SPL MixDream analog DAW summing

[IIRs]

The problem with many channels is that -- to put it easy -- calculating errors are getting more and more annoying

Here's a test you can try: pull some audio into your DAW (doesn't matter what it is) then duplicate it onto another track, and flip the phase. You should get a 100% perfect null between them, correct? Now paste in another two copies, and again flip the phase on one of them. Still a perfect null, right? Keep going, and tell me how many tracks you need to be running before these "calculating errors" start to creep in and you no longer get a perfect null.

I got bored and stopped with 102 stereo tracks still perfectly nulling another 102 phase reversed copies...

<edit> actually the 204 summed tracks (each peaking individually at about -6dBFS) read approx -133dBFS on the master rather than -infinity as when they are all muted. Thats good enough for me...

 

[IIRs]

And another test in case you still aren't convinced: create a mix ITB, then pull your rendered file back into the project and flip the phase. Perfect null?

You need to be a bit careful with this test of course, as any modulation type effects (including modulated reverb tails) will not be identical on subsequent passes. You would need to print these effects first.

 

[IIRs]

Maybe its what makes golden ears mastering engineers.

Or perhaps "golden ears" comes from an awareness of how subjective human hearing is, and how much amazingly sophisticated processing is going on in our brains to interpret the signals we are getting. We don't hear music as a single fluctuating waveform: our brain splits it up into ensembles of distinct instruments, and assembles them into a 3 dimensional soundstage. And every time you listen to a certain piece of music you will notice something different, even if nothing has changed at all.

Proper double blind testing is the only way to determine if there really is an audible improvement. And as it happens, I think listening to a commercial radio station is a good example: you don't know which mixes were analog and which were digital, and the presenter probably doesn't know either (nor care for that matter!). Would you be able to pick out the analog mixes with any confidence?

 

[IIRs]

I thought this was very interesting, sorry if its been posted elsewhere:

[="http://www.youtube.com/watch?v=BYTlN6wjcvQ"]YouTube - Audio Myths Workshop[/]

 

[IIRs]

And this: [="http://www.witnit.org/archives/2006/06/1_creating_your_1.php"].:. WitNit .:.: 1. Creating Your Life[/]

 

[jammster]

IIRs, I enjoyed your Audio Myths Workshop post, I just finished watching the whole thing.

Something comes to mind. I hope you don't mind that this is a bit OT.

The way in which we decide to go about our mixes, be it either method of summing/mixing, it is our choice in the way we go about our craft. One will take it personally when cornered on either front. When confronted one will be offended, and I would like to think that does not reflect the nature of Recording.ORG. This is a big wide world we live in, lets realize that this topic is an artistic preference. The WWW is certainly changing things, and I would like to think diversity could be honored here as well.

To me what it really boils down to is this: You have grown to be comfortable with what you use. I will testify there is something stark and void in an all digital recording, that fact that it is so perfect and does not vary and comes with a computer attached weather you like that or not. I think that it has become more popular with the use of the internet and use of MP3's as well.

Take a moment and think of all the sudden there was no other medium other than MP3.

I think I would leave the world of recording for good if there was, I mean, if that is all people want for their music collection. I find MP3's like genetically modified foods, they have temporarily blinded consumers into their quality, or lack of. But, if that is what the consumer wants that is all we get, no more CD's, no more high definition. It seems to me were on a slippery slope

The lack of some noise and variation that would come with ease in the world of analog tape recording is slowly/quickly leaving us forever. Some will agree with the use of digital mix bus and some will disagree, the fact remains that they are both very valid ways to work, like painting with watercolor vs. acrylic vs. oil.

I really don't think it matters which you use, its the colors in the painting that we are looking for. An artistic choice of any medium is valid, none is better or superior until you attach yourself to the outcome and notice which colors and methods you prefer. Beauty is in the eyes of the beholder. But, some will insist there is no difference between the mediums. I must say to me personally there certainly is a difference, and some of that by the way has to do with the method of the work itself, not always in the ways you hear it.

Ask yourself how the computer has been influencing your music. It is possible to work without the use of a computer, at least in the tracking? Yes, most certainly. You can use an old tape deck and a mixer, and why should that be shunned, or laughed at? I think that those that laugh are being immature, a sound is a sound and an artist can do what they prefer, its the engineer who's job it is to make the recording as good as possible, and probably has less emotional attachment about which medium is utilized.

Remember that as an artist (weather you think of yourself as one or not) you have the right to chose whichever medium you want, or even switch them up for different effects, the possibilities are endless and you can utilize them all to their full potential if you wish. And why would you want to rule the wide world of variation down to one medium anyway? Just remember that inspiration comes in many ways and at different times and that a computer does not need to be attached for them to be beautiful and effective.

 

[IIRs]

Again, I have no argument with that. If you have a personal preference one way or the other, thats fine by me. My argument is with the assertion that there are mysterious voodoo errors in digital summing, despite the lack of any evidence to support the claim.

 

[AudioGaff]

Perfect nulling is perfect nulling. I don't listen or record nulling. Perfect nulling doesn't mean that there are not artifacts that one can hear. There are many recordings when you can find with complete less than perfect nulling and the audio results sound fantastic.

Don't need to use nulling, doesn't matter if there are calc errors, or if the summing issues are real, my ears and experience can tell me that many times ITB mixes have less pleasant results than those that can be done with analog mixing. There are many times when you can really tell.

Do this as a test. Import a song from a CD into your DAW. Save the project. Close your DAW. Reopen your DAW, now render/mixdown that to a two channel wav file. Burn that song into a CD. Listen to the CD and the CD you made. Do you hear a difference? You may or may not. Try different songs from different CD's as results can vary. If you hear a difference, then you hear a difference.

You also have to put things into perspective. ITB is not just ITB in the sense that how your DAW will process your mix using plugs, whcih A/D converters and D/A converters are used, the levels of tracks. It all equals the final ITB results.

In addition, it also depends on your listening environment and ears. I'm one of them cave men that did quite a bit of mix recording on DAT tapes and tracked using Black face ADAT's. I also owned the best of Apogee and have used the mastering stuff from Weiss, so I have a pretty decent amount of direct experience of hearing and knowing good digital from bad. It has gotten better over the years that is for sure, and I have no doubt that one in the not too far future, ITB will one day be the glorious thing we have all have been waiting and hoping that it would be. But that day is not today.

That doesn't mean that we go backwards or we just don't use it. You work with what you have.

A thermal blanket will keep you warm, and is amazing for how well it manges energy for as thin as it is a material, and is cheap for what it does, but given the option of choice, if I was really cold I would prefer a fluffy fur coat most of the time.

Why?

Because as humans there is more to us that just being warm. There is a need and desire to also feel warm...

 

[IIRs]

Perfect nulling doesn't mean that there are not artifacts that one can hear.

Huh?? Again, what artifacts? Provide some evidence for this assertion please.

Perfect nulling means the numbers are identical, therefore so is the sound. I can sum 204 stereo tracks with artifacts down below -130dBFS. Are you telling me you can hear floating point errors that are 60dB below the level of 16 bit dither noise? You would need to prove that in blind testing before I would treat that claim as anything other than laughable....

 

[Jeemy]

Surely nobody is claiming they can hear summing artifacts hidden below a noise floor. Perhaps mistakenly, I think people are saying that the results of such calculations impart a clinical or artificial sheen to the sound.

i.e. when you have 16 tracks in a DAW, using calculations to reach 24-bit digital audio and sum the high, low and mid frequency content into 2 channels, there is a flavour imparted by the summing process. In layman's terms, your argument is that the summing process imparts no flavour at all, and that such a thing is in fact not possible?

Fine, flip it. The digital summing process provides no flavour or resulting loss of quality, ambience, vibe, etc at all.

So is it not then reasonable to propose that if the same 16 tracks are converted to analog by good A/D converters, and then the summing is performed by an analog mixer or summing box, the circuitry by its very nature sums the high, low and mid frequency content in a different way, and some folk here and elsewhere contest that this provides a fuller and more open, spatial mix.

Perhaps the contention that analog is the more 'accurate' method is in fact not correct. Perhaps that contention should be that an inaccurately-summed analog mix feels and sounds more real to most people? and that if you move your mix process to a more accurately summed and digitally perfect mix, there is a loss of imparted warmth which results in it seeming lacking?

I don't contest that this is only part of the whole and that when one takes an excellent entirely digital system, its impossible to tell the difference from an excellent entirely analog system from song to song. However when we aren't working with entirely excellent systems, circuitry indubitably imparts flavour to audio, whether productive or destructive.

However I don't think its fair to denigrate the statement that for a semi-pro, project or professional studio with good D/A converters, one might find that switching to an analog desk or summing box provides a leap in quality to your mixes. I can believe thats true. I'm sure you're right if all other things are equal, but thats just the point, they never are. Whether its down to artistry, budget-imposed limitations or simply your preferred working methods, I don't believe a categorical mathematical argument can prove that an analog summing chain would not assist people getting better results.

 

[IIRs]

Perhaps the contention that analog is the more 'accurate' method is in fact not correct. Perhaps that contention should be that an inaccurately-summed analog mix feels and sounds more real to most people? and that if you move your mix process to a more accurately summed and digitally perfect mix, there is a loss of imparted warmth which results in it seeming lacking?

This is exactly the point I made 2 pages back:

Analogue summing adds subjectively desirable colour to an otherwise too perfect digital sound.

My follow-up question was: why is that superior to summing digitally then running the whole mix through a colourful 2-channel unit? Remember: (a*x)+(b*x)+(c*x)=(a+b+c)*x

Perhaps you think that intermodulation effects will colour the sound in a different way if the whole mix is processed rather than simpler stems? In which case, why is analog summing preferable to tracking through a colourful preamp?

The test that I am very curious to run (but lack the expensive summing box to do so) would involve summing x number of stems via a summing box, then summing the same stems ITB and running the resulting stereo mix through 2 channels of the same device.

If anyone could reliably identify which mix was which under blind test conditions I would accept that analog summing had merit.

 

[Jeemy]

Ah yes I missed that follow-up point, and while I think its a pertinent question, I'm not qualified to say either in mathematical terms or practical. However although I may be wrong, I'd put forward that your equation needfully oversimplifies and although its true, the summing of x waveforms amounts to more than a simple a + b sum.

I also don't see why one would preclude the other; why you wouldn't still be using your colourful 2-channel unit to process the resultant 2-bus stereo mix; this would be a good idea, if appropriate, regardless of which type of summing. So I don't think anybody is saying that analog summing is preferable or an alternative to a colourful preamp; just that it is a valid alternative to digital summing with possible benefits.

But I guess, and again I may be wrong, that what we are actually talking about is not just analog summing, but converting 16 channels of digital audio to analog rather than 2 converters handling the entire conversion process. This may well mean (a total guess on my part) that part of the improvement is imparted by an imperfect converter converting a simpler wave (16 times) rather than 2 imperfect converters working on much more complex waves?

How large these benefits are is what your test would prove and it may be something one of the proponents using the Mixdream or similar can produce, although I understand subjective volume levels would be a factor to watch out for. I'll probably step out from the conversation now, I'm very interested in it all but not experienced enough to do more than try and clarify things for myself and watch from the sidelines.

 

[Boswell]

This is an interesting discussion, and to add more ingredients to the pot, I'll throw in the following:

(a) In an analog summing setup, if the D-A, A-D conversion and the summing processes were perfect and the channel gains all perfectly matched, the result would sound identical to a correctly-computed digital summation. But nothing is perfect, and in fact, we tend to like imperfections of the right sort, for example, low-level second-harmonic distortion increasing with signal amplitude. These imperfections are often called "flavour" or "colour". The trick is how to manage the acceptable or even wanted imperfections while not introducing unwanted ones.

Furthermore, in a typical pop/rock/folk music mix, we often like the different sound sources in the recording to have their own separate colours. This is mostly achieved by suitable choice of microphone and pre-amp for the various instruments and vocals, but more can be applied at the mixing stage on a track-by track basis. Once the tracks are balanced (for level) and summed, for the mix/summing engineer simply to slap a dynamics/EQ/effects unit on the stereo bus does not give us the same feel to the resulting mix. Instead, it is down to the art of the mastering engineer to apply subtle sculpting to the final mix to bring the recording up to release standard.

Out of interest, a parallel thread to this one discusses the use of M-S techniques at the final mix and mastering stages to give an added dimension for imparting effects, EQ and colour that would be either very difficult or impossible to achieve by processing solely on the L-R stereo bus information.

(b) The process of external analog summation is more than simple addition of signals. For example, the panning of tracks to their intended place in the stereo field is in itself not as simple a process as first appears. Most analog mixers and summation units implement one or more “panning laws” that govern the amplitude modulation that is applied to the L and R mix contributions as a track is panned. A general consensus might be that a track panned centrally should have a 4.5dB gain reduction applied to both the L and R components referenced to 0dB and infinity when panned hard to L or R.

Hence it is necessary to specify what functions exactly the external summing device implements. If all the level balancing and panning is performed in the DAW, with each DAW track being output in stereo so its place in the stereo field is determined, the summing unit can work at the same fixed gain for each input and use independent summing of L and R outputs. Note that in this case, the D-A converters may be operating a long way below their maximum outputs, but their noise floors all add the same amount. By contrast, many types of external summing unit also perform balancing (i.e. have faders or rotary amplitude controls for each input channel) and also panning. In this way, each mono DAW track needs only a single D-A converter, which can be operated close to normalised, resulting in a lower noise floor in the final mix. This technique does mean that the mixing and panning are neither stored in nor are under control of the DAW. These two cases are the extremes, and schemes that fall between the two are often used.

(c) I'm certain I could pick out a mix done by hand-crafted analog summation on top-drawer equipment from a straight ITB digital summation, but probably because they would not be the same mix. With reference to (b) above, I don't think it's likely that they ever would be the same mix. Maybe we should put more effort into options for digital summing, and not simply say that the machine always adds up correctly.

 

[IIRs]

, the summing of x waveforms amounts to more than a simple a + b sum.

Summing is summing. It really is just adding up a string of numbers, and is by FAR the simplest calculation your DAW has to handle during mixdown.

 

[IIRs]

But I guess, and again I may be wrong, that what we are actually talking about is not just analog summing, but converting 16 channels of digital audio to analog rather than 2 converters handling the entire conversion process. This may well mean (a total guess on my part) that part of the improvement is imparted by an imperfect converter converting a simpler wave (16 times) rather than 2 imperfect converters working on much more complex waves?

You can test that theory using the same method Ethan Winer used in his audio myths workshop: play a full mix out of your DACs, and record it back via your ADCs. If your theory that the converters cannot handle a complex signal is correct you will be able to hear that degradation in the re-recorded result. If you can't hear a difference, do the same thing again, and again, and see how many generations you need to go down before you start to hear a difference.

 

[Jeemy]

Well just leave that theory, its not mine, somebody else can look after it. All the test would prove is same-in, same-out, not whether 16 direct D/A outs are better overall than 2 digitally summed.

But with regard to this equation: (a*x)+(b*x)+(c*x)=(a+b+c)*x

When you mentioned it, you weren't talking about DAW summing, or have I misunderstood? You were talking about analog summing and saying it should be no different if the summing box sums 2 channels from summed DAW 16-channel audio, than if the summing box sums 16 seperate stems. If analog is imperfect and this is a more complex equation then that cannot be true. The equation is mathematically true for digital domain only?

 

[IIRs]

But with regard to this equation: (a*x)+(b*x)+(c*x)=(a+b+c)*x

... If analog is imperfect and this is a more complex equation then that cannot be true. The equation is mathematically true for digital domain only?

Actually it would depend on the nature of the imperfections. It they were linear (like an EQ) then my statement would be equally true of analog and digital systems. It they were non-linear (like distortion) then it would be equally un-true of analog and digital.

So it comes down to determining the precise nature of those analog imperfections. From the SPL Mixdream specs:

"Freq. Response: ‹1Hz-220kHz (+/- 3dB)".

But that doesn't tell us much about what's going on in the audible range. If there are +/-3dB swings in frequency response within the 20Hz to 20KHz range, that would certainly be audible and would impart a definite colour to mixes. But the colour would be exactly the same if you ran 2 channels, or 8 stems, or 48 individual tracks.

"THD+N Ratio: -104dB
(20Hz-22kHz, Input +10dBu, all channels active)"

That's pretty clean, wouldn't you say? Distortion definitely would be different for a whole mix compared to seperate stems. But distortion at those levels...?

 

[IIRs]

I'm not actually sure that we are disagreeing about anything, except perhaps:

(c) I'm certain I could pick out a mix done by hand-crafted analog summation on top-drawer equipment from a straight ITB digital summation, but probably because they would not be the same mix. With reference to (b) above, I don't think it's likely that they ever would be the same mix.

Logically: how would you know which mix was which? (assuming someone else mixed them of course!) And have you ever tested yourself?

I think it should be possible to do the test properly, if a little laborious: you would need to first create the mix using your summing box gains and pans (so that your mix desicions could be influenced by any analog gain staging non-linearities), then solo each of your stems, bus them internally as well as routing them to the individual outputs, and use meters to match levels and pan positions to the desired level of accuracy. If you have a DAW that lets you type in parameter values directly you should be able to match these to within 0.1dB fairly easily.

Of course, all modulation effects would need to be printed first, so that playback was identical for every pass.

Note that in this case, the D-A converters may be operating a long way below their maximum outputs, but their noise floors all add the same amount.

This got me thinking: if you were running 16 analog outs as 8 stereo stems, that is indeed 8 different noise floor instead of 1. They would all be uncorrelated (I think?) and would therefore each add 3dB to the overall noise floor. So 7*3=21db extra converter noise from this setup. Have I got the maths right...?

Maybe we should put more effort into options for digital summing, and not simply say that the machine always adds up correctly.

Not quite sure what you mean by that. care to elaborate?

 

[Jeemy]

I'm not very sure how to quote things. I also know I said I would step down from this. However, I'm really enjoying learning from what you are telling me, so if I may ask a few more questions, hopefully its clear which parts I am quoting from:

re: +/- 3dB swings and nature of imperfections: surely a 3dB variance per channel across 16 channels equates to a 3x16dB variance total, or some equivalent, summed, across the mean? so 2 does not necessarily equal 8, 16 or 48 in this example - as if the imperfections were predictable, they would not be variances as quoted but true, and predictable, imperfections.

Please can you clarify exactly what you mean by linear and non-linear imperfections in analog and digital systems, for both myself and others reading this.

When you say:

Actually it would depend on the nature of the imperfections. If they were linear (like an EQ) then my statement would be equally true of analog and digital

How can this be? Why would the imperfections or variances be identical in both systems? Surely what we are discussing in theory is the reasons why the two would differ? I'm curious to know why the situation arises where given 2 very different signal chains, we are contesting that all factors are equal?

Assuming for a second that factors are not all equal, that summing equations produce audible artifacts, disregarding whether analog or digital are better, yes, Boswell, please elaborate - where do the digital options start making themselves known? Do they impart flavour also?

 

[AudioGaff]

Good points Jeemy. Saved me from typing many of the same comments.

True, I can't hear digital summing artifacts down in thenoise floor, but I can hear a difference from a mix done in Cubase 5 or Sonar 8.5 using my Tascam DM-4800 as the ITB playback, compared to the same mix summed on my Tascam DM-4800 using the same levels and mix moves. How can that be. They both do "simple math summing"

Mix on a Mackie 8-bus then mix on a SSL, Neve 80 series or API and tell me if it is different? Heck, Mix on SSL and then Mix on API. Still hear a big difference?

Using the ITB argument, then they should be the same. After all, each one can perfectly null, they all are at least 20-20k, all likely within +/- 3dB, good THD specs.

More to the point, perfect nulling means very little in the scheme of things. I am a real electrical and an audio engineer that understands most all things digital audio as well as analog. While I usually like to know, I don't have to know, or care about the the reasons, or the details, or the science or the technology. I hear what I hear, and I like what I like.

I, like many others, feel ITB mixes suffer from something that we perceive as unpleasant or not up to par as we think and feel that they should be. It is something that is consistant from song to song, project to project. If we use diffferent DAW software and/or different converters, it is still there. Interesting enough, it would seem that analog summing through very good connerters offers a third midway kind of result. I have experienced that indeed, sometimes it was an obvious improvement. Sometimes it made no difference, but it never seemed to make things sound worse. With analog summing you also can add the capability to insert other great analog gear while you sum. When you do that, it can easily lean towards being better than just ITB most of the time.

The whole thing could very well be related to your audio background. If you have had a lot of experience with great analog mixing, then you have that as a reference. If you have not had that, then your not even really qualifed to discuss it without at least having shared some kind of frame of reference.