Summing ITB/OTB Of course? Nonsense?

[ghettosynth]

We had an interesting discussion going over in the S1 features thread that was getting off topic so I thought I'd pick it up over here.

Reading on this it's hard to really get what people appreciate. There seem to be several claims about ITB summing that don't pass the sniff test.

The first thing that struck me is in a comment about DIY summing that the entire point of OTB summing is to avoid the DAWs summing mixer, i.e., that using passive summing networks isn't, in and of itself, about adding tone, rather, it's about avoiding math.

https://www.gearslutz.com/board/low-end ... boxes.html

Ok, well, I'm trying to stay open minded about this, so, is there anything to this idea?

In the same thread several claim that any "tone" derived from this is really just a function of the makeup gain, i.e., saturation on the 2-bus, that the summing itself doesn't contribute to the sound.

At any rate, here's someone who feels strongly about analog summing, he even gives some samples to listen to.

http://www.farmelorecording.com/in-the- ... l-summing/

[Gamma-UT]

When I see people talking about wire-wrapping a passive mixer out of a bunch of cables and resistors, I'd be very surprised if there isn't an audible difference between that and straight digital summing. Unless I'm missing something that SPICE could confirm, without taking care to match impedances and stuff, surely everything is going to ring like a bell every time a big transient hits?

I can see how that might add some perceived mojo for some although it might infuriate actual mixing-desk designers who wonder why they added a bunch of fandy-dandy op amps and buffers to their products in order to maintain a clean, transparent signal path. "Hey, we could have thrown a bunch of 100k resistors in a box and charged an arm and a leg for it and people would have loved it."

[ghettosynth]

When I see people talking about wire-wrapping a passive mixer out of a bunch of cables and resistors, I'd be very surprised if there isn't an audible difference between that and straight digital summing. Unless I'm missing something that SPICE could confirm, without taking care to match impedances and stuff, surely everything is going to ring like a bell every time a big transient hits?

I can see how that might add some perceived mojo for some although it might infuriate actual mixing-desk designers who wonder why they added a bunch of fandy-dandy op amps and buffers to their products in order to maintain a clean, transparent signal path. "Hey, we could have thrown a bunch of 100k resistors in a box and charged an arm and a leg for it and people would have loved it."

No doubt. I'm thinking that I should get into the passive summing box business. The biggest issue there is I'm not sure that I could contain my snark when selling my products. I think that a passive mixer might have been the second or third project that I built as a kid. I could say that I have decades of experience in the analog summing game.

But ok, I'm not the smartest guy on the planet, so let's keep an open mind...

I'm trying to think about whether there's anything to this "DAW summing alogrithm" issue that people talk about. What is this "summing algorithm?"

This is pretty much what my limited skillset would come up with, or thereabouts. I might think about whether doubles made sense, I might not.

https://www.reddit.com/r/audioengineeri ... algorithm/

I love this though, vocals are more open and blah blah blah.

http://johnvestman.com/DAWSUM.htm

[Gamma-UT]

I'm trying to think about whether there's anything to this "DAW summing alogrithm" issue that people talk about. What is this "summing algorithm?"

This is pretty much what my limited skillset would come up with, or thereabouts. I might think about whether doubles made sense, I might not.

https://www.reddit.com/r/audioengineeri ... algorithm/

I've always assumed it's pretty much the loop shown at the top. I don't know for sure, certainly. But, even with 32bit FP, you've got a 23/24bit mantissa and 8bit exponent. Within a DAW, because the practical number ranges are pretty constrained, the number of bits you actually use of the exponent is going to be what? 4 bits? So, your mantissa is sliding around over a 4 or 5bit range maximum and your five/six LSBs are below the audible threshold anyway. It doesn't seem I'm going to be losing anything audible with that resolution.

[Uncle E]

I've tested the Folcrom and PM8, the small change in sound is certainly not worth bussing out for. Also, someone is already making a version of them for $200 and there are people home brewing them out of DSUB snakes for around $100.

[camsr]

Ok, well, I'm trying to stay open minded about this, so, is there anything to this idea?

If I recall correctly, old protools systems used integer math to sum. If the bitdepth was too low, detail could be lost by the integer summing. So the solution was to DA convert all the stems, mix in analog, and AD the result. This would prevent truncation (and at the same time add noise, which is a barrier to this idea having any merit).

If you are using floating point technology (most everyone is), I doubt there is much appreciable benefit. Reason being floating point has better resolution at low amplitudes, so the truncation doesn't occur.