strange results from channel fader grounding experiment

[JohnRoberts]

It is difficult to give a complete answer without knowing the design, but in general if inputs are only connected to the sum bus when assigned, connecting more will multiply the summing opamps input noise by N+1 times, as well as summing together the noise floor of all the input channels, which will increase +3dB for every doubling in number of channels summed..

In addition to this ideal summation of input noise, and expected amplification of sum amp noise, the difference between the sum amp ground reference and all the input channel ground references will also get amplified by that N+1 term (this is the hum you hear increasing or decreasing with ground tweaks).

There are scientific and brute force approaches for dealing with ground reference differences. Realistically we can't expect ground in the master section and a few feet away at a distant input channel to be exactly the same voltage. The scientific solution is to consider that distant ground as another signal that gets summed to the master then netted out from the hot portion of signal. Brute force, is to throw heavier wire at all ground connections, but this will only make incremental improvements not get us completely noise free.

While one could make every bus fully balanced and differential, that would roughly double the circuitry and cost, not insignificant in a console. A fair compromise to generate a single ground reference bus and use that sum for all buses, works pretty well...

At the end of the day it depends on how much cost complexity you can tolerate... brute force is not obvious or simple when dealing with very tiny voltages, but more accessible to DIY experiments.

JR

[jrp]

Thank you. Still, it surprises me that it makes such a difference if the channels are turned on, or not. Hum is being more prominent on channels further away from the master, but only very slightly. Also i noticed a big change in higher frequency noise.

The short end connected to chassis gave the best results in any case expect when only the first 4 channels were active.

I might want to try a connection to chassis for every fader, see what happens. This is brute force propably, but it costs me nothing but time.

A fair compromise to generate a single ground reference bus and use that sum for all buses, works pretty well...

Sorry, this part i do not understand. Are you describing the normal unbalanced bus concept found in this and most other mixers?

[JohnRoberts]

Originally Posted by jrp

Sorry, this part i do not understand. Are you describing the normal unbalanced bus concept found in this and most other mixers?

We can create an additional ground reference bus, by summing low value resistors (say 10 or 20 Ohms). One resistor from each input channel ground, and finally one from the master section ground. This summed (average) ground connects to the + opamp terminal in a common virtual earth sum amp. This is clearly not impedance balanced and only perfectly differential when all channels are assigned. But all channels assigned is the worst case for this ground reference noise amplification, so when you really want complete cancellation. Less perfect cancellation when less stems are assigned, and less noise gain is in effect, is a tolerable compromise.

Ideally using a second leg for every bus would be even better, and roughly 2x as expensive.

JR

[jrp]

Ah, this sounds like a very fine idea!
I might give it a try, although it involves quite a bit of working.
I know it is impossible, but i would love to hear an opinion if it is worth the trouble for this desc...
But why the resistors? Why not connect the channel grounds directly and feed them to the summing amps?

[JohnRoberts]

Originally Posted by jrp

Ah, this sounds like a very fine idea!
I might give it a try, although it involves quite a bit of working.
I know it is impossible, but i would love to hear an opinion if it is worth the trouble for this desc...
But why the resistors? Why not connect the channel grounds directly and feed them to the summing amps?

it probably isn't worth it.. buy a new console...

But to answer your questions about using hard wire vs. resistors to create a differential amp.. just think about it... If you connect a bunch of wires together, what are the chances that you will get an equal, perfectly proportionate combination of voltages from all the different ground nodes (no chance).

The values I suggested (10-20 Ohm) are low enough to not contribute any thermal or johnson noise, but large compared to your existing ground system so they won't create loops or other mischief.

JR

[jrp]

i was sleeping over this, it suddenly became clear when i started thinking about the ground potentials as signals.

I am not going to get a new console anytime soon...
I have become an enthusiast when it comes to room treatment and diy electronics. All on the cheap more or less. But it is non comercial, i live on being a musician and teacher.

I like the M600 for what it can do for me. I got it in perfect condition, great faders, everything working.
And by now i am really starting to get to know this thing. I can take it all apart and get it back together, something i would never have guessed in the past.

It sounds a lot clearer after what i did to the master section.
I don´t like the contoll room section, strange old switching opamps, looses quite abit against the master section. Will see where i get with this, maybe i redo it completly...

No, for now i want to see where the brutal force methode will lead me. It is already very quiet, but i like to dig into this, just for the fun and sport of it.

[Matt Syson]

Hi
There is probably more 'mileage' in establishing a mix ground setup, especially if you have some spare pins available on module connectors (or 'lash it in' if it is a hardwired beast).
There should be 3 or 4 'grounds' on any setup.
Chassis, Audio ground, switching ground and MIX ground and all 4 should remain separate up to the point where they HAVE to be linked, at one place only.
This is of course a massive 3 D puzzle and you now have to understand what WAS done, create a new plan in your head (or on paper), then implement it.
Is it 'commercially viable'? probably not, but is it an interesting challenge, YES!
It takes a leap of faith to strip out and reconfigure an existing grounding system. I have done it a few times on various desks. Make sure you FULLY test and record measurements of Hiss, Hum, distortion, crosstalk and frequency response as messing about can make some aspects worse if you are not on top of it.
Matt S

[jrp]

Not sure if i am up to this, but just to make it clear:


Chassis ground - shielding. I/O connector shielding is included in this.

Audio ground - signal ground on all the channels.
In my case provided by a Bus along with the Supply voltages. I had already been thinking: Wouldn´t it be a better design to power each Module (or maybe sets of 2-4) with it´s own wire, rather than chaining them? Would this not help a lot to establish the same reference on all channels - witch as i have learned is important when summing the channels.

Mix ground - I am interested in learning about this. Is it what John Roberts had described? 10R resistor from every channel ground to the summing amp + pin in an unbalanced mixbus?
Or something else?

Switching ground - Not sure what this is. Ground from relays if they are used and powered by the audio supply? The ground where turned off or bypassed circuit parts are switched to?

What about Leds? When i build some own design i always keep those seperated.

[Matt Syson]

Hi
That's about it.
LEDs would come under 'switching' of relays etc.
That's the plan, your mission is to implement it!
Matt S
PS the mix ground is what John mentioned.

[JohnRoberts]

Not to scare you, but it isn't always that simple. In a large system, running separate grounds only tied together at a single point, can encounter problems from inductance in large ground systems at HF (RF).

I lean toward simple highly parallel ground system for shielding and power return, with audio ground (?) treated mainly as a signal. In a large console it is impractical to treat the entire path as fully balanced differential, so there can be many different local audio grounds, that we use differential circuitry to reference into or away from. Every (most) active gain stage can be configured as a differential to keep ground references straight coming or going.

We need to not think of ground as single voltage.. ground is a concept, and audio is two signals, or often one moving one relative to another that isn't moving so much.

JR

[Matt Syson]

Hi
And as John alludes to, the differences of DC reference (the tens of millivolts that appear due to power current flowing), power line interference voltages and RF 'noise' voltages tend to behave and appear differently.
Since there are far more 'RF' interference sources in recent years (compared to early 1980's) substantial ground plane systems seem to have an 'edge' on RF rejection, using the 'mass ground' concept. Keeping 'received' RF from connectors and the chassis out of the audio ground is a good plan.
Matt S

[jrp]

Yor information has made a lot of things clearer to me. stuff i was thinking about and trying to understand for quite some time.

This insight does not scare me, but it makes me consider that i should be realistic about what makes sense and what i can do with my experiance and possibilities (both limited...).



First i need to learn some more:

In a large system, running separate grounds only tied together at a single point, can encounter problems from inductance in large ground systems at HF (RF).

A 24CH 16Bus Stereo Mixer is a large system?
Because something in me thinking of waterpipes pictures a star will give the most even distributin of whatever. I don´t want to be naive and run into something i canot tame.

I lean toward simple highly parallel ground system for shielding and power return

Parallel in this context meaning a: independent (parallel) power connection on each channel instead of a Bus or b: Channels connected in parallel to the bus like already done in my console?

My console takes the shield from ballanced i/o connections and takes it directly to the channel PCB ground. I now consider changing that. (propably the point easiest understood)

Does it make sense to try this on 4 channels only and see what happens? Or will the results be masked so much because so many factors contribute to the whole system and in the end things first promising or useless might look very different?

audio ground (?) treated mainly as a signal

I would have to look at the pcb if those grounds are easy to seperate (and i highly doubt that)
But where does power return end and signal ground start?
Signal Ground : input impedances, pots, pads?
Power return: In an opamp we have dual supply, i figure current flows from one to the other. Where does ground come into play? Every power conditioning cap or decoupling cap?

[JohnRoberts]

Originally Posted by jrp

Yor information has made a lot of things clearer to me. stuff i was thinking about and trying to understand for quite some time.

This insight does not scare me, but it makes me consider that i should be realistic about what makes sense and what i can do with my experiance and possibilities (both limited...).

Another reality is that you can't turn a VW into a porsche by moving where the nuts and bolts inside are attached. We are often better off just starting from scratch.

First i need to learn some more:


A 24CH 16Bus Stereo Mixer is a large system?
Because something in me thinking of waterpipes pictures a star will give the most even distributin of whatever. I don´t want to be naive and run into something i canot tame.

Size is relative.. I've handled over a 100 stems in a L/R mix, so 16 is relatively modest. Physical size also matters. Power supply distribution over feet instead of inches is proportionately more influenced by distributed impedance and reactance.

Parallel in this context meaning a: independent (parallel) power connection on each channel instead of a Bus or b: Channels connected in parallel to the bus like already done in my console?

Look at the classic pin 1 XLR path to ground. In my highly parallel world that pin 1 gets bonded to the chassis which ulimtately goes to earth ground and power supply ground (to return microphone phantom power current).

Likewise the shield on a TRS insert jack ideally "should" get grounded to the chassis. I say should because I suspect many are not, fearful of dreaded ground loops and ground corruption. If the signals being sent from and received into that unbalanced insert jack are treated differentially wrt to that specific node voltage, it doesn't matter what actual ground voltage is at that single point.

Things like this are not trivial to change if not in your original console design.

My console takes the shield from ballanced i/o connections and takes it directly to the channel PCB ground. I now consider changing that. (propably the point easiest understood)

OK to connect to both chassis and board ground AS LONG AS SIGNALS ARE PROPERLY DIFFERENTIAL.

Does it make sense to try this on 4 channels only and see what happens? Or will the results be masked so much because so many factors contribute to the whole system and in the end things first promising or useless might look very different?

By all means experiment... it helps to have a plan with some understanding of what you expect... In the example of adding a ground reference bus, a partial implementation will only give a partial benefit.

I would have to look at the pcb if those grounds are easy to seperate (and i highly doubt that)
But where does power return end and signal ground start?
Signal Ground : input impedances, pots, pads?
Power return: In an opamp we have dual supply, i figure current flows from one to the other. Where does ground come into play? Every power conditioning cap or decoupling cap?

I would not expect that after the fact redesign of PCB layouts to be easy, or even possible for some cases. I had the luxury of designing from scratch and still made my share of mistakes.

JR

[jrp]

thanks! Ok, for now i am gonna leave grounding as it is for the most part.
I/Os on the board are opamp balanced, the stuff connected is using INA/DRV134 or transformers in most cases (almost all my outboard is diy)

I am somehow curious to check out the "brutal farce" method, tying all channel grounds together with a thick cable going straight to the PSU.
From all i learned in this thread this is the least educated approach. But as you mentioned earlier, it is the most accessable when working with a tiny lab and on the cheap, having little experiance. And, most important, it is easily undone.
I have read about such experiments from other people that turned out good.

You can´t turn a VW into a Porsche, but it can sound a bit like one (in that case you seriously need to visit a garage).