Dithering when recording buss output to track

[BackAtTheLab]

I read the white paper on how the 48-bit mixer handles wordlengths internally and how we shouldn't worry about any but the most infinitesimal degradation in the summing and dithering of the long internal words down to 24-bit output. Very informative and reassuring. I have further questions, however: I like to bounce groups of channels down to stereo in the mixer (as opposed to using "bounce to disk") by assigning groups of tracks and auxes to a buss, and then using that buss's output to feed a track (through another buss), which I record. I got the impression from the white paper that the output of that buss would be 48-bit, unless it feeds a physical output. Did I misunderstand? If the output of the buss is an interface output, is it then dithered automatically? Or does a dithering plug-in have to be inserted on it's output? Are the outputs of master faders the only things that get dithered automatically (if indeed they are)? If a buss's output is assigned both to a master fader and to another buss for the purpose of recording it on a track, is that track truncating the 48 bits to 24 (or 16 if it's a 16bit session)?

If the buss outputs are indeed 48 bit and I am recording their output onto new tracks, perhaps I should insert dither onto those busses while I am bouncing their output to avoid just truncating (especially in a 16 bit session which I am sometimes forced to work in). Then I would turn that dither off, since those busses are also feeding a master fader, which has its own dither.

I'm either truncating those bounces, dithering unnecessarily, or everything is cool. I need some informed feedback.

[pyramidhead]

those busses are also feeding a master fader, which has its own dither.

The master fader tracks does'nt have an automatic dithering process. You have to assign it.
The truncating process is useful to bounce for example a 24 bit session to a 16 bit mixdown.

[BackAtTheLab]

The master fader tracks does'nt have an automatic dithering process. You have to assign it.
The truncating process is useful to bounce for example a 24 bit session to a 16 bit mixdown.

The master faders' outputs are dithered down to 24 bits from the internal 48 bit word length, I believe. I don't think they just chop off the 24 least significant bits. Check out the white paper. Look at the diagram on page two. This is a quote from p. 3 paragraph 1:

Within the Pro Tools TDM mixing environment, when the dithered
mixer is used, dither is added to the 48-bit signal before it
is reduced to 24 bits to be placed back on the TDM bus and
sent to the DACs. Dither is not added on a track by track
basis, as this would produce unwanted accumulated noise.
Dither is only added once at the Master Fader output of the
mix bus so the total system dither noise is 3 dB at -144 dB.

You do need to assign dither to that output if you want to end up with a 16bit file where the 8 least significant bits weren't simply lopped off. That much is very clear. What is not addressed in the white paper is the question of what happens to the outputs of busses (not master faders) that are routed to the inputs of tracks for the purpose of bouncing.

[el Daniel]

The bus outputs are 24 bits dithered when the dithered mixer is used, so you don't have to insert any dithering plugin when bouncing to track in a 24 bits session.

The master faders are always present just before the bus output, it doesn't matter if you have created the master or not, it's the last gain stage in the 48 bits enviroment just before dithering and truncation to 24 bits at the bus output. If it is not created in your session it is there anyway at unity gain.
You have to have in account that the inserts at the masters faders are post fader and post dithering and truncation so the signal arriving to these inserts is the 24bit dithered signal.

Then, if your session is 16 bits and you want to bounce to track, you have to assing the outputs of your tracks to a pair of busses, then create a master fader for these busses and insert a 16 bits dithering plugin in the master fader(i.e: all your tracks outputs -> bus 1-2, master fader output -> bus 1-2, 16 bits dithering plugin in one of the master fader inserts, recording track input -> bus 1-2).

Therefore, your mix is routed to the bus in a 48 bits wordlength, it is summed at 48 bits, then is dithered and truncated to 24 bits, it goes to the 16 bits dither plugin inserted in the master fader, and the 16 bits dithered signal gets in the recording track.

I hope all this makes sense.

Daniel.

[BackAtTheLab]

The bus outputs are 24 bits dithered when the dithered mixer is used, so you don't have to insert any dithering plugin when bouncing to track in a 24 bits session.

The master faders are always present just before the bus output, it doesn't matter if you have created the master or not, it's the last gain stage in the 48 bits enviroment just before dithering and truncation to 24 bits at the bus output. If it is not created in your session it is there anyway at unity gain.
You have to have in account that the inserts at the masters faders are post fader and post dithering and truncation so the signal arriving to these inserts is the 24bit dithered signal.

Then, if your session is 16 bits and you want to bounce to track, you have to assing the outputs of your tracks to a pair of busses, then create a master fader for these busses and insert a 16 bits dithering plugin in the master fader(i.e: all your tracks outputs -> bus 1-2, master fader output -> bus 1-2, 16 bits dithering plugin in one of the master fader inserts, recording track input -> bus 1-2).

Therefore, your mix is routed to the bus in a 48 bits wordlength, it is summed at 48 bits, then is dithered and truncated to 24 bits, it goes to the 16 bits dither plugin inserted in the master fader, and the 16 bits dithered signal gets in the recording track.

I hope all this makes sense.

Daniel.

Ah. This is the type of response I was hoping for. Yes, indeed, it does make sense and it confirms what I suspected. I wanted some reassurance, however, and you have provided it. If any board administrators are listening, I recommend that this information be added to the white paper, or at some other handy location.

So. Master fader inserts are post fader AND dither? If they are post dither, it would seem to me that whatever processing that occurs in those inserts would add to the wordlength, resulting in an output that needs to be dithered/truncated again. Why would the inserts be post dither?

I feel like part of my question is answered, but as is often the case, the answers bring more questions.

[el Daniel]

So. Master fader inserts are post fader AND dither? If they are post dither, it would seem to me that whatever processing that occurs in those inserts would add to the wordlength, resulting in an output that needs to be dithered/truncated again. Why would the inserts be post dither?

Plugins have their own architecture, in TDM the signal going to the plugins has to leave the TDM bus so the input in the plugins is 24 bits, and the calculations inside the plugin depend on the pluging architecture not on the Protools mixer. Many TDM plugins have 48 bits processing and dither their output to 24 bit when returning the signal to the TDM bus. There are others that doesn't dither their output or the processing is not double precission... Therefore, the wordlength in the output of a plugin is always 24bits.

So you are recording the 24 bits signal being in the output of the last plugin and no re-dithering is needed because it is already a 24 bits wordlength no 48.
If the inserts in the master fader were prefader you couldn't dither to 16bits because, although the signal in the output of the plugin is 16 bits, it would be summed with the 24 bits dither added when the signal left the bus and you would have again a 24bit wordlength that would be truncated to 16bits without dither.
You can try this by inserting the 16 bits dither plugin in an aux track whitch have the inserts prefader, instead of a master fader (track->bus 1, input aux=bus 1, dither plugin in the aux, output aux=bus 2, recording track input=bus 2). Compare the result using both ways, and you'll see the diference. The session has to be 16bits and the dithered mixer has to be installed, with the non-dithered mixer there is no difference between both methods because no dither is added to the 16 bits signal. Try to do it lowering the fader of the track with the audio file about 70dB, so the signal recorded will be very low level, then rise the gain of the recorded files with the audiosuite plugin to a normal level to be able to listen to the results clearly.

Cheers,

Daniel.

[el Daniel]

[QUOTE]

If the inserts in the master fader were prefader you couldn't dither to 16bits because, although the signal in the output of the plugin is 16 bits, it would be summed with the 24 bits dither added when the signal left the bus and you would have again a 24bit wordlength that would be truncated to 16bits without dither.

I mean "If the inserts in the master fader were predithering..." not prefader.

[BackAtTheLab]

[QUOTE]

I mean "If the inserts in the master fader were predithering..." not prefader.

Not that I disbelieve you, but this doesn't seem to matter. If you used a plug-in to dither to 16bits and then that were lengthened to 24 bits and then chopped back down to 16--wouldn't you end up with the same dithered 16 bits?

So...double precision plug-ins all dither their output to 24 bits? The double resolution is only internal to the plug-in? So every single time you instanciate a plug-in you are dithering? Never mind...I just checked the white paper and it is so. Well, I guess I shouldn't be surprised. It's not as though the 48 bit precision can give you more resolution than the 24 bits you started with. That resolution just keeps us from losing resolution during summing and processing, if I understand correctly.

To clarify: the TDM bus is 24 bits. The 48 bit resolution only comes into play during summing and processing inside double precision plugs and exists in order to preserve the 24 bits we started with from end to end, yes?

Dave

[el Daniel]

Not that I disbelieve you, but this doesn't seem to matter. If you used a plug-in to dither to 16bits and then that were lengthened to 24 bits and then chopped back down to 16--wouldn't you end up with the same dithered 16 bits?

You will end up with the same dithered 16 bits if 24bits dither is not added after the plugin as happen if the plugin is inserted in the master fader because it is the 16 bits dithered signal with 8 zeros added, so truncating that signal only remove these 8 zeros. But if the 24 bits dither is summed with the 16bits dithered signal it becomes a 24 bits signal with not only zeros at the end and distortion is added when truncated to 16 bits.

So...double precision plug-ins all dither their output to 24 bits? The double resolution is only internal to the plug-in? So every single time you instanciate a plug-in you are dithering? Never mind...I just checked the white paper and it is so. Well, I guess I shouldn't be surprised. It's not as though the 48 bit precision can give you more resolution than the 24 bits you started with. That resolution just keeps us from losing resolution during summing and processing, if I understand correctly.

To clarify: the TDM bus is 24 bits. The 48 bit resolution only comes into play during summing and processing inside double precision plugs and exists in order to preserve the 24 bits we started with from end to end, yes?

Right, and you can't get more resolution than you have at the begining, so just preserving the resolution you started with is not such a bad thing.

See the plugins as if they were hardware inserts, the signal leaves the mixer and goes to the insert through a physical output and return through a physical input to the mixer, and to be able to pass through the outputs and inputs the wordlength has to be 24 bit.

Daniel.

[BackAtTheLab]

Right, and you can't get more resolution than you have at the begining, so just preserving the resolution you started with is not such a bad thing.

See the plugins as if they were hardware inserts, the signal leaves the mixer and goes to the insert through a physical output and return through a physical input to the mixer, and to be able to pass through the outputs and inputs the wordlength has to be 24 bit.

Daniel.

Fair enough. Thank you for your responses. Signing off.

Dave