Question about OPA power reference in mixer designs

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

boji

Well-known member
Joined
Jan 6, 2010
Messages
2,375
Location
Maryland, USA
Ian:  "At Neve, we used a brass bus bar for B- (=AGND) which was tapped at regular intervals along its length. "
Some formats give OPA's their own 0v rail as distinct from audio reference (Pin #3 / Pin #5 in 500 series).

Other designs have a single reference for OPA and audio, sometimes even when there's led's and mute fets sharing the same power source.

Further still,  some discrete OPA's don't have an 0v pin or I've read they aren't always needed. 

Are these single / dual  references arbitrary, or Is it that some designers put more trust in decoupling/filtering caps than others?

I ask because in my power distribution, I've added a seperate bus bar for OPA power reference as distinct from signal ref (tied together much later), and would like to know if there's any advantage to doing it this way.  Either method would have decoupling and smoothing caps everywhere.

Thanks in advance for your opinions/comments,
-Boji
 
boji said:
I've added a seperate bus bar for OPA power reference as distinct from signal ref (tied together much later), and would like to know if there's any advantage to doing it this way.  Either method would have decoupling and smoothing caps everywhere.
Using separate bus for 0V and audio gnd is supposed to "isolate" audio gnd from interference from decoupling currents, but looking at the bigger picture it does not make much sense, because for better separation, currents have to return exactly where they originate from. Returning current via 4 meters of bus is not my idea of a short loop.
In particular I could compare the performance of Soundcraft mixers that were very similar in design and functionalities, but one sported the separate bus scheme. It turned out that the one that had the separate bus was not any better; actually longitudinal noise was less controlled on the latter. The way insert and direct outputs were connected introduced more noise variation than in the simpler one. I decided to use both bus in parallels and that gave an overall better performance.
However, there is a distinct advantage in having a separate reference bus for the summing amps, as it samples the grounds of each module. Even better, a separate reference for each bus. Ultimately that's differential summing ("balanced bus").
In the past, very often input grounds had a separate bus, making sure different connections would not create too much variation of longitudinal noise; today, adherence to the AES recommandation of tying Pin 1 to chassis takes care of that.
 
ruffrecords said:
I cannot find anywhere a definition of longitudinal noise (except in the context of a moving vortex lattice).

Cheers

Ian
Longitudinal noise (translation of the french term) is noise created by circulation of currents along the resistance (impedance) of the ground path. Main contributions are PSU ripple dumped into "ground" by decoupling caps, individual channel audio currents and currents resulting from gear with different "earths".
 
boji said:
Some formats give OPA's their own 0v rail as distinct from audio reference (Pin #3 / Pin #5 in 500 series).

Other designs have a single reference for OPA and audio, sometimes even when there's led's and mute fets sharing the same power source.

Further still,  some discrete OPA's don't have an 0v pin or I've read they aren't always needed. 

Are these single / dual  references arbitrary, or Is it that some designers put more trust in decoupling/filtering caps than others?

I ask because in my power distribution, I've added a seperate bus bar for OPA power reference as distinct from signal ref (tied together much later), and would like to know if there's any advantage to doing it this way.  Either method would have decoupling and smoothing caps everywhere.

Thanks in advance for your opinions/comments,
-Boji
I don't recognize what am "OPA" is.

I tried using a brute force "bus bar" ground and or power distro exactly one time in my very first console but quickly realized it was not a productive approach. (In fact I was finishing a design somebody else started.)

I may sound like a broken record but maintaining signal integrity inside consoles is best handles by treating audio signals differentially (i.e. signal hot relative to signal cold).

For a modest number of channels (small mixers) you can brute force busing, but for larger mixing platforms differential busing can preserve signal integrity.

JR

PS: every active stage inside a console is an opportunity to treat the audio signal differentially forward referencing, or back referencing between different local audio ground references (AKA 0V).
 
Many thanks! It belongs in theTablets of Stone, with Ohm and Kirchoff's laws.
Abby: Also many thanks to you. Felt lucky last night that AES left part one free to download, as I went on your initiated 'deep dive' at about 4am. 

Wish I was better at math.  I shall experiment with and then without the opa ref-- 'doubling up' as you say, to extricate the 'gremlins'.  Also provide the aca's with their own returns + a balanced test,  to listen/look for differences in noise floor.  Exceedingly helpful and informative. <3

If someone had not invested in such unnecessarily large bus bars,  that someone would have found room for the extra aca reference lines on the iso block.  ::)

Moamps: thanks for the link to part II!! 

Every active stage inside a console is an opportunity to treat the audio signal differentially forward referencing, or back referencing
JR:  I think Mr. Self has a chapter on this in his small sig handbook. I will review that today, thanks for the reminder. 
Edit: Surely we wouldn't want two dozen psu send and return lines, each going to each channel/group/aux card? Hmmm.

Very late Edit:  Wish I could see a balanced/differential power application as it relates to mid/large format console at the level of the layout/physical to compare it with a 'brute force method'.  Something tells me there'd still be need for lots of low resistance copper bus or super-wide (inches not mils) trace runs.

 
boji said:
If someone had not invested in such unnecessarily large bus bars,
You'll never regret!

Surely we wouldn't want two dozen psu send and return lines, each going to each channel/group/aux card?
If they were all floating vs. each other, that would be a great thing. In some sensitive equipment, each module has its own isolated DC/DC converter, so they have the benefit of distributed power and those of independant sources. The cost may be prohibitive for a mixer, but I think the main reason it was never (?) done is because mixer designers hate radio transmitters.  ;)
 
Main iso with initial choices of distribution, you can see it won't be difficult to tie the top two together like you did Abby:
5efqhgj.jpg


After mentioning the improvements to be had making aca's independent of main iso power/refs, thinking of adding yet another iso block under the existing ones, but down on the end, under patch panel section where aux, pfl/afl/group aca's will congregate.
 
boji said:
Main iso with initial choices of distribution, you can see it won't be difficult to tie the top two together like you did Abby:
5efqhgj.jpg


After mentioning the improvements to be had making aca's independent of main iso power/refs, thinking of adding yet another iso block under the existing ones, but down on the end, under patch panel section where aux, pfl/afl/group aca's will congregate.
Since you have two bus available, use one for power distro and  the other to bus the module ground at the reference point of the channel's output stage.
 
JohnRoberts said:
I may sound like a broken record but maintaining signal integrity inside consoles is best handles by treating audio signals differentially (i.e. signal hot relative to signal cold).

For a modest number of channels (small mixers) you can brute force busing, but for larger mixing platforms differential busing can preserve signal integrity.

JR
I am getting confused by (apparently) different uses of the term bus. In many cases a  bus is a signal mixing bus and I understand your point about differential buses in that context. 

But in additions to that there are still a whole bunch of AGNDS that need returning to the power supply and a bus  is still necessary for this. Longitudinal noise can still be created in this bus so does it not make sense to minimise it by making the bus as low resistance as possible?

Furthermore, in mixers where differential mixing buses are impractical, does not a big fat AGND bus bar make sense?

Cheers

Ian
 
boji said:
JR:  I think Mr. Self has a chapter on this in his small sig handbook. I will review that today, thanks for the reminder. 
Not a new concept but IMO a good one... (the concept, I have no idea about his book).
Very late Edit:  Wish I could see a balanced/differential power application as it relates to mid/large format console at the level of the layout/physical to compare it with a 'brute force method'.  Something tells me there'd still be need for lots of low resistance copper bus or super-wide (inches not mils) trace runs.
That's why we put consoles inside metal chassis....  The metal box can serve as a low Z ground for shields and whatever (no copper bars involved).

JR
 
ruffrecords said:
I am getting confused by (apparently) different uses of the term bus. In many cases a  bus is a signal mixing bus and I understand your point about differential buses in that context. 
Omnibus=> To all. the term is perfectly adequate for power distribution rails.

But in additions to that there are still a whole bunch of AGNDS that need returning to the power supply and a bus  is still necessary for this. Longitudinal noise can still be created in this bus so does it not make sense to minimise it by making the bus as low resistance as possible?
If the channels were used separately, it would not matter. The problem arises from mixing the signals they are outputting, the way we usually mix. We do it in a way that supposes the references of all signals are the same, an equipotential, which it is not.

  does not a big fat AGND bus bar make sense?
That is indeed the common cure for trying to make the reference an equipotential. It has its limits, since the level of magnetic interference is proportional to the length of the bus, whatever their resistance.
 
abbey road d enfer said:
If the channels were used separately, it would not matter. The problem arises from mixing the signals they are outputting, the way we usually mix. We do it in a way that supposes the references of all signals are the same, an equipotential, which it is not.
But surely even if the channels are separate it does matter because longitudinal noise due to signal currents will appear as crosstalk
That is indeed the common cure for trying to make the reference an equipotential.
My point was that It may be common but where differential mixing is not an option it has to be worth doing.
It has its limits, since the level of magnetic interference is proportional to the length of the bus, whatever their resistance.
That is a new one on me. Why isn't it proportional to the loop area?

Cheers

Ian
 
abbey road d enfer said:
Omnibus=> To all. the term is perfectly adequate for power distribution rails.
If the channels were used separately, it would not matter. The problem arises from mixing the signals they are outputting, the way we usually mix. We do it in a way that supposes the references of all signals are the same, an equipotential, which it is not.
Exactly... calling something ground ASSumes it is at 0V potential.... Calling multiple nodes ground ASSumes they are all at the same exact voltage. I am surely repeating myself but we can use differential circuits to forward and back reference between these sundry ground potentials with their slightly different voltages. Of course they still need to be connected together, but now slight voltage differences don't affect audio integrity, obviating the benefit of brute force. (Abbey I know you know this, I just like repeating myself apparently.  ::) )
That is indeed the common cure for trying to make the reference an equipotential. It has its limits, since the level of magnetic interference is proportional to the length of the bus, whatever their resistance.
And current flows in ground paths also cause simple IxR voltage drops.

JR
 
Since you have two bus available, use one for power distro and  the other to bus the module ground at the reference point of the channel's output stage.
Would you mind being a bit more specific what point you had in mind at the output stage Abby? 
On each channel/group card, Interstage and output stage audio reference share a large pour (AGND) that ends its journey at three, (triple wide) blade connectors meant to plug into backplane.  This reminded me of an earlier post where I was trying to decide where to pair AGND to channels, and I think Monte McGuire (referencing below picture) thought version 'A' was better if he was forced to choose, saying:

"Lots of copper will get you low resistance, but low impedance results from 'electrically short' conductors. Inductance is proportional to the length of a conductor divided by its width, so a piece of foil 1" or 2" wide might work better between console modules than a piece of heavy rod stock, since it will have lower AC impedance because of its lower inductance.
So, to answer your question, 'A' seems better somehow, but if you could use the metal front panels of the modules, then those would be better conductors"
5Zt76iu.jpg



So, If I understand what you had in mind,  It would be to NOT use the backplane AGND bus or piece of foil as monte suggested, but to tie in with some wire directly from the big bus bar to each individual edge connector's AGND output, these being the 'reference points' of each channel?  This would include severing the AGND trace shared among channels on each backplane too, I believe.
Thanks for expanding on your suggestion, if only to avoid a mistake on implementation.
-Boji


 
Back
Top