Best way to add an attenuation pad to DIY opamp based LDC mic

[Glenn Wardle]

Two sort of connected questions with this...
Firstly, does anything other than the LDC sensitivity change when you reduce its polarisation voltage? - and secondly, is there an acoustically 'best way' to add switchable attenuation. The C414 uses both approaches, it drops the polarisation voltage from 60v to 20v for the first 10dB, then drops 100pF of C across the capsule for a further 10dB.
My design will have an infinitely variable polarisation voltage so I don't need to slug any C across the capsule - is that a preferable approach?
I will be using two impedance converters so dropping the polarisation voltage is easier than adding C.

 

[Voyager10]

I think that should be fine. The Austrian Audio OC818 and Neumann TLM107 pads both reduce polarization voltage when the pad is operated.

A too high polarization voltage pulls the diaphragm towards the backplate through electrostatic attraction, and in theory this will increase tension and change the resonant frequency. I've not seen any measurements of this though, nor do any capsules have a required minimum voltage for correct operation, as far as I know.

 

[k brown]

Doesn't the opamp have much higher max input volt than a FET; is a pad really needed?

Has @rogs tested this?

 

[Khron]

Doesn't the opamp have much higher max input volt than a FET; is a pad really needed?

It is, at least for the sake of the preamp the signal goes into.

 

[kingkorg]

.

A too high polarization voltage pulls the diaphragm towards the backplate through electrostatic attraction, and in theory this will increase tension and change the resonant frequency. I've not seen any measurements of this though, nor do any capsules have a required minimum voltage for correct operation, as far as I know.

Not resonant frequency, but limit bass response. There's a sharp knee of sorts though. For example a diaphragm will be fine up to say 70v, between 70v and 75v low end will be limited, at 75-80 it will be close to collapsing and plosives might push it to stick to backplate, at 80 it will stick immediately. I experienced arcing at about 120v.

The main takeaway is 0 to 70v it will sound exactly the same, no change in any aspect besides obviously S/N. Diaphragm compliance is dominated by the tension, electrostatic attraction force can be ignored.

Some capsules have very tight and thicker diaphragms, they will not show any of these effects up until collapse or arcing. Center terminated diaphragms are less prone to these effects. They all have more limited low end response to begin with, compared to CK12 which has huge low end, but very limited voltage that can be applied to it.

Resonant frequency of the diaphragm becomes irrelevant once it is put in place simply because air cushioning behind it and center termination in some capsules. Everything behind the capsule is actually there to linearize, and kill that resonant point, and flatten the response.

Resonant frequency is used only as a measure for adequate diaphragm tension so it doesn't colapse if loose, or limit low end if too tight. Resonant diaphragm frequency of say 1000hz has no implications or relation to 1000hz area sound wise, once diaphragm is put in place. Also, once put in place it is no longer 1000hz. Especially if center terminated.

The myth spilled over from ribbon mics, where ribbon is not damped, and resonant frequency in the low end directly translates to boost at ribbon's resonant point.

 

[abbey road d enfer]

Two sort of connected questions with this...
Firstly, does anything other than the LDC sensitivity change when you reduce its polarisation voltage? - and secondly, is there an acoustically 'best way' to add switchable attenuation. The C414 uses both approaches, it drops the polarisation voltage from 60v to 20v for the first 10dB, then drops 100pF of C across the capsule for a further 10dB.
My design will have an infinitely variable polarisation voltage so I don't need to slug any C across the capsule - is that a preferable approach?
I will be using two impedance converters so dropping the polarisation voltage is easier than adding C.

Dropping polarisation voltage reduces electronics S/N by the same amount it drops sensitivity, so there is a point where it becomes questionable.
OTOH, adding a capacitor across the diaphragm has two effects: it drops sensitivity but also increases distortion, and at the same time, reduces KTC noise.
Both these effects happen when selecting very large attenuation, though. I don't think in normal conditions is matters much, but you may have a problem when recording very loud sources.

 

[rogs]

Doesn't the opamp have much higher max input volt than a FET; is a pad really needed?

Has @rogs tested this?

As Khron says, it might be useful to include a pad for some uses.... The OPA1641 op-amp allows for 'rail to rail' output swing (and no asymmetrical clipping points, as there often are with FET front ends! )
So - with a DC supply of up to c.24v (depending on overall system current drain - and the actual phantom power supply impedance ) that's a possible max. output of c. +20dBu...... That's quite a big input voltage swing for your average mic pre-amp input!

I've been tidying up my OPIC stripboard layouts - again! - to make them just a bit easier and neater to put together - the latest incarnation is OPIC 43
That includes an adjustable voltage multiplier output..... By adjusting the value of R4, and fitting a single pole switch across VR1 (stripboardO5 and P4 are appropriate locations?) you could include a pad selected for whatever value you choose by simply dropping the polarisation voltage.
4000 series CMOS can operate on supply voltage any where between 3 and 18V, so there should be quite a wide range available.

Interesting to note that the nice simple oscillator created by using a single CMOS Schmitt inverter will change frequency as the IC supply voltage is varied. (The Schmitt trigger thresholds - which determine the frequency - are affected by the DC supply).

The frequency increases as the supply voltage is dropped. Not normally a problem..... With an oscillator always running in excess of 100KHz, even a simple passive 1Meg/100nF low pass output filter should provide over 90dB of 'ripple' rejection.....

 

[abbey road d enfer]

Interesting to note that the nice simple oscillator created by using a single CMOS Schmitt inverter will change frequency as the IC supply voltage is varied. (The Schmitt trigger thresholds - which determine the frequency - are affected by the DC supply).

Have you checked the variation of output vs. supply voltage?
I guess when the supply voltage is approaching the recommended minimum, there is a point where the output voltage drops significantly and is not very stable.
This may be a problem when dialing high attenuation. For example, dropping the output from 70 to 7V is impractical because that would imply lowering the supply voltage below 3V.
In order to provide even 12dB of attenuation would require dropping to about 3V. One had better use an intermediary tap of the multipler. Using the 3rd tap would result in a useful 6dB atteuation.

 

[rogs]

Have you checked the variation of output vs. supply voltage?
I guess when the supply voltage is approaching the recommended minimum, there is a point where the output voltage drops significantly and is not very stable.
This may be a problem when dialing high attenuation. For example, dropping the output from 70 to 7V is impractical because that would imply lowering the supply voltage below 3V.
In order to provide even 12dB of attenuation would require dropping to about 3V. One had better use an intermediary tap of the multipler. Using the 3rd tap would result in a useful 6dB atteuation.

I must admit I'd only really thought this kind of mod might be suitable for a 10dB pad ... that should keep the supply voltage at around 5V --- above the 4000 series minimum.
As you say, a tap would probably be a better solution for larger pad values.

Many of my project notes relate to stripboard layouts, rather than PCBs...... Very much on the 'experimental' side of DIY mics, rather than the 'kit building' side of the hobby ... So really just ideas that may perhaps suggest ideas to other experimenters?

Like many others, I've had some frustrating layout problems with inductor based voltage multipliers, and using CMOS logic as a basis for charge pump options for the task seems to be more consistently reliable...

Taking your 'tap' concept even further, I'm thinking it might be an idea to experiment at the other end of the CMOS 4000 series supply voltage range (c.18v) to create a higher final output? ..
Possibly around 120V, which might make a 'tapped' option for voltage based capsule pattern selection viable ? (Rememering to fit 200v caps where necessary of course!).

 

[abbey road d enfer]

Taking your 'tap' concept even further, I'm thinking it might be an idea to experiment at the other end of the CMOS 4000 series supply voltage range (c.18v) to create a higher final output? ..
Possibly around 120V, which might make a 'tapped' option for voltage based capsule pattern selection viable ? (Rememering to fit 200v caps where necessary of course!).

The problem is that you would need a voltage divider to provide the intermediate voltages. I don't know what is the multiplier's output capability.
Simulation says that it should work with a 50 Megohm divider chain but it needs to be confirmed in practice.
Simulation also says that powering with 18V just makes it for 120V.

 

[jp8]

Tapping the lower polarization voltage from 1st or 2nd stage of the voltage multiplier is one solution. But for a -20 dB pad, that will still be too high if you do not lower the supply voltage of the CMOS stages. But why not just tap the voltage from the internal supply? This is how I typically do it, and it works fine.

In the circuit depicted below, I used a voltage divider to generate 6V from the 18V CMOS supply (24V goes to the impedance converter). When I have an impedance converter that runs on 18V, I will either use two series connected zeners (6.2V + 12V) and tap from the 6.2V, or use a resistor divider. Any noise is filtered by R8/C10. Polarization voltage goes from there to the capsule via the usual X-GigaOhm resistor.

Potmeter P1 adjusts the polarization voltage and is optional, but could come in handy to match the gain of two mics to form a matched pair. Build the circuit as compact as possible from SMT parts, use a separate digital ground plane, use local decoupling, and run audio signals as far away as possible from the oscillator on the PCB and you will find absolutely zero digital spikes on the output. Even without filter caps on the outputs.

Jan