Voltage Summing Questions For Console 2-bus

[Matt C]

I've got a couple questions for the group about voltage/passive summing.  Since virtual earth summing is so much more common it's been hard to find good info that's applicable.

I've got a console that uses voltage summing to combine the 8 mono groups into the main 2-bus (in the schematic below it's labeled "matrix").  I don't really want to rebuild the whole thing with a different topology, but I'm looking at it and wondering if there's anything worth tweaking in order to optimize performance.

I was looking at the 33k summing resistors and thinking they seemed high, but then realized I don't really know what I would be gaining by lowering the value of these.  The insertion loss, and the gain needed by the following amplifier (and corresponding noise) only depends on the number of channels connected, which isn't going to change.  I know higher value summing resistors will reduce cross-talk between channels. But is there anything to gain from using smaller summing resistors, say 5-10k? Less Johnson noise? I should say the group outputs that are feeding this summing bus are discrete op amps with just a 10 ohm build-out resistor, no pan pot, so they have pretty low Zout and can drive reasonably heavy loads.

My other question is regarding the volume pot in the schematic, labeled "Matrix Master".  Obviously it works, but I've never seen any other examples of this implementation and I'm wondering if it looks particularly offensive to anyone, or if I should just leave it alone.

 

[JohnRoberts]

I've got a couple questions for the group about voltage/passive summing.  Since virtual earth summing is so much more common it's been hard to find good info that's applicable.

Lots of info, but good info is always scarce.

I've got a console that uses voltage summing to combine the 8 mono groups into the main 2-bus (in the schematic below it's labeled "matrix").  I don't really want to rebuild the whole thing with a different topology, but I'm looking at it and wondering if there's anything worth tweaking in order to optimize performance.

The fixed gain stage is likely to be the weakest link (easiest to improve) but impossible to determine from a generic IC5 description.  From observation it appears to be a unity gain stable op amp running at a fixed gain of something well more than unity... (I can't accurately read the feedback R value but at least 10k so more than 20 dB gain).

I was looking at the 33k summing resistors and thinking they seemed high, but then realized I don't really know what I would be gaining by lowering the value of these.  The insertion loss, and the gain needed by the following amplifier (and corresponding noise) only depends on the number of channels connected, which isn't going to change.  I know higher value summing resistors will reduce cross-talk between channels.

actually higher impedance resistors are more susceptible to crosstalk. (I did one console with 2.4k bus resistors for that reason).

But is there anything to gain from using smaller summing resistors, say 5-10k? Less Johnson noise? I should say the group outputs that are feeding this summing bus are discrete op amps with just a 10 ohm build-out resistor, no pan pot, so they have pretty low Zout and can drive reasonably heavy loads.

Unlikely to make an audible difference compared to other stuff you can improve. The lack of differential treatment for grounds could be an issue if the layout is large enough for ground at the gain stage to be different than ground at any of the send pots. Any ground difference will be amplified 20dB or more (still can't read the resistor value).

My other question is regarding the volume pot in the schematic, labeled "Matrix Master".  Obviously it works, but I've never seen any other examples of this implementation and I'm wondering if it looks particularly offensive to anyone, or if I should just leave it alone.

This whole thing looks offensive to me but I long ago stopped getting that worked up about console design minutiae. Now spell bus wrong and watch out tho...

The low hanging fruit is to improve IC5 (maybe? I don't know what it is).  Most IC op amps don't like +/-24v rails. So that is probably a DOA. If you are willing to go under the hood, you can reduce the dominant pole capacitor since it is running at a closed loop gain of 20dB (or more). Of course you would have to lose the feedback cap too. 

Since the apparent attractiveness to using this topology is to get the characteristic sound of some popular DOA du jour, in the make up gain stage socket,  maybe don't improve that at all... maybe don't listen to me (what would I know?).


JR

 

[Matt C]

The fixed gain stage is likely to be the weakest link (easiest to improve) but impossible to determine from a generic IC5 description.  From observation it appears to be a unity gain stable op amp running at a fixed gain of something well more than unity... (I can't accurately read the feedback R value but at least 10k so more than 20 dB gain).

It's a DOA (specfically a Yamaha 80100), Feedback R is 18k so gain is around 25dB. What about it makes this the weakest link here?

The lack of differential treatment for grounds could be an issue if the layout is large enough for ground at the gain stage to be different than ground at any of the send pots. Any ground difference will be amplified 20dB or more (still can't read the resistor value).

I think there's a section on this in Douglas Self's book, maybe I'll see if it's practical to implement here

The low hanging fruit is to improve IC5 (maybe? I don't know what it is).  Most IC op amps don't like +/-24v rails. So that is probably a DOA. If you are willing to go under the hood, you can reduce the dominant pole capacitor since it is running at a closed loop gain of 20dB (or more). Of course you would have to lose the feedback cap too. 

The DOA is potted but I guess I could always replace it with a homemade 990

 

[ruffrecords]

You appear to have 14 inputs so the loss is 14 times but you also have the 10K master pot which is like 3.3 more 33K inputs giving a total of 17.3. The loss is therefore 17.3 times which is 24.76dB hence the 25dB gain make up.

This bus impedance is 33K/14 in parallel with 10K which turns out to be just under 2K. The noise due to the bus resistance is therefore around -120dBu so the noise contribution of the bus resistors is not an issue. Reducing the bus resistors would increase the interaction between controls and mess with the law of the 10K pots. Changing them all to 10K would only improve the bus noise by 3dB.

The pot is a handy property of passive mixing. It means you save a gain stage. I use it a lot in my tube mixer designs with 47K bus feed resistors.

Cheers

Ian

 

[Matt C]

Thanks Ian. Seems like maybe it's not worth messing around with this portion of the console.

I should note that I'm planning on replacing the 8 groups' 10k level pots with simple on/off toggle switches (just for ergonomic reasons) but I don't think that makes any difference when talking about noise.

 

[JohnRoberts]

It's a DOA (specfically a Yamaha 80100), Feedback R is 18k so gain is around 25dB. What about it makes this the weakest link here?

Because everything else is passive...

My point is the feedback cap means that it must be unity gain stable which gives up some 20dB of loop gain margin.

But maybe you like the sound of it?  You don't need to please me.

I think there's a section on this in Douglas Self's book, maybe I'll see if it's practical to implement here

maybe determine if there is a problem first, don't fix it if it isn't broke.

The DOA is potted but I guess I could always replace it with a homemade 990

Or ignore me...  8)

[edit] but if you upgrade the DOA you can brag about that... [/edit]

JR

 

[Matt C]

My point is the feedback cap means that it must be unity gain stable which gives up some 20dB of loop gain margin.

But maybe you like the sound of it?  You don't need to please me.

maybe determine if there is a problem first, don't fix it if it isn't broke.Or ignore me...  8)

[edit] but if you upgrade the DOA you can brag about that... [/edit]

It sounds fine to me but I was looking at this this more in terms of noise reducing or any other more practical considerations, not searching for the magic tone. 

If I'm aiming to find out if the grounding scheme is a problem, does that just become a simple matter of probing the local ground to look for noise? Or am I missing a piece of the puzzle?

 

[JohnRoberts]

It sounds fine to me but I was looking at this this more in terms of noise reducing or any other more practical considerations, not searching for the magic tone. 

I would be tempted to drop in a modern IC uber-op amp, but you will need to scrub a few volts off the PS rails. (No suggestions since I haven't used any of them, but the data sheets are spectacular).

If I'm aiming to find out if the grounding scheme is a problem, does that just become a simple matter of probing the local ground to look for noise? Or am I missing a piece of the puzzle?

The answer is in the ultimate noise floor coming from the bus amp with all channels turned down.

Is the bus noise floor hiss or hum? If hiss that is a good sign (could be reduced by lower noise amp), if hum, there may be room for improvement to the bus topology (like differential reference of channel grounds). 

BUT I repeat, don't fix it if it isn't broken.

JR

 

[Matt C]

If I want to use a modern IC there's the OPA604.  It's the only one I've found that can handle the +-24V rails, and seems to have good specs (at least when compared to something like the TL072)

I'll have to test the existing noise a little more.  The console's noise performance is more or less acceptable for me (recording loud music that drowns out all but the most offensive noise), but definitely not great. That's why I'm exploring some possible ways to clean it up a bit. 

 

[JohnRoberts]

If I want to use a modern IC there's the OPA604.  It's the only one I've found that can handle the +-24V rails, and seems to have good specs (at least when compared to something like the TL072)

being better than a TL072 is not a design criteria for a bus amp. May not even be better (quieter) than that old IC5... The old ne5534 is lower noise and decompensated for minimum gains of 3x so already a better candidate IMO, but there are many candidates newer and better than 5534. 

Sticking with +/-24V rail limits your choices to mainly DOAs or specialty op amps, ruling out the new generation of uber-low noise stuff that is mostly running on standard 36V process (or so I hear)..

JR

PS: That OPA604 looks like a nice op amp for lower gain applications, but not my first choice for a fixed 20dB gain stage.

I'll have to test the existing noise a little more.  The console's noise performance is more or less acceptable for me (recording loud music that drowns out all but the most offensive noise), but definitely not great. That's why I'm exploring some possible ways to clean it up a bit.

 

[alexc]

Tha yamaha 80100 potted module performs quite well in my 18:2 sum box which uses the non-inverting opamp config for voltage summing.

The front end is balanced with 27K each phase (from memory) feeding some some moderately low input Z traffos. (haufe).
Any suitable output traffos is good.

The end to end THD+N was super low ..  I did compare the yamaha to some diy doa's in this app and the yamaha beat the others in terms of thd and hum-noise margin ...  by a small amount. 

After testing with some modern uber-opamps, there was a measurable improvement .. but not that much.
ie. ne5534 and then with the super lme types 

They all sounded pretty much the same to my ears  (which are not the greatest they have been!)

I use that sum box for my diy builds in conjunction with a behringer ultralink pro (8chan split/mix unit)  and a good fx unit 

I swear the behringer sounds really good and measures very, very well  (at 40% of the cost) ...  the two together just works!

Also built a similar box with virtual earth summing with the opamp inverting mode ...  didn't measure as well, although the diff could be said to be small enough to be 'other factors' ....

Build tally  so far :  voltage summing : 2    current summing :1  and counting.

doa based summing is pretty easy to get great results - tube summing is a lot harder!