Protools Mixer

[Nika]

Guys,

After hearing reports from people that this was an audible "issue" within Protools, and confirmation from a customer of mine who was involved in one of the famed "PT vs. Oxford" tests, I set up an extreme stituation in Protools only to establish if there WAS a difference, checking both phase and listening, and in an extreme circumstance there was a "difference" between choices A and B. I'm not sure how this difference plays out in more complex mixes with more complex routings, etc. But I was able to identify the difference between choice A and B in an A/B/X test in a test scenario I built for this.

Nika.

[Extreme Mixing]

Nika,

Very cool infromation! How extreme was your test? And what about the 56 bit acumulator that Dave Lebolt talked about? I guess that's 56 bit per mixer chip with a 24 bit sum being passed from the first mix chip to the second. So much for the 48 bit summing with a 56 bit acumulator for pristine number crunching.

I'm assuming that we are better off keeping our levels at the stereo bus fairly near the top to get the best results. Should the louder elements of the mix be directed towards the mixer that gets summed to 24 bits amd passed along? I suspect that lower level information would suffer the most.

Has any of this changed with TDM II?

I'll be checking back. Again, great information. Don't let us misuse it.

Steve

[dave-G]

Gratitude Sidebar:

I just came across this topic, and in reading all the way through it thus far, It seems clear that Nika must not sleep very much.

I find it amazing that he finds the time and energy not only to investigate all this stuff himself, but to articulate it and share it so regularly, thoroughtly and educationally with us boobs.

Thanks, Nika. If there was a Gabe Wiener award, you'd have to be a nominee. . .

And thanks to Toby. I don't think I've ever read more concise descriptions of this stuff.

-dg

[Zep Dude]

By the way, a real world set up, assuming 2 aux sends per track, is 14 tracks plus 4 aux buses and a single master fader. Hey that's only 46 into two. I thought we were supposed to get 59 sources!

Unfortunately, I don't see any possibility of ever getting one of my mixes onto a single chip unless I switch to small ensemble live jazz recordings. [img]images/icons/frown.gif[/img]

When I have a chance I'll try some mixes on one vs two chips to see what exactly is the level of degradation.


So is this the answer to the long standing debates regarding the weakness of the Pro Tools summing bus? The issue of dither which everyone has fixated on is not really the culprit; it's the summing of the mix across multiple chips!

Toby, is this something that can be modified in the software or is it a hardware issue?

[Nika]

Zep,

It's only how many faders you have feeding a particular output. Multiple aux sends on each track don't affect this. It's only how many faders get to an output. If you have 14 mono audio tracks, 6 stereo audio tracks, and 5 stereo aux returns you're only at 36.

Nika.

[Zep Dude]

Nika,

Toby states it's "any source." My experience is from sitting there and adding tracks and sends and watching the system usage window. I see more of a drain when adding a send to a track than when adding a track itself. Strange. Try it yourself. With no sends, you get a ton of tracks, start adding sends and you're dead.

[Nika]

Toby,

It sounds to me like every time you create a destination a mixer gets opened. So if you create a destination of a bus and send 3 different tracks to that bus, it has to sum them. This requires some sort of mixer.

But that mixer is separate from the mixer used for the main outs. It is also a separate mixer from the mixer used for any other busses. It is also a separate mixer from the mixer used for any of the other physical outs (headphone outs, etc.).

Now what gets tricky is whether or not there can be multiple "mixers" on the same DSP chip. If there can then your situation is just as valid: Open up 14 faders on a track and it creates a 14 channel mixer. Then bus them all around and it creates a bunch of mini sub-mixers. Then try to add some more tracks to the first "mixer" and it has to spill it onto a separate chip because the first chip is full of all kinds of little mixers, and it doesn't dynamically reallocate them. So yes, a chip can handle a 59x2 mixer, but unless you create all 59 tracks at once while you start the session it is very likely that you won't get all 59 tracks on the one chip, no?

Did that make sense? This is getting more and more interesting the more I open it up. I'm curious as to whether or not I'm right.

Nika.

[meltedmediamusic]

Nika, thanks again for expanding this very interesting topic. As I'm not quite following this 100%, I'm not sure if my comments will make sense.

If mixer 1 gets summed and truncated (or dithered depending on which mixer you use)and fed to mixer 2 on the 2nd DSP, which is mixed and then summed with the 2nd mixers outputs, wouldn't it be safe to say that the tracks (channels) of mixer 2 to be of better audio quality due to only being truncated (dithered) once?

Assuming this to be correct, wouldn't it stand to reason that by creating a large template mixer to be opened at the start of each session, and recording your most important tracks on tracks that feed mixer 2 directly, you would obtain supperior mixes to ones done in the opposite manner (most important tracks feeding mixer 1).

As most of my projects involve a great number of tracks, especially vocal oriented projects, I always use the 64 channel setup as my default setup, even when I just need to record a few tracks.

Could this be the source of perceived lack of mix depth and imaging so many are talking about?

Tell us more, please!

Kenny Meriedeth
Melted Media Music

[Chris Townsend]

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

It sounds to me like every time you create a destination a mixer gets opened. So if you create a destination of a bus and send 3 different tracks to that bus, it has to sum them. This requires some sort of mixer.
<HR></BLOCKQUOTE>

Sure does. And of course each separate mixer takes additional cycles to do master fader arthimetic and other overhead, so obviously one 40 input mixer will take less DSP than four 10 input mixers.

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

Now what gets tricky is whether or not there can be multiple "mixers" on the same DSP chip. If there can then your situation is just as valid: Open up 14 faders on a track and it creates a 14 channel mixer. Then bus them all around and it creates a bunch of mini sub-mixers. Then try to add some more tracks to the first "mixer" and it has to spill it onto a separate chip because the first chip is full of all kinds of little mixers, and it doesn't dynamically reallocate them. So yes, a chip can handle a 59x2 mixer, but unless you create all 59 tracks at once while you start the session it is very likely that you won't get all 59 tracks on the one chip, no?
<HR></BLOCKQUOTE>

I think that is possible, although the mixer is going to try to keep the tracks on one DSP by moving other mixers, because it more efficient that way.

I'd also like to add that it is not particularly important that there is a measurable difference between different mixer track orderings. For instance, just about any floating point mixer is going to have a similar issue where you will get slightly different results depending on the order that the tracks are mixed (i.e. ideally you would want to sum the quietest tracks first). That is just inherent fact of floating point. Nonetheless, what really matters is whether the difference is even remotely audible. When the stereo mixer is not spanning chips the single truncation/dither noise floor is around -144dB. When mixer is spread across two DSPs two dither/truncate stages will occur, which will increase the noise floor by about 3dB to -141dB. Considering that the noise floor of even the best D/A converters is about -120dB this mixer noise floor source is completely swamped by the converters.

[Digi Engineering]

Some more good questions, guys. I'll try to answer. (And thanks Chris for the extra info)

1) Greg - the dithered stereo mixer on HD can go up to 56 x 2 on a single DSP.

2) Zep/Nika - regarding the DSP usage, again Nika has the right idea. Each mixer destination needs its own "stereo mixer". The stereo mixer plug-in itself only knows how to take a bunch of inputs and sum them to a single stereo output. For example, 24 tracks sent to output 1-2 creates a 24 x 2 mixer. Then, if you add a few sends to bus 1-2, another mixer must be created for bus 1-2. Route a couple of tracks to output 3-4, yet another mixer for output 3-4 - you get the idea. (Note: the surround mixer is different, so that it can route a single track to a 5.1 output for instance) The important point is to remember that the mixers will "grow" in such a way that keeps the "growing" mixer on the same chip until it can't fit. In our example, the main output 1-2 mixer will grow to 59x2 on a MIX system, while the smaller bus 1-2 and output 3-4 mixers will get moved to another DSP as you add more inputs to the output 1-2 mixer.
Another thing to remember - the master fader is "inside" the stereo mixer for each stereo pair (i.e. the same chip). The mixer's summing bus (at 56 bit) has a gain applied by the master fader for that output pair (which is done on the full precision signal), then the signal goes back out to TDM at 24-bit to whatever destination you are routing to (like output 1-2, bus 1-2, etc).

3) (I like putting numbers on these to make it a bit more readable ;-) Steve/Kenny- there is no reason to do anything like put "louder elements of the mix be directed towards the mixer that gets summed to 24 bits". When the mixer does actually get spread over DSPs(which only happens with more than 59 tracks all routed to a single destination, remember), each input gets treated the same. Each separate mixer (like 59 x 2 on the first DSP and another 59 x 2 on a second DSP, for a total of 118 x 2) gets routed to a 3rd summing mixer, whose job is just to take the outputs from the first 2 and sum them together. That way all tracks have the same thru delay when being mixed across different DSPs. And, there isn't any track that has a "better" path to the final output.

4) Hope this clears things up a bit more. I wonder if the other digital mixers have had such scrutiny? ;-)

-Toby