Stereo Dithered Mixer

[priorytools]

It can't be post fader because the fader is pre-insert on the master
Matt

I think the insert on a master fader is post fader. .

[Matt_G]

Sure, but it should be taken into account that the best noise floor each track could have after tracking is the noise floor of your converters and, since that noise is much louder than dither noise, that would be the noise your compression and limiting will bring up.

100 24 bits dither generations per track would be needed to equal the noise of your great -120dBFS noise floor converters and therefore building up the noise floor by only 3dB. So given a session with 100 plug-ins in each track (I know there are only 5 inserts) your final noise floor would be only 3dB louder than in a session with no plug-ins at all, independently of the amount of compression and limiting you use.

Cheers,
Daniel

This still only applies to the plug-ins that actually do dither to 24bits on output. I also don't buy the whole 'I like to think of TDM plugs like hardware inserts', Pro Tools is a computer based DAW designed to run plug-ins so you can run everything in the box if you wish, so why base it on hardware? It doesn't have to have the same physical limitations as outboard digital hardware. HD is supposed to be 'High Definition' so why have 24bit I/O on the plug-in bus? It just doesn't make sense when LE has the abilitly to hand over at higher precision & without clipping.

I do agree about the bigger problem not being accumulative 24bit dither, the bigger problem is the potential to clip or distort through the 24bit plug-in I/O & I still believe that 24bit truncation distortion can also affect audio quality if the plugs are not dithered correctly.

Matt

Matt

[Aisle 6 Recording]

I'm with you Matt.
We pay a premium for HD with the capability of 48bit precision when all the while the the signal is going 48/24/48 etc every time you add a TDM plug. When the cheaper LE system hands over at the same word length right the way through it's mixer.
Come on Digidesign the TDM system is being left behind and we owners are paying for it at every turn. Plugs are double or even triple the price and the core system is 10 times that of an LE system. We are sold that idea on the premise that we are professionals. Or are we just SUCKERS!!!
Scott.

[bionic]

Suckers?

I'd love to hang around, but I am off to work on my Mac Pro LE system with lovely floating point headroom ; )
Joking aside, I recently did some tests with Matt on identical PT sessions running in host mode RTAS on my rig and then TDM versions on his HD rig. Mine had virtually no distortion while the TDM version sounded choked and distorted - obviously this is based on a few things and already we all know there are things to compensate for on the TDM side, but I was actually surprised at how big the difference was...

...and I am coming from a much more musical then technical place here. The difference wasn't just technical (even though it is in reality) - it affected the music in a negative creative way. It felt more laboursome to come to the same end as the LE side of things.

I feel a bit cut on the behalf of all TDM users. The price that is paid for those systems isn't being realised on the fundamental back end at the moment. Computers are flying ahead in terms of processing power compared to the core cards. The main thing is latency compensation and the I/O support on TDM, but then again, most of my plugs have no latency now and my I/O path is so simple, I haven't had any major issues yet.

Take care everyone,

CJ.

[RadioFilter]

I've been experimenting with this and have measured the noise floor creeping up as I add plug-ins. Fine, I can accept that. I've noticed some particular problem plug-ins, though, namely ReVibe (!!) and TL EveryPhase. When I insert a ReVibe on an Aux I notice a sizable jump in the noise floor. Using four ReVibes adds so much noise it's unacceptable. I'm stunned to find this with a Digidesign plug. Maybe I'm doing something weird so I plan to continue experimenting.

[Brandonx1]

I guess this scenario makes sense, but then that dither would come undone after the first plug-in is inserted on the master & would need to be re-done after each plug-in. Meaning all the hardwork to dither your 48bit signal to 24bits is useless if it truncates through the proceeding plug-ins. It still doesn't prevent the audio truncating to 24bits on individual channel plug-in inserts either.

Hi Matt,
How many points is dither added in the dithered mixer? Let's say in a mix you have audio tracks, sub master that is an aux,and a master bus.
After the last plug-in on the audio track, 1 point of dither. On second thought, no dither hear. No distortion is introduced by going from 24 to 48 bits, correct? 1 point of dither before the first plug in on the aux , and 1 point of dither before the last plug in on the master. That's a total of 3 points of dither in a standard mix.

This doesn't seem like anything too agregous to me. If this fixes distortion in the TDM mixer I can deal with a very very little bit of noise.
Best,
Brandon

[roberts]

Basically the more plug-ins you have in your session the more truncation distortion is being added to your audio. For one or 2 plug-ins this is less of a problem but for bigger sessions this weak link will become much more audible.

Please don't state your opinion to be facts.
That statement is totally not true, I mix large scale sessions with well over 250+ plug-ins on a single session
I never hear any distortion ...

[Triode71]

I think it's time Digidesign upgraded the precision of the plug-in bus, if the mix bus can run at 48bit why can't this be maintained right through the plug-in bus as well? I understand that it would use more DSP but sonic precision does come at a cost. At least we should have the option to turn this on just like the stereo dithered mixer plug-in.

Matt

Matt,

I think you're right. I truly think the mixer has a good ways to go-- not that it hasn't come a long way, mind you! Digi has obviously been working hard to make things right. I hope they continue to do so. I would love to see the option to use a 48-bit-wide TDM bus. Ideally, only at the very, very last instant should *anything* be dithered & truncated back to 24-bit.

Of course, there's one problem here, and it's the old "elephant in the room": not all TDM plug-ins are processing at 48-bit internally. Somehow, to make things transparent and to make the feature usable, the programmers might be forced to re-write code to deal with this. It may be easy, or it may be more difficult; not being a programmer, I don't know. If any Digi engineers could chime in on this, it could help us all decide whether to get excited about the possiblity of a 48-bit TDM bus, or maybe just re-adjust our Christmas wish list to ask for more incremental and subtle improvements.

It would be great if a Digidesign engineer could jump in here and help us understand these problems better. I'd like a 48-bit-wide TDM architecture myself, but how do you do it? I see lots of problems on the way to a 48-bit TDM architecture.

For example, let's assume we suddenly have a new 48-bit TDM architecture that:

A) receives a 48-bit word from the 48-bit mixer to be handed to a TDM plug-in, and
B) also transmits a 48-bit word output from a TDM plug-in to the 48-bit mixer.

Assuming all the above,

1.a. Would the plug-in designers/ programmers now be forced to develop newer versions of 48-bit TDM plugs that *don't* dither down to 24-bit?

b. Would all TDM plug-ins processing at 24 bits now have to multiply by two to get a 48-bit word to output to a 48-bit TDM bus?

c. Would TDM plug-ins have to be re-written to handle the 48-bit word input?

d. If not, how would existing 24-bit TDM plug-ins handle a 48-bit word input? Wouldn't there have to be truncation from the 48-bit mixer to the 24-bit TDM plugin's input?

2. Another approach could be to have each plug-in equipped with an output toggle adjustability comparable to the tick boxes on plug-ins like Maxim, or Waves' L2/ L3. If the user had checked the "Use 48-bit-wide TDM Bus" option in "Preferences", let's say, he would then check "48-bit output" on all his plug-ins. Would this be feasible?

3. On the other hand, perhaps the plug-in designers and the Digi guys would have to work together to develop some protocol for the plug-ins to speak to the TDM bus (or the DSP chip, or whatever) to tell it whether 24-bit or 48-bit wordlengths are being used within the TDM plug-in and therefore how this signal should be input/output to/from the TDM bus, and then adapt accordingly. Would this be feasible?



Anyway, I hope it can be done, because a 48-bit fixed-point signal path throughout would probably just sound better. Higher resolution, if it's reasonably feasible, even at the cost of some DSP cycles, should at least be an option in a system which calls itself "HD".

[Matt_G]

Please don't state your opinion to be facts.
That statement is totally not true, I mix large scale sessions with well over 250+ plug-ins on a single session
I never hear any distortion ...

Dear Roberts, these are indeed the facts whether you can hear the truncation distortion or not is irrelevant because it's maths 101. The facts are this... the plug-in bus has a 24bit I/O architecture & is not internally dithered between any of the plug-in interconnects (including the output back to the mix engine) not even with the 'Stereo Dithered Mixer' plugin installed. So unless the plug-in manufacturer includes 24bit dither on it's output, you are indeed getting 24bit truncation distortion period (regardless of whether you can hear it). This is less of an issue if you use native plugins such as the 32 or 64bit RTAS plugs, due to the rounding down of floating point maths, but for TDM fixed point maths every truncation point from 48bit to 24bit should be dithered to prevent truncation distortion.

Ok, here is the caveat... 24bit truncation distortion is very small & even with numerous instances it can be hard to perceive if your monitoring isn't up to the task or if you don't know what to listen for. It's not going to leap out at you like 'traditional' digital distortion (which we all hear on most top 10 records... lol) it's more of a vagueness or slight veiling of the sound which can make your mixes sound a little lifeless or 2 dimensional. Remembering that the more DSP (plug-ins) you use the more distortion you are adding (regardless of dither). Sure digital is more convenient & if done correctly can sound quite clean, but if I had the choice of 250 plugs or 10 high quality analog processors I know which would sound sonically superior. I'm sure at least some of your 250+ plug-ins would be dithered correctly anyway, some smart developers dither the output of their plug-ins correctly, such as URS, Waves (not all) etc. Also as I've already said it's not as much of an issue if some of your 250+ plugs are RTAS.

But ultimately (for the purest digital audio signal), it's best to preserve the highest precision where possible, right up until the final physical output/s where the wordlength must be reduced to 24bit for the D/A conversion. This makes more sense because you don't have to deal with truncation distortion or accumulative dither noise. This is the direction that a lot of DAW developers are going these days. Sonar, Sonic Studio, Sequoia/Samplitude etc. are all doing 64bit floating point resolution throughout the whole audio path including plug-in interconnects. Of course if the plugin you use is only 24bit or 32bit in the case of native systems, then it's up to the plug in manufacturer to dither it correctly or up to you to use double precision plug-ins where possible to avoid these issues. I've heard a 64bit path & it certainly has sonic benefits over Digidesign's HD model.

What adds fuel to the fire is how easy it is to distort TDM plug-ins because of the 24bit interconnect limitation. At least in Pro Tools LE you have an extra 48db of headroom on the plug-in bus due to the 32bit floating point interconnects.

Matt

[scottgreiner]

Mark Haliday
Engineer, Producer

Mark, are you related to David and Johnny?