Tracking levels ...

[andychamp]

Originally Posted by Mayor999

I think by resolution he meant that the physical distance the fader travels has less of an effect than at other parts on the channel... (The distance between -80 to -60 is the same from -5 to 0)

Exactamundo, my friend (because of the metric system).

[Dean Roddey]

That's not really an issue in a DAW either though. You can just type in the level you want if you feel you can't get it exactly by dragging the fader, though that's unlikely to be necessary.

[andsonic]

I just wanted to add my 2 cents to what Jules is saying. When I cut my teeth, my mentors were all engineers who began their careers in the early '70's. The " all in a line " technique was mandatory!

When I use this technique, I always get good levels.

... and most importantly to me: it makes mixing easier! No battles between instruments and great drum mixes right off the bat. This technique is a major shortcut to great sounding mixes.

At a EQ master class I went to a couple years ago with George Massenburg, Rupert Neve, and Malcom Toft on the panel, this was said: (pardon my slightly faulty memory)
Question: What is the first eq move you would make when recording?
GM: I would move the mic.
RN: Then I would change the mic
MT: Only then would I EQ.

Get it right at the start. It will make the whole project easier.

[Dean Roddey]

But getting the sound right has nothing to do with the level on the track. It will sound the same whether the fader is at 0 or -12. And you don't ahve to commit to a level that you thought was good during tracking but which now seems too low during the mix.

[TheArchitect]

Originally Posted by andsonic

I just wanted to add my 2 cents to what Jules is saying. When I cut my teeth, my mentors were all engineers who began their careers in the early '70's. The " all in a line " technique was mandatory!

When I use this technique, I always get good levels.

... and most importantly to me: it makes mixing easier! No battles between instruments and great drum mixes right off the bat. This technique is a major shortcut to great sounding mixes.

At a EQ master class I went to a couple years ago with George Massenburg, Rupert Neve, and Malcom Toft on the panel, this was said: (pardon my slightly faulty memory)
Question: What is the first eq move you would make when recording?
GM: I would move the mic.
RN: Then I would change the mic
MT: Only then would I EQ.

Get it right at the start. It will make the whole project easier.

I fail to see how recording something at less than optimum levels for the sake of keeping faders in a line can be called "getting it right from the start". A "name" with a quirky habit gets a little press suddenly it becomes the defacto standard? Sad but not uncommon copycat behavior. People really need to think for themselves.

[Jules]

Originally Posted by TheArchitect

I fail to see how....

IMHO You are focusing in on some sort of (imagined) old boys snobbery and missing out on the benefits of this practice.

I will explain some of them in follow up posts

[TheArchitect]

Originally Posted by Jules

IMHO You are focusing in on some sort of (imagined) old boys snobbery

And you would be wrong.

The "old boys" as you say are the ones doing this, not me. I fail to see how intentionally screwing up gain structure so my faders are in a nice row at the final mix makes any sense at all. Other than "I saw someone else doing it" and name dropping I haven't heard any reasonable explanation as to why this is anything but nonsense.

[delcosmos]

Originally Posted by Jules

I will explain some of them in follow up posts

Great!

Thank you Jules.

Delcosmos.

[BlakeJohnsonCSR]

So I'm trying to understand both sides here...maybe you can better help me through my current situation.

I am tracking drums right now and I am not taking the approach of getting a good mix with all the faders at unity. And it is kind of funny because I had a guy come into my studio last week to do a drum session and he had very low levels on his overheads and room mics, yet he said he didn't care about levels because he was ultimately satisfied with the sound of it.

So a week later I am doing a different drum session, and using a different approach. Is it bad that the inputs on my AD-16x are indicating clipping levels yet it shows no evidence of clipping within Pro Tools? I don't hear any clipping either.

[synthoid]

Originally Posted by TheArchitect

And you would be wrong.

The "old boys" as you say are the ones doing this, not me. I fail to see how intentionally screwing up gain structure so my faders are in a nice row at the final mix makes any sense at all. Other than "I saw someone else doing it" and name dropping I haven't heard any reasonable explanation as to why this is anything but nonsense.

It's even worse than this. The faders are in a nice row according to whatever was in the head of the tracking engineer. If the mix engineer is not the same person, or if he, you know, has an opinion of his own about the mix, then he might be forced to disturb the neat row by, you know, pulling a fader up. And that would suck, because who wants to see one of those faders out of line at the end?

-synthoid

[peeder]

Another thread on this (among many)

Tracking - Levels

[u b k]

Originally Posted by TheArchitect

I fail to see how intentionally screwing up gain structure so my faders are in a nice row at the final mix makes any sense at all.

i gotta say, if i push all the faders up to ~unity and the mix is mostly there, that seems to me the very epitome of a gain structure that is *dialed*: optimum output level from tape, coming into the sweet spot of the fader with no trim, all hitting the mix rails in the right place.

i truly am having a hard time understanding how this can be seen as a bad thing.




gregoire
del
ubk
.

[peeder]

Originally Posted by u b k

i gotta say, if i push all the faders up to ~unity and the mix is mostly there, that seems to me the very epitome of a gain structure that is *dialed*: optimum output level from tape, coming into the sweet spot of the fader with no trim, all hitting the mix rails in the right place.

i truly am having a hard time understanding how this can be seen as a bad thing.

If you're using analog tape OK, but that isn't necessary in digital, you have >30db of penalty-free leeway in 24 bit digital, and you don't have tape helping you out with saturation and distortion etc. So you make that up by driving your analog gear pre-ADC to the point you want it, which may be low and clean or hot and raging...and if you need to trim or gain up some stuff when you do faders up, well then, you've just done your job.

[kooz]

Originally Posted by Jules

(its a buzz getting a great sounding vibe going in the studio, having another engineer drop in and watching their split second eyebrow raise as they look at the faders and realize that what you have going on is 'all faders in a row' - its kind of an "engineering machismo" thing to do

Yeah, it does get people thinking. "what does he know that I don't?" and "how the hell is he getting everything so evenly balanced?" As our esteemed mod has pointed out, it's the gain pot most often. I once had an engineer with major label credits walk into a tracking session I was doing of a gospel act, using 36 of the console's 40 mic inputs, bussed to the 24 track and see that my monitor faders were all at unity. He looked like he was in a hurry when he walked in, but he stayed for the session, and even helped tear down.

[Dean Roddey]

i gotta say, if i push all the faders up to ~unity and the mix is mostly there, that seems to me the very epitome of a gain structure that is *dialed*: optimum output level from tape, coming into the sweet spot of the fader with no trim, all hitting the mix rails in the right place.

But gain staging the analog input chain is what is being discussed. That's all before the fader. If you have some part that's going to be way off in the distance and at -24dB or something, then to do that with the fader at 0, you couldn't really possibly be getting an optimally staged signal through the analog gear on the way in, I don't think. It would have to be way lower that is optimal. And if you have any processing on the track, it's going to be working on really sub-optimal bits as well. Good gain staging would mean getting a good signal through all those things so that they have plenty of signal to noise ratio, then pulling down the fader to what's appropriate for the track in the context of the mix, or so it would seem to me.

[peeder]

Originally Posted by Dean Roddey

But gain staging the analog input chain is what is being discussed. That's all before the fader. If you have some part that's going to be way off in the distance and at -24dB or something, then to do that with the fader at 0, you couldn't really possibly be getting an optimally staged signal through the analog gear on the way in, I don't think. It would have to be way lower that is optimal. And if you have any processing on the track, it's going to be working on really sub-optimal bits as well. Good gain staging would mean getting a good signal through all those things so that they have plenty of signal to noise ratio, then pulling down the fader to what's appropriate for the track in the context of the mix, or so it would seem to me.

In practice, modern ADC noise floors are much lower than most chains, and you can dial in what you want and not worry much about it either way. Take a look at how much leeway there is in your chains, and how, after (digital) loudness matching, your converters sound at different levels.

Both "sides" of the level argument are wrong: hotter isn't always better, and colder isn't always better. Goldilocks knew this.

[Dean Roddey]

But even in the 24 bit world, if you track with peaks at -6dB, you have a million levels of resolution. If you track at -24dB, you are down to like 268K levels or something like that, a good number of which are useless because they are noise.

I just don't see why it would ever be a bad thing to track all tracks at reasonable -12db to -6dB peaks, so that you have a good, warm signal through the analog chain, plenty of bits for the plugs to work on, plenty of dynamic resolution, and plenty of gain to play with, should you decide later in the mix that something needs to be louder than you though it might need to be (perhaps because you notched out some EQ from it in the end which lowered its effective volume considerably.)

I just can't imagine how that would ever be the wrong thing to do.

[peeder]

Originally Posted by Dean Roddey

But even in the 24 bit world, if you track with peaks at -6dB, you have a million levels of resolution. If you track at -24dB, you are down to like 268K levels or something like that, a good number of which are useless because they are noise.

You're still back on the "resolution" fallacy we have to stamp out from time to time. In modern dithered digital systems, there is no concept of "resolution", there is only dynamic range. If the noise floor of your input is higher than that of the converter, which in 24 bit is and probably will always be higher than that of the digital format, then you are losing nothing printing at whatever level the signal is at.

Nothing.

Nada.

Zilch.

Bupkes.

I just don't see why it would ever be a bad thing to track all tracks at reasonable -12db to -6dB peaks, so that you have a good, warm signal through the analog chain, plenty of bits for the plugs to work on, plenty of dynamic resolution, and plenty of gain to play with, should you decide later in the mix that something needs to be louder than you though it might need to be (perhaps because you notched out some EQ from it in the end which lowered its effective volume considerably.)

I just can't imagine how that would ever be the wrong thing to do.

It would be the wrong thing to do if the gear upstream from it had gain that wasn't as clean as digital gain you might apply post-conversion, and you wanted that cleanliness.

[Dean Roddey]

You're still back on the "resolution" fallacy we have to stamp out from time to time. In modern dithered digital systems, there is no concept of "resolution", there is only dynamic range. If the noise floor of your input is higher than that of the converter, which in 24 bit is and probably will always be higher than that of the digital format, then you are losing nothing printing at whatever level the signal is at.

I don't think that's true. The number of bits you store determines the number of avilaable amplitude levels to represent that signal and therefore the amount of dynamic resolution *stored*, not processed. In a floating point summing engine, during the processing, then yeh there's effectively an infinite range of gradations. But we are talking about the signal as converted on the way in.

So if you record at -24dB, you are down 4 bits from the top. Every bit down from the top halves the number of available sample values you have to represent the amplitude of the signal when sampled. So 4 bits down is 16 times fewer available amplitude levels.

So if you record the signal once at -24dB peaks and again at -6dB peaks, and it's the same performance, the one at -6dB peak will have many times more possible sample values to work with, and that will increase the dynamic resolution, because in both cases you've taken the same range of incoming signal and in one you've overlaid it over a million gradations and in the other you've overlaid it over 256K gradations.

I find it hard to understand how both of those could have captured the same amount of dynamic resolution to disk. And, during the processing, the plugs have (in this case) 8 times more gradations of dynamics in the original incoming signal to work with, which is different from the fact taht the floating point engine allows them create almost infinitely fine gradiations as the output of their processing.

It would be the wrong thing to do if the gear upstream from it had gain that wasn't as clean as digital gain you might apply post-conversion, and you wanted that cleanliness.

But that's all part of correct analog gain staging, sure. If the hardware CAN'T do it, then of course don't do it. But if it can, and most good pro equipment can, why would it ever be wrong to do that, or go compromise that just so you can have a set of faders all in a pretty row?

[peeder]

Originally Posted by Dean Roddey

I don't think that's true. The number of bits you store determines the number of avilaable amplitude levels to represent that signal and therefore the amount of dynamic resolution *stored*, not processed. In a floating point summing engine, during the processing, then yeh there's effectively an infinite range of gradations. But we are talking about the signal as converted on the way in.

So if you record at -24dB, you are down 4 bits from the top. Every bit down from the top halves the number of available sample values you have to represent the amplitude of the signal when sampled. So 4 bits down is 16 times fewer available amplitude levels.

So if you record the signal once at -24dB peaks and again at -6dB peaks, and it's the same performance, the one at -6dB peak will have many times more possible sample values to work with, and that will increase the dynamic resolution, because in both cases you've taken the same range of incoming signal and in one you've overlaid it over a million gradations and in the other you've overlaid it over 256K gradations.

I find it hard to understand how both of those could have captured the same amount of dynamic resolution to disk. And, during the processing, the plugs have (in this case) 8 times more gradations of dynamics in the original incoming signal to work with, which is different from the fact taht the floating point engine allows them create almost infinitely fine gradiations as the output of their processing.



But that's all part of correct analog gain staging, sure. If the hardware CAN'T do it, then of course don't do it. But if it can, and most good pro equipment can, why would it ever be wrong to do that, or go compromise that just so you can have a set of faders all in a pretty row?

Read Paul Frindle's explanations of this issue...a lot has to do with the interpolator in the DAC recovering resolution that is hard for people to imagine is encoded in the bits. People, confused by the poor waveform displays in their DAWs, don't understand how much an interpolator can do with very little information stored. As Dan Lavry says, you need only two points to determine every other point on a straight line.

e.g. here:

Q for Paul Frindle

[MarkRB]

All faders at zero is "correct" gain staging when using a desk as a front end, at least as far as I am aware. Please correct me if I am wrong.

When you consider a desk channel has the same inherent noise floor at all levels below 0dB it makes sense to only add as much noise from the mic-pre as you need by leaving the fader at zero and essentially mixing with your pres.
If you have your fader at -10 you will have 10dB less channel noise but an additional 10dB of gain to reach the same level on your DAW/Tape. It seems kind of obvious to me but maybe I'm missing something.
Likewise if you know something is going to be quiet in the mix why add additional noise from the pre or channel just to get the VU bouncing near zero.

This is obviously less relevant these days as most people go straight from a pre to the AD.

As I said, this is how I understand it but I am an idiot.

[Dean Roddey]

Read Paul Frindle's explanations of this issue...a lot has to do with the interpolator in the DAC recovering resolution that is hard for people to imagine is encoded in the bits. People, confused by the poor waveform displays in their DAWs, don't understand how much an interpolator can do with very little information stored. As Dan Lavry says, you need only two points to determine every other point on a straight line.

You are talking about something different. You are talking about frequency, I'm talking about amplitude resolution. Yes, given two sampled points, you can interpolate a curve between them. But if you are capturing sample points that are 8 times more strongly quantized, then the curve you are creating is between those more heavily quantized points. The Nyquist stuff only talks about recreation of the curve from the points, it doesn't have anything to do with how accurately the sample points are captured in terms of their amplitude.

If you take the same signal and capture it with 8 times fewer possible amplitude levels, then each sample point will have to be pushed further from the actual amplitude level because those are the only sample points you have to work with. So it's dynamic (amplitude) resolution that's I'm talking about here, not frequency.

[Matt Grabe]

First of all, thanks for all of the replies! This has been very helpful.

A few people mentioned that there are many other topics regarding this subject. All I could find was hot vs. cold, bit resolution, gain staging, etc. Similar to my thought, but I was curious about tracking towards a mix in-particular. I do apologize if I missed a thread on that.

Regarding gain staging and optimal use of the analog front end ... I'm not so curious about the gain staging of the analog front end. I assume that those who track towards a mix still use their analog front end (preamps, compressors, eq's) are their optimal gain staging, and then use their tracking fader (or some other form of attenuation at the end of the chain) to balance out their tracks to tape/ProTools.

Also, the intention would not be to have your mix complete with the faders in a line, and if a fader had to be raised then all hell would break loose .. but rather a "nice general" balance is had with the faders in a line.

Personally, I've never tracked this way. I probably won't until I have some time to try it on something other than a paying client to see how it goes. However, I do like the idea of putting all the faders in a line and hearing a nice blend.

Continue to debate! Good information all around ...

[peeder]

Originally Posted by Dean Roddey

You are talking about something different. You are talking about frequency, I'm talking about amplitude resolution. Yes, given two sampled points, you can interpolate a curve between them. But if you are capturing sample points that are 8 times more strongly quantized, then the curve you are creating is between those more heavily quantized points. The Nyquist stuff only talks about recreation of the curve from the points, it doesn't have anything to do with how accurately the sample points are captured in terms of their amplitude.

If you take the same signal and capture it with 8 times fewer possible amplitude levels, then each sample point will have to be pushed further from the actual amplitude level because those are the only sample points you have to work with. So it's dynamic (amplitude) resolution that's I'm talking about here, not frequency.

No I _am_ talking about amplitude resolution. Read the link I added and the others like it...

[steveschizoid]

I thought the 24 bit range (from 24 0's to 24 1's) described the entire range form absolute silence to the point of clipping. I know my understanding is feeble, so be gentle. If a given performance, and the resulting samples derived from it, all fall within that range and above the noise floor of the convertor, why would it make any difference where within that range the samples fall? A sample corresponding to a louder sound (based upon my understanding) would be a higher number, not more numbers. The latter is what you seem to be saying Dean. If that was the case then it would seem that we should discourage dynamic performances because the quieter sections would be described digitally with less resolution than the louder sections.