Digital Summing

  • Thread starter Thread starter RhythmRmixd
  • Start date Start date
RhythmRmixd

RhythmRmixd

New member
I've read several posts discussing analog vs. digital equipment, the positives and negatives to each side. I've noticed the term "summing" comes up a lot when reading these posts, and how those who have experience using analog consoles and equipment tend to make the argument that analog summing is better than digital summing, which is very believable to me given the fact that analog has much experience and history, while digital recording is still fairly new and has yet to be perfected. I remember hearing a quote in a post that "digital summing just sounds wrong".

My question is, what exactly is "summing"? Is it simply the process of recording your multitrack mix down to a stereo 2 track mix? Would good summing be defined as your 2 track stereo mix sounding as identical as possible to your multitrack mix?
 
RhythmRmixd said:
Would good summing be defined as your 2 track stereo mix sounding as identical as possible to your multitrack mix?
Click to expand...
No. Not quite. If there was no summing at all on a 16 track mix, then you'd need 16 speakers to hear it. Even though you haven't gotten to the finished to track mix yet, the tracks still have to be summed to play through your stereo monitor speakers (or headphones, ect).

In short, summing is taking two or more channels and combining them into one channel. For a stereo mix, you have two summing processes (one for the left and one for the right). This has to happen any time you play back through two speakers.
 
So your saying summing occurs just by listening to your multitrack mix through a pair of speakers, even if you haven't bounced the seperate tracks to a single, stereo mix. I wonder why digital audio (as far as what I've read) would have a more difficult time getting this to sound as good as analog. Maybe the D/A conversion to the speakers? Just a guess.
 
RhythmRmixd said:
So your saying summing occurs just by listening to your multitrack mix through a pair of speakers, even if you haven't bounced the seperate tracks to a single, stereo mix. I wonder why digital audio (as far as what I've read) would have a more difficult time getting this to sound as good as analog. Maybe the D/A conversion to the speakers? Just a guess.
Click to expand...

Well, not all listeners agree on this. Personally I have no opinion because I haven't heard a controlled test. However, if the effect exists, there are various theories that attempt to explain it, ranging from the math used to accomplish the summing, or the accumulation of digital errors that inherently exist on each track being more noticeable in a mix, to something more prosaic like our ears being biased towards the pleasing distortions of an analog console.

I don't think the DA conversion is typically blamed, since that occurs either on the stereo bus if digitally summed, or on each track if analog summing is used, and is thus present in both situations--arguably more present with analog summing.
 
RhythmRmixd said:
So your saying summing occurs just by listening to your multitrack mix through a pair of speakers, even if you haven't bounced the seperate tracks to a single, stereo mix.
Click to expand...
Yes.

RhythmRmixd said:
I wonder why digital audio (as far as what I've read) would have a more difficult time getting this to sound as good as analog.
Click to expand...
Math accuracy problems. Let me give a simplified example. Let's say for the sake of the example that a computer can only handle 2 digit numbers. Now say that somewhere in the summing process, the computer has to divide 54 by 4. The correct answer to 54/4 is 13.5, but since the computer can only handle two digits it has to settle for 54/4 = 14.

14 is almost the same as 13.5, but not quite so the sound is a little bit off. In real life the computer can handle much more digits than 2, but there is still a limit to what it can handle and the same basic problem occurs.
 
I find there's a noticeable difference in digital summing via Cubase (SX 2) vs. my digital console.
 
Chibi Nappa said:
Math accuracy problems. Let me give a simplified example. Let's say for the sake of the example that a computer can only handle 2 digit numbers. Now say that somewhere in the summing process, the computer has to divide 54 by 4. The correct answer to 54/4 is 13.5, but since the computer can only handle two digits it has to settle for 54/4 = 14.

14 is almost the same as 13.5, but not quite so the sound is a little bit off. In real life the computer can handle much more digits than 2, but there is still a limit to what it can handle and the same basic problem occurs.
Click to expand...

DAWs are generally using 32 bit float to do the math. Now if we accept that 24 bit audio is sufficient to capture more dynamic range than we need, then doing calculations at 32 bit float gives the process a massively greater degree of precision than it needs.

There are some technical posts around somewhere that explain this in greater detail, I can try to find them if you like.
 
RR,

> analog summing is better than digital summing, which is very believable to me given the fact that analog has much experience and history <

This is not a very good reason to accept a "theory" as believable. That's like arguing against antibiotics when they were first introduced because blood-letting has much "experience and history."

If there are any artifacts or errors caused by digital summing they can be expressed as some amount of distortion. Even with 16 bits that distortion is so low as to be inaudible. Of all the quality issues people have to fret over, I put summing errors at the very bottom of the list.

The big problem with people's "opinions" on summing is that it's pretty much impossible to set up a controlled A/B test.

--Ethan
 
Ethan Winer said:
The big problem with people's "opinions" on summing is that it's pretty much impossible to set up a controlled A/B test.
Click to expand...


In the sense that no test is truly perfect, that would be an accurate statement.

But I do think a very relevent subjective listening test could be done easily. Suppose I were to take a typical session that's already mixed with all of the waveforms in my editor, effects, etc. and do the following:

* Render each individual track as it's own stereo mixdown, keeping the mix bus at unity. This way any stereo effects, panning, etc. and relative volume is already reflected for each track.

* Take each of these newly-mixed tracks and burn them to CD / DVD, or put them on a removable hard drive. Now load them up to a new session with everything at unity and mix / render the session.

The resulting mix will be your "digital summing" example.

* Now take these same files and sum them the same way using an analog console of your choosing ... again with everything at unity. This resulting mixodown will be labeled your "analog summing" sample.

The fact that each of the tracks is rendered to it's own stereo track with all effects, panning, eq already applied, you're removing any of these as variables, so as to better isolate the summing as the variable.

The main variable I still see left is the fact that the analog samples are going through added a/d and d/a conversions. You could even out that playing field by re-recording each of the "digital summing" files by going out of your DA / and back in through your A/D before you load them up on your digital editor and render that sample. That way both have gone through the conversion processes and you are, again, better able to isolate summing as the variable in your audio comparison.

A test like this could be done in little more than a couple hours with the proper setup, and would be a relevent a/b comparison for most people. The fact that no one has done this yet shows either a general laziness or lack of imagination (or maybe just interest) amongst the audio community -- myself included, mind you. :D
 
Chess,

> Render each individual track as it's own stereo mixdown ... <

Yes, that would work. I'm not even worried about an extra A/D/A conversion because that adds very little distortion. But all of the "tests" I've seen people do are far less scientific. Usually they mix a song in their computer, then send all the tracks out to a mixer and mix it again. With that kind of setup there are far too many variables to come to a meaningful conclusion.

> The fact that no one has done this yet shows either a general laziness or lack of imagination (or maybe just interest) amongst the audio community <

The reason I haven't done such a test is because I just don't care. I know that digital summing works fine, and even if analog summing offered some teensy weensy improvement, the huge drop in convenience overshadows any slight "improvement" in quality. Not that I accept that an improvement is even possible in theory...

--Ethan
 
Blue Bear Sound said:
I find there's a noticeable difference in digital summing via Cubase (SX 2) vs. my digital console.
Click to expand...

Interesting - I thought they used the same algo....maybe not.

Can you upgrade the software in that bad boy if/when a better algo comes along?
 
Digital Summing Is Bad Because It Is Not Analogue!!! Even Though Digital Is Numerically Based Derivitive Of Analogue Sound Frequencies And Should Add Up Better, Analogue Is Sumthing Else So Therefore It Sums Better.

Add It Up--digital Makes No Sense!!!!!
 
Hardware and software will always be different. In the case of the Sony console versus Cubase SX, there are a lot of varied factors there. First, one mix was done on one, and one on the other so right off the bat things were different. Second, they most likely do use different algorithms. This does not even mean that one sounds better than the other. Just that they sound different. In the Sony console, much more than just "digital summing" is applied. The converters inside it are differnt, the signal path is differnt, conversions may or may take place at different points in the signal chain etc.... There are just too many differences to be able to do a truly "Scientific" comparison. But then again I don't care about scientific comparisons. I only care about how the two different things would impact me and my workflow. Its like comparing preamps. Everyone says you have to get the outputs between different preamps at the exact same level blah blah blah. I agree that is useufl, but just as useful to me is just to set the knobs between different preamps at the same settings. Is one preamp louder or fuller at the same setting? Those are things I would want to know. Which one runs out of headroom first? How do they sound distorted? Anyhow, thats what it is like for me comparing algorithms. I want to know how the summing "feels" since that is how I mix, by how the music "feels"
 
chessrock said:
The fact that no one has done this yet shows either a general laziness or lack of imagination (or maybe just interest) amongst the audio community -- myself included, mind you. :D
Click to expand...

Actually Lynn Fuston did something like that already. It was called the AweSum DawSum CD its on sell at mercenary
 
Teacher said:
Actually Lynn Fuston did something like that already. It was called the AweSum DawSum CD its on sell at mercenary
Click to expand...


I think Lynn is a great guy, and I appreciate what he does for the audio community ...

... but I checked out that CD, and frankly didn't find it valuable in the least. It was very far from the kind of test I described.
 
mshilarious said:
DAWs are generally using 32 bit float to do the math. Now if we accept that 24 bit audio is sufficient to capture more dynamic range than we need, then doing calculations at 32 bit float gives the process a massively greater degree of precision than it needs.
Click to expand...
For individual calculations that's true, but rounding errors are cumulative, and by the time your s/w has performed the thousands of manipulations necessary in summing, artifacts can and do creep into the lower order bits causing distortion
 
I'll be on the lookout for the Sumbus Test which I found from a link here years ago. A guy took individual drum tracks and loaded them into both Digital Performer and Logic (With everything set at unity, and identical pan settings in both programs). What a dramatic difference between the two resulting summed files (Logic was a LOT warmer sounding, while the DP files were very brittle sounding).

Also worthy of note, just bouncing files offline within a program can (and will) often yield different results as bouncing in realtime within the same application (I have witnessed this first hand at my nephews studio (utilizing Sonar).

To my ears I have always prefered the way analog mixers sum (Most of them anyways). But any minimal degradation in quality I might get from digital, is so insignificant, it is well worth the trade to have all the wonderful flexibility and automation of a digital setup.
 
You must log in or register to reply here.
Back
Top