 bits, yes you should use a dither plug-in on the output of the mix.

Dave

I did not know this. So, you're saying that the dither in the menu in DP is only applicable to fades and non-real-time processing? So, to dither your final 16 or 24 bit output, you'd have to consciously select "dither" on the dialog, right? And if you don't, it doesn't dither.

Or are you saying that you have to use a plugin to dither on the final output?
This is unclear to me. There are several dithering possibilities, and I'd like to know the do's and don'ts of each of them, including when it is done automatically, and what requires manual intervention to either dither or not dither.

Geez... just a little while ago I thought I fairly well understood dithering in DP. Now, I'm not so sure.

Shooshie

You must add a dither plugin if you want it in your final bounce. It is NOT automatic and there are no "dialogs" anywhere.
The dither option in the menu has nothing to do with bouncing with/without dither.

Well, yeah, I got what Dave said. But your interpretation is ever-so-slightly off, which can make all the difference. This may take a bit of work, but let me see if I can straighten this out so that we understand what it does and what it does not do. First, here is the Manual on Dithering:

• Dither
  Digital Performer’s mixing and processing engine employs 32-bit resolution. When 16-bit or 24-bit digital audio first enters Digital Performer, its bit depth (resolution) is raised to 32-bits. From then on, as the audio is mixed and processed, it remains at 32 bits until it leaves Digital Performer. The Dither command, when checked, makes Digital Performer use dither whenever it must reduce the bit depth of audio. [NOTE THE FOLLOWING STATEMENT FOR REFERENCE IN A MOMENT] For example, if you bounce your mix to disk as a 16-bit, 44.1 kHz AIFF audio file to later burn a CD, Digital Performer converts your mix, which internally is being processed in 32 bits, to 16 bits during the bounce operation. If the Dither command is checked, dither is employed to help reduce the inherent quantization distortion that occurs when audio bit resolution is reduced.
  page 592

The Dither command it is referring to is the first item on the Audio menu:

• 

This is the Dither command referred to in the above paragraph of the DP manual. I believe this must be the command that Dave is talking about which only affects non-real-time processes. That would include conversions, would it not? It also includes Bounce To Disk, according to the statement in the manual that I asked you to notice. That's not a real-time process, either.

Here the subject comes up again in the section about File Conversion:

• The Convert Sample Format section of the Audio File Conversion, as shown below in Figure 74-1, allows you to change the sample format of an audio file or soundbite. You can change to 16 bit integer, 24 bit integer, or 32 bit floating point. Like Digital Performer’s other file-based “constructive” DSP processes, sample format conversion can occur in the background.
  
  Dither
  Digital Performer incorporates dither when converting to a lower bit depth— rather than truncating the extra bits — to ensure the smoothest possible conversion.
  Page 862

• 

Now, if you go back to page 861, you will see the (above) dialog that I thought had a Dither checkbox in my previous post. It does not, for I was mistaken. It appears that dither is automatic when DOWNsampling. (rather than truncating the sample) So you don't need a checkbox in the dialog. But the statement on page 592 that says "if the Dither command is checked." Does that mean that you can turn off dithering in a file conversion by unchecking the command? It does not refer to the Audio Menu when discussing file conversions on page 862, but if you had just come from page 592, you would be led to believe that it's the key to whether or not the conversion uses Dithering.

And yet it makes it sound like it's automatic. The manual says DP uses dither when downsampling. It does not say that it depends on the Dither command (Audio Menu) being checked. I presumed earlier that there was a checkbox on the conversion dialog, but since there is not, I am led to believe that it's automatic: When downsampling there is dither. When upsampling there is not. This is probably a non-real-time process which Dave was describing, but does the Dither command in the Audio menu have the ability to turn it off? I'm guessing that it does, and that the manual just fails to mention it in this section.
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Then there is the Masterworks Limiter interface, whose interface has a checkbox. You can read about it on page 54 of the Plugins Guide:

• Changing bit depth with dither
  Bit depth or resolution is the number of ones and zeros used to describe a digital audio sample. Commercial CDs use 16-bit samples. MOTU software has the ability to record 16 bit, 24 bit, or 32 bit floating point samples. Under some circumstances, it may be desirable to have lower sample resolution, such as 8 bit web audio or 12 bit sampler playback.
  
  It is possible to change the bit depth of digital audio. Reducing the bit depth adds quantization distortion. To overcome this, dither can be applied. Dither is a small bit of noise that is applied the digital audio as it is quantized. This actually reduces the amount of distortion introduced by quantization, and has a more pleasing sound, particularly for softer passages in material with a wide dynamic range. The disadvantage of adding dither is an increased noise floor. Noise-shaping spectrally shapes the dither noise so that it is less noticeable.

Here's the problem. There is the Audio Menu dither, and the MW Limiter dither. Either can be on or off. Examine the following pictures. Look for the Dither checkbox on the MW Limiter interface, and the Dither command in the Audio Menu:




Next, there is the Quan Jr. Plugin:

• QUAN JR.
  Quan Jr. is a simple dithering and quantizing plug- in. It is useful if you want to use the dithering features of MasterWorks Limiter, but do not need the extra features and added overhead. This is especially important for mixing in surround if you are running low on CPU power.
  
  Quan Jr. also makes a handy low resolution effects generator. Turn off dithering and set the quantization to three or four bits and get ready for some grunge.
  For more information on dithering and quantization, “Changing bit depth with dither” on page 54.

The same possibilities of menu vs. plugin exist here, too:


In the cases where both the menu and the plugin are set for DITHER, are we getting two ditherings? If one is on and the other is off, are we getting any dither? One dither? Is one taking care of real-time playback, and the other for Bounce to Disk? Are both coming through when bouncing to disk? Would you use the Quan plugin to dither when recording a real-time "bounce" file? Conversely, would you use the Audio Menu's Dither command when actually Bouncing to Disk? And in the latter case, would you need to turn off the plugin? I'm sorry, but I think I have a right to be a little confused.

Exactly what's happening here? These aren't earth-shaking questions, but the answers are essential to precision, so I think we need to be precise when discussing it, and not take for granted that any of us knows anything. The facts are important. I can only hope that I've been getting it right all these years!

Shooshie

[Shooshie]

Once again, Dave posted while I was composing my reply. But I'll let it stand.

Shoosh

[FMiguelez]

I see what you mean, Shoosh.

The dither in the Audio menu, its non-existence in the Convert Audio File dialog box, its presence in those plugins, and the old aparent errors of information in the manuals since at least DP4, certainly haven't helped at all to understand how dither works in DP



I'm glad Dave, who has been touched by unicorn magic, has shed some light on the topic