internally reducing a track fader?????

[curve666]

Hi folks,
I read this over at ProSoundWeb and need help understanding how it impacts PTLE (or "Alsihad" as the forums refer to PT as):

"Perhaps the digital resolution is not reduced in all DAWs, but it has been well established within the recording community that internally reducing a track fader will greatly reduce its audio quality."

I don't understand this at all clearly and am left to speculate. Could someone help me suss out what this means, how to avoid loss of resolution etc.
thanx,
dna

[tele_player]

Firstt, read this:
Digidesign White Paper

Then, consider that it's referring to the TDM mixer, which uses 48-bit integer accumulators, unlike the PTLE mixer, which uses a 32-bit floating point accumulator.

The TDM system explicitly addresses the loss of resolution when channel faders are reduced. 32-bit (even floating point) isn't sufficient to do the same in the PTLE mixer. But from what I can tell, most or all native DAWs use a 32 bit accumulator for mixing.

We'd need a real computer math whiz to explain the implications of this. I'm a software engineer, but my field of expertise doesn't involve arithmetic calculation. A guy who's usually is real good at explaining the nuts and bolts, is Nika Aldrich. He didn't have an answer for this, at least not that I've seen yet.

[Nika]

It's pure bunk for starters. Let us do one thing to help in the dialogue, though. Let us shun the term "resolution" and find a more descriptive term. There are four characteristics of a waveform in the audio industry - amplitude, phase, frequency, and level of the noise, or dynamic range. Any waveform can be described using only these four terms. There is no characteristic of a waveform called "resolution." If you go to a mathematician or a wave theorist or a digital audio specialist and you start talking about the "resolution" in a waveform they are going to look at you like you have three heads. Any degradation of the signal will be manifested in one of those four ways. Now, which way is the supposed manifestation of the track faders?

Let's think about it. When you turn a fader down you decrease the number of bits used to describe the signal (same number of bits, actually, but we force more of them in the front to be "0," so we are effectively using fewer). As we know, decreasing the number of bits increases the amount of error in the signal. You have more rounding error. This manifests itself as greater noise. Ergo, the decrease in a fader's position increases the noise, right? Not really. The noise is at a constant level - -96 or -144 or whatever dB below full scale. All we are doing is bringing the fader down closer to the level of the noise. We are not actually raising the noise. This is no different than in any other situation - as you turn a fader down, or a knob down, or a level down you turn the signal down, closer to the level of the fixed noisefloor of the equipment.

The intuitive approach is that dropping a fader decreases bit depth. It really decreases the dynamic range by lowering the signal closer to the noise floor of the system. Is this bad? No. All systems have a noise floor and your job as an engineer is to understand that and work with it so you keep you signal as clean as possible. Keep in mind, though, that the noisefloor on most of these systems is at around -138dB and the human dynamic range is more like 100dB, so you've got about 40dB of dynamic range to kill there before it becomes audible.

Did I answer the question? Let me know. And tele - restate your question if I didn't answer it and I'll take a stab at it.

Also, can you provide the link to where you found that quote so I can go comment over there?

Cheers!
Nika

[tele_player]

Nika,

OK, loss of resolution is misstated.

I'm not sure what quote you refer to, but I think I bounced this idea off you on one of the boards a while back, trying to determine: if the Digi White Paper cited above is correct (re: lowered channel faders), then how can a 32 bit mix buss be correct?

I'm definitely not trying to argue, but I'd love to understand this detail better.

[Nika]

I'm not sure what quote you refer to,

The quote is the one referenced by the first poster - that he got off of another forum.

but I think I bounced this idea off you on one of the boards a while back, trying to determine: if the Digi White Paper cited above is correct (re: lowered channel faders), then how can a 32 bit mix buss be correct?

First, I don't remember much about the paper. The truth of it, however, is that a 32 bit floating point mix really can't be correct because proper dither cannot really be used. The only way to do so properly would be to trick the 32 bit system into believing it was a fixed point system - but then we really couldn't say that it was doing "floating point math." This is where floating point mixers have a problem - there is just no proper way to dither a floating point signal after its calculations. On the other hand, the level of this error stays very low, and the rest of what I said above is true. It's just that the low level "noise" is replaced by a low level "distortion" and that distortion stays at a fixed level, even when you lower the faders.

Did this answer the question? If not we'd better attack from a different angle.

I'm definitely not trying to argue, but I'd love to understand this detail better.

No arguing perceived over here!

Nika

[tele_player]

Excellent, I'm clear on this now, and you reminded me of the dither/float issue.

[jeffwuollet]

So, would mixing from an automated analog console with the outs of a TDM system to a mastering deck increase resolution as opposed to internal software mixing in a bounce? Forgive me if I missed the point entirely.

[Nika]

So, would mixing from an automated analog console with the outs of a TDM system to a mastering deck increase resolution as opposed to internal software mixing in a bounce?

First of all, if you read back you'll see that we're avoiding the term "resolution" because it doesn't mean anything. There is no such thing in a waveform as "resolution."

As to whether or not mixing it through an analog desk will decrease the noise, the answer is no. Mixing to an external desk will increase the noise. The fixed noise level in the Protools system is around -138dBFS. The noise level on a mixing desk is much worse - perhaps -110dBFS. So keeping it internal will keep the noise level lower. Even though you drop a fader and lower the signal in relation to the noise, that noise is very, very low. If you run it externally you will do the same thing when you eventually drop a fader - drop the signal closer to the level of the noise, but the noise will be much higher.

Nika