My Signal Neutral rant (Gain Staging)

The forum for petitions, theoretical discussion, gripes, or other off topic discussion.

Moderator: James Steele

Forum rules
The forum for petitions, theoretical discussion, gripes, or other matters outside deemed outside the scope of helping users make optimal use of MOTU hardware and software. Posts in other forums may be moved here at the moderators discretion. No politics or religion!!
Post Reply
User avatar
Shooshie
Posts: 19820
Joined: Sat Oct 16, 2004 10:01 pm
Primary DAW OS: MacOS
Location: Dallas
Contact:

My Signal Neutral rant (Gain Staging)

Post by Shooshie »

A post in another part of our forum reminded me of a rant I've been meaning to write about. So... here it is:

If I could get audio gear manufacturers and software developers to do ONE thing together, it would be to have a notch or gauge or even just a line showing the setting for a signal-neutral output. That is, an output that neither amplifies nor attenuates the signal. That may be tricky for amplifiers which of course... amplify. But even there, there is a natural level at which the amplifier works best with minimum distortion. That line.

If we could adjust all software and hardware so that they are passing the most natural signal with the minimum processing, it would speed up our work and make our final products so much better. We're always having to experiment to compensate, and you don't always know where the level is coming from. For example, I recently plugged in a headphone amp, then turned up the volume to a natural level. But I was hearing hiss and noise like crazy. That didn't seem right, as this was a decent headphone amp, though certainly not a candidate for "finest." It was just an adequate amp for 8 independent headphones, but it couldn't have been sold with this much noise in it.

The answer, of course, was to balance my input levels, and to do that, I had to go back up the line a ways. Since MOTU made CueMix response to your computer volume, I've enjoyed that feature, but I've never quite known where things really stand, doing it that way. Anyway, I got the line volume turned up to a more natural level, and then I was able to turn down the headphone amp's volume, and the noise went back to the floor, where it belongs. (A little noise is acceptable in a headphone; you're not going to be recording it.)

I've tried to count the number of level adjustments from track to ear, and that's actually sometimes very hard to do. You've got the track fader, levels in many of your plugins — which varies from track to track, then the level of the output fader, Apple's Audio-MIDI Setup, the Mac's Volume, CueMix (and I have no idea how many ups and downs there are in CueMix), then you output it to something: amp, powered speakers, or whatnot. I may have missed a few, or perhaps some of those things I mentioned control the same level streams. I don't know. That's part of what makes it so hard.

If any of those levels is jacked too high or too low, it sets off a series of compensations that may fix it, or you may have already peaked and created transients that cannot be fixed by merely pulling them down. So, a mid-line, a "natural" line, or a "signal-neutral" setting would be such an advantage, but everyone would really need to do it, or it wouldn't work.

Is this a windmill I'm tilting at? Is it even possible? You'd think if it were practical or possible, it would have been done. A lot of gear has a midline, but you set it there, and it's not usable. Maybe there's something up the chain that's set too high. It just seems like this could be done without much trouble. So... why isn't it?

Just my rant of the day.

Shooshie
Last edited by Shooshie on Sat Sep 26, 2015 3:42 pm, edited 1 time in total.
|l| OS X 10.12.6 |l| DP 10.0 |l| 2.4 GHz 12-Core MacPro Mid-2012 |l| 40GB RAM |l| Mach5.3 |l| Waves 9.x |l| Altiverb |l| Ivory 2 New York Steinway |l| Wallander WIVI 2.30 Winds, Brass, Saxes |l| Garritan Aria |l| VSL 5.3.1 and VSL Pro 2.3.1 |l| Yamaha WX-5 MIDI Wind Controller |l| Roland FC-300 |l|
User avatar
Shooshie
Posts: 19820
Joined: Sat Oct 16, 2004 10:01 pm
Primary DAW OS: MacOS
Location: Dallas
Contact:

Re: My Signal Neutral rant

Post by Shooshie »

By the way, it works both ways: loud and soft. We all know what happens when you turn your volume up too loud. But if you have any level set too soft, it has an effect similar to dialing down the bit-rate at which you're processing audio. 24 bits makes possible a lot of audio levels, but if you have one level set at a very soft volume, it's like knocking off maybe 12 of those 24 bits, or even more. From that point on, you're using 24 bits to describe just a few levels of volume. You'll be amplifying noise. Clipped transients on one end, clipped resolution on the other; you have to admit that finding that mid-line is pretty important if you want the full extent of digital resolution for your audio.

Shoosh
|l| OS X 10.12.6 |l| DP 10.0 |l| 2.4 GHz 12-Core MacPro Mid-2012 |l| 40GB RAM |l| Mach5.3 |l| Waves 9.x |l| Altiverb |l| Ivory 2 New York Steinway |l| Wallander WIVI 2.30 Winds, Brass, Saxes |l| Garritan Aria |l| VSL 5.3.1 and VSL Pro 2.3.1 |l| Yamaha WX-5 MIDI Wind Controller |l| Roland FC-300 |l|
User avatar
stubbsonic
Posts: 4601
Joined: Fri Dec 22, 2006 12:56 pm
Primary DAW OS: MacOS
Contact:

Re: My Signal Neutral rant

Post by stubbsonic »

It may be that for much of what goes on in a digital signal path, minor ups & downs might not matter very much. But with complex signals that go in & out of various domains, having a "gain neutral" or "unity" kinds of indications would be super-helpful.

And this general topic has come up so many times recently. For example, I got an effects pedal, and there are many places in the pedal where gain can be added, but not a word about how to know if/where clipping could occur. And worse, the kinds of knobs where from 1-10, maybe 5 or 6 is gain neutral. Erg.

Is that kinda what you mean?
M1 MBP; OS 12, FF800, DP 11.3, Kontakt 7, Reaktor 6, PC3K7, K2661S, iPad6, Godin XTSA, Two Ibanez 5 string basses (1 fretted, 1 fretless), FM3, SY-1000, etc.

http://www.jonstubbsmusic.com
User avatar
HCMarkus
Posts: 9713
Joined: Tue Jan 10, 2006 9:01 am
Primary DAW OS: MacOS
Location: Rancho Bohemia, California
Contact:

Re: My Signal Neutral rant

Post by HCMarkus »

Oh yeah. I totally get it and concur, but recognize that, with processes that effect sound, unity gain may vary based on the nature of the audio input.
User avatar
Shooshie
Posts: 19820
Joined: Sat Oct 16, 2004 10:01 pm
Primary DAW OS: MacOS
Location: Dallas
Contact:

Re: My Signal Neutral rant

Post by Shooshie »

stubbsonic wrote:It may be that for much of what goes on in a digital signal path, minor ups & downs might not matter very much. But with complex signals that go in & out of various domains, having a "gain neutral" or "unity" kinds of indications would be super-helpful.

And this general topic has come up so many times recently. For example, I got an effects pedal, and there are many places in the pedal where gain can be added, but not a word about how to know if/where clipping could occur. And worse, the kinds of knobs where from 1-10, maybe 5 or 6 is gain neutral. Erg.

Is that kinda what you mean?
Basically, that's what I mean. Technically, I mean a marking, groove, or "catch" at the point where you literally maximize your potential for a given bit rate. That would mean the most headroom with the greatest amount of the bit rate overhead actually used. A point where you're most likely to handle any transient that comes through, while utilizing all your 24 bits in describing the audio in a given song — if that's possible. I understand the impracticality of saying that one-size-fits-all, because it doesn't, but if that's the starting point for every level setting in the signal chain, it's going to be very easy to adjust it for any situation.

For example: let's say that you're recording some really hot audio, and you are in CueMix, so you reach for the Trim control to turn it down. You've just lowered the potential signal for all the tracks you adjusted. Those mics are now squashed. Later you record a flute and piano in a church. You forget to bring the trim back up. Now you're amplifying the noise in that hall, that mic, because you had already digitally squashed it, and now you're bringing up the signal elsewhere. Not good.

We just need a way to set our whole signal chain at the most neutral level, without having to go through literally every spot in the entire computer where that signal may be passing a level control, every time we start over. That was fairly easy to do on an analog board, but in the digital realm there are so many hidden level controls that we often forget they even exist.

The only place that it doesn't matter so much is in the trim controls for plugins. That's fairly late in the chain, and it's only working on one track at a time (unless it's the master). Everywhere else, it matters.

I've always assumed that maximum signal was the best place to set digital controls, if it's not clipping internally. Within DP and its realm, that's certainly true. I can't speak for the rest of the Mac.

Shooshie
|l| OS X 10.12.6 |l| DP 10.0 |l| 2.4 GHz 12-Core MacPro Mid-2012 |l| 40GB RAM |l| Mach5.3 |l| Waves 9.x |l| Altiverb |l| Ivory 2 New York Steinway |l| Wallander WIVI 2.30 Winds, Brass, Saxes |l| Garritan Aria |l| VSL 5.3.1 and VSL Pro 2.3.1 |l| Yamaha WX-5 MIDI Wind Controller |l| Roland FC-300 |l|
User avatar
stubbsonic
Posts: 4601
Joined: Fri Dec 22, 2006 12:56 pm
Primary DAW OS: MacOS
Contact:

Re: My Signal Neutral rant

Post by stubbsonic »

There were some things in your last post that didn't make sense.

You mentioned bringing up noise in the hall, but that would be the case whether the mic was hot or attenuated. The flute and the background sound would be inseparable. The only thing would be just the lack of resolution which, in 24 bits is not as concerning. I usually reserve the term "squashed" for compression, but I think you just mean attenuated. So the signal is just quieter and can be turned up (intact).

The question of moving signal through various stages and not losing track of places where gain can be gained or lost is a matter of procedure. Controlling gains on inputs, of course, is the main place for tracking/mixing live.

But for example if signal was running through a compressor, and you were "squashing" the signal (in the compression sense) and you wanted to make sure you were coming out with basically a "gain neutral" output, how would the compressor possibly know? It would have to listen for a while, and find out how much gain reduction was happening over time and then tell you where to put the output knob. Not impossible though.
M1 MBP; OS 12, FF800, DP 11.3, Kontakt 7, Reaktor 6, PC3K7, K2661S, iPad6, Godin XTSA, Two Ibanez 5 string basses (1 fretted, 1 fretless), FM3, SY-1000, etc.

http://www.jonstubbsmusic.com
User avatar
Shooshie
Posts: 19820
Joined: Sat Oct 16, 2004 10:01 pm
Primary DAW OS: MacOS
Location: Dallas
Contact:

Re: My Signal Neutral rant

Post by Shooshie »

stubbsonic wrote:There were some things in your last post that didn't make sense.

You mentioned bringing up noise in the hall, but that would be the case whether the mic was hot or attenuated. The flute and the background sound would be inseparable. The only thing would be just the lack of resolution which, in 24 bits is not as concerning. I usually reserve the term "squashed" for compression, but I think you just mean attenuated. So the signal is just quieter and can be turned up (intact).

The question of moving signal through various stages and not losing track of places where gain can be gained or lost is a matter of procedure. Controlling gains on inputs, of course, is the main place for tracking/mixing live.

But for example if signal was running through a compressor, and you were "squashing" the signal (in the compression sense) and you wanted to make sure you were coming out with basically a "gain neutral" output, how would the compressor possibly know? It would have to listen for a while, and find out how much gain reduction was happening over time and then tell you where to put the output knob. Not impossible though.
I don't think we're on the same wavelength here. What I said about turning the signal down at input (or anywhere else before it reaches the tracks) is indeed a potential problem. When you turn it up later to compensate, you raise the noise of all devices up to that point in the chain. The flute and background noise in the hall may come up proportionately to each other, but not to the noise of the interface. Further, if they are recorded at that low volume, and brought up in the mix, you've expanded a low-resolution sound, like enlarging a small JPEG to fit the screen. It depends on which levels we're talking about, of course, and where the noise enters the system. Maybe to make the example more realistic, I should have added a PreAmp in there, which is still in the analog domain, and used its levels.

But the confusion between what I'm saying and what you're interpreting is typical of that same confusion we get when we search for the offending level: what can we rule out? What could be causing the noise? Where in the chain would it have to be?

I have set up new speakers on at least one occasion which sounded like they were producing a lot of noise. It turned out to be up the chain. That same signal had not noticeably affected the previous speaker setup, which ran through a tube amp, but it affected the new powered speakers at the time. While fixing it wasn't all that hard to do, it made me groan for the recordings I had made with the offending levels of the past, because the audio wasn't at its optimum level for high-quality sound.

If only one level is at a sub-optimum setting, it forces compensation later in the chain. I wouldn't be complaining about it if I had not gone through it so many times. Even now, with 20 years of digital audio experience, I sometimes wonder if I've got all my levels set as well as possible.

Anyway, I think my logic is correct. Maybe my example wasn't the best, but it still works if you consider the self-noise of any device in the system.

As for your example of a compressor, I'm not saying that it should know what its output should be. That's your job, and you set it with your ears. I'm talking about all those levels that we rarely touch when we're recording and mixing, at least not after we've gotten our levels and have finalized what will be our input levels. But maybe I'm tilting at windmills. It's something I think about a lot when recording and mixing. That's all.

Shooshie
|l| OS X 10.12.6 |l| DP 10.0 |l| 2.4 GHz 12-Core MacPro Mid-2012 |l| 40GB RAM |l| Mach5.3 |l| Waves 9.x |l| Altiverb |l| Ivory 2 New York Steinway |l| Wallander WIVI 2.30 Winds, Brass, Saxes |l| Garritan Aria |l| VSL 5.3.1 and VSL Pro 2.3.1 |l| Yamaha WX-5 MIDI Wind Controller |l| Roland FC-300 |l|
User avatar
stubbsonic
Posts: 4601
Joined: Fri Dec 22, 2006 12:56 pm
Primary DAW OS: MacOS
Contact:

Re: My Signal Neutral rant

Post by stubbsonic »

Yea. I'm struggling to understand this.

Setting input levels at a pre-amp is just what we do. As you said about the compressor output, "that's my job".

Getting good gain structure in a whole system is tricky, to be sure.

But it seems like there are three main "sections" or chains:

1. Pre-digital, analog path: source (mic, gtr, etc.), pre-amp, monitor up to the interfaces analog input & controls (pre-AD)

2. Digital path: once in the machine- all the routings, effects, from the drive, from the input, through channels, effects, to outputs, etc.

3. Post-digital, analog path: after comes out of the digital realm-- through a DAC, into a mixer out to an amp or powered speaker, etc.

There are logical ways to manage the gain structure through each of those "sections" but I think you are onto something about manufacturers and developers doing what they can to clarify unity gain when it is possible, and optimum level when practical.
M1 MBP; OS 12, FF800, DP 11.3, Kontakt 7, Reaktor 6, PC3K7, K2661S, iPad6, Godin XTSA, Two Ibanez 5 string basses (1 fretted, 1 fretless), FM3, SY-1000, etc.

http://www.jonstubbsmusic.com
User avatar
Phil O
Posts: 7231
Joined: Thu Jul 28, 2005 10:01 pm
Primary DAW OS: MacOS
Location: Scituate, MA

Re: My Signal Neutral rant

Post by Phil O »

I think what you’re talking about is gain staging.

There’s this:
https://en.wikipedia.org/wiki/Gain_stage

and this:
http://www.soundonsound.com/sos/sep13/a ... headed.htm

I'm sure if you google "gain staging" you'll find tons of stuff on it. It's been the bane of engineers for a very long time. There's a real science to it, but it requires lots of measurements and fussing. Many do it empirically. Others are oblivious to it. If it really bugs you, Shoosh, it should. You'd think by now someone would have come up with some simple tools to optimize gain staging, but to the best of my knowledge it's still just fussy work.

Phil
DP 11.23, 2020 M1 Mac Mini [9,1] (16 Gig RAM), Mac Pro 3GHz 8 core [6,1] (16 Gig RAM), OS 14.3.1/11.6.2, Lynx Aurora (n) 8tb, MOTU 8pre-es, MOTU M6, MOTU 828, Apogee Rosetta 800, UAD-2 Satellite, a truckload of outboard gear and plug-ins, and a partridge in a pear tree.
User avatar
Shooshie
Posts: 19820
Joined: Sat Oct 16, 2004 10:01 pm
Primary DAW OS: MacOS
Location: Dallas
Contact:

Re: My Signal Neutral rant

Post by Shooshie »

Phil O wrote:I think what you’re talking about is gain staging.

There’s this:
https://en.wikipedia.org/wiki/Gain_stage

and this:
http://www.soundonsound.com/sos/sep13/a ... headed.htm

I'm sure if you google "gain staging" you'll find tons of stuff on it. It's been the bane of engineers for a very long time. There's a real science to it, but it requires lots of measurements and fussing. Many do it empirically. Others are oblivious to it. If it really bugs you, Shoosh, it should. You'd think by now someone would have come up with some simple tools to optimize gain staging, but to the best of my knowledge it's still just fussy work.

Phil
Thank you, Phil. Gain Staging is the term I was looking for. Not long ago, I probably would have written it that way. Yesterday I circled and circled around this "hole" in my memory. I knew it was there, but I couldn't see it. Gain Staging was the term in that hole. Now it's all in place; thanks.

The SOS article describes it much better than my rant, but they don't really talk about what happens when you set some stage too low, and how that effectively cuts off the flow, reduces the effective bit rate (not literally, but by relegating the relevant signal changes to a very small band), which cannot be corrected later. If somewhere among your stages you have a very low setting, followed by a very high setting to compensate in the next stage, you may not realize that it's there, and your audio may be sounding pretty low-res without your realizing why.

Phil's links are exactly what I'm talking about. Read them, not mine.

Shoosh
|l| OS X 10.12.6 |l| DP 10.0 |l| 2.4 GHz 12-Core MacPro Mid-2012 |l| 40GB RAM |l| Mach5.3 |l| Waves 9.x |l| Altiverb |l| Ivory 2 New York Steinway |l| Wallander WIVI 2.30 Winds, Brass, Saxes |l| Garritan Aria |l| VSL 5.3.1 and VSL Pro 2.3.1 |l| Yamaha WX-5 MIDI Wind Controller |l| Roland FC-300 |l|
User avatar
stubbsonic
Posts: 4601
Joined: Fri Dec 22, 2006 12:56 pm
Primary DAW OS: MacOS
Contact:

Re: My Signal Neutral rant

Post by stubbsonic »

I think you meant "Bit Depth" not "Bit Rate" (but that really is nit-picky).

I'm sure you know that every 6 dB of level is 1 bit of resolution. So if you have 24 bits, you could be down 48 dB and then you would be at the fidelity equivalent of 16 bit audio. I have heard of engineers pulling their hair out to get their signal to peak at -1 or -2 dB. I think this is fine for when you are dealing with files, mixing, and mastering. But it is impractical for setting live mic levels.

I think setting safe levels (so you aren't torturing the talent with too much level adjustment) and peaking at around -10 or -6 seems like a pretty reasonable thing. Of course, this will depend on the source material and instrument or singer.

When you describe a situation where you take something with a solid 20 bits of level (-24 dB) that you normalize, you are probably talking about an inaudible difference in resolution. Often the noise in your pre-amp would be boosted if you bumped the gain to get more resolution.

Most important to get the sound you want, get the performance you want, give yourself (and the talent) some wiggle-room headroom-wise, and focus on the music.

You know all this. I'm preachin' to the choir.
M1 MBP; OS 12, FF800, DP 11.3, Kontakt 7, Reaktor 6, PC3K7, K2661S, iPad6, Godin XTSA, Two Ibanez 5 string basses (1 fretted, 1 fretless), FM3, SY-1000, etc.

http://www.jonstubbsmusic.com
User avatar
Shooshie
Posts: 19820
Joined: Sat Oct 16, 2004 10:01 pm
Primary DAW OS: MacOS
Location: Dallas
Contact:

Re: My Signal Neutral rant

Post by Shooshie »

stubbsonic wrote:You know all this. I'm preachin' to the choir.
I'm the choir, all right, but I seem to have shown up for practice without my hymnal!

Shoosh
|l| OS X 10.12.6 |l| DP 10.0 |l| 2.4 GHz 12-Core MacPro Mid-2012 |l| 40GB RAM |l| Mach5.3 |l| Waves 9.x |l| Altiverb |l| Ivory 2 New York Steinway |l| Wallander WIVI 2.30 Winds, Brass, Saxes |l| Garritan Aria |l| VSL 5.3.1 and VSL Pro 2.3.1 |l| Yamaha WX-5 MIDI Wind Controller |l| Roland FC-300 |l|
User avatar
Phil O
Posts: 7231
Joined: Thu Jul 28, 2005 10:01 pm
Primary DAW OS: MacOS
Location: Scituate, MA

Re: My Signal Neutral rant

Post by Phil O »

Shooshie wrote:...If somewhere among your stages you have a very low setting, followed by a very high setting to compensate in the next stage, you may not realize that it's there, and your audio may be sounding pretty low-res without your realizing why...
...and if somewhere among your stages you have a very high setting, you may introduce distortion, and that's the nitty-gritty of it (pun intended). In fact, in the analog world (and even your digital equipment has analog stages), distortion is always present. It's just a matter of how much. Every amplifier has an area of best linearity/low distortion. Ideally you'd like to be in that area and I think that's the holy grail you coined signal neutral. To make matters worse, it's often a balancing act between distortion and noise floor. I understand you pain. :(

The good news is, the digital domain gives us a lot more wiggle room. No more noise reduction hardware. :woohoo:

But we have to deal with dithering. :banghead:

Philippe
DP 11.23, 2020 M1 Mac Mini [9,1] (16 Gig RAM), Mac Pro 3GHz 8 core [6,1] (16 Gig RAM), OS 14.3.1/11.6.2, Lynx Aurora (n) 8tb, MOTU 8pre-es, MOTU M6, MOTU 828, Apogee Rosetta 800, UAD-2 Satellite, a truckload of outboard gear and plug-ins, and a partridge in a pear tree.
User avatar
Shooshie
Posts: 19820
Joined: Sat Oct 16, 2004 10:01 pm
Primary DAW OS: MacOS
Location: Dallas
Contact:

Re: My Signal Neutral rant (Gain Staging)

Post by Shooshie »

Yeah, you said it, bro.

I've been dealing with it for years and years and years, but I'd just like a little help! When I add a new piece of gear, or change a configuration, or (heaven help us) start one from scratch... I'd just like a little guidance along the way. THIS setting works best full blast. THIS trim works best at the Unity Gain tic mark. THIS level works best somewhere in the 3 to 6 area. Yadda, yadda. I mean, I don't remember settings I made 5 years ago, nor why I set them there. Was that a special occasion? Was that meant to be "set and forget?"

I've been enjoying this new Project Notes window in DP9. I've been writing out all my cabling, connectors, channels, and operations of things like talk-back mic, which I still do with a regular channel, though I'd like to start using the Talkback Mic feature of CueMix as I used to do many years ago on another MOTU device and another version of CueMix that worked quite differently than the current one. There are so many connections, so many things that can change from project to project. You move this cable to device QRS, because you need all 8 channels of that one on XYZ. You ran out of 50' cables, so you connected two 25' cables, so the one that used to be purple at the microphone is now red. (I use color-coded cables) And so forth. The Project Notes really help with these kinds of things. Now, I wish we could specify chunk-specific project notes and general file notes. It's currently File-specific, not chunk specific. But I'm changing the subject.

To restate my original rant, I wish there were some indicator on every level control, hardware or software, that helps us see and set the optimum range for that particular level so our gain-staging can go quickly without so much testing. One who has tested and perfected his levels from mic to tracks to ears has acquired a set of knowledge about his own system that is valuable, and that can take hours (or years) to acquire.

Shooshie
|l| OS X 10.12.6 |l| DP 10.0 |l| 2.4 GHz 12-Core MacPro Mid-2012 |l| 40GB RAM |l| Mach5.3 |l| Waves 9.x |l| Altiverb |l| Ivory 2 New York Steinway |l| Wallander WIVI 2.30 Winds, Brass, Saxes |l| Garritan Aria |l| VSL 5.3.1 and VSL Pro 2.3.1 |l| Yamaha WX-5 MIDI Wind Controller |l| Roland FC-300 |l|
Post Reply