Gareth, I see your name mentioned on multiple occasions in Douglas's book in the console section. An honor to have you aboard.
Gareth Connor said:
A couple of observations:
Lack of DC blocking caps - Input level pot wiper & CMOS switch chip.
Crosstalk on the CMOS switch - reduce it by running the change-over in a series-shunt configuration.
C7 does block output offset from U1. As an aside, that is one of the compromises of using the THAT1200, the output offset is spec'd at 10mV. I should have added them after IC1A and after U2.
No shunting at the switch is a typo, I had it there at one point but must have gotten deleted when I moved parts around. I'll adjust that for Rev 4.
When you switch audio in a virtual-earth configuration you need an inverting buffer - it's not an annoyance but a necessity.
In this case, for me, it's both, as now I may need an inverter someplace later. In the grand scheme of things, not so big a deal, really. Probably better to label it "mildly inconvenient compromise to ensure best performance."
Why would polarity-flippng at the inputs be cheesy?
It's not really, and present a rather simple solution to the issue. I just have a programmer's aversion to workarounds, don't mind me.
Is there a particular reason that the "trim" control is a full-range pot (attenuates to -infinity) and not a trim, i.e. +/-10dB (or similar) around nominal operating level?
It seemed to be the simplest configuration at that location parts and noise-wise. There's no guarantee that what comes in there is actually at +4, especially if someone decides to patch a hot compressor inline before the input instead of at the insert point. The one drawback I see is that I'm amplifying the noise of the pot right off the bat and the noise doesn't decrease as signal level is adjusted to be lower. Something to think about, I guess.
Prototype / breadboard the circuit to prove that it will meet expectations of noise, distortion, crosstalk, DC-isolation, switch-click, pot-wiper noise, etc. What is presented is a good starting-point, but there is a lot more to do before it is ready to make it into a a console.
Duly noted. I am breadboarding specific bits of the circuit as well as simming as I go, but that only takes me so far. The problem with breadboarding I find is that especially noise performance is not indicative of on-PCB performance. Plus, you really need to stuff it into a box to keep the interference out. That is not to say it isn't useful, though.
Henke said:
You could make IC1A an inverting stage and wire the "Input trim" as a rheostat (+end resistor) to vary the gain instead. That means you can move the CMOS-switch to the IC1A input and get rid of IC1B.
That is a possibility, although that limits me to a unity gain configuration at the lowest. It would be nice to have some attenuation. Perhaps the real answer in this case is to move the trim and gain block to after the switch buffer, thus allowing trim of the bus input also. The problem I have with that solution, though, is that it encourages poor gain staging. If the bus is too hot it makes more sense to pull your faders down instead of reducing the bus input trim to the channel.
JohnRoberts said:
Instead of one NF resistor, imagine two. One connected to R10 and one to R11 at the input side of the switch. This way the switch is inside the feedback loop so only driving the input of the opamp (easy lifting).
Maybe I'm not visualizing that correctly (although I'm pretty sure I am), but doesn't that still place the TG at the same circuit node? The only thing I see that doing differently than the way I have it drawn is that during the switch transition the feedback pairs will be both out of the circuit and not yet shunted, thus causing a brief open loop gain, right during the time where charge injection shows up: to me, that says big click instead of little click, even with a feedback cap.
The 1k feedback R is low by itself, two in parallel is much too low for comfort.
Good catch. 4K7 like the rest will work fine here.
Of course you must bread board this and tweak it for silent running, perhaps DC blocking on output side of switch and on input resistors. Since 5532 are not very low offset parts.
I can't imagine needing blocking on both sides, but we'll see when I test it. 5532 is not very low offset indeed. I'm trading offset for cash, but if I need a blocking cap after every 5532 the economics may not be sound.
Speaking of blocking caps in the audio path, I'm aiming for a -3dB point of 10Hz, which makes these things rather spendy as film caps go. I know everybody says electrolytics are the devil, but how much are they the devil? A nice design goal would be the oft quoted 2hz-200Khz for good phase from 20hz-20Khz, but that really needs caps that get into tube amp blocking cap type expenditure if I stay with film. My budget is $25k for channels, which sounds like a lot until you realize that just doing the faceplates, PCBs, and quality faders alone is probably 1/3-1/2 of that budget right there for 32-48 channels.
Brian Roth said:
ANYWAY...bottom line on the design. A 4053 "triple SPDT" combined with a summing node (and inside of the feedback loop) of an opamp makes a switch node that spec'ed well in what haas been a major signal path at the studio. In that configuration, noise and THD were essentially the same as the opamp according to my ancient Amber test equipment, and no one has complained.
I haven't run the switch setup through the prototype test yet, but Douglas Self's numbers for a similar switch in the configuration I have drawn look to be basically negligible, and his part was at least an order of magnitude worse Ron.
...you can see the little Panasonic relays on an Amek 9098 module. They are the DIP-10 items scattered around the circuit card.
Those are great shots, thanks! Yeah, looks like those little guys are all over the place.
-Matt
