DIY RF Condenser Mics

[rogs]

.... Would you mind sharing the type of low tension capsule you used?

I would if I could! ... I purchased 2 edge terminated capsules on Ebay, here in the UK, about 5 years ago. I've not seen the item repeated since.
They were advertised as 3µm devices, and I soon discovered that they didn't like to have any more than 50VDC applied to them.
(Applying 60V made them collapse-- although they did recover, once that source voltage was removed).
They were also very sensitive.
That suggested to me that they were probably low tension membranes, and I thought that made them ideal candidates for RF mic experiments.
Sadly, membrane tension is not a parameter that is often mentioned in published specs -- although the use of thin membrane material, and a high capsule sensitivity, might be useful indicators? ......

 

[gyraf]

..shouldn't we be able to extrapolate tension from given change-in-capacitance-vs.-polarization-voltage?

that is, if we knew material and dimension?

 

[rogs]

..shouldn't we be able to extrapolate tension from given change-in-capacitance-vs.-polarization-voltage?

that is, if we knew material and dimension?

I would think it might at least be possible to make some educated guesses? ......
I've certainly discovered, for example, that most of the cheap K.67 type Chinese capsules I've tried are more sensitive than the K.47 style, given the same polarisation voltage (usually 60V to start with).
The edge terminated types seem to vary quite lot in sensitivity... Without a centre termination, I'm guessing you probably need a thicker membrane - or a higher tension - to allow for a reasonable polarisation voltage to be applied without causing problems?

 

[dx7]

I did an interesting experiment. Since I have a double membrane I decided to connect the rear membrane in place of C4 and try different capacitors in parallel with the rear membrane to create a small capacitance difference between the front and the rear membrane. The result is a figure 8 pattern(!)
When I place only 10pF in parallel to the rear membrane I already get a very badly tuned T1 (low Q) resulting in a relatively bad signal to noise ratio as output. Only when I use 3pF or less I can tune T1 to a nice high Q factor and get a good signal to noise ratio.

When I change the tuning of T2 I get a louder or softer output signal but it has no effect on the overall signal to noise ratio. A very resonant T1 seems to be key to a high quality output signal. I already saw similar behaviour when I was working with only one membrane and a fixed value capacitor for C4.

The capacitance of a single membrane is around 74pF.

 

[rogs]

I did sketch out and try the various pattern options - including the figure of 8 - early on in the project (See here: https://groupdiy.com/threads/diy-rf-condenser-mics.71586/page-17#post-921808 ) and was quite pleased with the results from the fig.of 8 set up.
I've always found that the value of C8 across T2 secondary can make a significant differnce to the 'Q' of T2, and the best results seem to come where there is a very specific 'sweet spot' for the tuning of T2...... I hadn't realised that signal to noise ratio is not affected by the Q of the T2 settings though....
That does rather suggest a dominant noise source maybe from the oscillator itself.... and the higher the Q of the T1 tuning, the more linear the oscillator output ?
I'm suprised to learn that the high Q of T2 doesn't improve the signal to noise ratio though?..... I'd always assumed it did !
The oscillator design may well benefit from some 'tweaking' ... It's a very basic design, based on a Colpitts configuration, but I'm not sure it's optimised for best performance, especially for either linearity and crystal drive level.
As with all aspects of this project, I'm always keen to learn what others discover from their own experiments....

 

[dx7]

I did sketch out and try the various pattern options - including the figure of 8 - early on in the project (See here: https://groupdiy.com/threads/diy-rf-condenser-mics.71586/page-17#post-921808 ) and was quite pleased with the results from the fig.of 8 set up.
I've always found that the value of C8 across T2 secondary can make a significant differnce to the 'Q' of T2, and the best results seem to come where there is a very specific 'sweet spot' for the tuning of T2...... I hadn't realised that signal to noise ratio is not affected by the Q of the T2 settings though....
That does rather suggest a dominant noise source maybe from the oscillator itself.... and the higher the Q of the T1 tuning, the more linear the oscillator output ?
I'm suprised to learn that the high Q of T2 doesn't improve the signal to noise ratio though?..... I'd always assumed it did !
The oscillator design may well benefit from some 'tweaking' ... It's a very basic design, based on a Colpitts configuration, but I'm not sure it's optimised for best performance, especially for either linearity and crystal drive level.
As with all aspects of this project, I'm always keen to learn what others discover from their own experiments....

Thanks for the link to the patterns and the idea for the switch.

It might well be that, in my situation, T2 is less resonant than T1 and that therefore It only has an amplitude effect. It needs more testing on my side to come to a solid conclusion.

 

[dx7]

I was mistaken, tuning T2 has a big effect. It’s just that when you tune it both the signal and the noise go up but the signal goes up a lot more than the noise goes up.

I’ve tried to improve the oscillator but it hasn’t reduced the noise so far.

 

[rogs]

I was mistaken, tuning T2 has a big effect. It’s just that when you tune it both the signal and the noise go up but the signal goes up a lot more than the noise goes up.

I’ve tried to improve the oscillator but it hasn’t reduced the noise so far.

That concurs with what I have found.. selecting C8 to maximise the 'Q' of T2 can help provide some 'noise free' voltage gain, but it's not 'perfect' - as it were - so any noise generated within the T1 assembly will be amplified - to a lesser extent - as well.

The oscillator output itself is not that linear, which is typical of many Colpitts oscillators, but the Q of the inductor assemble and the inclusion of R3 does help to linearise the signal presented to the bridge. T1 is used as part of the oscillator feedback network (as opposed to the more usual second capacitor in the 'standard' Colpitts.
It's the crystal drive level that may need some re-thinking...... I have no way of measuring it accurately (I don't have access to an appropriate current probe ) and I suspect it might be a bit high.
It hasn't caused any problems so far on the versions I've tried out, and the larger crystal dies do seem to be a bit tougher than the tiny ones!
Definitely one aspect for some further experiments, I suspect....

Referring back to your observations on my patterns and switch idea, I should perhaps mention that although the patterns idea is valid (I have tried all 3) the idea of being able to switch between them is not really a sensible option in practice....
Each different pattern will present a different capacitive load to the system, and will thus require the tuning of the inductors to be modified for each mode, to optimise performance. So the switch idea as suggested is probably not really viable......

 

[micolas]

I've done my first step in the process: I've built the microphone (v.5.0) using the Tzt RK-12 double membrane from aliexpress and only connected the front membrane (and temporarily removed the cable from the rear membrane). I've done capacitance testing and came to 74 pF for C4 for maximum output (just before phase reversal of the output). In this situation I've simply used the 10mHz crystal to stay with the original design and see how it behaves.
In general the microphone works very well.
As a test I've compared the output of the RF microphone to a Neumann km183 and have the following observation: Where the km183 has self noise that is more similar to pink noise the RF mic has background noise that is more similar to white noise. As a result the RF mic's self noise is louder than the km183's self noise from approximately 3kHz onwards. This is a rough measurement but clearly observable. Would there be a way to improve this?

Is this the capsule you used?

https://a.aliexpress.com/_EGUHxmB

It's not quite a CK-12 style capsule, but ignoring the microphone's self noise, how does the sound compare to the C-12 or other microphone families?