Protools Mixer

[Nika]

I have a question for those that really know Protools well. If you can help me out with this that'd be great. I'll put all of the questions in italics and number them. For the sake of these questions we are talking about the Mix systems and the non-dithering mixer.

If I use 32 voices or less in Protools only one DSP chip gets used to do all of the summing for the Protools mixer.

If I use more than 32 voices (and up to 64) in Protools then I have to go under the "playback engine" settings and select that two DSP chips get used for the Protools mixer. The questions here are all about how this works.

I am assuming that each DSP chip sums it's portion of the mixer and then they get summed together on one of the two chips. 1.Is this correct? 2.Does this mean that each DSP chip is only capable of summing 32 tracks?

OK, now let's discuss what happens if I'm set on "2 DSP's, 64 voice". 3.Does any form of truncating happen there if I only use less than 32 voices? 4.What would happen if I created a 4 track mix and assigned the voices to be 1, 2, 3, 4; would this mix the tracks differently than if I assigned them to voices 1, 2, 40, 41? It certainly looks possible that the different voices would get summed on different DSP's, truncated to 24 bits, and then summed together if I did it this latter way, which would be a different result than if the 4 tracks were simply summed and truncated at once.

OK, now let's discuss aux busses and how they might change this paradigm. 5.If I have my mixer set on 1 DSP (32 voice) and I have more than 32 total channels, including aux busses, how does this mixer handle this? 6.What would happen if I just created a mixer with 80 aux tracks and no audio tracks, how would the mixing get done on the DSP's? 7.Does having auxes on your mix change the number of DSP's that get used for the mixer? This goes back to the first questions, namely 8.how does the summing happen if multiple DSP's are used for a session?

Most of what I do is classical music using 12 tracks or less, so I haven't gotten much into the large format mixer templates that can get set up. If someone would explain these things to me I'd be most appreciative.

Thanx!
Nika.

[Greg Malcangi]

Hi Nika,

The only thing I can remember reading on this subject posted by Digi was just over 2 years ago in the "Sound Quality" thread.

Somewhere else in that thread Dave Lebolt said that the "largest mixer that can be created on a single DSP with this mixer is 47x2, as compared to 59x2 with the undithered version that we send out today".

Like you I'd like to know exactly what goes on behind the scenes.

Greg

[Nika]

Avery just went down for a nap, so I'm going to do some testing.

Nika.

[Nika]

Results of testing:

Using the non-dithered mixer there is definitely some truncation going on in different ways for the same mix depending on which tracks you have where.

To answer my own questions:

1. It appears so.
2. According to Greg's quote from Dave Lebolt, no. It's more like 57 tracks. I didn't get into the specifics of quantity of tracks - just the effects - in my testing.
3. YES!! It really depends on how many TOTAL tracks you have (including aux's). I was able to set up an 8 track mix with the 32 voice engine and get it to perform differently depending on how many aux channels I had and in which order I called new tracks up. (Create 4 audio tracks, then 32 auxes, then 4 more audio tracks, then some more auxes. Then do it again, but switch the orders and different results).
4. It didn't seem to follow which voice they were on, but WHAT ORDER they were called up into the session.
5. If you exceed the number of total tracks that the mixer can handle on one DSP chip (57?) then it starts to put them onto a separate channel and sums them individually, causing an extra truncation process, before a total summing.
6. Ref. above. It would sum the first 57(?) together and truncate to 24 bits, then sum the next 23(?) together and truncate to 24 bits, then sum the results and truncate to 24 bits. I suppose it's also possible that it sums the 23(?) together and truncates to 24 bits and then sums that with the untruncated 48 bit sum of the remaining tracks, eliminating one of the truncations.
7. Yes.


I am pleased that I have now isolated that you can get two mixes to function differently depending on the order that you call the tracks up into the screen!

Nika.

[Mark Haliday]

Very interesting...

Three questions :
- Without going into too much detail, what kind of testing procedure do you use ?
- Any recommendation based on your findings ?
(like lets do only 8 tracks mixed to mono again... [img]images/icons/wink.gif[/img] )
- any idea where the dithering is happening in the dithered mixer ? (before summing ?, at each insert slot ?)

One thing I noticed by ear, plugins *really* make a difference. And I suspect that the way the bit depth is returned to 24 bit is partly responsible for this.
This could also partially account for the Sony plugs sounding so good and taking so much DSP.

[dstagl]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:<HR>Originally posted by Nika:
It really depends on how many TOTAL tracks you have (including aux's). I was able to set up an 8 track mix with the 32 voice engine and get it to perform differently depending on how many aux channels I had and in which order I called new tracks up. (Create 4 audio tracks, then 32 auxes, then 4 more audio tracks, then some more auxes. Then do it again, but switch the orders and different results).<HR></BLOCKQUOTE>

Did you try closing and reopening the session? Does that make a difference. If I creat 4 audio tracks, then 32 auxes, then 4 more tracks, close the session, and reopen it, does it change depending on the order I have my tracks setup in the mixer from left to right? If so is there an ideal ordering of tracks?

Dave

[dstagl]

OK, I just thought of a workaround for getting your tracks up in the order you want them when you reopen a session. You can always deactivate all tracks, and then reactivate them in the order you want them in. So what is the ideal order?

Dave

[Kev]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:<HR>Originally posted by dstagl:

You can always deactivate all tracks, and then reactivate them in the order you want them in.
So what is the ideal order?
<HR></BLOCKQUOTE>

Is this enough or do you have to assign the voices in order of priority? Or is it the edit window top to botom, if you get my drift.

So once you have decided you have indeed got some control of priority, like dstagl asked , how do you go about chosing an order?.. and what effect does subgrouping/aux have on it all if they down the priority?

All the tracks are important. [img]images/icons/confused.gif[/img]

[Digi Engineering]

We need to clear up some misunderstanding about what is being called the "Mixer". Remember that the playback engine on MIX (either 32 or 64 voices) only determines how many DSPs are being used to provide disk tracks into the TDM system. These "engine" DSPs (either 1 or 2 on a MIX system) don't do any mixing; they just provide a way to get the disk tracks (either 24 or 16-bit, depending on the bit depth you have selected) into the TDM system. Then, the "Mixer plug-in" is responsible for summing signals to any pair of outputs. The non-dithered stereo Mixer plug-in can sum up to 59 inputs to a pair of outputs on a single MIX dsp (68 x2 on the HD hardware) It doesn't know anything about where these signals have come from - they could be normal audio (disk) tracks or aux inputs - any 24-bit source from TDM. There is no truncation until the final outputs (for example, the signal leaves the mixer plug-in at output 1-2 to go directly to the engine dsp connected to the 888/24 interface). In the dithered mixer case, dither is added at the final output, before truncating back to 24-bits.This should answer Nika's questions 1,2,3, and 5, . Question 4 also - again, it doesn't matter to the Mixer plug-in what voices are being used, so the answer is that nothing is different as far as the Mixer is concerned. Regarding aux tracks (questions 6 and 7), remember that the Mixer doesn't see any difference between audio (disk) tracks and aux tracks . Re: question 8, when you need to sum more than 59 sources (68 on HD) and use more than 1 DSP, we simply "cascade" the mixers into a final summing stage, at which point the final result is truncated back to a 24-bit output. Nika is basically correct in his assessment (aside from the numbers should be 59/21) of point 6 above. Remember that every pair of outputs being used will "instantiate" a different Mixer plug-in, though they can share the DSP with other Mixer plug-ins. For instance, routing 10 tracks to outputs 1-2 and another 10 tracks to outputs 3-4 will need 2 different Mixer plug-ins - one would be a 10x2 mixer for outputs 1-2, and another 10x2 mixer for outputs 3-4. The mixer always grows by using up the maximum DSP for a single output pair, so you always grow up to 59x2 (68x2 on HD) before "splitting" across DSPs. It's possible that the order of adding tracks, etc. can affect when the "splitting" does need to occur, but remember that any individual output pair doesn't "split" until it has more than 59 inputs (or 68 on HD). Hope this helps!
-Toby Dunn, Digi SW

[Nika]

Toby,

Wow! Very helpful!

Sorry, guys, I've been off recording for 3 days and missed all of your questions. Toby answered them, it appears.

One last point - if you create a mixer and it gets big enough that it creates a second mixer (say, 60 tracks), the tracks all appear to end up assigned to a particular mixer. If you then delete tracks 2 through 59, leaving only tracks 1 and 60 it still sees this as two mixers and truncates appropriately.

As to the question of shutting down and reopening the session, I have no idea. But duplicating a track might be a viable alternative?

Cheers!
Nika.