How do I calculate coupling capacitor value for Alice OPA circuit.

[uJustinRee]

I am attempting to design my own little OPA Alice-based circuit and am stumped trying to figure out what the C2/C1 coupling capacitors should be in my schematic. I trust the original circuit that 47uF is sufficient, but I'd still like to understand how to derive it.

I understand the coupling capacitors are forming an RC high-pass filter, which necessitates a large capacitance to push the corner frequency as low as possible to keep it out of audio frequencies. My goal is to use the OPA1656, falling back on the OPA1642 if the quiescent current or some other factor makes it not possible.
Here's the datasheet for convenience: https://www.ti.com/lit/ds/symlink/opa1656.pdf

Looking through the datasheet of the OPA1656, I've tried calculating the corner frequency with the open-loop output impedance vs frequency graph in figure 6-40 (although I wouldn't think it is open loop), the differential input impedance, and dividing the 1G capsule biasing resistor by the theoretical gain at different frequencies to get a "reflected impedance"--I'm unsure what you would call this. None of these methods have ended up with a corner frequency that seems reasonable for either the OPA1656 or the OPA1642, so I don't think I'm going about the problem correctly.

I only really have experience with ideal op amps, and zero experience dealing with AC in op amps, so I'm lost on where to go from here. Any assistance from some circuit analysis gurus would be greatly appreciated.

 

[Khron]

No output impedance figures are necessary. Those 47uF output caps form an RC high-pass in conjunction with the impedance-setting resistors inside whatever preamp(s) the mic is getting plugged into.

Most of the time, that will be in the region of 1-3kohms, but those will also have (in the case of transformerless input ones) input DC-blocking caps in the 22-47uF range usually. So you'll want to pick a value that's manageable as physical size goes, but allows through enough low-end to not mess with phase within the (meaningful) audio range (say, above 40Hz). Aiming for a corner frequency of 1/10th of that is a rule-of-thumb figure.

 

[uJustinRee]

Oh alright, I see... thank you so much. Haha, that is so much simpler than I was thinking.

 

[rogs]

The quiescent current for each amp in the OPA1656 is listed as 3.9mA. So almost 8mA for the dual amp. About twice the current of the OPA1642.
Add the quiescent Zener current, and you are getting close to the phantom power limit.

What is your reason for adding the Zener? .... It is possible to select the values of R4 and R5 so that the maximum supply to the op amp(s) can never exceed the specified limit. The half rail supply will automatically correct itself of course.

I took a slightly different approach to keep the current low, minimise the noise and maximise the headroom.
I used a single OPA1641 op amp - so only about 2mA required there - and a single sided audio output to keep the noise to a minimum.
(Adding R7 and R8 into the signal path to derive the differential audio output will add noise).
Using a single sided audio output - with the differential audio output grounded, but still impedance balanced - will generally result in less noise.

I'm also not a fan of adding EQ into the mic amp itself. Passive single order filters tend to be too 'blunt' to do exactly what you need, and the extra resistors in the signal path will also add noise. I prefer to add EQ into the signal path at line level. (Some ideas on that here: CEQ mic EQ system )

As you have not mentioned a voltage multiplier (which itself will require more current from the phantom power supply) I'm guessing you're looking using a FETless electret capsule?

I tried out a much simpler circuit using a single OPA1641 for this task, and have been pleased with the results.
Some notes here: OPIC Impedance Converter

 

[Khron]

PS: That's some awfully low resistor values in the feedback network of the first opamp... Usually one would want to keep those in the (low) kohm ranges.

 

[abbey road d enfer]

I'm also not a fan of adding EQ into the mic amp itself. Passive single order filters tend to be too 'blunt' to do exactly what you need,

In this particular case, the filter is second-order, which is basically what's needed for compensting teh usual HF (actually high-mid) peak.

 

[uJustinRee]

rogs:

As you have not mentioned a voltage multiplier (which itself will require more current from the phantom power supply) I'm guessing you're looking using a FETless electret capsule?

I actually do plan to use an externally polarized capsule--the JLI Electronics TSC-1. Just below this part of the schematic is a voltage multiplier connected to the +15V rail that I left out of the post. If the OPA1656 turns out to use too much current, hey, oh well, that's why the OPA1642 is there as a backup.

rogs:

I took a slightly different approach to keep the current low, minimise the noise and maximise the headroom.
I used a single OPA1641 op amp - so only about 2mA required there - and a single sided audio output to keep the noise to a minimum.
(Adding R7 and R8 into the signal path to derive the differential audio output will add noise).
Using a single sided audio output - with the differential audio output grounded, but still impedance balanced - will generally result in less noise.

I'll look into using a single op amp output and the other suggestions you made... shouldn't be too hard to omit soldering a few resistors/shorting to ground after the fact. If the OPA1656 quiescent current is too much, I'll try the single audio out method you described with the OPA1655 single op amp version of the OPA1656 and compare with the OPA1642 differential circuit, OPA1641 single output circuit, etc.

rogs:

What is your reason for adding the Zener? .... It is possible to select the values of R4 and R5 so that the maximum supply to the op amp(s) can never exceed the specified limit. The half rail supply will automatically correct itself of course.

About the zener, I was under the impression that this is needed to regulate to a fixed voltage to supply the IC's, charge the capsule, and generally not have voltage be dependent on current draw.

Khron:

PS: That's some awfully low resistor values in the feedback network of the first opamp... Usually one would want to keep those in the (low) kohm ranges.

Thanks for the tip, I was mucking around in LT Spice trying to find the first setup that gave the frequency response I wanted. I'll try to raise those resistors to the 1-10k range.

 

[rogs]

rogs:
About the zener, I was under the impression that this is needed to regulate to a fixed voltage to supply the IC's, charge the capsule, and generally not have voltage be dependent on current draw.

I find it is useful to regulate the voltage multiplier DC supply, to enable a predictable capsule polarisation voltage to be generated.
Variations in that supply voltage will of course be 'multiplied', so can be quite dramatic.
I see no advantage in regulating the op-amp supply.... That seems to only serve to increase the overall current drain, and reduce headroom.
(The half rail 'ground' reference will always self adjust to suit the final supply voltage).

So long as the maximum voltage does not exceed the op-amp spec, the higher the supply voltage, the more headroom is available.
I found the values selected for the op-amp and voltage multiplier circuits in this variation seem to work well : OPIC LDC

 

[uJustinRee]

Here's my new circuit. I definitely don't believe in the OPA1656 dream anymore haha, OPA1642 will do.

I wasn't quite sure how much current the voltage multiplier would consume, but I came to an estimate of ~1.7mA based on your OPIC LDC, rogs. One thing I was wondering is how much headroom do I really need for for the op amps? It seems that any headroom over what I really need only serves to reduce the overall voltage across the capsule, giving smaller voltage swings to read as audio. My rough math for VCC based on a 1.7mA voltage multiplier comes out to ~19V, leaving the 1/2 voltage divider GNDREF at 9.5V, for an overall capsule voltage of 90 - 9.5 = 80.5V. For reference... with your OPIC LDC circuit I calculated 25.4V supplied to your OPA1641 and a capsule voltage of 84 - 12.7 = 71.3V.

Is there a rule of thumb for how large the voltage swings on the capsule end up being and how much voltage the capsule can handle before any adverse effects?

EDIT: durr, I attached the wrong image.

 

[abbey road d enfer]

Here's my new circuit. I definitely don't believe in the OPA1656 dream anymore haha, OPA1642 will do.

I would increase at least 5x the value of the selective FB R's and decreases C's accordingly. OPA1642 does not like being loaded with 470r.

Is there a rule of thumb for how large the voltage swings on the capsule end up being and how much voltage the capsule can handle before any adverse effects?

Not a rule of thumb, but math: the sensitivity is directly related to the polarization voltage, so if the sensitivity with let's say 70V bias is 10mV@94dBspl, it will be 7mV with 50V bias, o 14mV with 100V bias. Now you have to add the actual SPL to that, so at 105dBspl, the voltage with70V bias will be 11dB higher, or 35mV.
Remember that these are rms values, so the actual voltage swing is 2.828 times more.

 

[Khron]

Here's my new circuit. I definitely don't believe in the OPA1656 dream anymore haha, OPA1642 will do.


EDIT: durr, I attached the wrong image.

I'm not crazy about the switching of that filter around the opamp, though. Especially if the switch is a break-before-make type, the opamp will be briefly running open-loop, which will ensure a massive thump / pop on the output.

 

[uJustinRee]

I would increase at least 5x the value of the selective FB R's and decreases C's accordingly. OPA1642 does not like being loaded with 470r.

Not a rule of thumb, but math: the sensitivity is directly related to the polarization voltage, so if the sensitivity with let's say 70V bias is 10mV@94dBspl, it will be 7mV with 50V bias, o 14mV with 100V bias. Now you have to add the actual SPL to that, so at 105dBspl, the voltage with70V bias will be 11dB higher, or 35mV.
Remember that these are rms values, so the actual voltage swing is 2.828 times more.

Okay, I've changed the load to 4k7.

I'm not crazy about the switching of that filter around the opamp, though. Especially if the switch is a break-before-make type, the opamp will be briefly running open-loop, which will ensure a massive thump / pop on the output.

I did not think of this. However, I don't plan to operate the switch while the microphone is being used. The switch is going to be attached to the PCB within the microphone shell, so I'd have to remove the casing to get at it. I just wanted an easy way to get rid of the boost if I didn't like the sound of it.
JS202011CQN: https://www.mouser.com/datasheet/2/240/js-3050885.pdf

Thanks for all the feedback guys, it really has been enormously valuable.

 

[abbey road d enfer]

I did not think of this. However, I don't plan to operate the switch while the microphone is being used. The switch is going to be attached to the PCB within the microphone shell, so I'd have to remove the casing to get at it. I just wanted an easy way to get rid of the boost if I didn't like the sound of it

An easy fix is to connect permanently a 2.2Meg resistor between the output and the inverting input of the opamp.