Analog summing emulation idea

[Kiwburger]

I just had a worrying experience that has caused me to think up a theory why analog summing might be working it's magic (and how to emulate it). It seems this is a highly debated subject, and so far most people agree it's the noise and distortion of conversion and analog gear that explains the difference. This - I think - is a major new piece to the puzzle ...

I have a small studio rig in my home, away from my main studio, for messing with Cubase SX and experimenting with new toys and plugins. I use a Korg Z1, a Pentium IV with an Audiophile 2496, into a Mackie mixer and use Sennhieser HD280 headphones. Very useful, and I get to hear stuff I might miss with my monitors in the bigger room.

I was evaluating a reverb with a sampled hand clap, and I decided to try extreme left and right hand panning. I noticed that the exact same sample appeared to sound different from left to right. One side sounded faintly louder and brighter, but noticable. I was worried that maybe my hearing had changed a little.

Well - it turned out that my Mackie mixer (or possibly the Audiophile card) colors the sound slightly differently left and right. If I swap the cables around, I hear the difference in the other ear. (Phew).

Now this lead me to thinking. Analog gear depends very much on resistors and capacitors for level and eq and impedance matching, etc. We all know that components come in different quality levels, and different tolerances. Cheaper stuff will use low tolerance components. Higher end stuff might use 1% tolerance components or better. But still - tolerance is tolerance, and even 1% can have a significant effect that will be noticable.

The thing about DSP summing, mixing, eq'ing is that it is deadly accurate. Mathematically precise. Sample accurate. Boring even.

I've just realised that the best built analog summing device is going to nowhere near as accurate - even with 1% tolerance parts. That means that - like my Mackie mixer - left and right channels are going to sound slightly different. The pan will not be exactly zero at the centre detent. Even the eq 'sound' - even if no eq is activated - is going to be slightly different on each channel - within component tolerances.

Also - because it takes time for current to flow through longer circuits than throught shorter circuits, there will be subtle phase differences. Very subtle - I think the panning and eq effects will be more noticable - simply because of parts tolerances being fairly wide and fairly random.

As well as the noise and distortion - I think this explains the perception of better stereo imaging and more interesting mix sound.

With DAWs, there is a tendency to work to numbers. All the tracks we want centred, we probably set to exact centre. And with DSP, exact centre is exact centre. And the signal is perfectly equal each side. No wonder it's a little boring - and psycoacoustically, this has to sound a bit wrong. No mix of real musicians could ever be exactly all in the same spot.

I think, mix them up a bit.

Maybe somebody could make an Analogiser plugin. It could simply apply a random offset to panning and a random subtle eq change between each side. Maybe the odd phase shift by a sample. Maybe the plugin could be programmed so that these subtle changes are randomised - maybe using the plugin selection time to seed the randomiser. That way, you could insert the plugins on every track, and they would all be slightly different. Maybe choose a tolerance range. Or emulate various mixers - from SSL to Behringer perhaps for a wide range of extremes. A hint of saturation and even extremely low level white noise - also randomised, within user parameters.

Might work ...

I hereby submit this idea into the public domain - so any dsp developer can use this idea, for free or commercial use.

[egbert]

Aleksey Vaneev has written about this sort of thing - asymmetry etc - as being part of the analog sound. He talked about it in the context of the Marquis compressor IIRC.

[bduffy]

I don't know why they don't just grab an analog guru, á la Aleksey and make an analog host. Imagine Cubase, Sonar, FL )your fave host) with analog summing running through the whole mixer, native EQ's etc. Probably too resource heavy at this stage.

[lfm]

Isn't this partly what the mastering process is doing, apart from matching levels between tracks. Basically it's called exciters, which enters a lot of stuff which is getting the mix more alive.

Recording software in the better class let's you prepare a session with a set of tracks for different purposes that you usually use. So fix such a template with phasers, chorus and panning effects that are very suttle. Or put these on a bus, and use the send effects from the tracks to add this which is summed to the mix on the master bus. And you are set for a new session very quickly with this kind of effects.

I've seen such plugins that use random parameters somewhere. I cannot recall where right now. Some panning plugins had random mode.

[Kiwburger]

Aleksey certainly knows his stuff. I hadn't read his ideas on this before.

But it seems very obvious to me now. Analog components definately come in tolerances - and even 1% is fairly random.

Imagine your stereo signal runs through a few 100K ohm resistors - each of which could be anything from 99K to 101K ohms. The variation is voltage going to be significant.

Imagine a typical DAW mix with many tracks panned dead centre. Now mess up the pan of each one slightly. Then apply a dual band eq to each stereo track, and apply slightly different eq changes to each side.

The mix is going to sound like it has more space and depth. This is exactly what will happen if expensive analog summing is used - because component tolerances will effectively do just that.

Sure, the extra noise and distortion helps. But maybe that's not the major part of the sound? I think this random assymetrical tolerance thing might be a good thing.

It wouldn't add much, if any, overhead to a DAW to implement this analog-like behaviour. A user control knob, giving anything from clean digital to dirty analog would be cool. I could imagine a panel giving user control over the following:

noise
distortion
crosstalk
asymmetry - pan
asymmetry - eq
asymmmetry - phase

Could be interesing .. imagine a DAW mixer with presets for various digital or analog desks ... Neve, SSL, Mackie, Behringer, Paris, etc ..

[Kingston]

NNNNnnnnnnnnooooooooooooo...............!!!!!!!

not again!

Bloody summing fantasists.

placebo goddamit! PLACEBO!

[bmanic]

[fgfab]

placebo goddamit! PLACEBO!

No, I don't agree.

I NEED a 50/60 Hz hum generator.

It sounds so great in my mixes.

[nuffink]

Aleksey certainly knows his stuff. I hadn't read his ideas on this before.

But it seems very obvious to me now. Analog components definately come in tolerances - and even 1% is fairly random.

Imagine your stereo signal runs through a few 100K ohm resistors - each of which could be anything from 99K to 101K ohms. The variation is voltage going to be significant.

Imagine a typical DAW mix with many tracks panned dead centre. Now mess up the pan of each one slightly. Then apply a dual band eq to each stereo track, and apply slightly different eq changes to each side.

The mix is going to sound like it has more space and depth. This is exactly what will happen if expensive analog summing is used - because component tolerances will effectively do just that.

Sure, the extra noise and distortion helps. But maybe that's not the major part of the sound? I think this random assymetrical tolerance thing might be a good thing.

It wouldn't add much, if any, overhead to a DAW to implement this analog-like behaviour. A user control knob, giving anything from clean digital to dirty analog would be cool. I could imagine a panel giving user control over the following:

noise
distortion
crosstalk
asymmetry - pan
asymmetry - eq
asymmmetry - phase

Could be interesing .. imagine a DAW mixer with presets for various digital or analog desks ... Neve, SSL, Mackie, Behringer, Paris, etc ..

Even though a component may have a tolerance of 1% any decent electrictronic engineer will use it a circuit where the tolerance is not critical. Otherwise even well made equipment would have THD figures approaching 100%

[Aleksey Vaneev]

Marquis Compressor is indeed a bit assymetric left/right-wise when Soft/Sharp coloration is selected.

[Kebmaster]

There are the colourtone and valvetone plug ins which apparently emulate analog signal paths but to my ears the effect is so subtle it is not worth it!! I dumped hardware in 2004 to go completely native but found that my creativity went down the toilet!!! I became depressed and ill and very frustrated.
In December I bought an old Spirit SX desk and a Emu 1820m audiodock and now run seperate outs thru the desk. I still mix most of the percussion and effects thru the main stereo out, but seperate some things thru an additional 2 stereo pairs. The seperation and warmth is 1000% better. I am now writing to my potential.

[Kebmaster]

Remember, another heavily overlooked factor is cables and leads. More expensive cable tends to add less colour than cheap cable. Cheaper cable tends to sound more muddy and bass heavy.

[jupiter8]

At least there is no Placebo left for the rest of us.I believe this guy drank the whole bottle.

[farlukar]

I dumped hardware in 2004 to go completely native but found that my creativity went down the toilet!!!

Yeah... always blame the tools

[Kingston]

At least there is no Placebo left for the rest of us.I believe this guy drank the whole bottle.