I try to avoid console design questions since I no longer do it...mediatechnology wrote:You show all the 10k input mixing resistors grounded when that input is deselected. This means the mixing stage IC1a is operating at maximum noise gain all the time. If they are left hanging when the input is deselected, there's less noise if only a few inputs are mixed.
ricardo is right about that. I saw it originally but was focused on the 1646.
What you might consider is putting the 10K on the left side of the switch with the switch "pointing" to the right.
Have the commutator on the left, connected to the 10K Rmix, and the switch stators on the right-hand side.
One stator is connected to the virtual ground mix bus for On; the other grounded for Off.
When de-selected from the bus, the right-hand side of the 10K is grounded.
That configuration lowers crosstalk from open switch capacitance and not increase noise gain unnecessarily when a large number of inputs are de-selected.
You might eliminate one half the OPA2134 and use it on the other channel. As pointed out you can re-define the 1646 outputs to obtain polarity inversion.
JR and ricardo are the console designers so ultimately I defer to them.
emrr is correct about keeping a low impedance drive to the THAT1646.
Fortunately the PGA2320 is buffered and the output stages appear to have quite a bit of drive current capability with a 75 mA Isc.
yes back-grounding de-selected inputs keeps a constant noise gain but may be preferable for keeping the differential math in balance (to cancel ground noise errors). The operative question is 9X the ein of modern summing op amp worse than one mic preamp at nominal gain (rhetorical no)?