setting record levels into ProTools with the intent of mixing in ProTools

[Nika]

Joseph,

I'll work on a couple of pictures. In the meantime, to answer your other question:

To make sure you capture "the whole signal", just make sure that the noisefloor of your signal ends up hotter than the noisefloor of the converters. At first this may seem way too technical, like you're going to have to measure something.

If that you really want to make sure you're doing it right, then do this:

Hit record in protools with nothing turned on except your converters. What is the peak level of the noise generated by the converters alone? It should be around -120dB depending on the converters, etc. It might be around -85dB, and you can check it by clicking "command" and clicking on the "vol" in your expanded mix window twice. That will tell you peak levels.

OK, now what you have to do is get the self noise of your recording path to be hotter than that. Do this:

Plug a mic into a preamp. Turn it all the way down and just set it up in your studio with the doors closed and noone around. Now check your meters and turn it up on the preamp until the signal that it generates is hotter than whatever the peak level was of your converters by themselves. Whereever that point is is the minimum amount you need to turn it up in order capture maximum "resolution. Now go ahead and have someone sing into the mic and you'll probably realize that you have a LOT of headroom before you clip. So now you know the window of level that you have to play in, and it may be 30dB or more. That is the difference between your noise being as low as it can be, and your signal clipping.

OK, I don't expect you to do this before every session and on every track. I expect for you to have an IDEA of where you need to be and set your levels appropriately. That's where the term "engineer" comes into your title.

Let me know if this is unclear.

Thanx!
Nika.

[McGriffy]

Nika,

Fascinating factoid that converters use sign and magnitude instead of two's complement. I have wondered occasionally if such issues didn't come up in converter design. I remember as a kid reading about a system (Grey code?) where each time you increase the value by one only one bit changed. The idea was to avoid ever transitioning through invalid values. BTW, I was going to say that these kind of details don't really matter and are only discussed here out of curiousity, but Nika's story shows exactly why such things do matter. Perhaps not directly to every PT user, but certainly to those who design the equipement we all use.

Another point I was thinking of last night: You say that whenever we turn something down too low you lose information. Of course, we sometimes must turn something down to fit in the mix and this is normal. We also lose information any time we edit out unwanted parts of a file. But isn't it true that if the mixer is well designed and carries extra precision properly throughout, then you could turn a signal way down, then have it brought back up either by master fader gain or a compressor or something and not lose anything. That is, if the TDM mixer keeps 48 bits internally, your could drop soemthing 24 bits down, then boost it back up and get the same bits back. Or, if we leave it down a little, dither can still bring some of that information into our resulting 24 bits. But here's my question: In the final mix, say there is noise at -70 from a signal from a real microphone mixed near unity. If some other signal is mixed say 30dB down but also has 70dB of information, there will be information down below the noise floor of the first signal. Is the effective dynamic range of the mix 70dB or 100dB? I have often heard that humans are good at hearing signal underneath noise which makes me think that what matters is that you have look over 100dB down in this example before there is no information left. If I am right here and the mix has 100dB of information, this would explain why we need to keep the mix bus fairly hot for most mixes. Since some tracks will be mixed lower, there will often be a wide dynamic range there to capture even if no original signal had that much.

DMcG

[Nika]

DMcGriff,

The Protools mixer does not carry higher resolution throughout. From any place in Protools to any other place in Protools (busses, etc.) there is only 24 bit resolution. I think you see the issue.

If you take two signals and mix them - one is -70dBFS of raw noise. The other is 70dbSNR but it's peak is at -30dBFS then you will end up with 40dB of signal peaking somewhere between -30dBFS and -24dBFS and the remaining 30dB of signal ends up below the noise floor.

Various tests have shown that the ear can percieve some information below the noise floor because the noise ends up correlated with the signal and is therefore not truly random. I would argue that to some degree this is not really white noise, then, but no need to discuss semantics. Anyway, the furthest below the noise floor that I've ever seen a study validate that we can percieve information is 20dB. You can actually test this yourself using the "Signal Generator" plug-in in Protools and some creative engineering.

This all feeds into the comment I made earlier that you should ALWAYS maximize your mix bus as much as possible as it seems to be the weakest link in most digital mixers.

Does this all make sense?

Thanx!
Nika.

[blairl]

What's your line of thinking on the following scenario:


Let's say I use a mic and preamp to set the minimum level in order to capture the maximum resolution and I do end up with 30 db of headroom. What if I decide to use up the 30 db of headroom by increasong the preamp signal accordingly. Would this then allow me at the mixing stage to lower the level on that channel in Pro Tools by 30 db without loosing information?

[Nika]

Yes.

But there is no difference between recording it in with 30dB of headroom and recording it in with no headroom and turning it down 30dB. As long as you're hot enough there is no benefit or detriment to adjusting your input level up any farther short of clipping. And as long as you're quiet enough there is no benefit or detriment to adjusting your level downwards short of going before the noise floor.

At some point you have to decide at what level your analog equipment functions best at. My mic pre's may sound best at 20dB of gain. Does that fit into the window?

Does this answer your question?

Nika.

[bassmac]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:<HR>Originally posted by Nika:
When you turn down the faders you get low level distortion, essentially reducing your dynamic range by lowering your bit depth and raising the noisefloor (both end up happening).<HR></BLOCKQUOTE>

Nika, Is there any difference, or perference between using the mixer fader versus a plug-in gain adjustment to lower a level?

Thanks for sharing your vast knowledge with us.

[ckevperry]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:<HR>Originally posted by blairl:
What's your line of thinking on the following scenario:


Let's say I use a mic and preamp to set the minimum level in order to capture the maximum resolution and I do end up with 30 db of headroom. What if I decide to use up the 30 db of headroom by increasong the preamp signal accordingly. Would this then allow me at the mixing stage to lower the level on that channel in Pro Tools by 30 db without loosing information?<HR></BLOCKQUOTE>


No, you won't lose any information. Protools has 48 bit internal precision allowing you to turn a 24bit signal down 144dB without loss of any information.

The protools MIX BUS is the same as any other digital mixer. Oxford, Mackie, R100 or whatever. I haven't personally done it, but this has been tested.

I believe it's the plugins, sends and bussing that don't carry out the fancy whiz bang high resolution. Turn a signal down on these, and the bottom bits of signal would be stripped off.

Interesting aside: Do we ever really truncate at 24 bit? There are those who believe that since all analog gears noise floor is above the noise floor of 24bit converters (levels properly set mind you), that this "self-dithers" the signal. Tis an interesting thought.

[Nika]

OK, that's a tough one and has to do with dither.

If you are using the dithering mixer then the Protools mixer is dithering the result of the gain reduction on all of the channels.

If you are using a plugin that dithers then dithering happens after the gain reduction on the channels that you instantiate that plug-in on.

One way or another you want dither (not noiseshaping). Are you using the dithering mixer? Do you know which plugins dither? Once you figure that out you're most of the way toward answering this for yourself in your own application.

BUT.....not all dither is equal, and I'm not sure who is using what technology to do their dither. It's possible that one version of dither (say a plug-in manufacturer's vs. Digidesign's own) is better. This, however, is getting PRETTY tweaky. I'm just trying to answer your questions in complete detail, though I'm not sure that this level of tweakdome is really of utmost concern for most mixing applications.

I hope this helps.

Nika.

[Felix]

nika- i think bassmac was asking in a more general sense.
for example, lets just assume we're not using the dithering mixer.
i figure that since there is already processing going on in the plug-in, may as well add level changes into the task basket. that way it all gets done in one shot before being brought back into the mix via whatever dither process the plug uses. i usually do this with McDSP or Waves plugs. so this keeps most faders at unity most of the time. this just seem logical to me. now as to whether you should add a small plug (like a McDSP B1) just to change levels... i don't go this far. [img]images/icons/shocked.gif[/img]

[Nika]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:<HR>Originally posted by ckevperry:
No, you won't lose any information. Protools has 48 bit internal precision allowing you to turn a 24bit signal down 144dB without loss of any information.

The protools MIX BUS is the same as any other digital mixer. Oxford, Mackie, R100 or whatever. I haven't personally done it, but this has been tested.<HR></BLOCKQUOTE>


Kevin,

This is not correct. The summing accumulator in Protools is 48bits, but all datapaths are only 24 bits including channel -> mix bus according to a Digidesign Development Partner engineer:

"The Pro Tools mixer does indeed use 24bit data paths - but employs at least 48bit to accumulate the result of the addition of the channels. The result of all this is that the performance is good with properly dithered 24bit input signals with faders at 0dB exactly, but degrades rapidly if there is any gain or loss either on the faders or in the plug-in environment. This is because the actual mults and data paths are in fact only 24bit precision despite the 48bit summing buss performance. This is largely as a result of limitations in the 56000 processors they use."

And the Protools mix bus does NOT sound the same as other digital mixers, despite the claims of Mick Guszawksi. The differences are clearly audible when a session is set up properly. I have been involved in one of these tests.

Nika.