3 way rotary switch for LCR pan - what to get?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
You are right, it is not clear from the block diagram but I think the CAPI product is a virtual earth mixer. The block diagram just does not show the 47K feedback resistor around each 2520. In the support thread it says all the amps are arranged as ACAs which is APIs obscure term for virtual earth (Active Combining Amp).

Cheers

Ian
 
leigh said:
I can't fully tell from the block diagram there, but I would infer that their summing scheme is a passive sum + boost. (So, a non-inverting opamp stage, and not a virtual earth summing scheme.) Can someone who knows more about API stuff chime in?

If that is the case, then Ian's LCR switching scheme earlier in the thread would be better than the one I posted, since his won't have crosstalk between the L and R busses. If a channel is switched to Left bus only in his scheme, then its Right bus summing resistor gets connected to ground.

In my scheme, a signal switched to the Left bus can hop across those 2 middle resistors right into the Right bus - unless it's attached to a virtual ground summing amp.


Ok, this should help. I dug deeper into CAPI's site and found this:

http://www.capi-gear.com/catalog/images/gallery/PCBs/2-ACA-Bo/CAPI-2-ACA-Bo-Rev-B.1-schematic.pdf

 
Good find. Well,  that is an inverting stage, and rather than follow it with a second amp to flip the polarity back, they are simply doing that with the output transformer.

However, I am surprised to see that R12 is such a high value (100 K).  For a regular op amp, you would want to balance the current into the two inputs, by feeding them with the same source impedance. The impedance presented to the "L ACA Input" will be the resistance of all the summing resistors in parallel.  If you are following the summing scheme that they lay out in the block diagram, that means four stages summing, with 47K summing resistor each. Those four in parallel make about 12 K - pretty far from 100k.

So, does it behave as a virtual earth summing amp still? I don't know offhand, I'd have to parse that out.

I would be sure to check the math if you are not following the API scheme exactly (i.e. summing only 4 channels into one of those amps). You might want to lower the value of R12 if you are summing 16 channels into that stage. But that's just a guess.
 
leigh said:
Good find. Well,  that is an inverting stage, and rather than follow it with a second amp to flip the polarity back, they are simply doing that with the output transformer.
It is followed by a transformer which does the polarity flip. Unusual I agree.
However, I am surprised to see that R12 is such a high value (100 K).  For a regular op amp, you would want to balance the current into the two inputs, by feeding them with the same source impedance. The impedance presented to the "L ACA Input" will be the resistance of all the summing resistors in parallel.  If you are following the summing scheme that they lay out in the block diagram, that means four stages summing, with 47K summing resistor each. Those four in parallel make about 12 K - pretty far from 100k.
The summing resistors are dc isolated from the -ve input so they do not contribute to bias current. The -ve input bias is determined by the 28K feedback resistor. To equalise bias current R12 should also be 28K. Howvwer, as the output is ac couple to the transformer, a few millivolts of offset is not a real problem.
So, does it behave as a virtual earth summing amp still? I don't know offhand, I'd have to parse that out.

I would be sure to check the math if you are not following the API scheme exactly (i.e. summing only 4 channels into one of those amps). You might want to lower the value of R12 if you are summing 16 channels into that stage. But that's just a guess.

It does behave as a summing amp. The gain is less than unity (28K/47K) because API typically had gain in the output transformer to increase the maximum output level.

Cheers

Ian
 
Thanks for having a look at this, Ian. What you say mostly makes sense to me, except for this part:

ruffrecords said:
The summing resistors are dc isolated from the -ve input so they do not contribute to bias current. The -ve input bias is determined by the 28K feedback resistor. To equalise bias current R12 should also be 28K. However, as the output is ac couple to the transformer, a few millivolts of offset is not a real problem.

I hadn't been thinking about the DC bias current – to be honest, I have only a faint idea of how that applies here. I was thinking more in terms of the (AC) source impedance of the summing node, which is why I was talking about the impedance of all the summing resistors in parallel.

In API's scheme (according to the block diagram posted previously), they are summing 4 channels, each with a 47k summing resistor, into an ACA stage. So the summing node's source impedance is 47k / 4 = 11,750.

(Compared with what I'm used to working on, that's a high source impedance for a summing bus. The Trident S65 that I've spent the most time with under the hood has 32 channels of 12k summing resistors, giving a 375Ω bus source impedance. I understand that ideas about what makes the "best" summing circuits have changed over the years, so I'm not saying one is right and one is wrong, just pointing out the difference.)

Anyways - why doesn't the ACA summing amp have its non-inverting input connected directly to ground? You're saying "to equalise bias current R12 should also be 28K", but instead it's 100k. Yet 28k is closer to zero ohms than it is to 100k... so why didn't they just go straight to ground there?
 
leigh said:
Thanks for having a look at this, Ian. What you say mostly makes sense to me, except for this part:

I hadn't been thinking about the DC bias current – to be honest, I have only a faint idea of how that applies here. I was thinking more in terms of the (AC) source impedance of the summing node, which is why I was talking about the impedance of all the summing resistors in parallel.

In API's scheme (according to the block diagram posted previously), they are summing 4 channels, each with a 47k summing resistor, into an ACA stage. So the summing node's source impedance is 47k / 4 = 11,750.

(Compared with what I'm used to working on, that's a high source impedance for a summing bus. The Trident S65 that I've spent the most time with under the hood has 32 channels of 12k summing resistors, giving a 375Ω bus source impedance. I understand that ideas about what makes the "best" summing circuits have changed over the years, so I'm not saying one is right and one is wrong, just pointing out the difference.)
As the actress said to the bishop, size doesn't matter for a virtual earth summing bus source impedance. It does matter in a passive mixing system where you want to use a mic pre for gain make up. In a passive mix bus internal to a mixer it does not matter too much either provided the final bus impedance is a suitable source for the gain make up amp. The only difference it makes in either system is in noise performance and as a rule, the contribution of bus impedance to overall noise is insignificant. The noise in an 11K resistor is about -110dBu.
Anyways - why doesn't the ACA summing amp have its non-inverting input connected directly to ground? You're saying "to equalise bias current R12 should also be 28K", but instead it's 100k. Yet 28k is closer to zero ohms than it is to 100k... so why didn't they just go straight to ground there?
If you look at the schematic  you will see the +ve input is brought out to the connector and IIRC it is labelled 'shield'. I think this make be an overall bus shield used to reduce noise pickp on the bus itself in the mixer. I think that is why it is not connected directly to 0V. I don't know anything about the internal bus details of API mixers so it's just an educated guess.

Cheers

Ian
 
Back
Top