How does a Digital Mix Buss WORK?

[dbbubba]

I started a thread a few months back about my problems with DP4's mixbuss and really didn't learn anything new or remedy my situation.

In the quest for understanding why mixing in the box just doesn't sound as good as mixing on an analog console I tried to find out how a digital mix buss actually works. I want to know what is going on with the signals! How are all of those 24 bit, 44.1K or 96K bit streams blended together to produce the final product?

So far, I have not found any of the companies that "manufacture" software or digital hardware to go beyond telling me how to make features work. Their standard reply is, "We really can't answer questions about subjects that technical."

This bugs me because I can pick up my phone and call the actual guys that designed several well known British and American summing circuits and discuss how things work with them. In other words, I am pretty comfortable with how analog summing works and why it sounds good or bad.

DOES ANYONE HAVE A CLUE HOW DIGITAL SUMMING WORKS?
DOES ANYONE KNOW ANY OF THE SOFTWARE DESIGNERS THAT HAVE WORKED ON DIGITAL SUMMING SOFTWARE?

It just seems toi me that digital "mixers" do a lot of great stuff, but they don't do the one thing that they are actually named too well.... MIX.

Danny Brown

[minister]

here's a couple things:

DIGIDESIGN WHITEPAPER

read Paul Frindles comments here: Illegal Signals in a DAW he designed the dyunamics sectioin in the SSL G and the the EQ mods to fit the G and later series. now he works for SONY OXFORD.

also, check out these books : The Art of Digital Audio

Principles of Digital Audio

[juicemaster1500]

the answer to your question is;

not well enough.

I'm convinced it's an inherent problem with;

MATH

as such numbers as pi and that other number from 'the da vinci code' 1,6 etc.. isn't possible to express with;

NUMBERS

there's gonna be rounding errors,

as sound waves are round in shape,

not square,

and the summing we hear in daw's is a compromise

not necessarily bad, but not perfect.

[George Necola]

juice.. you sound like johnny B's brother;)

[juicemaster1500]

I'm not as hard core as johnny b,

but I'm tired of reading links to a white paper made by a manufacturer, trying to prove that their product is 'perfect', something which will never be the case.

a daw is a great compromise , as is analog

nothing else.

I'm hoping I will get old enough to experience a recording format with fewer compromises and artifacts, a recording format more true to music,

as that's what I care about, not gear.

[dbbubba]

JUICEMASTER.....

You should have been there the night in 1975 that I installed and first got sound out of an all tube United Audio console (it was kinda' like 16 UA 610s.) It had something like 124 12ax7a tubes in it!

Steinway 7"
two U87s
all tube UA console (model number?)
Crown DC300a
JBLs

I still recall this as a defining moment in my fifty years of hearing!

danny Brown

[juicemaster1500]

yes, recordings made in that time period often had a lot of mojo happening,

and a warm sound,

yet we're now living in another era, and I try to look forward, hoping for a different main stream recording format, something which will allow for the warmth of vinyl, and the clarity of cd's,

a format were the listener is listening to the actual music,

not a recording format ,

I can never shake the feeling of listening to a cd or listening to a vinyl or listeing to tape , the recording format gets in the way of the music, on a much more severe level than the speakers or amplifiers does.

maybe music would start to sell again too, if there was a shiny new product hitting the stores, instead of 25 year old cd's, that nobody cares about.

[dub3000]

warmth is a pretty hard term to describe, but i think it's probably because: most analog gear has a specific character to it, plus some hiss. hiss is basically dither. and dither makes stuff sound good.

good digital stuff has always been: reproduce it perfectly. not necessarily best sounding, but as close as possible to the original. maybe you don't want a pristine treble, you want it smooth and warm and round.

please stop complaining about digital mix busses. they just add numbers up, same as a mix desk, but without stray impedences and resistive loads and tape smeared transients and slew rates on ICs and buffers everywhere. it's going to end up on a CD at some point anyway, so no point complaining about 44kHz artifacting at mixdown and your transient peaks being rounded up to 22uS. if it's that big a problem for you, go back to analog everything.

now, digital eq ... that's a different story...

cheers

--dan http://remaincalm.org

[George Necola]

back on topic.. does someone know how it is internally workin?

[dub3000]

the short simple version:

background on digital: sound can be represented digitally as a series of numbers. the sample rate is how often you check the sound to see what number should represent it. 44.1kHz means you "sample" the sound 44100 times per second. with ideal (nonexistent) equipment, this will let you cleanly reproduce everything up to half that, so sounds up to about 22kHz. realistically it gets a bit gnarly up there because real world filters are non-ideal but most people can't hear anything sensible above 18kHz or so anyway. there is some very tricky maths at work here, look up "nyquist" in google if you're interested.

anyway, you've got a series of numbers. these numbers represent a sound. in this case, one channel of sound. let's say you do that again. you've got a second series of numbers now.

all the digital mixer does is add the numbers together. you get a new number then, and that's the output. the only magic is that it needs more resolution to store the final number, because if both of the channels are at full range, you'd need twice as much range to store the sum of it. this is what is going on when you see stuff about 48bit integer mixing or 32bit floating point mixing - that's how the number is stored. you get a perfect sum of the signals like that.

analog devices aren't like that at all. if i remember correctly, transformers have a transfer curve that drops to 0 at DC (so inaudible subsonic stuff will get removed, which is good) and all sorts of strange stuff happens at high frequencies. some good info here: http://gearslutz.com/board/showpost....0&postcount=71

it's been a couple of years since my last ElecEng course though, please correct me if i'm wrong on anything here.

should i go into more detail?

--dan http://remaincalm.org

[edit: wow, that's sketchy, i really need more sleep. sorry, guys. the main thing is, if you have a streams of numbers representing channels, you just add all the streams together, that gives you a new stream which is your output. if you want it half as loud, you put a "multiply by 0.5" at the end of it. it's really simple]

[edit: more info added]

[wires]

yes, it's like dub3000 said; however, I must post a link to musicdsp, just because
it's just a great site: musicdsp FAQ
it's explained thoroughly here (for programmers)

HTH
jelle