Differential Grounds in Summing Bus

There's been a lot of talk here about this idea of treating ground/common differentially in a summing bus to reduce noise, and I understand the idea vaguely, but I'm having trouble picturing exactly how this would be implemented/retrofitted into a mixer. Can anyone point me to a detailed explanation of this, beyond all the little snippets I've dug up here on the board?

This is what I'm imagining, someone tell me if I'm way off base here:
- Create a new 0V/common bus independent from existing grounds.
- Connect this bus to each channel's ground (I'm guessing it should be connected as close as possible to where the + signal output is?) through a small resistor (I've read recommendations anywhere between 10r and 1k)
- At the summing amp, disconnect the op amp's + terminal from the existing ground, and instead connect it to this newly created ground bus

This book has a description:

D. Self Small Signal Audio Design 2010
p 451
Ground Cancelling Summing Systems

You might get lucky and find those two pages online.

Matt C said:

There's been a lot of talk here about this idea of treating ground/common differentially in a summing bus to reduce noise, and I understand the idea vaguely, but I'm having trouble picturing exactly how this would be implemented/retrofitted into a mixer. Can anyone point me to a detailed explanation of this, beyond all the little snippets I've dug up here on the board?

Click to expand...

Just to be clear it doesn't really reduce noise. What it does do is reject ground potential differences between channels and master section.

This is what I'm imagining, someone tell me if I'm way off base here:
- Create a new 0V/common bus independent from existing grounds.

Click to expand...

ASSuming you are talking about a virtual earth summing structure. You can call it anything you want but it needs to be a parallel bus, similar to the audio + but instead of signals, summing together the local channel audio grounds, or audio - get summed too. 

The outputs of these two buses get combined differentially, effectively compensating for of all the sundry local ground potential differences.

- Connect this bus to each channel's ground (I'm guessing it should be connected as close as possible to where the + signal output is?) through a small resistor (I've read recommendations anywhere between 10r and 1k)

Click to expand...

It needs to be sending from the local channel ground, typically near the fader/pan/etc.

- At the summing amp, disconnect the op amp's + terminal from the existing ground, and instead connect it to this newly created ground bus

Click to expand...

There are different ways to accomplish this... Cheap and dirty is to just sum together low impedance resistors from each channel local audio ground and connect to the + sum bus amp input ( the cheap and dirty approach often uses low z resistors for the ground bus, but they still need to be precision matched for differential accuracy.) In fact there needs to be one more same value resistor from the sum to master section local ground.

The less cheap and dirty approach is a same value bus and sum amp in parallel (followed by a differential combining amp).

I have used both ways. One perhaps not obvious downside is that the number of audio sends going to the audio bus need to be the same as number of local ground resistors being bused. If only a fraction of channels are assigned to the sum bus, the associated grounds need to be deselected too, for differential accuracy.

JR

[edit- I have cheated on the requirement to deselect grounds in some modest mixers (say 16 inputs or less). Hard wiring all grounds and living with the very small errors when small numbers of channels selected. When only a few channels are selected the noise gain of the sum amp is proportionately lower (for less error sensitivity). You could alternately back ground signal bus resistors when deselected, but that forces the bus amp to work at worst case noise gain for less than worst case situations. [/edit]

JohnRoberts said:

Just to be clear it doesn't really reduce noise. What it does do is reject ground potential differences between channels and master section.

ASSuming you are talking about a virtual earth summing structure. You can call it anything you want but it needs to be a parallel bus, similar to the audio + but instead of signals, summing together the local channel audio grounds, or audio - get summed too. 

Click to expand...

Yes, talking about virtual earth summing here. I understand your point about rejecting ground potential differences, but such ground potential differences can sometimes inject noise into the signal, which in this case would be rejected (to some extent)?

Matt C said:

Yes, talking about virtual earth summing here. I understand your point about rejecting ground potential differences, but such ground potential differences can sometimes inject noise into the signal, which in this case would be rejected (to some extent)?

Click to expand...

I guess it depends on how you define noise, and the nature of these ground potential differences. Inside a grounded metal chassis (with external power supply transformer), there should not be much mains frequency hum, RF,  or non-signal related ground noise. Note: I have seen situations inside consoles where noisy voltage regulators could corrupt grounds and that ground noise could be amplified by bus amp noise gain (if not corrected for by using differential sum bus processing).

The more common ground noise shows up as signal crosstalk between channels, an even larger concern for broadcast consoles (that may have completely unrelated signals running around inside) than for recording where small amounts of crosstalk are innocuous, but always a difficult figure of merit on the spec sheet that designers try to meet/beat.

Analog summing/combining is actually a mature technology. I stopped trying to make further improvements*** last century.

JR

**** I am repeating myself again, but the noise floor from a reasonably competent sum bus amp, will be well below mic preamp and ambient room noise floors.

Matt C said:

- Connect this bus to each channel's ground (I'm guessing it should be connected as close as possible to where the + signal output is?) through a small resistor (I've read recommendations anywhere between 10r and 1k)

Click to expand...

These resistors should be as low as possible in order to minimize their noise contribution. However they should be higher than the impedance of the source that drives them. This source has very low impedance since it's the console's 0V bus. Also they must be sufficiently high compared to the possible variations of contact resistance in the connectors - we're talking typically a tiny fraction of ohm. 100r always seemed to me like a good compromise, although going up to 10k would probably alter noise by less than 1dB.

abbey road d enfer said:

These resistors should be as low as possible in order to minimize their noise contribution. However they should be higher than the impedance of the source that drives them. This source has very low impedance since it's the console's 0V bus. Also they must be sufficiently high compared to the possible variations of contact resistance in the connectors - we're talking typically a tiny fraction of ohm. 100r always seemed to me like a good compromise, although going up to 10k would probably alter noise by less than 1dB.

Click to expand...

+1 if the Rs are too small the audio low bus is no longer a simple arithmetic sum of the different channel grounds but a parallel ground path with poorly defined current steering, ruining the utility for maintaining signal integrity.

JR

Thanks guys this is very helpful. 

Just to make sure I'm not misunderstanding; the only existing connection that needs to be broken/changed is the connection from the summing amp's + terminal to its local ground (which should now be through a resistor of the same value as the other new ground bus summing resistors).  On the individual channels I'm not breaking any existing connections, just sampling the local ground with a summing resistor going to the new dedicated ground bus.

correctamundo...

JR

That sounds right. Think of it as you're "lifting" that ground through 100R resistors. So when you're done, you should be able to measure some resistance between the two grounds.

squarewave said:

some resistance between the two grounds.

Click to expand...

Only one of them qualifies as "ground", the other is a bus (another confusing name).
The various sorts of "ground" (earth, chassis, audio ground, logic ground, Velvet Underground,,.,) are destined to be connected together somehow. The ground sensing bus is due to be connected to the non-inverting input of the VE mixer and definitely not to be "grounded".