DIY RF Condenser Mics

[Khron]

It was interesting to note that when testing an AMX10 PCB with two equal mlcc capacitors - and no capsule - for the bridge  components , the whole sub assembly is still microphonic.
Presumably something  to do with the IF cans although - as I mentioned above - I think the overall noise generating mechanisms may be complex to analyse.

The cans are a "maybe", but the MLCC's are a definite "yes" for being microphonic.

 

[rogs]

The cans are a "maybe", but the MLCC's are a definite "yes" for being microphonic.

I'm pretty sure the 68pF caps I used for the test were NPO, and not X7R or Z5U .....

I  thought I had read that class 1 MLCC caps weren't very microphonic?... although I may have that wrong...

 

[Khron]

Hmm... At least according to Wikipedia, class 1 ceramics (NP0 among them) should indeed not really be microphonic - i stand corrected.

 

[rogs]

It's been a couple of months since there's been any more posts on this topic ....
I know that Gerard was hoping to complete a  version  using Khron's RF.AMX8 pcb layout -  as time permitted - and Ruud had already tried out an early version using his original prototype PCB ..
But whether anyone else has had a look at trying the idea out, I'm not sure?.

Meanwhile, I've rationalised  many of the ideas put forward by the various contributors to the topic, and have come up with a 'repeatable' version based on Khron's RF.AMX10 PCB layout.

I've made some notes on it here:

www.amx.jp137.com

Hopefully, they will inspire others to have a go ...I'd love to see what variations  other people might suggest.....

 

[Gerard]

Hi Rogs,
Great set of pages on your website. And thank you again, and Kron and Ruud and everyone else who has contributed, for all the dedicated work. My own professional work as a mechanical engineer has taken me away from home to another country for some months before I could complete my mics; I was still waiting on some parts when the assignment came my way. I look forward to completing the mics when I am home from time to time.

 

[Khron]

I'm just glad to have been able to put my (admittedly limited) skills to good use And i'm slightly sorry i can't really do a whole lot more in this direction, as it stands 

 

[shot]

Fantastic job Rogs for putting up a webpage with all the info for this mic! This makes so much easier to digest all 18 pages of this journey.
I pulled a trigger and ordered pcbs and inductors. I thought to give it a try since I have BM800 body unused (two of them in fact!) and I'll experiment with some leftover capsules I think I have somewhere.
I don't need another mic but this is so tempting!

Keep up the good work!
I'll report how it sounds (or does it sound at all ) when pcbs arrive. It'll be a long wait since I went with low cost shipping option on jlcpcb, but I'm not in a rush.



Luka

 

[rogs]

Fantastic job Rogs for putting up a webpage with all the info for this mic! This makes so much easier to digest all 18 pages of this journey.
I pulled a trigger and ordered pcbs and inductors. I thought to give it a try since I have BM800 body unused (two of them in fact!) and I'll experiment with some leftover capsules I think I have somewhere.
I don't need another mic but this is so tempting!

Keep up the good work!
I'll report how it sounds (or does it sound at all ) when pcbs arrive. It'll be a long wait since I went with low cost shipping option on jlcpcb, but I'm not in a rush.



Luka

Glad it helps! --- I have to admit even I would be hard pushed to recall all that's happened along the way .. Lots of changes as the experiments developed, and there was more input from different contributors.
It was time to rationalise the project.....

I've now built 3 exact copies of the  latest version, and they all perform virtually identically -- something which didn't always occur with the early prototypes!
There are slight differences between capsule sensitivities - maybe 1 or 2dB  - but that's about it!.

Look forward to seeing how you get on in due course...

 

[abbey road d enfer]

I've made some notes on it here:

www.amx.jp137.com

The info available on the coils renewed my interest in simming the circuit. Finally got a meaningful sim.
There are two aspects of performance that need more investigation. One is the final tuning. It would be useful to know how you proceed and what the final values are for the inductors, once they are tuned. I know it's somewhat difficult, because measuring them implies disconnecting them, but that would be a significant marker.
The other element is the capacitor that "antagonizes" the capsule (C4 in the schemo). From the beginning I have been skeptic about its presence. I understand that it's a legacy of attempts to use a typical FM detector.
The reason I object to this topology is that, in an extreme case where the capacitor would be of the same exact value as the capsule, there would be zero output because the bridge would be perfectly balanced. Then, any variation of equal value but opposite sign in sound pressure would result in a rectified signal, thus very distorted. I understand capacitor C4 must be chosen different enough of the capsule's value for an acceptable performance, but how much different?
Instinctively, I would think that dispensing with C4 would give a better result in terms of distortion, so I simmed it, and so far, the conclusion is: it works, but it takes a significant re-alignment of the inductors. Counter-intuitively, omitting C4 shifts the alignment down. After re-alignment, it seems the sensitivity is not too different; again, that's according to sim.
I hope you, or someone else had the time and patience to do these experiments. I would have done it but I'm in the process of moving so most of my test equipment is not usable.

 

[rogs]

The tuning procedure is quite straightforward, and I've found only needs to be repeated twice for best results.
Repeating further has so far proved unnecessary.

The microphone under test is arranged as per the top image on this page: http://www.amx.jp137.com/index-perform.html
The pre-amp in my set up is a Sound Devices USBPre, and the output from this is run through a software spectrum analyser, to allow a visual - as well as an audible - maximum audio setting to be observed.

With both cores set in roughly mid position, a 1KHz audio tone is applied to the transducer held right against the headbasket. 
It is set to simulate a typical speech level directly at the mic capsule.

T1 core is adjusted for maximum audio output. This can be quite low - depending on the uncalibrated T2 setting - but should be obvious.

T2 is then adjusted until a maximum audio output position is located. This point will be a dramatic 'peak'  output. There can be more than one peak discovered as the core is tuned, but there is always one dominant peak  which coincides with the minimum second harmonic distortion setting observed on the spectrum analyser.
It is not difficult to locate this dominant peak practically.

It is then desirable to check that the T1 setting is still at maximum, and then repeat the T2 calibration as well.

As you say, it is possible that the capsule and associated bridge capacitor could in theory be exactly the same value. In practice I have never encountered this. There appears to be a very wide range of 'bridge centre' voltages which  can be accommodated by adjusting T2 to optimise the FET bias setting.

The T1 setting is set for maximum oscillator volts into the inductor. The resonant setting  allows for the lowest oscillator current - at around 5mA.

The T2 setting seems to be more complex.  It is apparent that the optimal FET bias setting is not necessarily at maximum resonance of  T2.

On all prototypes so far (around 10) it has been possible to find the optimum output setting for a number of different capsule/ C4 values.  Even when the measured value of C4 has been very close to the capsule value, it has always proved easy to find the optimum setting for the associated FET.

The 5u3H inductors are specified as having a tunable range of 3 to 7.5 uH.
As long as the associated parallel capacitive loading falls with the range of 33pF to 82pF -- which covers most options in this case - then the inductors seem to perform as specified.

You describe a configuration that omits C4 --- are you thinking along the lines of the second sketch in the attached drawing - or have I misunderstood?

 

[abbey road d enfer]

As you say, it is possible that the capsule and associated bridge capacitor could in theory be exactly the same value. In practice I have never encountered this.

I think I haven't been clear enough. It is almost impossible to have a capacitor of the same exact value as the capsule, so tehre would always be an output, but this output would be quite distorted if the values were matched within about 1 pF.

You describe a configuration that omits C4 --- are you thinking along the lines of the second sketch in the attached drawing - or have I misunderstood?

That would be the one attached here.
I don't think the second arrangement would work reasonably; it would be a slope detector, which is known to be problematic in terms of linearity.

 

[rogs]

I think I haven't been clear enough. It is almost impossible to have a capacitor of the same exact value as the capsule, so tehre would always be an output, but this output would be quite distorted if the values were matched within about 1 pF.

I clearly haven't experienced anything that close then!...
As I say, even with values apparently within a few pF of each other, I've always managed to find the 'dominant' tuning peaks I've come to expect, with respectable harmonic distortion figures  (less than 0.05%) 

That would be the one attached here.
I don't think the second arrangement would work reasonably; it would be a slope detector, which is known to be problematic in terms of linearity.

It was simple enough to try that quickly - simply removing C4.  Sadly it hasn't worked well. A serious drop in sensitivity (something like 30dB) and a very broad range of 'tuning' where there is no real change in amplitude, suggesting to me that we are no longer within the required resonant tuning range using these inductors.
There might be some improvement by connecting across the whole winding, rather than just  half? .... but in the present PCB layout the centre tap is grounded, so that will have to be tried with a modified PCB in due course.

 

[abbey road d enfer]

I clearly haven't experienced anything that close then!...
As I say, even with values apparently within a few pF of each other, I've always managed to find the 'dominant' tuning peaks I've come to expect, with respectable harmonic distortion figures  (less than 0.05%) 

It was simple enough to try that quickly - simply removing C4.  Sadly it hasn't worked well. A serious drop in sensitivity (something like 30dB) and a very broad range of 'tuning' where there is no real change in amplitude, suggesting to me that we are no longer within the required resonant tuning range using these inductors.

I would think the main issue is that this connection changes so much the tuning it begs for changing the 47pF cap to a much lower value, probably about 10pF. That's because now, the dominant tuning is transferred to the 1st tank, that sees the capsule's capacitance instead of being virtually shorted.

There might be some improvement by connecting across the whole winding, rather than just  half? .... but in the present PCB layout the centre tap is grounded, so that will have to be tried with a modified PCB in due course.

I don't think so. Even with the half winding, there is largely enough gain, but the resonance is shifted so much lower it becomes impossible to tune the circuit using only the core adjustment.
There may be a problem with my proposal, though. Since the dominant alignment is controlled by the capsule's capacitance, proper alignment may require a change in inductance value or the use of a capacitor in parallels with the capsule (e.g. for SDC's), which would reduce sensitivity.

 

[rogs]

As I've mentioned before here - and in my notes - the idea behind this project came from Peter Baxandall's 1963 paper
(see here: http://www.jp137.com/lts/Baxandall.RF.mic.pdf ).

In that paper, Baxandall's initial experiments were carried out using an unbalanced bridge, and a conventional diode detector. 
He then observed limitations with that approach, and moved onto using switched detectors and a balanced bridge. 
That, I suspect, would be the logical next step for this project to take..... Ricardo already made that observation in this thread some time ago.

What Baxandall didn't use in his unbalanced prototype was an FET infinite impedance detector, which I found to be  a significant improvement over  a standard diode detector - if only because it  essentially removed the loading from the tuned secondary of T2.
With the reality of unbalancing the bridge (if necessary ) simply requiring the addition of a single small extra capacitor, and the wide range of adjustment for optimum FET biasing from T2 tuning, I think  this simple design has worked out pretty well - certainly way better than I first imagined!

I do think that further  improvements are probably best attempted by moving on to Baxandall's balanced bridge concept.
Whether anyone feels like having a go? -- there's certainly the expertise  on this forum!

I think I'll probably give it a miss for the present ..... but I've said things like that before in haste !... 

 

 

[shot]

Oh my! I made this mic!
It's not completely done but I managed to hear some sounds from it.

Unfortunately, I thought I had correct FET in my inventory but I was wrong. And since I didn't want to wait for days to get the correct FET delivered, I tested it with what I had.

I have put J111 and got a nice signal out of the mic, but at too low level. Tried to roughly calibrate T1 and T2 while speaking into the mic but still it was too quiet. Then I tried J201 and got a bit more level out of the mic (I had to recalibrate inductors) but this still seems too low level. I had to crank the preamp on my audio interface to the point it brought too much noise of its own. There's probably some noise from this FET, but I can't distinguish it from the preamp noise at this point. And it seems unstable since some noise an level fluctuations were present. But that's just me testing the mic with wrong FET so please disregard those observations. It's just me being impatient to hear what I've built. Guess I'll have to wait for J113 to be delivered. I cannot buy it locally so it'll be a week or so to wait for it.

I also had similar situation with Q1. I thought I had BC549, but it turned out that I was out of it so I put BC547B instead. I'll get the correct transistor tomorrow. Could it be that this one is actually the cause of low output level? Any other transistors I can try here?

In the BOM on amx.jp137 website it's not specified which variant of transistors should be installed (A, B,C...). For Q2 and Q3 I've installed BC557B with hfe matched to around 278. Should I use C variant that has higher hfe?

Even if it's not working as it should yet, I really like the sound of this mic! It has very well defined low end. And I must admit that I totally dig it's distortion in this uncalibrated stage of the build! Sounds very interesting. In a way it's different from other mics that I've built but still high end, even though it was very cheap to build it.

I'll swap the Q1 for the correct one and wait for the FET to arrive.
In the meantime, should I also put BC557C instead of BC557B?



Luka

 

[Ironworks]

It's been a couple of months since there's been any more posts on this topic ....
But whether anyone else has had a look at trying the idea out, I'm not sure?.
Hopefully, they will inspire others to have a go ...I'd love to see what variations  other people might suggest.....

I've been following your idea for a while, came across it while researching RF mics.
Excellent project!
I will be build a couple of these as soon as time permits.
BTW, on the matter of microphony  I would suggest this is the coils themself's, I have had this occur in some manufactures FM transmitters that used a similar style of coil in the oscillator. In our design we eliminated it by employing a custom sealed toroidial coil in a shielded box . This of course is not an option here!

 

[rogs]

I have put J111 and got a nice signal out of the mic, but at too low level. Tried to roughly calibrate T1 and T2 while speaking into the mic but still it was too quiet. Then I tried J201 and got a bit more level out of the mic (I had to recalibrate inductors) but this still seems too low level. ...

The J111 is significantly different from the J113 in 2 important parameters - IDSS and Vgs, so I'm not surprised  that  it has not been possible to calibrate the FET to the optimum point.
The J201 should be a bit better - although it does have a lower IDSS  than the J113. It may be possible to improve the performance using that device by increasing the values of  R4 and R10  to - say - 10k to reduce the current in the FET. (You would also need to reduce the value of  C9 and C13  to 1nF to correct the HF cut of frequency)

I have tried a 2N5457 which seems to work OK . It's just that the J113 is easier to find, here in the UK.

I also had similar situation with Q1. I thought I had BC549, but it turned out that I was out of it so I put BC547B instead. I'll get the correct transistor tomorrow. Could it be that this one is actually the cause of low output level? Any other transistors I can try here?

The main difference  between the BC547 and the BC549 is the noise figure, which is probably not really a significant consideration in this instance. The difference is Hfe between the B and C variants may make some difference , although I have tried several  BC549C with Hfe differences of more that 50, and have not seen any change in oscillator amplitude measured at the emitter.
I would suggest that most NPN small signal BJTs will probably perform OK in this application...

In the BOM on amx.jp137 website it's not specified which variant of transistors should be installed (A, B,C...). For Q2 and Q3 I've installed BC557B with hfe matched to around 278. Should I use C variant that has higher hfe?

In this emitter follower configuration, matching the Hfe is probably more important that whether you are using B or C versions.

In my prototypes I am using the B version. The higher Hfe of the C version will effectively increase the input impedance to the devices, and thus - in conjunction with C5 and C10 - affect the extreme LF response of the mic  - marginally.
I would suggest that increasing the value of C5 and C10 to 220nF would have a much greater effect on the low  frequency response than using BC557C rather than a BC557B.
....In either case, you're probably only increasing extreme LF 'rumble' noise anyway!

The values shown on the website schematic represented a version that I found to give repeatable results , and could be built with components that were easily available from a single supplier here in the UK.

One further point.  I would strongly recommend applying a fixed tone at the microphone input to calibrate, rather than  trying to find the best settings from a speech input.

There should only be one 'sweet spot' for any given bridge/FET combination. However, there may well be other smaller 'peaks' discovered within the overall tuning range.
The optimum tuning point is to be found within a pretty small range of rotation (around 10 -15 degrees) . So the use of a fixed tone makes things much easier.
If you have the option of running the pre-amp output into a spectrum analyser, that makes thing even simpler!

 

[rogs]

BTW, on the matter of microphony  I would suggest this is the coils themself's, I have had this occur in some manufactures FM transmitters that used a similar style of coil in the oscillator. In our design we eliminated it by employing a custom sealed toroidial coil in a shielded box . This of course is not an option here!

I had reached the same conclusion - that the microphony was from the coils themselves! 
It was only noticeable under very high gain settings during testing, and in reality has not proved to be a problem.
The most significant factor regarding noise from this  circuit is the sensitivity of the capsules themselves. 
Using a selection of cheap Chinese 34mm capsules - K67, K47 and C12 types - I have discovered that the K67 types tend to be much more sensitive than the others . By a factor of around 10dB.
As the noise level from the circuitry remains constant, the signal to  noise ratio of the K67 types is thus around 10dB better than the K47 of C12 types.. and on par with my Rode NT1  (self noise 4dBA ...  sensitivity -29dB)

 

[abbey road d enfer]

In this emitter follower configuration matching the Hfe is probably more important that whether you are using B or C version.

Let's debunk a myth here. 
The two output transistors see different source impedances. The impedance from the drain of the FET is pretty much the drain resistor, but the impedance from the source of the FET is Re//Gm, which would typically result in about half.
With perfectly matched transistors with a hfe of 500, the output impedance at the emitters would be 9 and 7 ohms, with a hfe of 250, they would be 14 and 9.
You can see that a better CMRR can be achieved by having a transistor with hfe=500 in the drain path and one with hfe=250 in the source path.
Assuming the 47r resistors would attenuate the differences would be erroneous since CMRR depends on the actual absolute difference between branches.
To put things in perspective, Schoeps never worried about this, since HF CMRR is governed by the RF caps across the outputs, more than by the impedance of the active branches. LF CMRR is seldom an issue with mics.

 

[rogs]

The two output transistors see different source impedances. The impedance from the drain of the FET is pretty much the drain resistor, but the impedance from the source of the FET is Re//Gm, which would typically result in about half.

I hadn't realised that there would be two different impedances at source and drain in this configuration...

In this  infinite impedance  detector mode, I had always rather thought of it as a 'switch'  - with the oscillator waveform switching between an 'off' state and then 'on' into the linear region of the FET every cycle.
This process charges the associated  drain and source capacitors to a level determined by the changing amplitude of the 'on'  signal , which are in turn discharged through the equal value source and drain resistors, during each half cycle 'off' period.
As the source and drain are interchangeable in this configuration, I had always imagined that the drain and source impedances were the same?
From what you write, I've clearly misunderstood the concept!