Simple open-source condenser studio microphone with a modern twist

[RFDx]

Very nice build. Thanks for sharing.

Is there a specific reason for the bias resistor R13 to be only 100 MOhm? R13 sets the lower cutoff frequency. Why do you need a 1GOhm bias resistor in front of the noninverting op-amp? Btw. the ground in the +/-6V supply is "artificial". There's nothing "virtual" about it .

[Spirit532]

Is there a specific reason for the bias resistor R13 to be only 100 MOhm?

Nope. Just off the top of my head.

Why do you need a 1GOhm bias resistor in front of the noninverting op-amp?

Avoids DC charging of the input. Given the extremely high(many GOhm, and not particularly well characterized) input impedance of the opamp, it can cause a serious charge to build up on the input and just rail it. I've tried it without the resistor, and found that this happens.

Btw. the ground in the +/-6V supply is "artificial". There's nothing "virtual" about it .

https://en.wikipedia.org/wiki/Virtual_ground

[Marco]

The analog ground connection should ideally be in the middle of the bias string, the extra parasitic capacitance imbalances the input.

[Spirit532]

The analog ground connection should ideally be in the middle of the bias string, the extra parasitic capacitance imbalances the input.

Can you illustrate what you mean by this?

[Marco]

GNDA has 47uf capacitance to GNDREF, so for AC that's what it is. Which means the bottom opamp is doing nothing, it's single ended in disguise.

[Spirit532]

It's not doing nothing, it does provide a differential signal(that I've measured), but there is an imbalance due to the way they're linked. Solving this would require adding a buffer, which would likely blow the power budget. It's already stretched pretty thin with the high impedance takeoff.

[Marco]

Voltage wise it's creating a differential signal on the XLR connector, but it's creating it from an imbalanced connection to the capsule. A voltage differential signal an sich has no real value at mic level, it's not like the mic level inputs will saturate, it's cargo cult differential. If you made the top opamp a simple unity gain buffer, removed the bottom opamp and just connected the bottom coupling capacitor directly to GNDA you would have superior noise.

If GNDA was connected in the middle of the string of bias resistors instead of the bottom, the connection to the capsule would be balanced and the differential amplification would have some justification.

[Spirit532]

It's a differential signal going down the cable for common mode immunity, there's absolutely no point in doing differential pickup from the capsule.
All audio interfaces expect a differential input by default. That's what the second opamp is for. Yes, you can ignore that and take it single ended from ground.

[Marco]

It's not the differential signal an sich which gives you noise immmunity, it's the matched impedance and a differential input on the receiving end. You can have matched impedance on a single ended output just fine.

That's why this type of differential amplification is cargo cult, it looks at completely differential signal chains and confuses an unimportant artifact (the signal being differential to a reference voltage) and confuses it for being essential. It ain't, all you are achieving by recreating the artifact is increasing amplifier noise.

[Spirit532]

It's not the differential signal an sich which gives you noise immmunity, it's the matched impedance and a differential input on the receiving end. You can have matched impedance on a single ended output just fine.

That's why this type of differential amplification is cargo cult, it looks at completely differential signal chains and confuses an unimportant artifact (the signal being differential to a reference voltage) and confuses it for being essential. It ain't, all you are achieving by recreating the artifact is increasing amplifier noise.

It's partially true, but having it be even slightly mismatched differential will give an edge over completely single-ended. It matters less when the whole thing is shielded, and it is, but as this is an opensource project I'm expecting other, less experienced people to run into blunders like running cheaper unshielded cable, plugging it into questionable gear and such, in which case it does actually help.

[Marco]

Better make those 47 Ohm resistors very high precision then. In audio range the output impedance of the opamp is a bee's dick.

PS. AFAICS with something modern like an OPAx828, a 0.1% difference between the 47 Ohm resistors causes more impedance mismatch across the audio range than dropping the bottom opamp.

[RFDx]

Why do you need a 1GOhm bias resistor in front of the noninverting op-amp?

Avoids DC charging of the input. Given the extremely high(many GOhm, and not particularly well characterized) input impedance of the opamp, it can cause a serious charge to build up on the input and just rail it. I've tried it without the resistor, and found that this happens.

What I was wondering was why do you need 10 x 100 MOhm in series? You would get the same result with 200 MOhm for R13 and the opamp bias resistor.

https://en.wikipedia.org/wiki/Virtual_ground

That's right. A node that has the same voltage potential as GNDA but is not connected to it would be a virtual ground. The inverting input of U1B is indeed a virtual ground. The non-inverting input of U1B, which is connected to the (real) GNDA from the rail splitter, isn't.

[Zipdox]

Have you looked at the JLI electronics mic kits?
https://www.jlielectronics.com/diy-accessories/opa-alice-single-channel/
https://www.jlielectronics.com/diy-accessories/dc-dc-hex-inverter-pcb/
How does it compare?

Also, have you considered using two 6V zener diodes in series to create the virtual ground and 12V supply? That would result in a lower power-on time and reduce component count a bit.