|
Post by matt@IAA on Jan 17, 2020 17:01:56 GMT -6
popmann javamadSo as to not derail UA thread. I am curious as to what the “math” being described as in error or truncation is. I assume (perhaps naively) that the sum in digital domain is an actual sum of all tracks on a per sample per bit basis, and this is a lossless sum (ie, as you sum 16 bit tracks you get a 16 bit output). So no dither, no truncation. This means the rounding error from excel isn’t as described, because rounding is truncation of the number of decimals, which would be to me the same as reducing bit depth. It seems that this means that while it isn’t infinitely accurate as analog, it’s exactly as accurate as the bit depth and sample rate you already recorded at...and if that’s not good enough, then it’s not summing that’s the problem but digital. I am quite happy to be educated about this if I’m not understanding correctly. I don’t know what math would be involved. Teach me, please.
|
|
|
Post by ragan on Jan 17, 2020 19:18:45 GMT -6
I’m interested too. I don’t know that much about it.
|
|
|
Post by popmann on Jan 17, 2020 19:22:01 GMT -6
Not a math guy. Dad taught it my whole life, call it my silly rebellion.....this guy explains what he’s doing, though I dont know just how in depth he goes. www.airwindows.com/purestconsole/Though, this doesn't solve the shortcomings of software digital summing. There is no solution at single rate digital IME. It just changes and explains the math difference in combining signals....resistor network compared to what every DAW does. To solve the whole bundle you need to talk about the non linear saturations....and the timing between left and right channels, which IME, is what double rate improves, and further to quad. Then how pan law is implemented....the “daw code” was never an attempt to do what an analog mixer does, more than functionally...ie, combines signals into stereo.
|
|
|
Post by christopher on Jan 17, 2020 21:00:13 GMT -6
Something brought up and maybe relevant is the panning issue talked about in this guy’s videos. He encourages users to do their own experiments/don’t take his word for stuff... :
|
|
|
Post by matt@IAA on Jan 17, 2020 22:29:50 GMT -6
Not a math guy. Dad taught it my whole life, call it my silly rebellion.....this guy explains what he’s doing, though I dont know just how in depth he goes. www.airwindows.com/purestconsole/Though, this doesn't solve the shortcomings of software digital summing. There is no solution at single rate digital IME. It just changes and explains the math difference in combining signals....resistor network compared to what every DAW does. To solve the whole bundle you need to talk about the non linear saturations....and the timing between left and right channels, which IME, is what double rate improves, and further to quad. Then how pan law is implemented....the “daw code” was never an attempt to do what an analog mixer does, more than functionally...ie, combines signals into stereo. It just says something about sines and arcsine. I don’t understand what that has to do with addition. I need an ELI5. I don’t understand what you wrote. Please forgive me for being dense. As far as I understand, summing is literally that. It’s just addition. In AC terms 1 volt at 1k sine plus 1 volt at 1k sine is a 2 volt 1k sine wave. If that’s a resistor or digital, I don’t understand the difference. If you can capture a digital sample of an analog wave (sample >> nyquist frequency) it seems like a sum of the bit value of each sample would be all she wrote. Likewise with pan law..again, as far as I understand how analog works it’s just an attenuation curve. It simultaneously lowers the voltage level on one side as it raises the other. No nonlinearity. No saturation. Just a simple resistor pad. I mean if panning doesn’t work neither should the fader. They’re the same thing.
|
|
|
Post by popmann on Jan 17, 2020 22:52:32 GMT -6
No, pan law has nothing to do with saturation. I'm listing elements that a 48khz "standard DAW mixer" makes one sound different than another and/or fails to do similarly enough to analog. One doesn't inherently have much to do with the other. Some people are happy with different pieces being fixed/ band aided. Until budgets fell through the floor, most anyone who needed to mix did so on an analog desk. I maintain that a cheap Mackie will produce a better sum than any DAW you want to put up against it....and it's not like Mackie has some kind of euphonic goodness making saturation. It is my experience that they all work together to make the sound of a mixer.
I hear what you're saying that pan law adjustment is just a gain offset like a fader. I am only repeating what a designer of digital desks told me on that one....if I can hear and understand, I don't really care that much to get into the EE meets algorithm side of understanding.
|
|
|
Post by ragan on Jan 17, 2020 23:01:34 GMT -6
No, pan law has nothing to do with saturation. I'm listing elements that a 48khz "standard DAW mixer" makes one sound different than another and/or fails to do similarly enough to analog. One doesn't inherently have much to do with the other. Some people are happy with different pieces being fixed/ band aided. Until budgets fell through the floor, most anyone who needed to mix did so on an analog desk. I maintain that a cheap Mackie will produce a better sum than any DAW you want to put up against it....and it's not like Mackie has some kind of euphonic goodness making saturation. It is my experience that they all work together to make the sound of a mixer. I hear what you're saying that pan law adjustment is just a gain offset like a fader. I am only repeating what a designer of digital desks told me on that one....if I can hear and understand, I don't really care that much to get into the EE meets algorithm side of understanding. But the "EE meets algorithm side of understanding" is where the actual truth of the matter lies. It's the story of what's actually going on, regardless of what I believe I hear, no? Also, for whatever it's worth, I've tried getting the Magic Summing Goodies out of Mackies, RAMSAs, A&H and a few other budget mixers. Never delivered the goods for me. Either sounded functionally the same or sounded a (tiny) bit smeary. A given person might prefer that smear but it's certainly not what I associate with the classic desks I've had the pleasure to hear. Less width and depth and euphonic realism (with the Mackie et al), not more.
|
|
|
Post by popmann on Jan 17, 2020 23:20:29 GMT -6
Not really. I hear what you're saying....and sure....on some level-of course "that's the truth of the matter", but not a functional level for an audio engineer or musician recording themselves. I'm making the best recordings of my life. At home. And now completely in software.
It's to the point that I'm not sure I could unintentionally even fuck it up working too fast or something. The sound, I mean. I've got a track on here that I added a single EP track to (removing some faux strings I did in the demo)--and the whole thing is a the day one songwriting demo. I've left it, fought urges to recut anything to prove a point to myself. The sonics are there from "go" with this system. Fight another fight.
|
|
|
Post by ragan on Jan 17, 2020 23:26:09 GMT -6
Not really. I hear what you're saying....and sure....on some level-of course "that's the truth of the matter", but not a functional level for an audio engineer or musician recording themselves. I'm making the best recordings of my life. At home. And now completely in software. It's to the point that I'm not sure I could unintentionally even fuck it up working too fast or something. The sound, I mean. I've got a track on here that I added a single EP track to (removing some faux strings I did in the demo)--and the whole thing is a the day one songwriting demo. I've left it, fought urges to recut anything to prove a point to myself. The sonics are there from "go" with this system. Fight another fight. If I understand what you mean, then hell yeah. The proof of the pudding and all that. But you were talking about technical details which have objective answers (ie summing math, pan law, the ubiquitous "nonlinear" concept). That's the paradigm I'm speaking to when I say that the EE/algorithm side is where the actual truth of the matter resides. It's one thing to say, "Look, I'm getting great results and having a helluva great time making music and I'm just not really interested in the technical details". It's another thing to say "...[technical claims about what's going on under the hood]...but I'm not really interested in the EE/algorithm side." Or maybe I misunderstand you.
|
|
|
Post by drbill on Jan 18, 2020 0:03:07 GMT -6
I maintain that a cheap Mackie will produce a better sum than any DAW you want to put up against it.... That's not been my experience. For what it's worth.
|
|
|
Post by nick8801 on Jan 18, 2020 7:47:14 GMT -6
I always find it interesting when people say they can hear more separation in an analog sum. If that’s happening it must just be a bit of saturation bringing out the different elements in the mix. If you have a good ear and can mix well, it’s pretty easy to get separation from a DAW sum. There’s and old saying that our ears/brains can really only focus on 3 things at once. If you want separation, just make less stuff important at any given time in a mix. If you still need that saturation(which I do!) then just get a nice stereo outboard chain going, but I still think I can get cleaner mixes by staying in the box and not reconverting everything.
|
|
|
Post by matt@IAA on Jan 18, 2020 8:40:54 GMT -6
Sometimes imperfection can sound like separation. I think some of this is similar to the way double tracked and hard panned guitar tracks sound so much bigger. Taking the same effect in across an analog unit with ~1% variance - both level and distortion - on each channel is a lot of what people hear... I think. I don’t know. I almost wonder if this isn’t a similar effect as a digital dither. You add some slight degree of variance that matters only at the fringes. Kind of makes me want to go back and re-test summing and focus specifically on the quietest portions of the mix.
I still want to know what the actual problem was with early summing, because there’s quite a bit of people who say there was a problem, but it’s been fixed. ??
|
|
|
Post by trakworxmastering on Jan 18, 2020 10:01:33 GMT -6
I still want to know what the actual problem was with early summing, because there’s quite a bit of people who say there was a problem, but it’s been fixed. ?? As I recall, in the '90s the first DAWs were 16 bit audio with maybe 16 bit or 24 bit fixed depth audio engines and poorly implemented (or non-implemented) dither, so headroom and resolution in all of the busses was an issue. Also, AEs were just starting to learn how to use DAWs well. I think that's when digital summing got a bad name. Over time they upgraded to 32 bit float, then 64 bit float with proper dithering, making DAW summing much better than before. I think much of the disdain for digital summing is left over old information from back when it was true. Times change. I abandoned my analog summing mixer at some point after I upgraded to 32 bit float Pro Tools...
|
|
|
Post by mrholmes on Jan 18, 2020 11:12:12 GMT -6
I remember when the summing topic, truncation error etc. was big on GS. I asked a friend if we can run files with test tones for leveling via his AWS 900. We both came to the conclusion that summing is not a big problem on the computer... Artifacts that was the lack of the DAW sum.
|
|
|
Post by stormymondays on Jan 18, 2020 11:45:10 GMT -6
Re. Analog summing, many people were fooled by the change in panning laws when routing the mix to the hardware units. You need to be aware of the panning law of your daw and that of the external hardware (if any!).
|
|
|
Post by EmRR on Jan 18, 2020 15:31:49 GMT -6
Some of the problem with PT summing way back sounded like CPU/DSP bottlenecks leaving errors during bounces. I recall comparisons of bounces (or whatever PT calls it) versus real time routing to a track that's record enabled, and there'd usually be microscopic moments that wouldn't cancel all over the place, big spikes that would reveal an error in one path.
|
|
|
Post by matt@IAA on Jan 18, 2020 19:53:37 GMT -6
Some of the problem with PT summing way back sounded like CPU/DSP bottlenecks leaving errors during bounces. I recall comparisons of bounces (or whatever PT calls it) versus real time routing to a track that's record enabled, and there'd usually be microscopic moments that wouldn't cancel all over the place, big spikes that would reveal an error in one path. Makes me wonder what the difference would be between an online and offline render / bounce. Seem like this is like saying you can’t render video well digitally because real time rendering has quality issues.
|
|
|
Post by EmRR on Jan 18, 2020 20:03:59 GMT -6
Some of the problem with PT summing way back sounded like CPU/DSP bottlenecks leaving errors during bounces. I recall comparisons of bounces (or whatever PT calls it) versus real time routing to a track that's record enabled, and there'd usually be microscopic moments that wouldn't cancel all over the place, big spikes that would reveal an error in one path. Makes me wonder what the difference would be between an online and offline render / bounce. Seem like this is like saying you can’t render video well digitally because real time rendering has quality issues. The reports with PT are 15 years or so back, with real time record path preferred. I seem to recall multiple real time would cancel perfectly, multiple bounces would not against either a real time or another bounce. Those days and problems should be long gone.
|
|
|
Post by trakworxmastering on Jan 19, 2020 13:05:49 GMT -6
Makes me wonder what the difference would be between an online and offline render / bounce. Seem like this is like saying you can’t render video well digitally because real time rendering has quality issues. The reports with PT are 15 years or so back, with real time record path preferred. I seem to recall multiple real time would cancel perfectly, multiple bounces would not against either a real time or another bounce. Those days and problems should be long gone. Yeah I remember that back when PT Mix or Mix Plus TDM systems were in use. I think it was automation errors when bouncing to disk vs printing to a track. It's ancient history now. But it still helped give digital summing a bad name...
|
|
|
Post by johneppstein on Jan 19, 2020 13:52:03 GMT -6
Not a math guy. Dad taught it my whole life, call it my silly rebellion.....this guy explains what he’s doing, though I dont know just how in depth he goes. www.airwindows.com/purestconsole/Though, this doesn't solve the shortcomings of software digital summing. There is no solution at single rate digital IME. It just changes and explains the math difference in combining signals....resistor network compared to what every DAW does. To solve the whole bundle you need to talk about the non linear saturations....and the timing between left and right channels, which IME, is what double rate improves, and further to quad. Then how pan law is implemented....the “daw code” was never an attempt to do what an analog mixer does, more than functionally...ie, combines signals into stereo. It just says something about sines and arcsine. I don’t understand what that has to do with addition. I need an ELI5. I don’t understand what you wrote. Please forgive me for being dense. As far as I understand, summing is literally that. It’s just addition. In AC terms 1 volt at 1k sine plus 1 volt at 1k sine is a 2 volt 1k sine wave. If that’s a resistor or digital, I don’t understand the difference. If you can capture a digital sample of an analog wave (sample >> nyquist frequency) it seems like a sum of the bit value of each sample would be all she wrote. Likewise with pan law..again, as far as I understand how analog works it’s just an attenuation curve. It simultaneously lowers the voltage level on one side as it raises the other. No nonlinearity. No saturation. Just a simple resistor pad. I mean if panning doesn’t work neither should the fader. They’re the same thing. The thing about pan law is the same whether it's analog or digital, it's just that in digital one has the luxury of being able to select one's choice of pan laws easily, whereas in analog it's a matter of messy sunstitution of hardware resistors/pots, so it usually isn't a "choice", it's bakede into console selection.
Pan law isn't about pan "working", it's about how many dB down the output of the signal is on each side of the stereo buss when panned dead center as opposed to when it's hard panned.
|
|
|
Post by johneppstein on Jan 19, 2020 13:57:28 GMT -6
I maintain that a cheap Mackie will produce a better sum than any DAW you want to put up against it.... That's not been my experience. For what it's worth. It's been my experience that cheap Mackies don't have very well engineered summing busses. And I've owned a few.
|
|
|
Post by johneppstein on Jan 19, 2020 14:00:17 GMT -6
Not really. I hear what you're saying....and sure....on some level-of course "that's the truth of the matter", but not a functional level for an audio engineer or musician recording themselves. I'm making the best recordings of my life. At home. And now completely in software. It's to the point that I'm not sure I could unintentionally even fuck it up working too fast or something. The sound, I mean. I've got a track on here that I added a single EP track to (removing some faux strings I did in the demo)--and the whole thing is a the day one songwriting demo. I've left it, fought urges to recut anything to prove a point to myself. The sonics are there from "go" with this system. Fight another fight. Do you really think it's a matter of analog vs digital? Or perhaps it's simply that your mixing skills have got a lot better?
|
|
|
Post by lando on Jan 19, 2020 14:42:33 GMT -6
If anyone on the forum had a somewhat clean sounding summing box and could export one mix with that and one with digital summing this could be settled once and for all. With LCR panning to avoid differences in pan law and no random reverbs my guess is that it would prove digital and clean analog summing to be extremely similar, but I would love to be proven wrong.
|
|
|
Post by matt@IAA on Jan 19, 2020 15:03:02 GMT -6
To do what you’re describing you don’t need anything special. Just one resistor per channel. The whole thing can be done in an afternoon for a couple of bucks.
|
|
|
Post by stormymondays on Jan 19, 2020 15:51:14 GMT -6
If anyone on the forum had a somewhat clean sounding summing box and could export one mix with that and one with digital summing this could be settled once and for all. With LCR panning to avoid differences in pan law and no random reverbs my guess is that it would prove digital and clean analog summing to be extremely similar, but I would love to be proven wrong. But the summing box needs a panning law. Or if it has none, which would be most usual I think, you need to disable it in your DAW, which WILL change the mix. This is what has skewed this kind of test on "other forums".
|
|