...If you don't get much feedback on this, e-mail the guys at

They have given me great feedback on questions like this...and let us know what they say!

What you're asking about is a perceptual question. The popularity of analog summing has always been there, I've been doing it to my digital multitrack recordings since 1993, because I was using linear digital tape and Sphere/API/Neve consoles, going back to a digital 2 track DAT. Sure does warmup that nasty 16-bit PCM sound and of course the stereo imaging is great with the analog summing.

Many folks have found that when mixing ITB, the focus and stereo perception appears to be different from its analog counterparts, when compared side-by-side. But that is almost like apples and oranges, since mixing behind the console is frequently quite different from mixing in the computer. It's really hard to get identical results when dealing with different kinds of recording systems but can and has been done by others. The reason for the difference, I believe, is that when summing everything together in the computer, it's not like an analog process. It's very complicated math that must be executed extremely fast and in a certain mathematical order. And so, because everything is timed within the computer, summing everything together will have minute time differences from each other, causing a smear in the sonic image. When done digitally, everything has to be in its own time domain, which may not coincide with other sources. Do the math (I can't). Close enough for most. Not close enough for others. Plus, analog summing does have that extra analog mush factor which people seem to like, especially with vintage equipment, none of which mentioned here, like the API, Neve, SSL, analog summing networks, which will of course help to give you that flavor, which is hard to beat.

Sum times API. Sum times Neve.
Ms. Remy Ann David

RemyRAD wrote: The reason for the difference, I believe, is that when summing everything together in the computer, it's not like an analog process. It's very complicated math that must be executed extremely fast and in a certain mathematical order. And so, because everything is timed within the computer, summing everything together will have minute time differences from each other, causing a smear in the sonic image.

Nonsense. The maths is: track 1 + track 2 + track 3 .... + track n.

If there are "minute time differences causing a smear in the sonic image", thats a bug.

Do the math (I can't).

I'm sure you would be capable of adding together a series of numbers. Thats what "summing" means.

Plus, analog summing does have that extra analog mush factor which people seem to like, especially with vintage equipment

That seems more convincing to me... but if that's all that's going on why not sum ITB and then run the stereo mix through something "analog and mushy" instead?

I really don't think you are correct about your mathematical assumptions? Regardless of how it's added up together, there are tiny variations that occur in the process of mixing in the box, due to the inherent clocking. It's obvious to many engineers. In theory, it ain't supposed to happen that way but in practice, it does. I know what I hear and I know what other highly experienced engineers have verbalized. I've known plenty of electrical engineers that can come up with circuits on paper that do not work in practice. Put that in your fine British pipe and smoke it. This is beside the fact that many of us have been mixing through consoles most of our lifetime and within the past 10 years have been doing so in the computer as well and have noticed the differences. So how do you explain the differences?

Highly uncommon
Ms. Remy Ann David

RemyRAD wrote: Regardless of how it's added up together, there are tiny variations that occur in the process of mixing in the box, due to the inherent clocking.

What clocking? Samples are processed in blocks and then stored in buffers... "clocking" only takes place at the DAC stage. If the wrong samples are added together thats a bug, pure and simple.

So, IIRs what do you think the other top professionals are perceiving in the difference between analog and digital summing? If it's so mathematically perfect, why doesn't it sound like analog summing, which is also mathematically perfect in its analog way? We know that the smear and change of stereo imaging has to do with Time differentials. We know this because we can hear this and because we are professionals who know what we are hearing and we are hearing timing smear and tiny irregularities. The Time differentials come from all the equations that have to happen simultaneously in math. Everything in computers is clocked and with clocks you get timing errors and inconsistencies. Blocks? Blocks have to play with clocks and clocks play with blocks. I prefer Lego myself.

Designing an audio console out a Lego for you beginners
Ms. Remy Ann David

RemyRAD wrote: We know that the smear and change of stereo imaging has to do with Time differentials.

There are no timing differentials ITB. Its simply not possible, short of a software bug.

The concept of jitter is irrelevant when considering a wav file stored on a hard drive, right? Well, the same principle applies to buffers full of samples waiting to be clocked out of the DAC. That's why your DAW can offer the ability to render audio faster than realtime (unless you use Pro Tools of course!)

RemyRAD wrote: We know this because we can hear this and because we are professionals who know what we are hearing and we are hearing timing smear and tiny irregularities. The Time differentials come from all the equations that have to happen simultaneously in math. Everything in computers is clocked and with clocks you get timing errors and inconsistencies.

Don't patronise me Remy: I am also an "audio professional". I would never pick an argument with you on the subject of analog electronics, but when it comes to DSP you are obviously rather ignorant. As I said above, there is no such thing as jitter ITB: sure, all calculations are clocked at the processors' clock speed, but this is not relevant... did you ever need to change your CPU clock speed because you started working at a different samplerate? :lol:

The only thing that matters about the CPU clock is that it is fast enough to calculate each block of samples before it is required by the soundcard... then its up to the soundcard to clock these samples out at the correct speed, and if there is any time-smearing (ie: jitter) you need to look here for the cause.

So, you hear differences do you? In what circimstances? Presumably you mixed a track entirely ITB, bounced it through a single channel of a summing device, then split the exact same mix into stems and summed it externally with the same box. Then you imported both files and did an invert / sum test... what, you mean you didn't do that?

Ok let me guess: you can hear a difference between mixes done entirely ITB using generic plug-ins (and probably also using nearfield monitors in an untreated room) and mixes done with an SSL console in a large studio well equiped with tasty outboard... well, what a suprise!

Let me ask you this: you accept the ability of a computer to change the volume level of a file, right? That means multiplying every single sample by a scaling value. How about an EQ plug-in? Nothing wrong with that, right? If it uses an IIR algorithm that means a minimum of 4 multiplications and additions per sample. It its an FIR algorithm it will need rather more, or more likely it runs the signal through an FFT followed by an iFFT. Then there's convo reverbs... a simplistic implementation might require literally thousands of multiplications and additions per sample, while a more sophisticated algo will probably use another pair of FFT / iFFTs

Apparently the computer is capable of doing all these advanced mathematical processes without causing this smearing you speak of, but not a simple A + B + C etc addition? That's nonsense and you should know better.

IIRs is correct about the maths and how the computer implements it. Clock jitter at the calculation level plays no part in the result, and there is no concept of "smearing". However, I don't think Remy or anybody can ever actually perform an A-B test where the only difference is whether the additions are performed digitally or as an analog summation.

I'm not surprised that a mix done with analog summation sounds different from a mix summed in a DAW. For starters, there will be differences in the number of A-D and D-A conversions in the two cases and where the conversions come in the chain. There are also technical differences in the quantisation noise powers of summed D-A conversions versus that of a single conversion after digital summation.

I don't think it's profitable to argue that the maths is different or that the two cases should sound the same. There will be subtle differences the sound of the two cases just as there are subtle differences in the sound of different high-end converters all supposedly doing the same job.