Pro Tools 9!

[Wolfgang Eller]

Originally Posted by Matt_G

I'm talking about increased headroom between plug-in interconnects...

Why should it happen if your gain staging is correct?

...reducing the amount of points in the signal path where truncation to 24bits occur without dither & for the few (TDM) plug-ins that do have 24bit TPDF dither on the output...

The signal can´t be better than 20bit anyway (cause you recorded it with less than 20bit dynamic), so what you truncate is noise. Who cares?

The output of every channel's inserts are also being truncated to 24bits before merging back into the 48bit summing mixer. Also the input to every Aux channel is also truncated to 24bit (reducing the Aux fader won't prevent clipping either). Again at the Master Fader any inserts added are more 24bit interconnects when using any TDM plugs. These are all limitations imposed by the dated hardware architecture which was necessary to get work done with a reasonable DSP overhead on older computers.

See answer above.

Let's face it TDM has had it's day, we needed it when computers were underpowered.

Sorry, but if you put tires, which last 350km/h, on your car which drives 200km/h doesn´t make it faster or brings you any benefits.

A single, recorded audio signal is not better than 20bit, the single channel in TDM is 24bit (which are real and not like the 24bit AD converters), the mixing bus has 48bit. So tell me: Were is a bottle neck in the mixers resolution where the (real) audio signal will be truncated?
And, if you do correct gain staging, where is not a pot or a fader to avoid clips?

Btw: When do we get the first 32bit floating point AD/DA converters?

[lode]

I´m sure you know your stuff when it comes to the math involved here.

However I know from firsthand experience that TDM plugs can clip, where the same plug as RTAS will not clip. Even without lighting the clip warning. So real world examples have showed me that the headroom in a rtas enviroment is better than TDM.

Regarding recorded levels:
I work solely with post, and having proper gain staging on documentaries is not feasible. You woudn´t believe how wildly audio will vary in levels and quality. I try to hammer the levels into a reasonable shape ready for broadcast, but that might mean boosting 24 db or reducing the input on the first plug 6 db just so it won´t clip when adding a LF cut.

Bottom line:
We need a system that can handle levels way beyond what is optimal gain staging.
Not because we don´t know how to operate the gain pot on the microphone amplifier (done by someone "upstream"), we need to be able to work fast and not worry about clipping in the mixbus.

[Wolfgang Eller]

Originally Posted by lode

...we need to be able to work fast and not worry about clipping in the mixbus.

You can clip a plugin if you´re driving too hot, you can clip the output DA converter but how can you clip the 48bit mixbus?

[Firechild]

Originally Posted by Wolfgang Eller

You can clip a plugin if you´re driving too hot, you can clip the output DA converter but how can you clip the 48bit mixbus?

A 32 bit floating point mixer has more headroom than 48 bit fixed. But as you keep watching your levels it feels more stable to work with a TDM mixer. Maybe because it is more expensive..I don´t know. I like my TDM stuff.

[lode]

I have no idea why the TDM versions would clip when the RTAS didn´t.
I just know that it happened.

And I had reduce the gain by absurd amounts to avoid the clipping. I then switched to RTAS, and the problem was gone. Poof.

I have never used TDM plugs if I can avoid it after that.

[Wolfgang Eller]

Originally Posted by lode

I have no idea why the TDM versions would clip when the RTAS didn´t.

TDM are fixed and RTAS are floating point.

[Matt_G]

Originally Posted by Wolfgang Eller

Why should it happen if your gain staging is correct?

As stated if you know how to drive the TDM system properly some of these things such as clipping can certainly be avoided. However have you tried to open someone else's LE mix session on a HD system & tried to play that back? Quite often there will be clipping going on which wasn't present when played back from LE & you'll soon have to adjust your whole mix again with 'proper' gain staging. This in turn effects compression thresholds down stream etc, etc.

The signal can´t be better than 20bit anyway (cause you recorded it with less than 20bit dynamic), so what you truncate is noise. Who cares?

Only partly true, the minute you add any DSP be it EQ, gain, compression, whatever it is, the wordlength will grow to greater than 24bits in a lot of cases which is why a lot of plug-ins have double precision internal processing to prevent clipping during the process. With any TDM plug-in this has to come back to 24bits again on the output so if the signal level generated in the processor is greater than 0.0dBFS it will need to be trimmed back & ideally re-dithered prior to wordlength reduction at the output again. If it's not dithered & simply truncated, this will add low level distortion & you're also chopping off bits that do hold useful low level information that pertains to part of the recorded signal & the result will be aliasing.

If you take the noise floor of a typical signal, amplify it through EQ, compression, limiting or gain & then lower the level again on the output so the peaks aren't clipping what happens is that the noise floor begins to come up along with all the other garbage that resides there. Do this over multiple channels or multiple passes subjected to 24bit truncation & you'll begin to hear the signal integrity degrade.

When you start mixing channels together this can grow the wordlength considerably which is why we have a 48bit summing mixer with a floating 8bit mantissa. Auxillary's are 24bit in so if you're bussing channels to an Aux you need to group all the channels & reduce the level going to it or you run the risk of clipping.

If the goal is high RMS levels than this can mean quite a lot of gain stage juggling in a typical HD scenario.

Again a lot of this can be avoided without all the 24bit bottle necks.

Sorry, but if you put tires, which last 350km/h, on your car which drives 200km/h doesn´t make it faster or brings you any benefits.

It would make it a hell of a lot easier for everyone if Avid let us have the capacity to drive at 350km/h all the time without having to slow down & speed up again between all the speed humps. We would end up at our destination quicker with less wear & tear no?

A single, recorded audio signal is not better than 20bit, the single channel in TDM is 24bit (which are real and not like the 24bit AD converters), the mixing bus has 48bit.

See above...

So tell me: Were is a bottle neck in the mixers resolution where the (real) audio signal will be truncated?

See above regarding plug-in truncation & above regarding Auxillary's. These can effect the real audio signal imo.

And, if you do correct gain staging, where is not a pot or a fader to avoid clips?

We seem to be going around in circles, yes I know you can avoid clips through careful gain staging & I haven't said otherwise. However you don't have to be as careful with a floating point system & this is my point. Why climb all the way to Mt. Everest on foot when you can take a helicopter straight to the top in minutes? With HD Native or LE you don't have to check & adjust every facet of your mix to prevent clipping all the time. Native is streamlined to maintain a hot signal level with precision & transparency right up until the master fader where you would then need to reduce the signal &/or limit it & dither once to bring it back to 24bit or 16bit. It is the cleanest & most efficient way to work... whether you can hear the difference with your ears, DAC, monitors & room is a subjective matter not the scientific one you're trying to prove.

Btw: When do we get the first 32bit floating point AD/DA converters?

??? You don't need the added headroom at a D/A as all the DSP has been done pre conversion so there is no benefit. Nor would you need it on the A/D because there would be no added benefit. The same can not be said for DSP processing where wordlengths invariably grow & benefit from the higher precision throughout the entire DAW.

Are we done?

[Wolfgang Eller]

Originally Posted by Matt_G

yes I know you can avoid clips through careful gain staging & I haven't said otherwise. However you don't have to be as careful with a floating point system & this is my point.

And my point is: You can do everything equally in PT TDM which can be done in PT native (just for the comparison between fixed an floating point).
Everybody who "can" work in TDM is able to do the same Nativ.

But not vice versa!

So back to my initial question: "You really want a support for "technicians" who are not able to set the correct input level?"

BTW: Could you imagine how a recording/mix with a SSL 4000 and a Studer sounds when a guy is engineering who has had nothing but a floating point DAW in his hand ever?

Is this what it´s called progress?

Edit: My question for 32bit float AD/DA converters was a joke.
Sorry for not understanding it.

[daeron80]

In my experience, the 48-bit fixed system of TDM is superior to the 32-bit float engine. Matt is exactly right - the hangup is at plug-in I/O. Once they implement 48-bit throughout the chain, so you have several bits of headroom above 0 dBfs at every point, no more 24-bit dither, it'll be a world-class digital mixer. I bet that's coming.

[Cliffy_Boy]

To really understand all the technical points above, you need to have worked with an LE system. Otherwise its just a theoretically argument in physics. The same session created in LE will most probably clip a TDM system while it was fine in LE. LE just allowed much higher gain stages prior to clipping which TDM doesn't. Period.