DIY RF Condenser Mics

[Khron]

Just check what this Gigachad did

You mean this?

https://groupdiy.com/threads/where-does-the-tone-come-from-in-a-microphone.85257/
Not that it had much (or anything) to do with this particular thread in the first place, anyway...

 

[Matt Nolan]

Just check what this Gigachad did

Does it have anything to do with RF mics?

 

[Coco Sloan]

Does it have anything to do with RF mics?

Sorry my bad. I was on the page DIY condenser mics and wanted to post this - but I guess I made a mistake and posted in RF mics.
The video is about testing lots of mics and their frequency response so I hoped it will help someone.
(My first post and it wrong - I will watch out in future)

 

[kingkorg]

Sorry my bad. I was on the page DIY condenser mics and wanted to post this - but I guess I made a mistake and posted in RF mics.
The video is about testing lots of mics and their frequency response so I hoped it will help someone.
(My first post and it wrong - I will watch out in future)

Happens all the time. Welcome!

 

[rogs]

I've been doing some more experiments with the alternative IF cans from Aliexpress that I mentioned here: https://www.jp137.com/lts/AMX10.Inductors.pdf
Those notes show how the original Spectrum coils are wound, with the primary wound onto a different 'tier' from the secondary.
I've taken a slightly different approach for T1....
• I wound the primary first - with 3 turns on the second tier and then 3 turns on the 3rd tier.
• I then overwound the secondary directly onto those same 2 tiers.... 10 turns on the 2nd tier - centre tap - 10 turns on the 3rd tier.
(I 've kept T2 as a Spectrum coil at the moment).

I thought I'd try it with one of my 'low tension' 3um edge terminated capsules (The one that doesn't like more than 50V DC across it!)
It also has a capacitive value of around 85pF, and has a sensitivity of c. -21dB.
The resulting mic is the lowest noise RF mic I've built so far!
It has a noise figure of 8.5dB(A) - so better than either Sennheiser or Rode.
With a signal/ noise ratio in excess of 85dB(A), and a sensitivity of -21dB, I need 8dB less preamp gain than with my Rode NT1, to record a signal at a specific level.
So, with a noise figure only 4dB worse than the NT1, the recorded noise level is about 4dB lower than that from the Rode, when the signal levels are matched.
Which is pretty quiet!
Of course, in the 'real world' these differences are largely academic (ambient noise is going to swamp either of those 2 noise figures, in most situations)
But I'm quite pleased with this result!

I think the next step might be to try and work out the kind of coil winding details that would be needed to work well with capsules of significantly lower capacitive values... Say in the range of 35pF to 50pF?
A quick calculation suggests they're likely to need inductances in the 15uH to 20uH range, to keep the oscillator values at a sensible frequency ?....
That's quite a lot higher than we're currently getting. How easy that would be to obtain with these coil formers, I'm not sure?.....

 

[abbey road d enfer]

I think the next step might be to try and work out the kind of coil winding details that would be needed to work well with capsules of significantly lower capacitive values... Say in the range of 35pF to 50pF?
A quick calculation suggests they're likely to need inductances in the 15uH to 20uH range, to keep the oscillator values at a sensible frequency ?....
That's quite a lot higher than we're currently getting. How easy that would be to obtain with these coil formers, I'm not sure?.....

It suggest increasing the number of turns by a factor ca. 1.5 for the windings that are connected to the capsule.
I know is only theoretical, but it shouldn't be too difficult to experiment.

 

[Gerard]

I came across this in another thread today. Not directly relevant, but interesting nonetheless in the context of winding coils: http://www.vintagewindings.com/tech%20swag/misc%20paper%20page.html
This reminds me a lot of stuff I read in the 1970s and was then 20-40 years old tech. In particular, winding large coils 'diagonally' (starting in a corner and building up layers from the first two turns) was something we did for valve coupling trafos). There may be other gems in there relevant to RF coils.

 

[ricm]

[...]
I think the next step might be to try and work out the kind of coil winding details that would be needed to work well with capsules of significantly lower capacitive values... Say in the range of 35pF to 50pF?
[...]

First of all, thanks for your great job. Very interesting and instructive.

Regarding the problem of adapting the transformer to the capacitance of the capsule due to the fixed frequency chosen, why not replace the oscillator circuit with a simpler one without a crystal where the oscillation frequency was determined by the bridge capacitance and the inductance of the transformer? Thus its frequency would naturally follow the changes in capacitance of the different capsules.
This change would allow also removing the second transformer (some rf chokes needed).
Simulating this change in LTSPice doesn't look bad (attached test asc file), but I don't know if noise (oscillator AM noise) can be the real problem here. Or maybe the simulation model is not correct (anyway, I haven't build this circuit).

 

[rogs]

First of all, thanks for your great job. Very interesting and instructive.

Regarding the problem of adapting the transformer to the capacitance of the capsule due to the fixed frequency chosen, why not replace the oscillator circuit with a simpler one without a crystal where the oscillation frequency was determined by the bridge capacitance and the inductance of the transformer? Thus its frequency would naturally follow the changes in capacitance of the different capsules.
This change would allow also removing the second transformer (some rf chokes needed).
Simulating this change in LTSPice doesn't look bad (attached test asc file), but I don't know if noise (oscillator AM noise) can be the real problem here. Or maybe the simulation model is not correct (anyway, I haven't build this circuit).

Peter Baxandall - whose 1963 WW article I posted here: https://www.jp137.com/lts/Baxandall.RF.mic.pdf - lists the reasons he abandoned FM for this type of project on the first page of that paper. It' s so difficult to keep the oscillator noise levels low, when using simple FM transmitters.
As you can read, Baxandall decided to move to an AM system (as did Sennheiser in the 1980s - some history on that here: https://assets.sennheiser.com/global-downloads/file/11061/MKH-Story_WhitePaper_en.pdf )

Another important observation from Baxandall is the capaitance shifts we are dealing with here.
From his paper: "Ordinary conversational speech at a foot or two corresponds to about 1 dyne/cm² alternating pressure, and this
causes, with a typical modern electrostatic microphone element, a capacitance change in the region of 0.001 pF..."
These are tiny changes..... we need all the help we can get in boosting those signals to usable levels, without being swamped with noise in the following amplification process necessary. Hence the use of a high 'Q' tuned inductor/ capacitor combination from the second transformer.
(High Q tuned circuits effectively allow for some 'noise free' gain ! )

It may be possible to try out alternatives... my project is only one experimental offering -- but I have tried both FM and AM with no second transformer, and have never got near to producing anything that wasn't unacceptably noisy.

You mention 'LT Spice'.... I'm not sure that kind of simulation software is much use in this type of project? ... there are simply too many unknowns in the inductor asembly construction for this type of software to produce anything accurate.
One of our most senior members - Abbey road d'enfer - tried some inductor assembly simulations earlier on in this thread
(round about post #257) . Some interesting results, but not very close to the 'real world' versions currently being tried at that time.
Comments about the theoretical aspects of this particular type of inductor assembly on other expert RF forums tend to bring about the same type of response ... too many unknowns to make an accurate 'simulation'
So up to the present it has just been a 'let's try this next' type of project ... Ideal for a hobby mic forum of course!

Be interesting to see if your 'simulations' help to create an alternative arrangement.
Could be you come up with an effective simple FM system that doesn't have too much oscillator noise, and doesn't need high 'Q' tuned inductors, and prove us all wrong!

 

[ricm]

Peter Baxandall - whose 1963 WW article I posted here: https://www.jp137.com/lts/Baxandall.RF.mic.pdf - lists the reasons he abandoned FM for this type of project on the first page of that paper. It' s so difficult to keep the oscillator noise levels low, when using simple FM transmitters.
As you can read, Baxandall decided to move to an AM system (as did Sennheiser in the 1980s - some history on that here: https://assets.sennheiser.com/global-downloads/file/11061/MKH-Story_WhitePaper_en.pdf )

Another important observation from Baxandall is the capaitance shifts we are dealing with here.
From his paper: "Ordinary conversational speech at a foot or two corresponds to about 1 dyne/cm² alternating pressure, and this
causes, with a typical modern electrostatic microphone element, a capacitance change in the region of 0.001 pF..."
These are tiny changes..... we need all the help we can get in boosting those signals to usable levels, without being swamped with noise in the following amplification process necessary. Hence the use of a high 'Q' tuned inductor/ capacitor combination from the second transformer.
(High Q tuned circuits effectively allow for some 'noise free' gain ! )

It may be possible to try out alternatives... my project is only one experimental offering -- but I have tried both FM and AM with no second transformer, and have never got near to producing anything that wasn't unacceptably noisy.

You mention 'LT Spice'.... I'm not sure that kind of simulation software is much use in this type of project? ... there are simply too many unknowns in the inductor asembly construction for this type of software to produce anything accurate.
One of our most senior members - Abbey road d'enfer - tried some inductor assembly simulations earlier on in this thread
(round about post #257) . Some interesting results, but not very close to the 'real world' versions currently being tried at that time.
Comments about the theoretical aspects of this particular type of inductor assembly on other expert RF forums tend to bring about the same type of response ... too many unknowns to make an accurate 'simulation'
So up to the present it has just been a 'let's try this next' type of project ... Ideal for a hobby mic forum of course!

Be interesting to see if your 'simulations' help to create an alternative arrangement.
Could be you come up with an effective simple FM system that doesn't have too much oscillator noise, and doesn't need high 'Q' tuned inductors, and prove us all wrong!

I have not said nothing about FM, I only suggested a different oscillator without a crystal, as Baxandall and Arends before did; I follow the simplest way, why fixing the frequency if we are not interested in frequency? The transformer and the capacitor bridge is a tuned LC tanq, let it oscillate, a crystal is not needed. Anyway, not FM demodulation, just to be clear.

 

[abbey road d enfer]

That's exactly what I wanted to say. Both rogs' original and ricm's proposal use bridge unbalance for detection, which is some kind of FM-in-reverse, where the carrier is fixed and the detector varies..
In both cases noise is dependant predominantly on the oscillator's amplitude noise, probably not so much on its phase noise.
I think this proposed circuit is worth trying in hardware.

 

[rogs]

I have not said nothing about FM, I only suggested a different oscillator without a crystal, as Baxandall and Arends before did; I follow the simplest way, why fixing the frequency if we are not interested in frequency? The transformer and the capacitor bridge is a tuned LC tanq, let it oscillate, a crystal is not needed. Anyway, not FM demodulation, just to be clear.

My apologies - I had read your comment "... where the oscillation frequency was determined by the bridge capacitance" as meaning FM........my mistake.

I have found so far that using a crystal to control the oscillator frequency is the best way to maintain the kind of accuracy the following calibrated filter system requires for consistent AM operation. Even modest frequency variations can cause problems with signal amplitude and phase distortion in the demodulated signal.
I'm sure it's possible to build a non-crystal oscillator that has sufficient frequency stability, but I think the DC supply would need to be very well regulated to cope with the variations found in the actual outputs of different phantom power supplies.
I'm sure there are several alternative arrangements possible - although whether these would need to employ different inductors or not, I'm not sure?
Using a modified Colpitts crystal oscillator seemed like the simplest option to me....certainly with the 'off the shelf' inductors I have been using.

I look forward to reading about how your experiments progress..... always good to see a different concept tried out!

 

[rogs]

That's exactly what I wanted to say. Both rogs' original and ricm's proposal use bridge unbalance for detection, which is some kind of FM-in-reverse, where the carrier is fixed and the detector varies..

Bridge imbalance is only essential for cardioid and omni patterns The 'figure of 8' configuration works well with a fully balanced bridge!
The output taken from the centre point of the bridge to the second transformer acts rather like the 'wiper' of a 'capacitive potentiometer'.
For both omni and cardioid that 'wiper' must not be allowed to pass throgh the balance 'centre point' - or the signal reverses polarity.
That limitation does not apply to the fig.of 8 option. It's the centre point that shifts in that configuration. So no polarity reversal.

 

[abbey road d enfer]

Bridge imbalance is only essential for cardioid and omni patterns The 'figure of 8' configuration works well with a fully balanced bridge!

That's not what my simulation shows.
With teh bridge balanced, i.e. both caps equal, any variation in either way results in the same positive output voltage.
You may have experienced differently because in a practical implementation it is impossible to have both sides having equal capacitance.

The output taken from the centre point of the bridge to the second transformer acts rather like the 'wiper' of a 'capacitive potentiometer'.

That's exactly how I see it, but the rectifier produces an output voltage that is proportioal to the absolute amplitude (although with an offset), so whatever teh deviation, an imbalance always results in a positive output.

For both omni and cardioid that 'wiper' must not be allowed to pass throgh the balance 'centre point' - or the signal reverses polarity.

I agree, but the notion of "reversed polarity" must be more defined. The output is always positive, even when the stimulus goes negative.

That limitation does not apply to the fig.of 8 option. It's the centre point that shifts in that configuration. So no polarity reversal.

That's not what I see - again in simulation.

 

[rogs]

That's not what my simulation shows.
With teh bridge balanced, i.e. both caps equal, any variation in either way results in the same positive output voltage.
You may have experienced differently because in a practical implementation it is impossible to have both sides having equal capacitance.

That's exactly how I see it, but the rectifier produces an output voltage that is proportioal to the absolute amplitude (although with an offset), so whatever teh deviation, an imbalance always results in a positive output.

I agree, but the notion of "reversed polarity" must be more defined. The output is always positive, even when the stimulus goes negative.

That's not what I see - again in simulation.

I must check my experimental results again ... As I recall, the polarity of the output signal depended on whether the capsule capacitance value was greater than or smaller than the bridge balance capacitor (C4).
Certainly, different mics I have built have had different output signal polarities.
So, although the change in capsule capacitance was always in the same direction for a given signal, the signal polarity seemed to depend on whether that value change tended to make the bridge more - or less - balanced.
Reversing the capsule connections made no difference to the polarity of the signal. Making C4 larger or smaller than the capsule value did.
I shall need to check that again....
EDIT: Just to confirm that if the value of C4 is greater than the capacitive value of the capsule, then the polarity of the output is reversed from the output when the value of C4 is smaller than the capsule....
The lower value gives an output signal that has the same polarity as my reference mic ( Rode NT1) when the circuit is wired as per my schematic.

 

[abbey road d enfer]

I must check my experimental results again ... As I recall, the polarity of the output signal depended on whether the capsule capacitance value was greater than or smaller than the bridge balance capacitor (C4).

Seems to me it's in contradiction with your post #593: "For both omni and cardioid that 'wiper' must not be allowed to pass through the balance 'centre point' - or the signal reverses polarity."
It implies that when the displacement exceeds a certain value - where the output is null - the polarity reverses. This is characteristic of a rectification.
It makes sense since AM detection is sensitive to the absolute amplitude. It means that the output signal is an image of the RF signal's amplitude.
In these conditions, the native fig-8 capsule would also result in "rectified" output.
Your experimental results probably show that there is a hidden detection process happening. Maybe it's simply the fact that both sides of a native fig-8 capsule are never strictly identical. The difference may be negligible, it's still much higher than the pressure-related deviation in normal conditions.

 

[rogs]

Seems to me it's in contradiction with your post #593: "For both omni and cardioid that 'wiper' must not be allowed to pass through the balance 'centre point' - or the signal reverses polarity."
It implies that when the displacement exceeds a certain value - where the output is null - the polarity reverses. This is characteristic of a rectification.
It makes sense since AM detection is sensitive to the absolute amplitude. It means that the output signal is an image of the RF signal's amplitude.
In these conditions, the native fig-8 capsule would also result in "rectified" output.
Your experimental results probably show that there is a hidden detection process happening. Maybe it's simply the fact that both sides of a native fig-8 capsule are never strictly identical. The difference may be negligible, it's still much higher than the pressure-related deviation in normal conditions.

You must be right of course -- in the case of fig-8 both 'bridge' capacitors are the capsule. If they both change capacitive value by exactly the same amount then the value of the total bridge capacitance would remain the same, and the carrier amplitude wouldn't vary.
I do recall that the 90° positions - where both capsulses did effectively receive the same signal - gave a pretty good null.
I must rebuild my fig-8 version and do some more experiements . I have been using one of Ari's 'flat K.47' capsules, which have quite well balanced outputs....
As you say, there must be 'real world' imbalances which allow the system to function.
I have created an amost perfect bridge balance by using a single capsule and fitting a variable capcitor for C4.... That's how I discovered the polarity reversal quirk.

 

[Voyager10]

I wonder whether, rather than relying on bridge imbalance, it would be possible to add (a little bit of) the oscillator output to the the T2 output, so the FET gets the (vector) sum of the two. So when the bridge is one side of the balance point, they add in phase, and when it's the other side, it will partially cancel.

 

[rogs]

I wonder whether, rather than relying on bridge imbalance, it would be possible to add (a little bit of) the oscillator output to the the T2 output, so the FET gets the (vector) sum of the two. So when the bridge is one side of the balance point, they add in phase, and when it's the other side, it will partially cancel.

In this system the JFET is being used as a self biased infinite impedance detector, which is essentially a non-linear funtion that detects changes in RF carrier amplitude.
I'm not quite sure how a change of phase relationship could be incorporated into this particular mode of operation?

But this is still a very experimental project, so just because I can't immediately see how the idea might be used, doen't mean it's not a 'light bulb' moment for someone else?.....

 

[rogs]

Thinking again and of course Abbey is right...
Even with the fig-8 configuration the bridge must not cross the balance point, even with maximum modulation.
I had got it into my head that the capsules would respond identically in opposite polarities to a single point stimulation.
They do respond oppositely of course -- but not by identical amounts ...or anything like.
As the fig-8 configuration is likely to result in both capsules being very similar values, it might be quite feasible to get the signal to reverse polarise if the input was strong enough....
Yes, speaking into one side will unbalance the bridge more and speaking into the other will unbalance it less (opposite polarity) but in neither case must the change in capsule capacitance be allowed to cross the 'centre line' as it were.
Speaking at 90° will apply the same signal to both capsules, and there is no output (the fig-8 side null point).... as you would expect.
I had clearly got the fig-8 concept confused in my mind .... my apologies...