Dangerous 2bus+

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
Are you summing 32 stereo busses? Also, are you processing the groups or just doing ouboard summing to avoid the DAW mix bus?

Maybe a DSP expert can offer info as to whether track count in a DSP mixer affects its behavior in any meaningful way.
I am not arrogant enough to self identify as a DSP expert but I dabbled in DSP and crunched a lot of microcode. Inside the digital domain some things are easier than others. Summing is a relatively easy one clock tic add. Fader moves are likewise a one clock tic multiply. EQ is a little more complicated with coefficients that can be difficult to represent using fixed point with enough resolution to prevent issues.

Modern processors have become massively faster than decades ago. One potential track count concern is are there enough clock tics available to perform all the math operations within one audio sample (tens of microseconds).

The designer/coder should do this math before you encounter limitations.

JR
Truthfully, my clients in all genres seem to prefer in-the-box mixes better than when I is summing. This even goes for pop, rock, r&b, jazz, folk, and classical music. When I sum I always mix with the summing setup before I start so my decisions are based on what I hear.
 
So I tried completely bypassing the dbox. There is a loss of clarity when running through the dbox. I've noticed it before with the summing but never tried the monitor controller by itself. Interesting.

The next question would be who actually makes a fully transparent monitor controller. Grace? Crane Song? Or just go passive and short cables?

The Grace is the best mainstream full featured analog monitor controller but it does have an impact sonically. Much less than the dBox or lower grade but an impact nonetheless, I would not use one in mastering for example. And I have used hundreds and been across AB tests quite a lot. The Grace built in DAC is very poor, best avoided.

I owned an Avocet for years and again have use them many times, not transparent IMO. Even the latest model.

If you need features like dim, speaker switching etc the Sound Skuptor is very good, sounds better than anything mainstream and commercial and is a fun build. No remote though (although there are some DIY options).

Then you get into esoterica like the Imperium which I'll try next week, the Topping stuff which I just bought to audition.
 
Well, no one. For everyone who likes something there is someone who doesn't. Passive isn't always transparent either. Working a passive monitor into a modern voltage transfer system is no easy task. The impedance of the attenuator has to be high enough to present a bridging load to the source and low enough to not be a problem for the power amplifier. There is no ideal value.

Agreed, passive is not the panacea. But if you understand the trade offs you can build something that works well within constraints. Good active is a better more general solution.
 
I'm currently using the Antelope Eclipse in my mastering suite as my monitor controller, and with my DAW I connect my interface to the Eclipse via SPDIF and use the Antelope's D/A rather than my interface's D/A. I do not like the headphone output of the Antelope, so I use a pair of analog outs from my interface directly into a 20-year-old Hafler power amp and then directly to my 'phones. I use a DAW fader for my headphone volume control, which works for me, but I wouldn't want that in a studio with guest engineers! The Eclipse has been great, except for the control panel software that won't run on any Mac OS past High Sierra.

Every monitor controller will sound different, but what is really correct anyway? I've had many interfaces and monitor controllers and the two that I personally felt sounded super clean are the Antelope Eclipse (via digital ins and its own D/A) and an analog Studio Technologies StudioComm. Even though the StudioComm uses VCAs (THAT 2180) I found it much cleaner than other VCA controllers, especially the SSL9000J monitor section! The Grace and Avocet (via their digital inputs) sound excellent, as well.

I have to work with one device for a few days and then switch to another before I can decide if one works better for me than another. Really, they all sound fantastic and any differences in mixing or mastering decisions that I would make due to the coloration of the monitor controller certainly won't affect the success or consumption of the music. There is way more sonic variation caused by speaker choice, component aging, listening position, bad acoustics, and a million other things than by the circuitry in a well-made monitor controller. Not to mention how much caffeine I consumed that day!

Shockingly, at the studio where I'm chief almost all of our visiting engineers and clients monitor a pair of stereo outputs of the DAW through two channels of either our SSL G+ or 9000J consoles. I always recommend that they patch directly into a stereo return to bypass the channels and routing, but nobody really seems to think there is a big enough sonic difference to care. These clients are making the pop and R&B records that are on the radio every day.

- Adam
 
I use a passive monitor in my shaker desk mastering console. All connections are balanced so I have a full H pad. It has a 10K input impedance and a 5K output impedance. It's a 30 position 12 deck Shallco with 1.5dB per step. The ATC monitors' amplifier input has a 50K input impedance so it's just about ideal. It took carful planning and it's still not as idiot proof as an active solution. In mastering 0.1dB level difference matters. Even bridging loads can have that much slop.
 
So I tried completely bypassing the dbox. There is a loss of clarity when running through the dbox. I've noticed it before with the summing but never tried the monitor controller by itself. Interesting.

The next question would be who actually makes a fully transparent monitor controller. Grace? Crane Song? Or just go passive and short cables?

BTW thanks for actually doing the test. It'snice to see someone actually make contact with reality rather than just repeating internet lore.
 
Every monitor controller will sound different, but what is really correct anyway? I've had many interfaces and monitor controllers and the two that I personally felt sounded super clean are the Antelope Eclipse (via digital ins and its own D/A) and an analog Studio Technologies StudioComm. Even though the StudioComm uses VCAs (THAT 2180) I found it much cleaner than other VCA controllers, especially the SSL9000J monitor section! The Grace and Avocet (via their digital inputs) sound excellent, as well.
I'm confident from the above that we hear things very differently, which is to be somewhat expected as I am a mastering guy and acoustics/speaker tuning guy and you are in the record / mix world.

To your question "What is really correct anyway?". This could not be easier to answer, we want an attenuated/amplified version of the input signal with the lowest possible distortion. And then add any features like mono, dim etc.

In John's test when the dBox is inline the sound is degraded. Not correct. This will lead to recording, production and mix decisions that will not translate to the real world as well as they could.
 
One thing I might add is I did the test with the dbox volume all the way and then attenuated in pro tools, basically using the dbox as insert.

What is interesting is the dbox sounded better with volume lower, i.e. things sounded worse with pro tools fader at -40dB and dbox max vs pro tools at unity and dbox -40dB. Could be a quirk of dbox, or it might suggest that keeping max bit resolution is preferred, and why analog summing can work better with the right gain staging.
 
One thing I might add is I did the test with the dbox volume all the way and then attenuated in pro tools, basically using the dbox as insert.

What is interesting is the dbox sounded better with volume lower, i.e. things sounded worse with pro tools fader at -40dB and dbox max vs pro tools at unity and dbox -40dB. Could be a quirk of dbox, or it might suggest that keeping max bit resolution is preferred, and why analog summing can work better with the right gain staging.

Oh, you lose your brownie points again.

You need to test in the manner it will be used. With attenuation (and without an extra converter loopback).

I use digital attenuation here right now, the "max bit resolution" trope is not real. Bits represent a sample levels, when bits are all multiplied for attenuation and correctly dithered the level drops and that's it, no change to resolution. The only gotcha is noise floor, also true in analog. Modern converters do not have noise issues. This may have been the case 40 years ago with some 3M multitrack but not now, check the converter linearity tests on Audiosciencereview.com

Analog summing is just distortion. Converter distortion, crosstalk, summing amp distortion, transformers if there are included.

As has been said many many times if computers couldn't add numbers I would not be able to send this message, planes would fall out of the sky and your bank balance would be poorly.
 
As has been said many many times if computers couldn't add numbers I would not be able to send this message, planes would fall out of the sky and your bank balance would be poorly.

As someone who has actually designed electronics to put planes in the sky I may have a thing or two to say on the subject :)

Every system design is a trade-off in determining what matters and what doesn't, there is no perfect. Computers can't actually add numbers correctly (perfectly). Banks can and have lost money not understanding this. And there are people out there designing and releasing software who quite frankly don't know what they are doing in terms of proper dsp for audio.

Also there was no extra conversion in my test, as that would obviously be another variable. My explanation may not have been clear.
 
Analog summing is just distortion. Converter distortion, crosstalk, summing amp distortion, transformers if there are included.

This is a test I would suggest that anyone could try. Compare digital vs analog summing with the same analog hardware. Take a summing mixer and run your mix though just ch 1-2. That way the summing is digital but you still get all the color of the summing mixer. Now split out the tracks across ch 1-16 or however many channels you have. Best to use stereo stems to avoid pan law differences. Every time I've done this you hear more separation with the stemmed out version. Subtle but obvious imo.

Now you might still prefer ITB with no summing mixer at all. That's ok. But it's not really a comment on analog summing, it means you don't like the sound of that particular summing mixer and its associated electronics, transformers, etc.
 
As someone who has actually designed electronics to put planes in the sky I may have a thing or two to say on the subject :)
audio combining?
Every system design is a trade-off in determining what matters and what doesn't, there is no perfect.
Digital math can be arbitrarily precise.
Computers can't actually add numbers correctly (perfectly).
I recall as a college freshman running computer homework on Hollerith punch cards written in Fortran language. The results of simple math back then was humorously imprecise (3x4=11.99) no doubt due to truncated precision of log tables used for multiplication.
Banks can and have lost money not understanding this. And there are people out there designing and releasing software who quite frankly don't know what they are doing in terms of proper dsp for audio.
Don't conflate human error with hardware errors. Analog combining has errors by definition related to op amp loop gain margin, digital combining does not. Of course analog design engineers are another source of errors.
Also there was no extra conversion in my test, as that would obviously be another variable. My explanation may not have been clear.
I am often misunderstood. ;)

JR
 
Digital math can be arbitrarily precise.

Take something simple like 0.1^2 in floating point, your code will fail if it relies on this to be 0.01 exactly. So digital has errors too, it becomes a question of what actually matters and how it is handled. Much audio software does not dither properly. At a certain point the errors add up and become audible, at a lower threshold than one might think. I was quite surprised to hear a difference in dithered vs undithered audio even at 24 bits.
 
I never wrote any floating point microcode... I did need to come up with a relatively fast square root calculation for a RMS value meter display. I did it but ended up not using it because the RMS and simple average pretty much looked the same with the same att/rel time constants, so decided to KISS.

JR
 
As someone who has actually designed electronics to put planes in the sky I may have a thing or two to say on the subject :)
This is fair :) But if you are going to (rightly) pull rank on me in this regard, then I'll have to do the same when it comes to having spent 28 years listening for a living, the last 6 in a state of the art environment.

Your example of running stereo through a single pair of summing inputs vs fanning out across say 16 can coexist with my explanation just fine. Distortion adds by square law IIRC, more converters, more summing inputs = more distortion. And these distortions are on each source and them summed for more "complexity".

This idea that 64 bit float environments can't sum 100 channels with enough accuracy does not seem plausible in the face of this obvious summing distortion mechanism. I've been lucky enough to receive many mixes over the years to master with an ITB sum and console or summer - API, SSL, Neve etc etc. The ITB almost always wins, and the reason IMO is distortion (don't always believe what you read on Gearslutz).

Music is truly magical, 2 + 2 can = 5, I've seen and heard it many times. But equipment is not, and with a decent bench and your skills you can parse these mechanisms easily with well designed tests.

Here's one - make a stereo WAV of CCIF IMD tones digitally generated (or 19.5k and 20.5k if you're brave) digitally. Duplicate this across 16 channels. Then sum digitally, run 1 pair through a stereo summing input in the manner you described, and finally sum all 16 channels through your summer. Level match the fundamental tones using an FFT and report the distortion level on each.
 
PS turn off your speakers if you do 19.5 and 20.5k, not down, off!
I like HF two-tone IMD testing to parse out nonlinearity related to NF (inadequate loop gain margin). Back in the day I modified my old SMPTE (60Hz and 7kHz) IMD analyzer to use 19kHz and 20kHz instead. It was pretty revealing on RIAA phono preamps because the RIAA EQ typically rolls off THD products, but the IMD products get boosted instead.

JR
 
I like HF two-tone IMD testing to parse out nonlinearity related to NF (inadequate loop gain margin). Back in the day I modified my old SMPTE (60Hz and 7kHz) IMD analyzer to use 19kHz and 20kHz instead. It was pretty revealing on RIAA phono preamps because the RIAA EQ typically rolls off THD products, but the IMD products get boosted instead.

JR
It's very effective and super easy to do with modern measurement gear. You were ahead of the curve!
 
This is fair :) But if you are going to (rightly) pull rank on me in this regard, then I'll have to do the same when it comes to having spent 28 years listening for a living, the last 6 in a state of the art environment.

No argument there, your listening experience and environment is certainly superior to mine, and why I enjoy your insights.

Distortion adds by square law IIRC, more converters, more summing inputs = more distortion. And these distortions are on each source and them summed for more "complexity".

Distortion can increase faster than signal, but the total would really depend on the gain staging and source of distortion. It's possible you might actually have less in total when spread across more channels. More channels could be more, same, or less, it depends. The only real conclusion would be that it would most likely be different.

This idea that 64 bit float environments can't sum 100 channels with enough accuracy does not seem plausible in the face of this obvious summing distortion mechanism. I've been lucky enough to receive many mixes over the years to master with an ITB sum and console or summer - API, SSL, Neve etc etc. The ITB almost always wins, and the reason IMO is distortion.

Was this also unanimous among your colleagues? Studies have shown some people prefer minimal distortion, some a little 2nd order, and some a little 3rd order. If it's really about distortion we might just be in different camps. Admittedly my favorite records and mixers tend to involve consoles.

My general opinion is this:
A = ITB
B = Summing mixer but just channels 1-2
C = Summing mixer spread out channels 1-16
I've heard A sound the best and I've heard C sound the best, but never B. C is always better than B.
This implies there might be something more complex going on that is not fully understood.
 
Was this also unanimous among your colleagues? Studies have shown some people prefer minimal distortion, some a little 2nd order, and some a little 3rd order. If it's really about distortion we might just be in different camps. Admittedly my favorite records and mixers tend to involve consoles.

I don't have much interest in what my colleagues think! The list of people whose ears I trust is very very short.

Re consoles, causation/correlation. Many of my favorite records involve consoles but that is because of when they were made, the console is not the thing that made it great.

I think distortion is much simpler to understand in production than most people seem to think. Simple/sparse arrangements generally benefit from 2 bus distortion, busy/complex/200 channel mixes generally do not.
 

Latest posts

Back
Top