DIY RF Condenser Mics

[rogs]

As I mentioned in my notes earlier, I can quite understand that the oscillator may well need some  revision.
The Colpitts oscillator 'block' was taken  as is from an online source - not sure where now? -  so it may well need to be tweaked.

I did fit R4 to allow the oscillator to run without the inductor connected, and in that mode you can measure around 6vp-p  at the emitter.
For those without a 'scope,  I included  DMM  'dynamic DC'  readings at  Q1  when the oscillator is running without T1 fitted:

Q1 collector 33.2
Q1 base 28.8
Q1 emitter 29.3

So the current  is not too high in 'dynamic' mode.

I should perhaps try again to AC couple the input to T1 -- I didn't have much luck the first time I tried doing that...

I look forward to seeing your suggestions in due course.

 

[RuudNL]

Maybe it is an idea to use an external oscillator, to see what the advantage could be of a higher oscillator frequency?
I will see what I can do to maybe optimize the circuit a bit.
(I just did a test with a 12 MHz oscillator. I got 9 V p-p at the output with a supply voltage of 12 V and a current of 2 mA.)

 

[Gerard]

Ruud: Thanks for the heads-up about the microphone body size required. I have a number of BM800 bodies, intened to be donors for 'Alice' or Homero Leal's 'Pimped Alice' circuits; but they are not long enough for this PCB. I'll get some of the bodies you suggest.

Rogs: Yes, this thread has sparked much more interested and positive response that my previous attempt. A big thanks to you and Ruud, and others who have responded, for rising to the bait with such enthusiasm. I'm looking forward to building a couple of these, even in their present state of refinement; maybe more, if the final design proves to be particularly good. Unlike you and Ruud, I have not done much circuit building in the past decade; I don't have a big supply of components, so I must temper my enthusiasm until the parts are delivered. 

 

[Gerard]

A bit off topic, but what a lovely capsule!

I just wonder how they achieve different patterns.

https://cvp.com/pdf/sennheiser_mkh_series.pdf

EDIT:
And i'll answer my own question, they use two back to back capsules.

KingKorg: Are you referring to the Sennheiser capsules or the one Rogs used? The Sennheiser ones are not two diaphragms with their backplates set back to back; rather they comprise a single diaphragm sandwiched between two symmetrical backplates, to provide a balanced output. As the diaphragm moves, the capacitance on one side rises proportionally as that on the other side falls.

 

[rogs]

Maybe it is an idea to use an external oscillator, to see what the advantage could be of a higher oscillator frequency?
I will see what I can do to maybe optimize the circuit a bit.
(I just did a test with a 12 MHz oscillator. I got 9 V p-p at the output with a supply voltage of 12 V and a current of 2 mA.)

I'm still waiting for my 12MHz crystals!  (should be here today?)  but in the meantime I built an external version of the oscillator in my schematic, to check out my original component selections....
Using a 12V supply, and including both inductors and their associated 'load' capacitors - I measure 20V p-p across C8, (without D1 connected ).

It's a complex load.....the crystal looks like  a series inductor/resistor/ capacitor  - in parallel with another capacitor! - and the Q of that is 'tamed' (or more accurately 'damped') a bit by  by R3.  I can  measure c.19V p-p sine wave  between base and ground.


At 10MHz the impedance of T1 secondary (which we are using here as a non centre tapped primary) is around 700 Ohms -- which is understandably slightly  lower than the value of 792 Ohms  specified for 7MHz operation. 
R5 is fitted to both allow control of the current flowing in the emitter circuit, and to reduce the Q of the load on the emitter, to help improve the linearity of the sine wave output - and  to permit more controlled tuning ! 

With  the oscillator fed from a 12VDC supply , I can measure about 11V p-p at the emitter, and as I  mentioned earlier this translates into around 20V p-p across C8 off load.  This is more than adequate to drive the rectifier pretty effectively.  In fact, I has tried using a BAT85 diode at one stage, but I think that's probably overkill in this case.

From a 12V supply, the 'loaded' oscillator draws about 10mA, and produces a pretty linear sine wave output ( it's a bit difficult to measure actual harmonic distortion at 10MHz with my simple test  equipment here!)

Powering the oscillator from phantom power will inevitably require a  compromise, as the effective DC supply is 48 V fed through 3.4K . The values selected will cause the oscillator to draw about 8mA when powered from  phantom power - and this seems to be a decent balance between current drawn and oscillator voltage output.

I think the only change I might make with a 12MHz crystal is to possibly reduce C1 to 100p to improve linearity.
What does remain to be seen if is the loss of  system 'Q'  from using the higher frequency renders any increase in voltage swing across the bridge lost in the reduced output ?....

I did try again to AC couple the input of T1 to the oscillator, but had way too much loss with any cap I tried ?.....

 

[RuudNL]

I am still waiting for the inductors. As soon as I have them, the experiments will start here!

I did try again to AC couple the input of T1 to the oscillator, but had way too much loss with any cap I tried ?.....

Did you try to lower the value of R4 when you used AC coupling?
It should be possible to AC couple T1 without losing too much signal. Anyway, there is no need for DC current through T1.

 

[rogs]

I am still waiting for the inductors. As soon as I have them, the experiments will start here!

Did you try to lower the value of R4 when you used AC coupling?
It should be possible to AC couple T1 without losing too much signal. Anyway, there is no need for DC current through T1.

R4 was originally included to allow the stripboard to be tested before the inductor sub assembly was added. I did try and reduce it  to increase the output volts ready for  my attempt at AC coupling to T1,  but of course that increases the current drawn by the oscillator. In addition, AC coupling this output to T1  produced a  much less linear sine wave, so I returned to direct coupling -  with  R5 in series to control the oscillator current. Much better sine wave!

Then of course you're back into the compromise required by the limitation of the current available  form the phantom power!

With the values I've currently selected for my prototype, you can measure a decent sine wave of c. 12v  p-p at each end of T1 secondary (with respect to ground).
The same sort of voltage appears across the secondary of T2 - but that will of course vary, depending on the values of the bridge imbalance.

The 5u3H inductors are not of course a perfect match for this task - just the closest I've found so far - but the 'step up' voltage gain available from both inductors use in the present configuration do seem to provide a useful amount of AM modulation to make the project worth experimenting with....

 

[Gerard]

• Inductors are type 5u3H  (10mm coils) from Spectrum Communications here in the UK.  (I notice the Ebay stock level of these devices is now 0, so you may need to buy from Spectrum directly - see here:  http://www.spectrumcomms.co.uk/Components.htm

In case people hadn't noticed, the eBay link given in rogs' write-up is to Spectrum Communications eBay outlet. I have written to the owner and he tells me he has lots ('thousands') of coils, not just the few listed on that eBay page. At the link rogs noted above, he also has many different coil specifications, for frequencies up to 50 MHz.

 

[RuudNL]

Damn, still no inductors in the mail today...  :-\

 

[rogs]

...and still no 12MHz crystals for me today either!       

I  note that Spectrum have now put some more stock onto their Ebay shop  although - as Gerard notes-  buying from Spectrum direct is another option with their higher stock levels - and slightly cheaper prices! 

There are useful details on both Spectrum's own site, and on their Ebay shop page with  information on the coil specs.

One thing I've found useful is the comment on the Ebay page about the range of tunable inductance for these 5u3H coils. -- 3.0uH to 7.5uH

And that got me trying out something new....... which seems to allow for another 6dB or so of improved sensitivity.

My prototype schematic has the capacitive loadings on both T1 and T2  set to around 47pF - which is the recommended load for 10MHz.

By  omitting C3 - changing C4 from 47pF to 68pF - and changing C8 from 47pF to 33pF - it is just possible to tune the inductors to around 7.5uH (the max limit).
As C3 was in effect across the 'bridge' , it acts  as a parallel  load to the bridge, and reduces sensitivity.
By removing it - and effectively changing the bridge  loading to around 33pF -  the sensitivity is improved.

Running at 12MHz should allow those coils to be retuned to around 5u3H , so I shall see if that improves the sensitivity any further. It could be that operating a 7MHz coil at 12MHz is a step too far .....we shall see.

So already there will be changes to the original schematic, in due course.

Once Ruud (finally!) gets his inductors,  there may be yet more improvements ...

All good experimental fun...... 

 

[RuudNL]

Just an idea...
Why not add an extra detector (D1, R7,C9), but now with the diode reversed.
Then connect the output to the (now grounded) leg of C5.
Theoretically, this would increase the output level ~6 dB. (And maybe even improve the signal to noise ratio.)
Also the load on T2 would become more symmetrical.

 

[rogs]

Just an idea...
Why not add an extra detector (D1, R7,C9), but now with the diode reversed.
Then connect the output to the (now grounded) leg of C5.
Theoretically, this would increase the output level ~6 dB. (And maybe even improve the signal to noise ratio.)
Also the load on T2 would become more symmetrical.

I did try grounding the centre tap of T2 secondary, and adding a second detector - connected to the presently unused Q2 base - so that the detector essentially became a full wave rectifier.
Unfortunately, the reduction of the 'free' step up voltage gain created by the higher Q in the resonance  of the non centre tapped secondary  T2/C8  dropped the sensitivity a lot.
I haven't tried introducing a second detector without the grounding centre tap........that might maintain the resonance of the secondary, so should be worth a try?.....

EDIT:  A preliminary test seems to indicate that adding a second detector diode and time constant network across T2 secondary - and removing the ground connection - still reduces the Q of the resonant  secondary network quite dramatically.
So I think any attempt to use a balanced detector may need to be a bit more sophisticated - which is a shame!

Certainly the most impressive improvements made so far involve maintaining the high Q tuning options of both inductors.

Removing C3 -  and changing the values of C4 and C8 - saw a dramatic improvement in sensitivity -  and thus signal to noise ratio.

Both my prototypes are now more sensitive than any of my dynamics - but not as good as most of my condensers..  (yet!  )

 

[RuudNL]

T2 is (roughly) loaded witt 10K/1nF.
I can imagine that the Q will be lower with two 'detector' circuits in parallel.
But maybe it is possible to increase the value of R7 (10K) while keeping the R-C time constant about the same.
A higher value will however raise the noise contribution of the PNP output stage.
It feels very frustrating that I can't test anything here, since the inductors still haven't arrived...
(No idea why it takes so long to ship something from GB to the Netherlands!)

 

[rogs]

Looking at it slightly differently, T1 secondary is presently only loaded with 10k/1nF for 50% of the time 

I was thinking about  grounding the T2 secondary centre tap again, and looking at possibly fitting tuning caps to each half of the winding separately -- then re-introducing  a detector to each half again.....  That might help restore some of  the missing 'Q' ?

As I've mentioned before, these inductors are not perfect for this task - merely the closest I've found. 
The loading on them in this configuration is complex though - way above my level of maths! -  I just think that it's probably only the noise free gain available from the tuning that makes this project even vaguely useful.
Without the the 'Q' of these reactive components I suspect the noise levels would render the mic unusable.....

It's annoying you haven't received the inductors -- they were dispatched for next day delivery here...
Mind you, Spectrum is only about 30 miles down the road ! 

 

[RuudNL]

Most RF microphones I know of, use an FM discriminator principle.
If this current approach doesn't lead to the success we want, it might be an idea to have a closer look at the FM discriminator principle. (By the way: most commercial microphones use some way of amplification in the circuit and since the current experimental design only uses emitter followers, it is IMHO no surprise that the output is only slightly higher than a dynamic microphone. As such, I think that is already a success!)
I have a RF microphone here, designed by the NRU (the technical department of the Dutch radio) in the 60's.
The design is very simple: only one transistor (BC109) en a ferroxcube core.
Oscillator is free running at ~8 MHz.
Unfortunately, no schematic or technical details are available.
I was told that the microphone capsules were made by Neumann, but the electronics were developed 'in house'.
The only thing I know, is that there is a trimmer to 'null' the DC output (before the coupling capacitor to the transformer) without any signal.
Output isn't very high, probably because they wanted not too much difference in output compared with dynamic microphones.

 

[rogs]

I've now tried out a 12MHz oscillator......Not really much different from 10MHz -- possibly 1 dB less sensitive?

Both crystal frequencies fall within the inductor tuning range -- and the fact that the 12M is further away from the specified centre frequency (7MHz) probably accounts for a slight loss of Q  even though it is tuned  more towards it's centre  5u3H value.
Swings and roundabouts here I think....

I have been able to improve the sensitivity by another 3dB or so by tuning the inductors manually against an external tone, rather than with a 'scope.
Even using a 10M 15p  divide by 10 scope lead will load the test point and  cause errors in maximum resonance tuning.

I've got the noise floor down to below -60dB, which I think is quite good for such a simple design.

Here's another  'raw' short speech sample: www.jp137.com/las/RF.AMX312M.sample.wav

-  (I think  the sibilance comes from the fact that the capsule is currently a cheap Chinese K67 clone with no EQ)

Next thing - to have another look at the balanced detector idea...

 

[RuudNL]

Excellent result for such a simple circuit!

 

[Khron]

Since the value in question is more a relative one than an absolute one, might an e-field probe be an option, in order to avoid loading down the circuit?

Seems like even a cobbled-together one works well enough to be (somewhat) useful
https://www.youtube.com/watch?v=2xy3Hm1_ZqI
https://www.youtube.com/watch?v=nImoQcoqkuQ

I have been able to improve the sensitivity by another 3dB or so by tuning the inductors manually against an external tone, rather than with a 'scope.
Even using a 10M 15p  divide by 10 scope lead will load the test point and  cause errors in maximum resonance tuning.

 

[rogs]

Since the value in question is more a relative one than an absolute one, might an e-field probe be an option, in order to avoid loading down the circuit?

Seems like even a cobbled-together one works well enough to be (somewhat) useful
https://www.youtube.com/watch?v=2xy3Hm1_ZqI
https://www.youtube.com/watch?v=nImoQcoqkuQ

Fascinating project! ...... in this case I'm not sure how useful it would be?
I can see a problem trying to isolate radiation of the variable RF output (from the IF transformer can) from that of the fixed oscillator..

In actual fact, it's not difficult to calibrate, by placing the mic  right up against a headphone transducer running a tone, and looking at the output of the preamp meter....

 

[rogs]

T2 is (roughly) loaded witt 10K/1nF.
I can imagine that the Q will be lower with two 'detector' circuits in parallel.
But maybe it is possible to increase the value of R7 (10K) while keeping the R-C time constant about the same.
A higher value will however raise the noise contribution of the PNP output stage.

I've now tried a number of combinations - including grounding the T2  secondary centre tap, and creating two separate half wave rectifiers , each tuned to one half of the secondary.  Only half the voltage swing for each half of course, so no real advantage there.

Isolating T2 secondary from ground and adding a second reversed diode with it's own CR network works OK,  but the return path for each half cycle  is then via the second diode and network ... so  losses there.  Again, no real advantage in voltage gain

Increasing the value of the resistors does of course up the output voltage swing - but increases the noise - as you've already mentioned.

I do think the idea of balancing the rectifier network and using both of the Schoeps outputs is a more elegant solution, but tests so far have only increased the component count, with no real gain in voltage swing - a least, not without  creating extra noise as well....

The most important improvement so far has come from removing C3 and changing C8.  Both of which improved the Q of the system -- that's where it seems  the most 'noise free' gain can be found!