|
Post by schmalzy on Jun 29, 2022 16:02:29 GMT -6
That's the way I'm interpreting the results as well: if you isolate only the signal-combining-step of summing only - combining voltages or combining bits is non-noteworthy.
Anything else you do along the way is what changes the result that people get via summing, etc..
I wasn't early as scientific as Dan or others here, but I tested a passive summing device for a while. Passive summing with clean makeup gain gave me nothing worthwhile. ITB summing going to clean line input gave me nothing worthwhile. Passive summing with gain from a Neve-style preamp's mic input gave me something. ITB summing going to line input gain from that Neve-style pre gave me something else. Passive summing with clean makeup gain then into the colored line input gain from that Neve-style pre gave me a very similar something else. The summing itself didn't seem to give me anything interesting or noteworthy.
I figured liked the passive summing into the mic pre because I liked hearing the makeup gain of the preamp's mic input over the preamp's line input. Looked up some specs on my preamp; yep, the mic input transformer is a much more colorful transformer than the line input transformer.
Moreover, I liked the result of using passive summing more when I: ran my drum bus to a hardware colored compressor and EQ then into the passive summing my bass bus to a hardware compressor then to passive summing and my guitar bus to a hardware line in/eq/compressor then to passive summing then the passive summing to my colored mic pre.
|
|
|
Post by svart on Jun 29, 2022 16:04:42 GMT -6
Won't work. By nature white noise is random so there would be no way to phase shift it 180 degrees to null because there would be no correlation. You need a repeating signal that can be correlated before polarity reversal can null it. I know you know more about this stuff than I, but if you recorded noise onto 16 tracks and used those tracks for all of your tests, random as the noise may or may not be, how is that any different than playing back actual music, for the purposes of such a test? It's still comparing two identical signals, (two identically random signals, if you will), but just 180 degrees out of phase from one another. The generator creating said noise might be random but, once it's recorded onto a track, it shouldn't matter what the signal even is beyond that point. All that should matter is that the return signal is phase flipped for the purposes of the test, no? In any case, it could be whatever kind of full frequency signal you wanted. I was just throwing it out there that it might not be unuseful to have something that is truly full frequency spectrum as a test signal. Also, to be clear, I actually think I tend to agree with your position that analog summing doesn't do much, if anything, in and of itself, and that it's the gain stages that are actually what makes people like analog summing. I'm mostly just discussing the topic out of curiosity and interest because it does seem to me that there could possibly be a better way to perform such a test than what was done in the video. Ok, I see what you mean. If you had one track of noise rendered but you duplicated it many times then yes, theoretically since they're now correlated, they should null same as a sine. A sine sweep would work like this as well.
|
|
|
Post by Quint on Jun 29, 2022 16:11:06 GMT -6
I know you know more about this stuff than I, but if you recorded noise onto 16 tracks and used those tracks for all of your tests, random as the noise may or may not be, how is that any different than playing back actual music, for the purposes of such a test? It's still comparing two identical signals, (two identically random signals, if you will), but just 180 degrees out of phase from one another. The generator creating said noise might be random but, once it's recorded onto a track, it shouldn't matter what the signal even is beyond that point. All that should matter is that the return signal is phase flipped for the purposes of the test, no? In any case, it could be whatever kind of full frequency signal you wanted. I was just throwing it out there that it might not be unuseful to have something that is truly full frequency spectrum as a test signal. Also, to be clear, I actually think I tend to agree with your position that analog summing doesn't do much, if anything, in and of itself, and that it's the gain stages that are actually what makes people like analog summing. I'm mostly just discussing the topic out of curiosity and interest because it does seem to me that there could possibly be a better way to perform such a test than what was done in the video. Ok, I see what you mean. If you had one track of noise rendered but you duplicated it many times then yes, theoretically since they're now correlated, they should null same as a sine. A sine sweep would work like this as well. Sure. A sine sweep would work too.
|
|
|
Post by matt@IAA on Jun 29, 2022 16:22:15 GMT -6
What I’m saying is… summing is just resistors tied together. No one would (or should) expect that a single resistor is going to have a sound. Passive and active summing deal with how the resulting combined signals are handled or amplified. It’s. A high gain operation. Limiting summing to resistors without the combining or makeup amp seems kind of like evaluating a mic pre using only a cable. I think you're missing the point of the discussion though. I get that you're saying makeup gain is a necessary component IF you're going to use passive summing. What is being discussed though is if the wire and resistors themselves actually bring anything to the party or if it all truly comes down to the makeup gain stages inherent to a summing mixer. If it's the former, then there remains some validity to the idea of summing in the analog domain. If it's the latter, one could just introduce those same gain stages at the same points in the signal flow and basically achieve the same thing. Makeup gain is required for passive or active summing. It’s part of it, no gain is like separating mic pre gain from a mic pre. The reason summing has a sound is from the gain required of the summing amp. Summing is potentially as heavy a demand on an op amp (or more!) as a mic pre. So to your point… no.. wire and resistors don’t bring much of anything to the party. Yeah, it’s probably closely approximated by digital summing, digital pad to drop the level the appropriate amount, then analog gain. I don’t think it’ll be exactly the same because of minor details in the circuit, impedances, imperfections between channels, cross talk, etc. but probably very close.
|
|
|
Post by Quint on Jun 29, 2022 16:29:43 GMT -6
I think you're missing the point of the discussion though. I get that you're saying makeup gain is a necessary component IF you're going to use passive summing. What is being discussed though is if the wire and resistors themselves actually bring anything to the party or if it all truly comes down to the makeup gain stages inherent to a summing mixer. If it's the former, then there remains some validity to the idea of summing in the analog domain. If it's the latter, one could just introduce those same gain stages at the same points in the signal flow and basically achieve the same thing. Makeup gain is required for passive or active summing. It’s part of it, no gain is like separating mic pre gain from a mic pre. The reason summing has a sound is from the gain required of the summing amp. Summing is potentially as heavy a demand on an op amp (or more!) as a mic pre. So to your point… no.. wire and resistors don’t bring much of anything to the party. Yeah, it’s probably closely approximated by digital summing, digital pad to drop the level the appropriate amount, then analog gain. I don’t think it’ll be exactly the same because of minor details in the circuit, impedances, imperfections between channels, cross talk, etc. but probably very close. That's what I've been saying.
|
|
|
Post by christopher on Jun 29, 2022 16:51:39 GMT -6
Its hard to argue about this stuff in little comment boxes and make sense.
I came to a similar conclusion as Svart on this stuff a few years ago: If you think of signal as data leaving our converters, then everything should null 100% unless it got corrupted along the way. It won’t get corrupted if you send it though anything designed to keep the signal as accurate as it can.
When things don’t null, why not? My conclusion if there’s almost no to negligible distortion added -as in the way of many pro audio designs- then it can get corrupted by only a few reasons:
- volume is not perfectly matched - volume at certain frequencies are not flat - phase smear - band limited somewhere - time domain slippage (not an issue when clocking from same clock)
None of these tell you how something sounds, right?
So if you can adjust gain better like a robot, you’ll null most any signal perfectly as long as the FR didn’t get jacked up. Remember to think of sound as a data signal.
Analog pots aren’t as easy to adjust with the precision of digital faders. Analog almost never nulls completely, esp because our fat fingers have no chance to ever dial it perfectly. But if we could I’m sure we’d be surprised how perfectly null-able all our gear COULD be. When I watch myself adjust a sine with an analog pot, my tiniest move jumps the signal 0.40 dB, whereas at the DAW fader I can dial in 0.01 at a time.
So if we get volume to match up as best as possible, and it still doesn’t null.. why doesn’t it null completely?
Go down the list I made above, correct those things and it should null completely… you will be left with the noise floor and whatever remnants from artifacts of corrupted signal.
We shouldn’t hear any music at all, if we do that’s our failure to match things up in volume, response, phase.
Final conclusion: we don’t use any equipment just to change any of that stuff that nulls. We use it because it sounds good. Whatever it does is in its topology, maybe we WANT to corrupt the signal. Maybe we want the Class A tone instead of Class AB. Or vice versa. We want the artifacts that are dancing around the noise floor when there’s no signal applied at all. And after a null, say it created some new harmonics from pushing things. If you subtract out the original signal, the harmonics should remain, but they are such low signal riding atop the source I’d expect them to be in the noise floor mostly. It’s not like we want to turn things into a fuzz box, we want almost imperceptible amounts of harmonic goodness, barely registering
|
|
|
Post by EmRR on Jun 29, 2022 17:05:23 GMT -6
Wire and resistors themselves actually bring NOTHING to the party. Simple version.
|
|
|
Post by matt@IAA on Jun 29, 2022 18:02:34 GMT -6
Except noise and capacitive crosstalk and rf interference etc. nothing good generally!
|
|
|
Post by seawell on Jun 29, 2022 18:51:54 GMT -6
Except noise and capacitive crosstalk and rf interference etc. nothing good generally! What would you say are the good parts of a well built summing box?
|
|
|
Post by plinker on Jun 29, 2022 21:17:51 GMT -6
Except noise and capacitive crosstalk and rf interference etc. nothing good generally! I've built all these aspects into my forthcoming analog summing modeler plugin. It uses neural nets-n-shit. You'll LOVE IT!
|
|
|
Post by OtisGreying on Jun 29, 2022 21:20:18 GMT -6
Well I guess now I’ll throw away my passive-resistor-and-wire-only 64 channel analog summing box that I’ve been building...
|
|
|
Post by seawell on Jun 29, 2022 21:22:36 GMT -6
Except noise and capacitive crosstalk and rf interference etc. nothing good generally! I've built all these aspects into my forthcoming analog summing modeler plugin. It uses neural nets-n-shit. You'll LOVE IT! Some Youtuber is going to run it through plug-in doctor and ruin your life 🤣
|
|
|
Post by plinker on Jun 29, 2022 21:27:07 GMT -6
I've built all these aspects into my forthcoming analog summing modeler plugin. It uses neural nets-n-shit. You'll LOVE IT! Some Youtuber is going to run it through plug-in doctor and ruin your life 🤣 DEAR LORD, NO!
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Jun 29, 2022 22:40:37 GMT -6
Some Youtuber is going to run it through plug-in doctor and ruin your life 🤣 DEAR LORD, NO!
That British guy with the audient interface and no monitors and Dan Worral criticized the Harrison 32C channel strip for not distorting. Dan Worral also tried to null the Harrison eq which is a semi parametric, almost equal power proportional Q EQ like API 550, Oxford Type 4, Slick EQ American type, to Reaper's ReaEQ, which is an almost constant q parametric eq with even less gain:q dependency than an SSL 4000 E. Like yeah you can null them to the dither in the Harrison strip but you'll never get the same results irl. Then he criticized the clean dynamics that have program dependent holds and releases to not distort and the limiter because... it's an instant volume modulator with a program dependent hold time and not a soft clipper like Molot's because it sounds better for nuking a drum. So of course Harrison is better for smashing down overshoots and words while Molot's limiter would clip them clean off. You'd need two instances of Molot to level and limit vocals and it would be more colored and cpu heavy than two instances of the harrison strip.
|
|
|
Post by plinker on Jun 29, 2022 23:29:47 GMT -6
That British guy with the audient interface and no monitors and Dan Worral criticized the Harrison 32C channel strip for not distorting. Dan Worral also tried to null the Harrison eq which is a semi parametric, almost equal power proportional Q EQ like API 560, Oxford Type 4, Slick EQ American type, to Reaper's ReaEQ, which is an almost constant q parametric eq with even less gain:q dependency than an SSL 4000 E. Like yeah you can null them to the dither in the Harrison strip but you'll never get the same results irl. Then he criticized the clean dynamics that have program dependent holds and releases to not distort and the limiter because... it's an instant volume modulator with a program dependent hold time and not a soft clipper like Molot's because it sounds better for nuking a drum. So of course Harrison is better for smashing down overshoots and words while Molot's limiter would clip them clean off. You'd need two instances of Molot to level and limit vocals and it would be more colored and cpu heavy than two instances of the harrison strip.
After I've made bank on the analog summing plugin, I'm going to create a plugin emulation of that British guy with the audient interface and no monitors.
...and Dan Worral too.
You won't believe it's not the real thing!
|
|
|
Post by seawell on Jun 30, 2022 0:23:15 GMT -6
That British guy with the audient interface and no monitors and Dan Worral criticized the Harrison 32C channel strip for not distorting. Dan Worral also tried to null the Harrison eq which is a semi parametric, almost equal power proportional Q EQ like API 560, Oxford Type 4, Slick EQ American type, to Reaper's ReaEQ, which is an almost constant q parametric eq with even less gain:q dependency than an SSL 4000 E. Like yeah you can null them to the dither in the Harrison strip but you'll never get the same results irl. Then he criticized the clean dynamics that have program dependent holds and releases to not distort and the limiter because... it's an instant volume modulator with a program dependent hold time and not a soft clipper like Molot's because it sounds better for nuking a drum. So of course Harrison is better for smashing down overshoots and words while Molot's limiter would clip them clean off. You'd need two instances of Molot to level and limit vocals and it would be more colored and cpu heavy than two instances of the harrison strip.
After I've made bank on the analog summing plugin, I'm going to create a plugin emulation of that British guy with the audient interface and no monitors.
...and Dan Worral too.
You won't believe it's not the real thing!
Will you please name it Mr. Doubtfire?
|
|
|
Post by plinker on Jun 30, 2022 0:34:43 GMT -6
After I've made bank on the analog summing plugin, I'm going to create a plugin emulation of that British guy with the audient interface and no monitors.
...and Dan Worral too.
You won't believe it's not the real thing!
Will you please name it Mr. Doubtfire? Day-yem!! Talk about Throwback Thursday!
|
|
|
Post by seawell on Jun 30, 2022 1:30:29 GMT -6
Hey while we're at it...someone post files and let's bounce a mix(no plug ins or anything) with different DAWs and see if they null 🤔
|
|
|
Post by thehightenor on Jun 30, 2022 1:54:50 GMT -6
In my very humble final analysis.
Mixers had a purpose, they had pre amps, dynamics, EQ, monitoring, routing and I/O.
Not that many years ago you had to have one to make records.
Then came along the DAW that took care of everything except preamps - those became available in endless rack and 500 series units and of course outboard EQ and dynamics for hybrid mixing.
(well actually an easy workflow got thrown out of the window along with the desk but that a different point)
What was left from the console?
It appears to me most people mistakenly presumed as I did “the sound of the console” was the mixers internal wired mix bus when it was actually the sound of its preamps, transformers, dynamics, EQ and class A buffered I/O.
Not the bits of wire internally connecting all the channels together.
Those that have a summer and use pres to add gain on the O/P can just do that to two channels and get the same result (I’ve null tested that and Dan Worrall is spot on)
I’ve long wanted a Neve 8424 (I don’t have the 30K sitting around to buy one) and Neve tried to convince me it has the “mix bus” sound of the Neve 80 series (I wasn’t buying into it at all)
I don’t want to restore the bits of wire internally connecting the channels of a console (they contribute nothing)
I want to restore the console workflow to my music making!
|
|
|
Post by jampa on Jun 30, 2022 2:48:57 GMT -6
At this rate, I expect DW will chime in soon
|
|
|
Post by matt@IAA on Jun 30, 2022 5:52:18 GMT -6
What would you say are the good parts of a well built summing box? It’ll sum the signals in a pleasant sounding way, which if it sounds different than ITB means distortion. The more signals you sum the more gain you’ll need, the more distortion you get, the more sound it has. I think crosstalk could probably be part of the glue people here too, along with whatever nonlinear stuff comes with the rest of the box (input circuitry etc)
|
|
|
Post by svart on Jun 30, 2022 6:42:14 GMT -6
Ok, I see what you mean. If you had one track of noise rendered but you duplicated it many times then yes, theoretically since they're now correlated, they should null same as a sine. A sine sweep would work like this as well. Sure. A sine sweep would work too. I did a little test with white noise. Made one track of white noise, duplicated it, flipped polarity and it nulled perfectly as expected due to the correlation. I also tried two channels of random uncorrelated white noise and flipping polarity did not do anything, as expected. For fun I wanted to see what kind of nulling would happen at various amplitude and phase mismatches. 0.01dB only nulled down to -65dB. 0.1dB only nulled down to -45dB. 1dB only nulled down to -25dB. 0.01 degrees nulled down to -80dB. 0.1 degrees nulled down to -60dB. 1 degree nulled down to -40dB. Which makes his nulling quite impressive since his noise floor is around -80 to -90dB. He does have some kind of signal around 15K-18K-ish which might be some nyquist noise from the converters folded over but I can't hear it.
|
|
|
Post by svart on Jun 30, 2022 6:55:29 GMT -6
Except noise and capacitive crosstalk and rf interference etc. nothing good generally! Not much of those happening with resistors. Resistors physically parallel wouldn't have much of an EMI field and they are much too short in physical length to be of any significant fractional wavelength of the audio band to be a good antenna. Traces probably have the worst effect on total noise ingress and crosstalk simply for being long and parallel to each other. The best rule of thumb for crosstalk is keep other traces at least 3x the width of your traces apart from each other. If you run ground between them, keep the traces at least 3x away from the ground and via the ground islands to the ground plane to short out the EMI fields. Wider traces might help a little since the resistance of the copper will be reduced and then dominate the RLC components. C might rise, but shouldn't be high enough to roll off audio, but it might actually help keep it a bit cleaner.
|
|
|
Post by svart on Jun 30, 2022 7:05:55 GMT -6
Also, I should note that the residual power supply hum/buzz we hear in the video is one of the biggest reasons I went mostly digital. My mixing console and all the outboard gear, preamps, effects, etc, might have had noise floors in the -80dB to -90dB range, but since most of it was powered from the same AC circuit, a lot of the rectified buzz was correlated. You sum a dozen outboard units, all with analog power supplies and you start to get noise floors in the -65dB range.
Add in the audio tracks with their noise floors, add some compression and all the other fun stuff you do to mixes and suddenly you can sometimes hear the buzzing at -45dB to -50dB on your final mix down track.
|
|
|
Post by Quint on Jun 30, 2022 8:14:47 GMT -6
Also, I should note that the residual power supply hum/buzz we hear in the video is one of the biggest reasons I went mostly digital. My mixing console and all the outboard gear, preamps, effects, etc, might have had noise floors in the -80dB to -90dB range, but since most of it was powered from the same AC circuit, a lot of the rectified buzz was correlated. You sum a dozen outboard units, all with analog power supplies and you start to get noise floors in the -65dB range. Add in the audio tracks with their noise floors, add some compression and all the other fun stuff you do to mixes and suddenly you can sometimes hear the buzzing at -45dB to -50dB on your final mix down track. I've fought like hell to get noise down every where I can, but, like you say, there still is residual noise buildup when you connect a bunch of analog gear together. Some of that shit will drive you mad. It's certainly one of the things that tempts me to go more ITB, but I doubt I'll ever go fully. I actually like a slight amount of hiss, like you'd get from a tape machine, so I just cover up low level hums/buzzes in that!
|
|