Protools Mixer

[Greg Malcangi]

Hi Toby,

Thanks for your post. Just so I'm sure I've fully understood, could you (or Nika) clarify a point for me regarding question 8.

Let's say I've got a PT|HD session with say 72 inputs (auxes and/or disk tracks) and I'm just outputting a single stereo pair.

1. Does the first DSP create a stereo mix of it's inputs, truncate (or dither) this to 24bit and cascade the result to the second DSP which then uses this stereo input as any other 24bit source and mixes in the remaining inputs (at 48bit) before truncating/dithering back to 24bit for output.

or

2. The 2 DSPs are cascaded together, in effect acting as a single DSP. IE. The summing only occurs once (at 48bit), after the master fader for the output pair and there is no summing or truncation between the DSPs.

I'm fairly sure from my interpretation of your message Toby that answer 2 is the correct one. Could you please confirm this (or not) for my slightly addled brain? [img]images/icons/smile.gif[/img]

Thanks,

Greg

[Nika]

Greg,

It is not choice #2, I'm pretty sure.

The results from one chip have to be added to the results from the other chip - but each chip can only send 24 bits out and accept 24 bits in. So truncation has to happen at a preliminary step in order to get the information from one chip to somewhere else for the final summing.

What sounds most likely is that chip #1 sums and truncates and sends to chip #2 which sums the results of chip#1 with all of the other tracks on chip #2, then multiplies the final result by the master fader level, and truncates the end result to 24 bit.

Obviously, in the dithering mixer the word "truncates" can be substituted with the word "dithers" in all of the places above.

Nika.

[Digi Engineering]

Nika is essentially correct. When going over 59 (or 68) inputs and spanning DSPs, the "sub-mixes" are output at 24-bits and summed together on the final "summing" DSP.
I would love to have some more feedback on whether or not the "spanning" DSPs creates any audible problems - for example, do a mix with less than 60 inputs, bounce it to disk as a reference (2 mono files to make it easy for importing into a listening test session). Then, add some "dummy" tracks just to spread the mixer across DSPs (to make sure that non-zero audio is actually spreading across DSPs, try "interleaving" the tracks, like putting a "dummy" track in between every "real" track). Then try importing the 2 mixes into a simple 4 track session, and do some blind A/B listening (using the solo latching/groups) between the 2 mixes. Thanks!
-Toby

[Nika]

Toby,

I can confirm that the difference is audible, at least on the undithered mixer.

Nika.

[Jonny Atack]

Excellent -- thanks for the great info, Toby, and the great questions, Nika (and Greg and Kev, etc).

Nika, what did you hear? How did it sound different?

Jon
Capitol Studios Paris

[Zep Dude]

My conclusion based on this information:

For a final mix of a large session occupying two mixer chips, it would be better to bounce some of the submixes to stereo and deactivate those tracks in order to reduce the Mixer Plugin to one DSP?

Actually to be more accurate, you would have to deactivate enough tracks to eliminate the second Mixer plugin, THEN bounce the still active tracks to stereo, deactivate the bounced tracks, then reactivate the remaining tracks (so long as you don't create a second Mix plugin while doing so) and finish the mix.

But wait, if you then change the level or pan of one of the bounced stereo submixes (to anything other than level=0, pan = hard left/ right) this then puts it back through the TDM bus a second time -just the same as if it were on a second DSP to begin with (is this correct?).

So once the submixes have been created, they must stay at level=0 pan=hard left/right.

I think I've got it, but please advise.

Thanks

[Peter Duemmler]

Zep,

AFAI understand it, you would gain nothing with this technique, because your "sub-bounces" are 24bit max., too.

Peter

[Nika]

Peter,

Yup.

Nika.

[Zep Dude]

That means I'm more confused than I thought about this.

Here is my understanding:

If you print a submix of 16 drum tracks down to stereo, that mix has now gone through the TDM one time and then truncated back to 24 bit.

So if upon playback you keep that sub mix at level=0/pan=hard left/right doesn't that prevent it from going through the TDM bus on the way to the master? Therefore when it hits the master fader it's been through the bus only one time just like all of the remaining unbounced tracks that are comming through for the first time?

[Greg Malcangi]

Zep Dude, using your example it shouldn't matter how many times the audio is being passed through the TDM buss. Your audio (sub-mix) is 24bit and the buss is 24bit, no processing is happening to alter the 24bit resolution, so there is nothing to truncate or dither. In other words what goes into the TDM buss under your conditions should be bit for bit identical with what comes out.

Thanks Toby and Nika for the useful info. I only have two more questions! [img]images/icons/smile.gif[/img]

1. Toby you compared PT|HD's 68 inputs per mixer chip with PT|Mix's 59, so presumably this is the number of inputs PT|HD can handle per chip with the standard mixer. How many inputs per mixer chip can PT|HD handle using the dithered mixer?

2. Nika I too would like to know what affects you heard from the additional truncation that happens when spanning more than one mixer DSPs. When I've got a bit of free time in the studio I'll run the test for myself, but I'd still be interested in your opinion.

TIA,

Greg