ACA setup... 12:1, 8:1 or 6:1?!?

I decided to post this here and not on Jeff's ACA/Bo build thread since I'm still in the fact finding stage.

My question is to Jeff or anyone who has experience configuring ACA's.

Jedi Jeff chose in his 32ch desk one inverting ACA per 8 channels, 4 ACA's to a card.
(Edit: if it wasn't clear, these ACA's are doubled for L & R, I'm just asking about one half the outputs.)
They sum to an ACA/Bo that inverts again and brings them all down to a 2 mix with fader and mults.

Since I'm going with 24 channels, should I use 3 of the 4 ACA's on the PCB (8:1) or drop inputs to 6:1 and use all 4 or...perhaps go with 12:1 and invert with only two ACA's?    :

I'm using 47k bus resistors throughout.

What setup in your estimation would perform best?

Also, do you think using an ACA/Bo for Aux buses is overkill?
It seems an appropriate fit when selecting Cue's and inserts prefade.

Thanks in advance for your responses!
-Boji

Google thinks ACA is a Nelson pass power amp? (or healthcare).

I will speculate from context that you mean "Active combining amp"

Back in my 1980 article about console design I speculated about using multiple bus amps to mitigate the inadequate open loop gain to manage large bus structures in only one bus amp (I also discussed other remedies). http://www.johnhroberts.com/des_art_1.pdf See fig 12 for the distributed busing. Please forgive me for spelling bus wrong back then, but that was almost 40 years ago and I was even more ignorant than now.

Not to be overly pedantic and actually answer your question, either way is probably OK (regarding audibility). On paper more is always better... the goal is to improve loop gain margin so 6 stems per bus is better than 8.  FWIW I have made 24 input mixers with a single bus amp and nobody asked for their money back.  (My biggest console had over 100 feeds to the L/R bus so that was not quite so easy).

You can simulate the results of 6 or 8 feeds per bus with a resistor to ground to make bench measurements, of course the actual topology (circuit) can have additional impact,,,

Good luck this is a pretty mature technology and there are other decisions with far more audible consequences than this one.

JR

On paper more is always better... the goal is to improve loop gain margin so 6 stems per bus is better than 8.

Click to expand...

Thanks so much for tagging your opinion to a personal example!  I wish I could plug in your GPS coordinates and attach a beer to a RC drone.  8)

there are other decisions with far more audible consequences than this one.

Click to expand...

If you could return the drone with a few xanax taped to it that would be great.

There is also a slight noise improvement. Summing 24 channels as 3 lots of 8 followed by one lot of 3 is a tad over 3dB quieter than summing 24 channels straight into 1. Check out Doug Self book for details.

Cheers

Ian

Thank you RR for the confirmation.

It seems a little counterintuitive since adding more stuff should inject more Johnson noise.

But I'll of course take your suggestions a read why it is so. Mr. Self tried to tell me a few times but I wasn't listening, apparently.

Thanks again gents!

ruffrecords said:

There is also a slight noise improvement. Summing 24 channels as 3 lots of 8 followed by one lot of 3 is a tad over 3dB quieter than summing 24 channels straight into 1. Check out Doug Self book for details.

Cheers

Ian

Click to expand...

24x is more noise** than SQTRT of ( (8^2)+(8^2)+(8^2) )...  but that assumes the bus amp is the dominant noise source.

In my experience mic preamp noise and/or even room noise is dominant.

JR

** the real issue as I wrote a few decades ago from large bus structures is phase shift and linearity degradation (distortion), but still not top of the audible list. I wish it was because I have analog summing solutions, but digital summation is pretty much perfect, so analog summation is moot.  :'(

have analog summing solutions, but digital summation is pretty much perfect, so analog summation is moot.

Click to expand...

I think exploiting the field effect of quantum particles, all rubbin' together and tossing each other around is conceptually more pleasing than adding and dividing 0 & 1's.  8)

I know you don't youtube John but mixes like these keep me motivated:
https://youtu.be/P_B0jCCdbUI?t=508

boji said:

Since I'm going with 24 channels, should I use 3 of the 4 ACA's on the PCB (8:1) or drop inputs to 6:1 and use all 4 or...perhaps go with 12:1 and invert with only two ACA's?    :

Click to expand...

Noise calculation of "bucket mixing" is not very hard to do. Ultimately it shows that buckets of 2 give the best theoretical noise performance, at the cost of ...cost!
Then it shows that the difference between 3 buckets of 8 and 4 buckets of 6 comes to less than a fraction of dB.
JR gave a brief hint at how this works (24x is more noise** than SQTRT of ( (8^2)+(8^2)+(8^2) ))
OTOH Ian mentions a 3dB improvement at the advantage of bucket mixing.
Yer pay yer money yer make your choice...

I'm using 47k bus resistors throughout.

Click to expand...

That may be the most significant factor in Johnson noise. With 24 stems, the bus Z is about 2k, which is probably a little on the low side for a typical 5534/2520 (or any summing amp with its input devices running at about 50-100uA collector current). A bucket of 8 would show a bus Z of about 5k, which is pretty close to the optimum for the aforementioned summing amp. That would give a double advantage to bucket mixing.

What setup in your estimation would perform best?

Click to expand...

I would think optimisation of the operating current in the summing amps is the most important factor. Unfortunately 2kohms falls just between 5534/2520 (7-10k) and über low noise opamps/JH990 (600-1k)  in terms of OSI (Optimum Source Impedance).
Anyway, you will find that longitudinal noise created by parasitic induced currents in the reference bus (ground, Ov,...) is very often dominant above Johnson noise.

Also, do you think using an ACA/Bo for Aux buses is overkill?
It seems an appropriate fit when selecting Cue's and inserts prefade.

Click to expand...

I don't understand this comment. Do you mean the operating level is different prefade?

abbey road d enfer said:

Anyway, you will find that longitudinal noise created by parasitic induced currents in the reference bus (ground, Ov,...) is very often dominant above Johnson noise.

Click to expand...

Sounds like a jumping off point into balanced summing buses... ?

Newmarket said:

Sounds like a jumping off point into balanced summing buses... ?

Click to expand...

Indeed. In fact balanced summing bus is too constraining, because strictly it means balanced drive and equal resistor values. What's needed is remote ground sensing bus.

Wow thanks Abby for your input.  Very much appreciated!

SQTRT of ( (8^2)+(8^2)+(8^2) ))

Click to expand...

Is SQTRT the same thing as SQRT ?

Sorry for the bad maths... for 8 buckets, SQRT of 3(n^2) gives me ~13.86. How you plug that in to get 5k?  :

Yer pay yer money yer make your choice...

Click to expand...

Gotcha, might as well go for the goods, since this is not a resale thing.

Unfortunately 2kohms falls just between 5534/2520 (7-10k) and über low noise opamps/JH990 (600-1k)

Click to expand...

So 2k is 24:1. given I go with 6:1 yeah that will be higher than 5k but hmmm....I've heard people drop in the 990's for ACA's so I wonder how they keep the OSI below 1k? (OSI...thinking TCP/IP...Layer what?)  ;D

I don't understand this [Aux prefade] comment. Do you mean the operating level is different prefade?

Click to expand...

Probably this is more about my lack of real world, larger band live recording...But I'm trying to imagine where auxes are at near-unity, faders set mix for cues so an Aux master volume isn't all that important, whereas dialing in a prefade custom mix, the bus might need boost or attenuation. An ACA/Bo would let me use the 10k faders I have, but I guess it seems a bit overkill to use tx's for cue.

boji said:

Is SQTRT the same thing as SQRT ?

Click to expand...

I would think so. ;D I guess Ian had a bout of typo.

Sorry for the bad maths... for 8 buckets, SQRT of 3(n^2) gives me ~13.86.

Click to expand...

That's the theoretical noise improvement calculation.

How you plug that in to get 5k?  :

Click to expand...

A bucket of 8 47k resistors plus a 47k FB resistor result in 47/9=5.2k

I've heard people drop in the 990's for ACA's so I wonder how they keep the OSI below 1k?

Click to expand...

I know people that have dropped ULN opamp (AD797, LME49990) in place of 5534, and wondered why they had no noise improvement.

Probably this is more about my lack of real world, larger band live recording...But I'm trying to imagine where auxes are at near-unity, faders set mix for cues so an Aux master volume isn't all that important, whereas dialing in a prefade custom mix, the bus might need boost or attenuation.

Click to expand...

I believe I understand you think you're going to operate the aux bus at a lower level than the mix bus; that is indeed somewhat true, but should not be desperately so. And most musos don't hear hiss, they're all deaf...

Thanks Abby!

So.. does anyone have an API aux or echo return schema I can learn from?
Not seeing any in the tech docs section.

If not, I'll experiment with an ACA/Bo card but I'd like to see how they did it back in the day since a few pictures I've seen look like they used a 2510 and no tx. Not sure how they sum or return if used as an fx loop.

boji said:

Wow thanks Abby for your input.  Very much appreciated!
Is SQTRT the same thing as SQRT ?

Click to expand...

yes my fat fingers  (I currently have 10 fire ant bites on my right hand... ).

Sorry for the bad maths... for 8 buckets, SQRT of 3(n^2) gives me ~13.86. How you plug that in to get 5k?  :

Click to expand...

it is just a relative multiplier so compared to 32x the op amp ein...

JR

Gotcha, might as well go for the goods, since this is not a resale thing.
So 2k is 24:1. given I go with 6:1 yeah that will be higher than 5k but hmmm....I've heard people drop in the 990's for ACA's so I wonder how they keep the OSI below 1k? (OSI...thinking TCP/IP...Layer what?)  ;D
Probably this is more about my lack of real world, larger band live recording...But I'm trying to imagine where auxes are at near-unity, faders set mix for cues so an Aux master volume isn't all that important, whereas dialing in a prefade custom mix, the bus might need boost or attenuation. An ACA/Bo would let me use the 10k faders I have, but I guess it seems a bit overkill to use tx's for cue.

Click to expand...

I've posted the schematic for the API E548 echo return module in the Tech Docs section.
It's in three parts as the schematic is an original Blueprint and very bad quality so I had to zoom in with my camera....

boji said:

Not sure how they sum or return if used as an fx loop.

Click to expand...

Typically, an FX return is nothing more than a simplified line input channel. It does not have a mic input, reduced routing/aux sends and simplified or non-existent EQ.

Sweet! Thank you Kevin for taking the time to do that! 

boji said:

Thanks Abby!

So.. does anyone have an API aux or echo return schema I can learn from?
Not seeing any in the tech docs section.

If not, I'll experiment with an ACA/Bo card but I'd like to see how they did it back in the day since a few pictures I've seen look like they used a 2510 and no tx. Not sure how they sum or return if used as an fx loop.

Click to expand...

Abbey gave you the gist of it. I typically make it a cut down version of the channel routing module because it needs access to many of the same buses. Only needs simple AUX sends to give the musos some echo in their cans. The front panel space saved can be used to include a level control pot plus maybe simple EQ (HPF at least). You  probably don't want to Solo it because if it mutes all the other channels there will be no send, but you may want or may not want to mute it when proper channels are Solo'd

Cheers

Ian

The front panel space saved can be used to include a level control pot plus maybe simple EQ (HPF at least). You  probably don't want to Solo it because if it mutes all the other channels there will be no send, but you may want or may not want to mute it when proper channels are Solo'd

Click to expand...

Helpful suggestions! Now that you mention it Aux bus HPF seems useful since they won't always be postfade and the channel's HPF's may not all need to be on.  Perhaps this is a chance to play with something active this time around.  Ok, truth be told I don't want to wire up any more of those tiny rotary switches!

I also need to sort out pan... I'll spitball some circuits hopefully this weekend.

boji said:

Now that you mention it Aux bus HPF seems useful since they won't always be postfade and the channel's HPF's may not all need to be on.

Click to expand...

You seem to be worried about the possibility to overload a reverb unit with bass. Normally, a half-decently designed reverb should have a HPF built-in its drive circuitry.
In practice, once you start to process the reverb sends, you end up using the full monty, EQ, delay, compression... I would leave the sends alone and concentrate on some facilities on the returns.

  I also need to sort out pan...

Click to expand...

Quite often I see penny-pinching on the pan circuit of groups and FX returns. Some thought should really be given; after all, these are on the main bus, which is supposed to be the noblest path in the mixer!