DIY RF Condenser Mics

[IvanP]

Thanks for reviving this thread and your interest in this project!
I've pretty much gave up on it since I first made one mic and this one is stellar fantastic, and on the wings of the success I've started another four mics. Unfortunately all four of those don't sound remotely good as the first one. Actually they are all unusable noisy! I've chased all components and in the end figured that it's all about capsules. All capsules are bought from the same seller that Rogs recommended in this thread but first one was bought I think about year prior to other four capsules. Guess they changed the batch of capsules that they sell now.
I have put a "bad" capsule in the first mic (the "good one") and it was also noisy after calibration. Actually I could never get the resonant peak to get as high as I could with the first capsule. Measured the capacitance of those "bad" capsules and they turned out to be as low as thirti-ish picofarads.
I don't have money to try and test different sellers and different capsules in order to chase one that is compatibile with this mic. So I gave up. Actually I've started fooling around with the idea to try and ditch RF board from one and do a p2p version of moded Oktava MK-012. Just to try it out, maybe it'll work with those low capacitance capsules I have in four mics.

But just to be clear - this RF mic when done with proper capsule is fantastic. It is extremely low noise with high output gain. I did a location recording for a film few weeks ago and when this thing is connected into a low noise preamp you can record miraculously quiet things with it. No noise at all! Hopefully someone will find out one day what capsule to get in order to get it working. Or what to change to get it working with those shitty capsules they sell now...



Luka

Luka, okay, I'm convinced it's because of the capsule. 30ish pF? Yikes! So I went ahead and measured all the LDC I got.

I currently have 4 of them:
- cheap RK-47 clone (2 wire, bought from a German seller, Ebay), 51pF and is the one I have currently hooked up to RF.AMX10, working great and the only one I tried yet.
- cheap RK-47 clone (2 wire, bought from China, AliExpress), 40pF and it arrived with this ugly wrinkle at the back membrane (attached photo, will attempt re-skinning at some point), untested.
- cheap RK-67 clone (3 wire, bought from China, AliExpress), 42pF and 50pF. I currently have the 42pF side is hooked up to an Alice v56, working great, haven't tried the 50pF side yet - untested.
- I was keeping this for later, but I did acquire an Advanced Audio AK-67 capsule (3 wire, for some reason I was convinced it's a single membrane, but well it's not, hence my investment in multipattern ), measuring 72pF and 76pF, untested. I got it specifically to see how a 'quality' LDC compares to a 'cheap' clone, but also to guarantee a capsule that meets the 65-90pF specs.

I am still waiting for my 3D printed mounts from JLCPCB (I was pleasantly surprised to find out they do 3D prints on the cheap as well, but they ship separate from PCBs, and will update on the quality), but when they arrive, I will do self-noise tests to see how each capsule performs with RF.AMX10 (10MHz or otherwise), and with Alice v56 (HiZ-Schoeps as benchmark).

I suppose I will do this with keeping the preamp gain at the same level, change the capsules, match C4, tune via 1kHz, record silence and then compare the noise floors of different capsules. It will be a hot minute, but I will get to testing them eventually.

I did acquire the cheapo clones as an upgrade from the electrets I've been playing around, but for testing LDCs as well. Understandably, neither do I have the funds to track down the perfect capsule (yet), but it might seem that for the same price, a branded/quality one might well be worth 3 bad clones. I even saw @Khron mentioning he was satisfied with his purchases from Advanced Audio in another thread.

Ivan

 

[rogs]

......

- Coils/crystal. I ordered the Spectrum 5u3HH coils and some cheapo K87 china capsules, and all the capsules tested in the 42-52pF range, which got me concerned at first, but the microphone worked honestly extremely well with the capsule, within 2pF for C4, and 10MHz crystal.
I did try out to swap out the crystals, I do believe there were mentions of Ruud using even a 12MHz with success. Placing the mic between my headphones playing a 1kHz tone, 52pF LDC capsule + 51pf C4, Behrry UMC404HD input gain set the same, both T1 and T2 tuned after each swap to max, I got:
10,00MHz, -42dB
10,70MHz, -36dB
11,06MHz, -28dB
So, obviously there is an increase in sensitivity, but I still haven't gotten into the realm of microphone testing.
There was a mention of a C3 'SOT' (also what does that stand for?) compensating for capsules with less capacitance, but it adds noise then, right? I also think someone mentioned that even the lower capacitance capsules will works, just will be somewhat less sensitive with no more extra noise? Speaking of noise...

- Distortion. What software did you use to determine your distortion (THD%) for RF.AMX10 and competitors? The other day I was trying to learn what capacitors are for, following this neat article: Practical Test & Measurement - Stop Worrying About Coupling Capacitors!
I made the tester and my goal was to measure THD in different types of capacitors, with the intention to prove or disprove the polyester vs polypropylene cap noise superiority. However, I myself own a UMC404HD which is my first audio interface and one I'm extremely happy (especially for the price), but I couldn't get THD under 0,2%, no matter what I tried with the components, tester or inputs, but then realized it's definitely the UMC404's preamp (apparently a Midas preamp ain't a Midas PRO preamp).

Rogs, have to tried to measure distortion with your UMC404HD? Their website doesn't specify THD levels (like Scarlett does) for the model, and I know Sound Devices are quality stuff. I don't know if the problem is in the tester I built or the limitations of a really-affordable sound card, because I would definitely like to give distortion testing a try.

I used REW to try and analyze cap distortion, but THD would always be way higher (THD 0,1-0,2% for caps assumed to be 0,01%), except for ceramics, which were about 0,2% - ceramics are definitely not recommended for audio signal paths.

Apparently, you could use REW to test microphone preamps (without the capsule) for distortion in this method. I'm not sure how, and haven't tried it, but maybe it could be useful in seeing where the problem lies, without the capsule.

- Padding. Is there a relationship, like equation, connecting the RP1 value to how much dB it will pad? I installed:
RP1 2,2k and got about 15db padding
RP2 1,0k and got about 20dB padding
but I didn't measure them too precisely, I will have to redo those tests.
I implemented Khron's idea to put the pins and the small jumper to act as a pad switch, but I put in a row of 4 pins instead, so that the jumper doesn't swivel when only on one pin (visible in the attached photo; padding is set off). Working great so far.
Speaking of which, @Khron, thank you for your PCB design contributions! I actually learned new things from your design, I'll send you a PM when I prepare questions and ideas about the PCB's design, and hopefully I can help contribute too.


- Multipattern. Rogs, you mentioned you don't think multipattern is feasible, but specifically in this, one oscillator design, right? Is there any point in perhaps making two oscillators run one dual-membrane capsule, connected to one JFET? I do believe there were mentions there might be interference from the other oscillator, but I'll be willing to test that out if you hook me up with a schematic. Alternatively, that is solved by using a housing bigger than a BM800 (i.e. I just got a fake U87 mic, and it definitely has enough real estate to ensure there is no interference.) I accidently ordered a three-wired capsule(front memb., back memb., with connected backplates) instead of a single membrane, so I'm invested now

Regarding the coils... As you may have already realised, there are 2 types of Spectrum 5u3H IF transformers.
The initial type supplied was called '5u3H' and the devices were simply marked '5u3'.
That was the type I used for the first year or so of the project experiments.

In 2020 a second 'high stability' version became available - type '5u3HH' - and those are marked '5u3H'.
The listed specification for those is slightly different, and they do indeed perform differently from the original coils, in some respects.
(Details of both types on this page: SPECTRUM COMMUNICATIONS )

The stock of the original coils is now exhausted apparently, so all current and further orders are supplied as type 5u3HH. In some ways these 'seem' to be better. More consistent, and the tuning 'seems' more defined, for example.
In other ways they are quite different from the original coils .
When used with capsules in the 90pF + range, the original coils worked better with an 8MHz oscillator.
That is not so true with the newer coils. They seem to function best with an 8MHz oscillator where the capsules (or to be more accurate, test capacitors!) are within the range of 120pF to 150pF. Much more difficult to find as 'real world' capsules!

It would be interesting to see if anyone else does attempt this project using alternative inductors?
As I have always said, my project is just one experimental offering to the original challenge presented by Gerard in his first post.
There are some expert design engineers and hobby experimenters on this forum. It maybe that other folk have already taken up Gerard's challenge, and have yet to publish their designs ... who knows?

Regarding your comment about C3, the initials 'SOT' stand for 'Select On Test'. In reality, I've never used C3.
It was originally intended for use with lower value capsules, to bring them into the 'tuneable' range of the inductors, but I found it tended to reduce sensitivity quite dramatically, but without reducing noise.

My distortion measurements were simply approximate 'ball park' comparison figures. I used a Neutrik MR1 audio generator, coupled to a Sennheiser headphone transducer, via a Hypex UCD100 class D amplifier.
The software was a freeware spectrum analyser from this site : Visual Analyser

I was able to establish that introducing the transducer/microphone components into the signal path increased the distortion figure by a factor of around 10.... From around 0.003% to 0.03%.
What I was not able to establish was how much of that distortion was introduced by the microphone under test, and how much by the transducer?
Hence only simple 'comparison' test results, using the same signal path, but with different microphones.

As for measuring 'distortion' introduced by different capacitor types, I shall leave that for others to comment on.
As far as I can deduce, here is much nonsense written when it comes to discussing this aspect of microphone construction, and I have no real expertise myself, when it comes to sorting out the science from the 'fairy dust'.
I have personally found very little measurable difference in any 'distortion' introduced by different capacitor types. I know other folk will probably disagree on that!

As you will have already deduced, the current PCB layout includes a number of revisions that Khron very patiently added to his original layout, as the project progressed. I remain very grateful to him for his contribution. Without it, I don't think many people would have tried to build examples of this project?

The pad resistors were added at one stage, but as they reduce the 'Q' of T2 secondary loading, I have tended not to use them, but rather simply reduce the oscillator amplitude, where a lower sensitivity is required.
That reduces the noise floor as well, while maintaining the system 'Q', which is an important part of this concept.

Multipattern... I did try out both a figure of 8 and an omni version of the AMX10, and they worked quite well - especially the figure of 8, which actually requires one less component than the cardioid version!
What did become apparent is that the introduction of different values of capsule capacitance means that each version requires different inductor tunings. So a simple 'switched' version is not really practical.
When it comes to introducing additional oscillators, then I don't personally think it's going to work.
As ex Calrec designer 'Ricardo' (Richard Lee) commented early on, he had tried this out some 30+ years ago, and had given up. Too many problems with 'birdies' (audio signal 'tweets' introduced by interference between 2 or more RF oscillators).
I can confirm that siting 2 identical AMX10 microphones - in BM800 bodies - next to each other with the body 'sleeves' removed renders both mics unusable. Audio frequency 'interference' from the small difference in RF oscillator frequencies. Important to remember also that AM modulation - like FM - generates side bands.
Personally I - like Ricardo - think that the use of multi RF oscillators within a single enclosure is not going to be a good idea.

This has been a fascinating project so far. But the detailed technical reasons behind why there seems to be quite a lot of variations in the quality of (apparently) very similar examples has yet to be established....
The actual maths behind the inductor/ capacitive 'heart' of the project remain elusive.
Academic folk who have been kind enough to take a look so far have all reported that there are too many unknown 'variables' to make any meaningful calculations.
So it is currently left for us experimenters to post what we discover..... And some of the observations do currently seem to result in some 'head scratching' on occasion

 

[IvanP]

It would be interesting to see if anyone else does attempt this project using alternative inductors?

I do recall people asking around how to construct such inductors, with no concrete instructions. Honestly, I was thinking of cutting open one 5u3HH coil one day and then try to recreate it. Just to double check - these specific coils are made from two loops of wire; first wire is center tapped, 10 + 10 loops (total 20), the second is 4 turns of wire (data taken from their site). Diameter of the wires is unknown, but from what I can see from the cracks, looks like something about 0,1mm, I'll check when I open it.

I'm a bit unversed in coils, but should both coils be wound the same way? Does it matter? CW or CCW? Or one is CW, other CCW?
Add some sort of metal casing, thread an iron core. I feel like I've manufactured worse. But no promises.
Oh and, what was the sweet-spot inductance range for capsules again? Somewhere around 7-10uH, was it?

When doing the calculations for the oscillator, do you calculate the series resonant frequency? f=1/(2pi*root(L1*C1)? Or both series and parallel resonant freq. and then stay inside the interval?

In reality, I've never used C3.

Duly noted.

As for measuring 'distortion' introduced by different capacitor types, I shall leave that for others to comment on.

From what I have gathered, sometimes you can get unlucky with a bizarrely noisy capacitor. But I was doing that mostly to learn how capacitors work practically.

The pad resistors were added at one stage, but as they reduce the 'Q' of T2 secondary loading

Got it. I was looking up how padding is solved in the Neumann U87, but that's still a bit too complex for me to grasp and apply, so I thought, hey this is pretty clever. How do you adjust your oscillator amplitude? Do you mean detune a coil or do you replace/adjust a component?

Multipattern... I did try out both a figure of 8 and an omni version of the AMX10, and they worked quite well - especially the figure of 8, which actually requires one less component than the cardioid version!

Oh, that's great news! I am going to try this soon then. You tried the config from your multi-pattern ideas, yes? Specifically figure of 8, where the membranes are connected to T1, center-tap to GND, and the connected backplates to T2 ending in GND, no C4 present?

Does that mean the two sides of the capsule should be matched in capacitance like a single membrane and C4? If so, it will be interesting to see how the imbalanced and cheap K67 will perform.

Personally I - like Ricardo - think that the use of multi RF oscillators within a single enclosure is not going to be a good idea.

Understood.

 

[rogs]

I'm a bit unversed in coils, but should both coils be wound the same way? Does it matter? CW or CCW? Or one is CW, other CCW?
Add some sort of metal casing, thread an iron core. I feel like I've manufactured worse. But no promises.
Oh and, what was the sweet-spot inductance range for capsules again? Somewhere around 7-10uH, was it?

When doing the calculations for the oscillator, do you calculate the series resonant frequency? f=1/(2pi*root(L1*C1)? Or both series and parallel resonant freq. and then stay inside the interval?

Like you, actual coil construction is not something I have any detailed knowledge of ...... Maybe one of the Ham Radio forums might be a better place to ask about winding polarities, etc..?

As for 'sweet spot' ranges I think that is something for experimentation.... The Spectrum coils are specified with 7MHz centre frequency. They can perform quite well at 10MHz as well, but specifically how the actual 'Q' is affected as you move away from the specified centre frequency I'm not sure.... Something that can maybe checked experimentally? ..
As I recall, as you move up into the 12 or 14MHz range, the 'Q' is reduced noticeably....

When it comes to oscillator calculations I found this tool very useful: Resonant Frequency Calculator

How do you adjust your oscillator amplitude? Do you mean detune a coil or do you replace/adjust a component?

I simply reduce the amplitude of the oscillator output, by reducing the DC supply volts the oscillator circuitry. That way you reduce sensitivity and noise, while maintaining maximum system 'Q'.
You can increase the value of R2 to achieve that. Remember though that, as you reduce the oscillator supply you also reduce the current, so the system DC value will actually rise. Not usually a problem as the JFET will always 'self bias' to the optimum setting, and the DC levels shouldn't rise by enough to exceed the JFET voltage limits...
You could also try increasing the value of R3 to reduce the oscillator amplitude feed to T1.... Or even increase perhaps R11 to reduce all DC levels?
Classic example of project experimentation !

.....You tried the config from your multi-pattern ideas, yes? Specifically figure of 8, where the membranes are connected to T1, center-tap to GND, and the connected backplates to T2 ending in GND, no C4 present?

Yes, I used the ideas sketch as a basis. Important to realise that the ground reference of both T1 secondary C.T. and T2 primary is only a reference point.
T2 primary fitted across the 'bridge' ( the part that makes the academic calculations so difficult!) can actually 'float' from an operational viewpoint, but I found that this central galvanically isolated inductive section performs better when it operates referenced around ground.
I found that the internal RF in that central section - when not referenced to ground - can do some strange things, over time!

...And the connections are as you describe.
Although I said that the 'figure of 8' version uses one less component than the cardioid, that rather depends on both sides of the capsule producing a similar amplitude signal.
In the first example I built - using a cheap 'RK12' style capsule - I found it necessary to 'pad' one side of the capsule with an additional capacitor in parallel, to equalise the signal amplitudes. In that particular example it was a 5pF cap fitted into the (otherwise vacant) C4 position.

 

[IvanP]

Okay, understood. Thank you rogs, you've cleared a lot of things for me. As soon as I get some free time, I will dive deeper in all we discussed and hopefully come back with some tangible data and/or observations. Until then!

 

[global-records]

Hi everybody. I'm new to RF circuits for mics, and somehow it never occurred to me to use them for conventional capsules.
But once a long time ago I experimented how far you can go by reducing the gap. Using epoxy and gaskets made of the same material as the membrane. it turned out to be a capsule with a capacity of more than 300 pF, and apparently the diaphragm could still oscillate freely, since the microphone worked, but only up to about 20 volts of polarization voltage and stuck from the slightest whiff.
Having not received any special advantages from this design, I abandoned it.
But now it became interesting, would such a capsule be ideal for an RF circuit? (we do not take into account the possible large distortions). Would the noise requirements of the circuit and the complexity of its adjustment be reduced? You can also do a weak damping to get even more output.
Has anyone here used diy capsules for this project?

 

[Ivan K.]

Hi, I suggest checking the web for Sennheiser MKH 405 / 416 and 805 / 816 schematics. I don't believe I've see capsule specs, but again, a web search may turn up something useful. These mics are basically using the capsule's varying capacitance as an FM modulator, so depending on your noise and sensitivity goals, the diaphragm should be close to the backplate. The main advantage of this arrangement is the lack of arcing, especially in humid and tropical climates. Building a low current, low THD discriminator circuit is not trivial. I find it amazing that the first examples of these RF mics (MKH 104, 404 & 804) used Germanium transistors and diodes and worked quite well, even though the carrier frequency was not crystal controlled.

 

[rogs]

The early RF microphones from Sennheiser did indeed use a form of FM modulation - more precisely 'phase modulation' - but they changed largely to AM modulation in the 1980s (Some details of the Sennheiser history HERE ).
This project uses AM modulation - at least my version of the project does! I found it simpler to implement than FM, especially as it doesn't need to be quite so concerned with oscillator phase noise.
As you can see from my project schematics, it doesn't get much simpler than this to implement the AM concept!

The idea of using capsules with a much larger capacitance is an interesting one.
As you can read throughout this thread, the difficulty in finding useful IF transformers suitable for this task has rather limited the range of suitable capsule capacitive values.
So far, I have only been able to locate 2 IF transformers that even come close to being suitable for this type of circuit.
The 5u3HH from Spectrum, and the Toko type KACSK3894.... and neither are perfect for the task!
The idea of using a capsule with a value exceeding 200pF is interesting.
With the Spectrum inductors, it should allow for crystal frequencies in the 6 to 8MHz range to be used, which are much closer to the specified 7MHz frequency for the inductors. I suspect it may be possible to get some pretty good results at those frequencies.
The concept of using low level voltage applied to a capsule - in this case only a couple of volt of RF - should allow for plates to be pretty close together. The sensitivity can be varied by the amplitude of the applied RF, and the lower that is, the lower the noise.
There will of course be the limitation of the 2 plates actually hitting each other with high input levels, if they are too close together.
It would be interesting to see if the lower noise created from a lower level of RF bias could be useful basis for a high sensitivity, low signal level mic.
With the low impedance characteristics of the RF mic reducing problems with humidity it might be highly suitable for nature recording, for example?

 

[kingkorg]

Im not really familiar with how RF mics work in detail, but achieving higher capacitance in capsules can be done in several ways. The capsule has it's passive capacitance which is pretty much the value that is measured outside the circuit with plain capacitance meter. The portion that "creates" the signal is small, they swing just a couple of pF.

Now, do we need higher passive capacitance? Or do we need higher amount of swing?

Anyways, larger surface of the capsule might be the answer, or even just using several capsules in parallel. Audio Technica 5040 comes to mind.

Not really related, but some great ideas here.

 

[gyraf]

yeah, but.. The passive part of the capacitance acts like an attenuator by forming part of a capacitative divider with the active part - which probably rules out simply paralleling capsule with a fixed (or trimmable?) capacitor..

/Jakob E.

 

[rogs]

Now, do we need higher passive capacitance? Or do we need higher amount of swing?

Both! ... RF mics work well with low tension capsules, simply because they do tend to be more sensitive, and you don't have to worry too much about them 'collapsing' as there is no serious amount of voltage bias across them.

The higher passive capacitance is desirable simply because the capsule forms one leg of the bridge, and is in series with another capacitor of roughly the same value.
That of course halves the value of the capacitive load across the inductor.
The Spectrum 5u3HH IF transformer has - as you may have guessed - a value of 5.3 uH.
For optimum use with the highest 'Q' at a frequency close to the published spec (say 8MHz) you need to load the inductor secondary with around 68pF. So the ideal capsule capacitance would be twice that - around 130 / 140pF.

The second inductor is placed across the 'bridge' to further confuse the issue, so the actual optimum values probably need to be determined experimentally....
The academic folk who have kindly already looked at the maths involved here have concluded that there are too many 'unknown' variables to be able to make any meaningful detailed calculations....

What would be really useful would be the option to select suitable inductors from a range of IF transformers.
Sadly, these seem to be getting almost impossible to find these days. (Understandably).
And from those still available, very few seem have the centre tapped secondary winding required.
The Spectrum 5u3HH can work pretty well, but you need to find capsules that suit it.
Ideally, it should be the other way round!

 

[gyraf]

Would it be somehow possible to specify what sort of behavior we'd really want from our transformer - so that we could maybe try to roll our own?

We wouldn't depend on small size, so we probably have better possibilities than someone aiming at miniature radios?

Any reason why air-core wouldn't work? Coupling?

 

[Khron]

Any reason why air-core wouldn't work? Coupling?

Coupling perhaps, but definitely adjustability.

 

[gyraf]

We can adjust by bending, radioists do that..?

 

[rogs]

Any reason why air-core wouldn't work? Coupling?

I don't think an air cored transformer would have a high enough 'Q' for this task? ... The current project is heavily reliant on the overall 'Q' of the inductor assembly being within the range of something like 60 to 80.

I'm no academic engineer, so I tend to derive my hobby projects by experiment.
It was Umashankar on the old google 'Micbuilders' forum who originally suggested maybe trying IF transformers for this task, and I started from that advice.

I recall it was problems with the details of the inductor construction that had led to Uwe Beis 2014 RF hobby mic design not being taken forward, so I thought I'd try and remove that problem by using 'off the shelf' inductors.
The Spectrum 5u3HH coils can work quite well, but that does depend on the capsule being within a fairly limited range of capacitive value .

For capsules with lower passive values - like SDC capsules for example - I don't think the Spectrum coils would give very good results.
From my own experiments - and without any 'theoretical' analysis - I would think an IF transformer style inductor with a primary impedance of around 1Kohm, a turns ratio of about 4 or 5 to 1, and with a centre tapped secondary might be useful.
It would need to have a Q factor of around 70 to 80 at 8MHz, and a inductive value in the range around 15 to 18uH.
I think that might work quite well with capsules in the 30 to 60pF range.

I've never found an 'off the shelf' IF transformer with that kind of spec, and I'm ashamed to admit I would have no idea about how to go about constructing one. So at the moment I just look for capsules in the 70 - 90pF range which seem to work quite well with the 5u3HH inductors...

 

[global-records]

Now, do we need higher passive capacitance? Or do we need higher amount of swing?

I think both will change equally.
For any type of converters, it is desirable to reduce the passive capacitance, and there are at least two ways:

Electrode surface profile and the performance of condenser microphones
Optimization of Capactive Microphone and Pressure Sensor Performance by Capacitor-electrode Shaping
(sorry if these documents have already been posted on the forum)

This will increase sensitivity, yes, but in some cases the capsule inherent noise may increase.

For RF circuits, sensitivity is much easier to increase by weakening the diaphragm and reducing the gap, so I don't think that passive capacity will hurt much.

For optimum use with the highest 'Q' at a frequency close to the published spec (say 8MHz) you need to load the inductor secondary with around 68pF. So the ideal capsule capacitance would be twice that - around 130 / 140pF.

Thanks for the detailed explanation, now I will have an incentive to assemble the circuit.

 

[kingkorg]

Well, the usual spacer is 40 microns. Reducing the tension and spacer will severely reduce the max spl. However, if the goal is to record low spl at distance, this might be the way to go.

 

[global-records]

Just thinking, I suspect that the ESR of the capsule will differ from ceramics of the same value. With aluminized mylar - even more. It is clear, because in LF-converters this is not critical.
The simulation shows the unbalance of the bridge with equal capacitors with slightly different ESR.
Is that a problem?

 

[rogs]

Just thinking, I suspect that the ESR of the capsule will differ from ceramics of the same value. With aluminized mylar - even more. It is clear, because in LF-converters this is not critical.
The simulation shows the unbalance of the bridge with equal capacitors with slightly different ESR.
Is that a problem?

I have always been advised that it's not really possible to create an accurate simulation of the inductor assembly for this project, simply because there is not enough published data on the Spectrum IF cans to allow for a simulation model with any degree of accuracy to be created?
I'm wondering how you have managed to overcome that, with sufficient accuracy, so that different ESR values become a significant variable in the simulation?
What I have discovered experimentally is that it is desirable to use a 'bridge imbalance capacitor' that is close to the capacitive value of the capsule - and ideally slightly smaller than that value - for optimum results.
But it doesn't want to be too close! If the varying capacitive value of the capsule is such that it 'crosses the centre line' of the bridge balance with audio stimulation, that can produce some very strange results! (The signal will attempt to continually change polarity, for example).

As I have always indicated, this project is very much an experiment, and I for one don't have the academic skills necessary to be able to make any precise theoretical calculations.
But if someone with the necessary mathematical skills can overcome that hurdle, and is able to present an accurate simulation, it would be fascinating to see how the concept might best be improved theoretically, as well as practically....

 

[global-records]

I don't know about the accuracy, the Coil64 program shows the input data for LTspice for the required Q-factor.
I have specified a series resistance for an example of just 0.1 Ohm.
The values of the real ESR are unknown for different types of capsules, my tester shows nonsense (apparently at 10khz) even for ceramic capacitors, but in comparison the capsules show an order of magnitude more. It would be useful if people with equipment measured this at megahertz frequencies. The result may be interesting.