opinions on bus switching

[juniorhifikit]

This may have been discussed/debated already, but I was wondering what the general consensus was on bus switching. Is it better to back-ground the summing resistor when a channel is switched off, so they're always representing the same load to the summing amp (consistent noise gain); or remove them from the summing node when the channels are switched off, thus improving noise gain at lower channel counts? I've seen it done both ways on different mixers.

From a consumer industry point of view, I could see where bragging rights of lower noise gain might be beneficial in the spec's, but is there a downside to the ACN's gain changing with channel count? JR, I'd imagine you probably have a preference from your console design days...

If I'm not making sense, here's something to illustrate what I'm talking about (summing amp would obviously have matching Rfb of 2K2 Ohms):

[mediatechnology]

JR's the expert but my vote would be to back-ground.

[JR.]

It depends....

How many channels?

how distant are the sources being summed?

From a quick glance you are asking the wrong question... how are you referencing forward the local to master section bus ground or audio low?
-----
The short answer is that back grounding means you get the worst case loop gain margin all the time, but if you tweak the sum amp so it is only stable at the higher closed loop gain (like decompensated 5534 is only stable above 10dB closed loop), you have to back ground, to be stable.

See isn't this fun... ?

JR

[juniorhifikit]

Mmmmm fun. Fortunately there are only 16 channels in a 1U chassis (5534 output) feeding Wayne's super cool 1570 + servo stereo summing amp. I've tested it stable from around 60 Ohms (32 channels) to 2k2 Ohms (1 channel). All analog grounds are tied to the ground plane on the main pcb (aprox 8" x 10", so no coat hanger wire necessary). Currently all summing resistors are back-grounded and it works great, I was just wondering if it would be advantageous to go the other way in order to be able to boast a better noise spec on paper. Playing the product spec' game is annoying, especially when necessary data is conveniently omitted (so the spec means nothing) and yet people base their purchase on these specs How long is a piece of string?

[JR.]

Mmmmm fun. Fortunately there are only 16 channels in a 1U chassis (5534 output) feeding Wayne's super cool 1570 + servo stereo summing amp. I've tested it stable from around 60 Ohms (32 channels) to 2k2 Ohms (1 channel). All analog grounds are tied to the ground plane on the main pcb (aprox 8" x 10", so no coat hanger wire necessary). Currently all summing resistors are back-grounded and it works great, I was just wondering if it would be advantageous to go the other way in order to be able to boast a better noise spec on paper. Playing the product spec' game is annoying, especially when necessary data is conveniently omitted (so the spec means nothing) and yet people base their purchase on these specs How long is a piece of string?

OK, without bringing myself up to speed on the inner workings of the 1570, IIRC that is a mic pre IC so low noise and capable of high gain. There was a popular bus topology back in the day using a transamp (a dedicated mic preamp module) in a similar fashion.

So you probably won't see a huge loop gain margin benefit from not back grounding (lower phase shift, lower distortion). Sum bus noise, is a widely over emphasized specification. Pretty much 99.99% of real sources that you will run, even one of, through that bus will have a higher noise floor. It doesn't matter if the bus noise is -95dBu or -115 dBu, if the mic pre is -72 dBu. As long as the sum amp is enough dB lower it truly is unimportant, but as an old console seller I recall the WFO test, where customers expect silence there (and everywhere).

I would suggest doing some simple measurements with 16x and 1 source connected and measure what the difference is (mostly phase shift, etc.).

I am a big advocate of differential bus topology. Back grounding the send, means you can leave the ground side bus always connected and the differential math remains intact. If you switch the send to open circuit, you need to lift the ground send too, so the sum bus math will track properly for the channels still attached.

There may be subtle differences with channel kill when de-assigned, crosstalk, etc, but this is again not as big a deal in recording as for broadcast where you may have multiple buses and multiple disparate stems running around inside the box.

JR