Summing Bus coupling cap in VE configuration

I'm reading Doug Self's Small Signal Audio book, and in his discussion of ACN / VE summing, he suggests the coupling capacitor (feeding the inverting opamp input) be sized for a high pass corner frequency using the bus source impedance.

It seems like everything I've read indicates that you typically use in *input* impedance of the following stage when sizing coupling capacitors.

Is there some kind of voodoo magic where the input impedance of the ACN / VE amp (inverting input + feedback network) mirrors the bus source impedance?

Normally would consider an opamp input to be very high impedance, but in this case, it's treated as very low.

Can anyone help set me straight?

Thanks,

Bob

It says in the book it's because the impedance is all of the mix resistors in parallel. So 8 22k is 2.75k.

But that cap is only necessary if you use an amp that has a significant offset current. You could get a fancier amp with low offset and use smaller mix resistors and that would probably be good enough to keep the routing switches from being "clicky" without the cap.

Makes sense. I understood that the bus source impedance were all of the mix resistors in parallel (with the feedback network).

What I didn't get was that the inverting input node "mirrors" this impedance as it's input impedance due to the VE topology. And that is what informs the RC high pass of the coupling cap. I agree with the right opamp can probably can the cap entirely.

Thanks for the response.

squarewave said:

It says in the book it's because the impedance is all of the mix resistors in parallel. So 8 22k is 2.75k.

But that cap is only necessary if you use an amp that has a significant offset current. You could get a fancier amp with low offset and use smaller mix resistors and that would probably be good enough to keep the routing switches from being "clicky" without the cap.

Click to expand...

In a VE amp the non-inverting input will be low impedance.  Each mix channel will see (roughly) it's respective mix resistor as a load. For example a 16 channel using 10k on each channel would have a bus impedance of 10k/16. But each channel would see (roughly) 10k so you would size an AC coupling cap based on 10k.

Yep, that makes perfect sense if coupling at the channel output into the mix resistors. Doug's reference in the book was in regards to coupling at the bus to the inverting opamp input. I didn't realize that junction presents as a low input impedance.

john12ax7 said:

In a VE amp the non-inverting input will be low impedance.  Each mix channel will see (roughly) it's respective mix resistor as a load. For example a 16 channel using 10k on each channel would have a bus impedance of 10k/16. But each channel would see (roughly) 10k so you would size an AC coupling cap based on 10k.

Click to expand...

The op amp is going to source / sink an equal and opposite amount of current as necessary to make the - input equal to the + input which is grounded. That's what op amps do. So it's the op amp output that is driving the bus to 0V through the feedback resistor which makes it low impedance at AC. All of those currents get summed and the result is an op amp output voltage that is directly proportional to that current.

Unfortunately a side effect of this is that if there is any noise or distortion on the bus, it gets amplified by the open loop gain of the op amp. So the bus needs to be shielded well and preferably thicker wire and something like a capacitor at the summing node is very sensitive to distortion. Fortunately the distortion of capacitors is a function of the voltage across them and that is very low at the summing node. And electrolytics these days are pretty good. But it doesn't cost much extra to get a bipolor cap there. Especially since it can be very low voltage. So a 6.3V low-esr (aka "audio grade") 220u or 470u bipolar electrolytic is good there.

Or skip it entirely as described previously. In fact looking at modern mixer designs, I don't recall seeing a cap there. The impdeance isn't high enough to cause a large offset so I'm not sure what Self is worried about there.

Is Self suggesting to use one big cap after the summing resistors?  As opposed to multiple smaller ones before each summing resistor?

Not a good idea imo.

john12ax7 said:

Is Self suggesting to use one big cap after the summing resistors?  As opposed to multiple smaller ones before each summing resistor?

Not a good idea imo.

Click to expand...

Usually, the summing resistors are DC-free; if they weren't the pan-pot and routing switches would be noisy/clicky. So the question is why a capacitor between bus and summing amp.
DC offset can become a nuisance, with clicks resulting from switching channels in.
It may not be a serious issue, because no one in their right man should engage a channel to an active track. But customers expect all switches to be noise free with no signal. They're right because they're the ones who pay.
And the summing amp may not be referenced to 0v. Many summing amps use a bipolar transistor in conjunction with an  opamp, and often the input is one Vbe off 0v.
I don't know if it's a good idea or not, but it may be necessary.

abbey road d enfer said:

Usually, the summing resistors are DC-free; if they weren't the pan-pot and routing switches would be noisy/clicky. So the question is why a capacitor between bus and summing amp.
DC offset can become a nuisance, with clicks resulting from switching channels in.
It may not be a serious issue, because no one in their right man should engage a channel to an active track. But customers expect all switches to be noise free with no signal. They're right because they're the ones who pay.
And the summing amp may not be referenced to 0v. Many summing amps use a bipolar transistor in conjunction with an  opamp, and often the input is one Vbe off 0v.
I don't know if it's a good idea or not, but it may be necessary.

Click to expand...

+1 the correct calculation is the effective parallel impedance of all the bus send resistors in parallel. This breaks one of my unwritten rules against using electrolytic capacitors in lower impedance nodes in audio paths. While modern caps are much improved, back in the 70s in bench tests I detected significant inductance in common electrolytic caps (enough to be responsible for measurable phase shift at 20kHz working into a few hundred ohms. I make no claims about the audibility of that phase shift, but any errors you can avoid are worth avoiding IMO).

Clicks while assigning or de-assigning bus sends should have zero impact on mix quality, but customers hear the clicks and ASSume an inferior audio path, even if de-clicking that path with extra caps makes it less linear. 

JR

abbey road d enfer said:

Usually, the summing resistors are DC-free; if they weren't the pan-pot and routing switches would be noisy/clicky. So the question is why a capacitor between bus and summing amp.
DC offset can become a nuisance, with clicks resulting from switching channels in.
It may not be a serious issue, because no one in their right man should engage a channel to an active track.

And the summing amp may not be referenced to 0v. Many summing amps use a bipolar transistor in conjunction with an  opamp, and often the input is one Vbe off 0v.
I don't know if it's a good idea or not, but it may be necessary.

Click to expand...

attempted to eliminate the input coupling capacitor on TSM mix bus.
it didn't work nor does it make any sense as there is no DC being switched.
but  I've seen it fitted in other consoles.
you console gurus must have an idea.

opamp bias current

gridcurrent said:

attempted to eliminate the input coupling capacitor on TSM mix bus.
it didn't work

Click to expand...

It should have, unless one or several inputs have DC on them.

nor does it make any sense as there is no DC being switched.

Click to expand...

As Gyraf mentioned, there's DC offset on the opamp's input, which makes switching noisy.

but  I've seen it fitted in other consoles.

Click to expand...

Because it makes sense. You don't want to have the whole mixer unusable in the case one channel sends DC.