ground summing bus...

[radiance]

I've a question regarding a ground summing bus as seen in this pic.

When making a summing mixer like in the picture only with multiple busses/groups, would there be a need for 1 ground summing pes PER group, or would 1 ground summing bus for all groups be ok?

Also, how can I determine the right value of these resistors to ground?
In the schematic above they are 1K but I've seen schematics where they used 500 ohm or even 100 ohm.

As I understand it right, the concept is to make a "virtual ground" that floats above normal ground, that collects hum/noise, the same hum/noise the "audio bus" also collects so it will cancel out due to common mode rejection...

 

[Guest]

It is obvious that if you want to cancel capacitive coupled noises as well you'll need the same gain and impedance. Also, resistances of pan pots has to be considered.

 

[keefaz]

Also, if you have input transformers to put in front of JE990's,
you could use balanced busses with quasi-balanced channel outputs,
surelly a good way to get rid of noise (require input xfmrs though)

 

[radiance]

No input transformers, no faders and no pan. Just a bunch of That 1243 line receivers (see pic) connected to a summing bus (well, that might be 20 line receivers actualy )

This summing mixer I'm planning to build will be within a 4 unit rack case and all connections will be very short so whether there's a need for a ground summing bus or not..I don't know.

I might go for a more simple solution like this...

Thanks all...

 

[JohnRoberts]

[quote author="radiance"]I've a question regarding a ground summing bus as seen in this pic.


When making a summing mixer like in the picture only with multiple busses/groups, would there be a need for 1 ground summing pes PER group, or would 1 ground summing bus for all groups be ok?

Also, how can I determine the right value of these resistors to ground?
In the schematic above they are 1K but I've seen schematics where they used 100R and no resistor to ground in front of the non inverting input of the summing amp.

As I understand it right, the concept is to make a "virtual ground" that floats above normal ground, that collects hum/noise, the same hum/noise the "audio bus" also collects so it will cancel out due to common mode rejection...[/quote]

The more correct way to think of this is as a simple differential amp. The signal is being sent forward as the + signal - the ground voltage. Any voltage at the local ground relative to the local master section ground is also superimposed on top of the signal at that channel so by subtracting the channel ground from the channel signal, leaves just the pure channel signal but now referenced forward to the master section ground.

The value of the resistors is not as important as the ratio. It needs to be similar to the ratio of resistors summing the audio signals into the (-) input. Indeed there does need to be one resistor from the ground bus to the master section local ground to perfect the forward reference differential. In general using a low resistor value here reduces Johnson noise contribution which would be in series with the + input and amplified by N+1.

If all of the buses are similar (unity gain inverting), you can get away with sharing a single ground bus, but keep in mind summing intergrity depends upon all of the channel sends being relative to that single channel ground point that is being sent to the ground bus, and if you hang a number of summing amps off that one node the opamp's input noise currents and bias currents will combine. So in general lower is better for that nominal resistor value. It is pretty common to share one ground bus for both L & R bus combining amps.

Note: the quality of this differential is compromised if signal is fed from a pot (like an aux send), which makes the differential unbalanced as the source impedance is added to the summing resistor. Likewise driving a L/R differential bus from pan pots will suffer similar source impedance related errors.

JR

 

[JohnRoberts]

[quote author="radiance"]No input transformers, no faders and no pan. Just a bunch of That 1243 line receivers (see pic) connected to a summing bus (well, that might be 20 line receivers actualy )

This summing mixer I'm planning to build will be within a 4 unit rack case and all connections will be very short so whether there's a need for a ground summing bus or not..I don't know.

I might go for a more simple solution like this...

Thanks all...[/quote]

That input differential references the outside world to your channel local ground (pin 1 on line receiver). This is where ground bus is fed from.

If your physical console is small some of these larger considerations can be ignored. The stray bus capacitance to ground is common when using shielded cable on some bus feeds. In a small PCB the bus capacitance may be trivial.

In a large console structure where there is lots of bus capacitance one approach I used to stabilize is to mimic a common 2 pole active LPF. One pole is the 91pF feedback cap. The other pole is the stray C to ground. The main difference from your fig 6. is that the feedback 10k resistor connects directly to the bus to the left of the 39 ohm resistor. that said I see little reason to go to that much trouble for a physically small mixer.

JR

 

[radiance]

Thanks a lot John! I really appreciate this. My goal is to really understand it this time. I've been reading the "handbook for sound engineers" more specificly the part about console design by Steve Dove and so far it's a bit dazzling but some parts made sense to me. I'll have to read this a couple of more times because I know there's a lot more info in there than I grasped so far.

Anyway, having read you answers a couple of times thoroughly there still some things I don't get

[quote author="JohnRoberts"]
The value of the resistors is not as important as the ratio. It needs to be similar to the ratio of resistors summing the audio signals into the (-) input. [/quote]

What do you mean by ratio? Is this the relationship between the resistor value and the number of channels?

[quote author="JohnRoberts"]
That input differential references the outside world to your channel local ground (pin 1 on line receiver). This is where ground bus is fed from.
[/quote]

So, should pin 1 (ref) from the line receiver be connected to the red dot or to the green dot? (see pic below)

Thanks!

 

[JohnRoberts]

Green dot... I guess calling them all grounds is a little confusing because we generally think of there only being one ground.

Inside a console there are many local grounds and proper use of differential circuitry maintains signal integrity as it gets referenced forward from place to place.

To wit. An input differential references the distant channel 1 signal source ground to the local channel 1 input ground. While these two grounds have different voltages on them the differential circuit properly extracts a signal voltage at the channel that is relative to that channel ground. A differential bus structure now references the clean signal forward between local input grounds and the master section ground. Finally an output differential or balanced output maintains signal integrity between master section ground and wherever you send the signal.

Regarding ratio. In your schematic what is shown is 15k input resistors and a 22k feedback resistor. If those 15k resistor were driven from a low source impedance, the forward reference resistor on the ground bus would be 22/15 x 1 or roughly 1.5k. In reality if the 15k resistors are driven from a pan pot with varying source impedance that 1.5k value drops as effective 15k rises. I would guess a good values is something like 1.2K but this is a guess without studying front end, and at best that depends on nominal pot settings so will vary in use.

Unless this pan pot is buffered you will not get perfect differential action so it is tuned for average source impedance at that point.

I like to joke that consoles are the most complex simple circuit. Simple in concept but complicated in execution.

JR

 

[Nishmaster]

To wit. An input differential references the distant channel 1 signal source ground to the local channel 1 input ground. While these two grounds have different voltages on them the differential circuit properly extracts a signal voltage at the channel that is relative to that channel ground. A differential bus structure now references the clean signal forward between local input grounds and the master section ground. Finally an output differential or balanced output maintains signal integrity between master section ground and wherever you send the signal.

I'm not sure I quite understand this. Let me type it out as I think I understand it and someone can let me know how wildly off base I am.

The diff amp - input is always held at roughly a 0V potential by the action of the amp, which becomes a ground of sorts? So what is the Rgnd for then? I'm assuming it's to hold the + to different than 0V. What is the functional use of this local (to the diff amp) ground then? Why wouldn't we just tie both the diff amp's + to ground without Rgnd (making it a standard inverting opamp) as well as the channel's ground?

I'm sure I'm missing something basic.

Regarding ratio. In your schematic what is shown is 15k input resistors and a 22k feedback resistor. If those 15k resistor were driven from a low source impedance, the forward reference resistor on the ground bus would be 22/15 x 1 or roughly 1.5k.

Why is this the case?

-Matt

 

[JohnRoberts]

[quote author="Nishmaster"]

I'm not sure I quite understand this. Let me type it out as I think I understand it and someone can let me know how wildly off base I am.

The diff amp - input is always held at roughly a 0V potential by the action of the amp, which becomes a ground of sorts? So what is the Rgnd for then? I'm assuming it's to hold the + to different than 0V. What is the functional use of this local (to the diff amp) ground then? Why wouldn't we just tie both the diff amp's + to ground without Rgnd (making it a standard inverting opamp) as well as the channel's ground?

I'm sure I'm missing something basic.

[/quote]

No the diff amp is a simple computing or calculating circuit.

It compares the voltage sources connected to it's two input resistors, and subtracts the minus voltage from the plus. It then multiplies that difference voltage by the ratio of input to output resistors (in this case 1:1) then delivers that voltage to it's output relative to the voltage present at the output reference terminal. The benefit of this differential action is that it corrects for ground potential differences between source and destination, effectively ignoring ground differences.

Regarding ratio. In your schematic what is shown is 15k input resistors and a 22k feedback resistor. If those 15k resistor were driven from a low source impedance, the forward reference resistor on the ground bus would be 22/15 x 1 or roughly 1.5k.
-----------

Why is this the case?

-Matt

If the 15k resistors are driven from a 0 ohm impedance the differential ratio (or gain) is 22k/15k (approx 1.5x). For the individual channel ground contribution to cancel out it needs a similar ratio or 1.5k/1k. Since the real source impedance coming from the pan pots is a few K more, the ratio is not 22k/15k but more like 22k/17k (guess). Therefore the xK /1k ratio comes down in like fashion.

JR

 

[Nishmaster]

If the 15k resistors are driven from a 0 ohm impedance the differential ratio (or gain) is 22k/15k (approx 1.5x). For the individual channel ground contribution to cancel out it needs a similar ratio or 1.5k/1k. Since the real source impedance coming from the pan pots is a few K more, the ratio is not 22k/15k but more like 22k/17k (guess). Therefore the xK /1k ratio comes down in like fashion.

Ok, I see why I was slightly confused by this. The schematic actually indicates 15k input resistor and 15k feedback resistor. The 22k indications are for 48 channels, mirror by similar indications on the individual channel blocks. So the 1k ground sensing resistors make sense then, aside from any unbuffered pot issues.

A couple of questions again (sorry John). Why choose 1K for Rgnd and not 2K or 22K or whatever else (still maintaining the 1/1 ratio needed here, of course)? Why the 1/15 ratio of Rsum and Rgnd?

Also, what are the purposes of the series 49.9R at the output and then the 49.9K shunt resistor? It seems to me that the output impedance then of U1A is 49.9 ohms and the input impedance presented from the fader section is then 49.9k. I guess without knowing what's happening with the fader section it's hard to guess, but wouldn't 4.9k be a better choice? Or is the difference negligible?

 

[radiance]

[quote author="Nishmaster"]
Also, what are the purposes of the series 49.9R at the output and then the 49.9K shunt resistor? It seems to me that the output impedance then of U1A is 49.9 ohms and the input impedance presented from the fader section is then 49.9k. I guess without knowing what's happening with the fader section it's hard to guess, but wouldn't 4.9k be a better choice? Or is the difference negligible?[/quote]

Is that not just part of the servo circuit? Can be left out if you use decoupling caps I suppose....

 

[Guest]

[quote author="Nishmaster"]

A couple of questions again (sorry John). Why choose 1K for Rgnd and not 2K or 22K or whatever else (still maintaining the 1/1 ratio needed here, of course)? Why the 1/15 ratio of Rsum and Rgnd?

[/quote]

Though input currents of modern opamps are very small anyway different resistances on inputs will create some zero shift. Also, as I mentioned before equal resistors would null out results of capacitive couplings.

 

[Nishmaster]

No, the 1M resistor and the related opamp and bits connected to it are the servo.

 

[radiance]

[quote author="Nishmaster"]No, the 1M resistor and the related opamp and bits connected to it are the servo.[/quote]

Yes, you're right. I did not see that 49,9K resistor to ground due to my small screen :?
I think these values are chosen rather arbitrary then? Like you said: who knows whats going on at the fader...

 

[Nishmaster]

[quote author="Wavebourn"]
Though input currents of modern opamps are very small anyway different resistances on inputs will create some zero shift. Also, as I mentioned before equal resistors would null out results of capacitive couplings.[/quote]

I'm not sure you understood my question. I understand the need to make both Rgnd labeled resistors identical (1/1 ratio) in this case, but I'm wondering what the rationale is for the specific value (1k). Why not 100R/100R or 22k/22k?

 

[radiance]

[quote author="Nishmaster"][quote author="Wavebourn"]
Though input currents of modern opamps are very small anyway different resistances on inputs will create some zero shift. Also, as I mentioned before equal resistors would null out results of capacitive couplings.[/quote]

I'm not sure you understood my question. I understand the need to make both Rgnd labeled resistors identical (1/1 ratio) in this case, but I'm wondering what the rationale is for the specific value (1k). Why not 100R/100R or 22k/22k?[/quote]

In my sony desk those rgnd resistors are 100R ...(they call it ACN ground resistors).

 

[JohnRoberts]

[quote author="Nishmaster"]

Ok, I see why I was slightly confused by this. The schematic actually indicates 15k input resistor and 15k feedback resistor. The 22k indications are for 48 channels, mirror by similar indications on the individual channel blocks. So the 1k ground sensing resistors make sense then, aside from any unbuffered pot issues.

A couple of questions again (sorry John). Why choose 1K for Rgnd and not 2K or 22K or whatever else (still maintaining the 1/1 ratio needed here, of course)? Why the 1/15 ratio of Rsum and Rgnd? [/quote]

I answered this several posts ago... making the ground bus resistors smaller reduces their self noise (Johnson thermal noise and opamp noise current) contribution. There is not significance to ratio between + network and - network in this application.

Technically for the bus to be balanced (not just differential) the + and - bus impedances should be the same, but for inside the chassis console applications, lower noise trumps balance in this case.

Also, what are the purposes of the series 49.9R at the output and then the 49.9K shunt resistor? It seems to me that the output impedance then of U1A is 49.9 ohms and the input impedance presented from the fader section is then 49.9k. I guess without knowing what's happening with the fader section it's hard to guess, but wouldn't 4.9k be a better choice? Or is the difference negligible?

The 50 ohm in series with output is to provide a nominal resistive source impedance to outside world, and to decouple cable capacitance from opamp for feedback stability. The 50k to ground is only needed to provide DC discharge path when cap coupling is used.

JR

 

[Nishmaster]

I answered this several posts ago...

Gah, yes you did. My apologies.

I think I'm getting a handle on the concepts. Thanks again for the guidance JR. Maybe someyear I'll build something useful instead of scratching my head. Once I see what's happening, it's intuitive to me (which is encouraging), but getting there is sometimes a rather challenging trip.

 

[radiance]

[quote author="Nishmaster"].... Maybe someyear I'll build something useful instead of scratching my head..[/quote]

For me the only way to ever build something like this is to just start building and learn the theory along the way :shock:

Learning pcb design as well which is great fun...