DIY RF Condenser Mics

[rogs]

Rogs, did you have time to try that yet? Any result or comment?

No, I haven't taken things any  further at present.
I'm still waiting for a dual membrane C12 style capsule to try this out with.
WGTcenter have sent me a K47 type in error - twice! (I'm still trying to get refund on the second one..... I think I shall be avoiding WGTcenter in future!)

Hopefully, I shall be receiving a dual membrane C12 style from Rayking within the next day or two.....

 

[rogs]

Just a quick update .....
I have now tried reversing the connections on my dual K47 prototype, and can confirm that Abbey is right -- it is differences between the capsules...

I shall experiment further when my C12s arrive....

 

[Gerard]

WGTcenter have sent me a K47 type in error - twice!

I found reference to another supplier of capsules in another GroupDIY thread; a Guosheng, who makes microphones under the 3U Audio label. The thread suggests he makes his own capsules in both single-sided (RK-47 type) and double-sided (CK12 type) and also SDC capsules. The messages there suggest his capsules are good. They are certainly more expensive than, say WGTCentre. See https://groupdiy.com/index.php?topic=70549.msg905992#msg905992 and the following two messages, and https://www.ebay.co.uk/itm/112996573095. I have written to him to ask what he can supply (the thread I quoted is not particularly recent) and what prices.

That thread also mentions other capsules, though I have not investigated them yet.

 

[Martin K]

Regarding the G. F. J. Arends paper in Omroeptechnische Mededelingen No.1 1961 as mentioned in Baxandall references;
A dutch radio historian kindly send me a copy.
I put it in Technical documents as "G. F. J. Arends paper, Omroeptechnische Mededelingen 1961".

Regards,
Martin K

 

[Gerard]

Regarding the G. F. J. Arends paper in Omroeptechnische Mededelingen No.1 1961 as mentioned in Baxandall references;
A dutch radio historian kindly send me a copy.
I put it in Technical documents as "G. F. J. Arends paper, Omroeptechnische Mededelingen 1961".

Regards,
Martin K

Thanks for posting. But something seems to be wrong; I'm getting "404 - Attachment Not Found"

 

[Gerard]

I see the re-upload. I have now been able to download it. Thanks.

 

[rogs]

I've now had a chance to try out Khron's latest RF.AMX10 pcb with different capsules.

The layout works fine - thanks again Khron! 

I did have some trouble with getting the noise floor as low as I had it with some earlier versions...but this turned out to be nothing to do with the PCB layout.

A bit more experimentation suggested that -  although any phase noise with the oscillator is not critical in this application - any amplitude modulated noise is. (Pretty obvious when you read Baxandall carefully of course! )

I've therefore made some component changes to try and improve the linearity of the oscillator, by increasing the voltage swing
of the oscillator itself, and simultaneously reducing the reactive load  presented by T1 primary by increasing the value of the feed resistor.  Seems to provide better - and more consistent- results.

Details of the changed values (no track pattern changes required) here:  www.jp137.com/lts/AMX10-2txt.pdf

EDIT : 1.7.19: Couple of further value 'tweaks'.  Version is now :  www.jp137.com/lts/AMX10-3txt.pdf

I've also tried out a double sided C12 capsule to create a figure of 8 version.  Seems to work pretty well - although there can be a need to add a small capacitor across one side of the capsules to make both sides match (discussed previously).

What I think maybe more complex is making a switchable multi-pattern version.....
Changing modes would require changes in  'balancing' the capsule capacitive loads, and this in turn would probably require the inductors to be retuned to optimise the settings for each mode  ....  Not good!

For the time being, I'm intending to try out a 'Mid-Side' mic, using 2 x BM800 bodies:
One double sided C12,  mounted as a 'side address' for the 'Side' mic --  and one 'end address' single sided C12 for the 'Mid' mic .

A bit cumbersome maybe, but at least I shan't need to worry about any RF 'birdies'  - which could occur with all capsules mounted in a single enclosure.....

 

[abbey road d enfer]

What I think maybe more complex is making a switchable multi-pattern version.....
Changing modes would require changes in  'balancing' the capsule capacitive loads, and this in turn would probably require the inductors to be retuned to optimise the settings for each mode  .... 

I suggest replacing the capsules that are not connected with fixed capacitors.

 

[rogs]

I suggest replacing the capsules that are not connected with fixed capacitors.

I did include that idea on my initial sketches for the concept (copy attached)  but I can't see how to actually achieve it in practice? The high(ish) Q of the tuned circuitry is pretty selective --- you can only move the tuneable inductor core  a few degrees (say 30) for the output to fall dramatically - over 10dB.

In a 'figure of 8' mode, each capsule is connected across T1 secondary in series..... So T1 load will typically look like about 40pF. (Assuming each capsule is around 80pF - typical for a C12 type)

The same capsule in ommi mode will have both capsules effectively in parallel - so around 160 pF.  That will require a similar value cap to be added to (almost) 're-balance' the bridge. ...So the T1 secondary load would then be around 80pF...

With a 10MHz oscillator, the resonant value of inductance would then need to be around 3.16uH , as opposed to  around 6.3uH for a 40pF capacitor....

Both those values do fall within the specified limits ( well almost!), but would need a very different setting for the inductor core...

Unless I'm missing something here?.....

 

[rogs]

Couple of further 'tweaks' to the RF.AMX10-2 schematic in post #326 above.  Ref. is now RF.AMX10-3

Copy here: www.jp137.com/lts/AMX10-3txt.pdf

 

[rogs]

I've tried out several different capsules in various prototypes, and one thing that has become clear is just how different the sensitivity is between the different types.
K47 and CK12 style capsules seem to have a similar capacitive values (around 80- 90pF)  and similar sensitivity.
K67 style capsules tend to have a lower capacitive value - around 65pF -  and are about 10 to 12dB more sensitive.

Problem is - using this type of RF circuitry - the noise floor remains pretty much constant, regardless of the sensitivity of the capsule. 
So the S/N ratio using a K67 is around 10 to 12dB better than when using K47 or CK12 types.
Using a K67, the noise floor is around the same as with a 'standard' FET Schoeps type circuit...
Using a CK12 or K47 type, the S/N ratio is around 10dB less.
Which is a bit disappointing, if you need a really low noise mic  because - IMHO - the K47 and CK12 capsules tend to sound better than the K67 type. 
I know there are a number of things which affect the sensitivity of a capsule - the  thickness and tension of the membrane for example.
With the lower voltages employed with RF bias type mics, one could afford to use a lower tensioned membrane...which might be more sensitive?...... It's just finding out which types have that, as part of their construction....

EDIT:  I should add that  simply listing a 'type' of capsule is probably oversimplifying some of the above comments. 
I have tried out 2 different 'C12' style capsules (i.e edge terminated).  The Rayking type from AliBaba is listed as 6 microns, and has a sensitivity  similar to a K47.
A 3 micron 'C12' style from a UK supplier (no longer available) has a sensitivity more like a K67.... so I think there is probably quite a range of sensitivities -  even among similar 'types'....

 

[gyraf]

..so is it the "dead capacitance" of the CK12-style design that swamps noise figure, or something else..?

Jakob E.

 

[rogs]

..so is it the "dead capacitance" of the CK12-style design that swamps noise figure, or something else..?
Jakob E.

I'm not sure I understand what you mean by 'dead capacitance'? ....
In the present configuration, the noise floor of the circuit varies considerably - depending on the inductor settings.

I tried removing the capsule and replacing it with a  similar value capacitor to the 'load' capacitor - in this case 68pf.

It was then possible to adjust both inductor slugs to find a very specific 'null' point where all the noise virtually disappears (It's then at least as good as my Schoeps circuit)
T1  and T2 both appear to be at their maximum 'tuned' settings at this null point.

Sadly, this is not the same null point that provides the maximum audio output with  a capsule fitted, with an audio source used for calibration.  There is an increase from the minimum noise floor by around 10dB or so at the maximum audio output setting.

The overall  signal/noise ratio is therefore determined by the sensitivity of the capsule employed.  --
My  3micron C12 and  K.67 samples are about 10dB more sensitive than the K47 - or indeed my 6 micron C12 from Alibaba.
And the signal/noise ratio of those latter two is obviously worse by the same factor.

I've not been able to discover so far what  causes this oscillator noise.  The oscillator output looks to be pretty sinusoidal - especially when measured at either end of the T1 secondary  winding.
The fact that it is possible to tune out most of the noise at the max 'tuned' settings does rather suggest that the signal at the gate of the infinite impedance FET is what is causing problems. 
Looking at the tiny 'residue' of the oscillator signal at both the drain and the source of the FET does show some non-linearity -  and the waveform at the drain is slightly different from the source.  At the minimum noise 'null' point these tiny (a couple of millivolts) residue signals appear to be at their maximum amplitude .

There may well be ways of further improving the application of the FET in this circuit -- it is acting as both rectifier and phase splitter - but I fear the technicalities of how best to progress  may be a bit above my 'pay grade', engineering wise.

As I have mentioned before, the circuit works way better than I had ever imagined when I first sketched it out - especially with high sensitivity capsules -  and I think we're now probably moving into the realms of the law of diminishing returns, trying to take such a simple concept very much further..

It would be nice to try and discover how the inductor tuning settings are  affecting the FET noise output though.....

 

[cyrano]

A question that popped up while reading about your progress: "Why is it Sennheiser only uses RF for shotguns?"

I mean, could it be related to what you're finding in capsule sensitivity? Smaller diameter? Smaller capacitance?

 

[rogs]

A question that popped up while reading about your progress: "Why is it Sennheiser only uses RF for shotguns?"

I mean, could it be related to what you're finding in capsule sensitivity? Smaller diameter? Smaller capacitance?

The Sennheiser white paper on the MKH story : (https://assets.sennheiser.com/global-downloads/file/11061/MKH-Story_WhitePaper_en.pdf)  makes some interesting observations on some of these points.  Because the impedance of an RF capsule is low - and independent of audio frequency - it's easier to manipulate frequency response without adding any noise. So extending the bass response of small capsules is much simpler.
In addition, the directivity of smaller capsules  is better than larger ones, making them more suitable for shotgun use. 
Add to that the advantages of RF condensers for outdoor use, and I think Sennheiser have merely decided to concentrate on  developing smaller capsules  for their RF mics.

One further point I have read (although I haven't been able to quickly find the reference in that paper)  is that the much reduced voltage across the capsule in RF usage allows for less tension in the capsule membrane  - and thus higher sensitivity - in capsules used in this mode.
Certainly, my 3 micron C12 capsules seem to have a lower tension than other capsules I've tried.  They certainly don't seem to like anything much above 50V  polarisation across them in 'conventional' usage!

 

[abbey road d enfer]

It was then possible to adjust both inductor slugs to find a very specific 'null' point where all the noise virtually disappears (It's then at least as good as my Schoeps circuit)
T1  and T2 both appear to be at their maximum 'tuned' settings at this null point.

Since the detection sensitivity is based on the unbalance of the bridgeI would think it's normal to assume the circuit does not operate at it maximum sensitivity when noise is minimum. Knowledge of the various RF voltages is paramount in analysis. Unfortunately quite difficult since probing is invasive.

Sadly, this is not the same null point that provides the maximum audio output with  a capsule fitted, and an audio source used for calibration.  There is an increase from the minimum noise floor by around 10dB or so at the maximum audio output setting.

Exactly for the same reason. In addition, the capsule's  acoustic radiation noise comes into play.

I've not been able to discover so far what  causes this oscillator noise.  The oscillator output looks to be pretty sinusoidal - especially when measured at either end of the T1 secondary  winding.

Since it's amplitude demod, I would think distortion and phase noise to be largely eliminated. It may even work with a square wave... Frequency stability is necessary though, since it modulates the tuned bridge output.

The fact that it is possible to tune out most of the noise at the max 'tuned' settings does rather suggest that the signal at the gate of the infinite impedance FET is what is causing problems. 
Looking at the tiny 'residue' of the oscillator signal at both the drain and the source of the FET does show some non-linearity -  and the waveform at the drain is slightly different from the source.

Indeed non-linearity is necessary; if not there would be no detection. I don't think noise results from these non-linearities. Indeed the actual voltage amplitude is what governs the detection efficiency.

  At the minimum noise 'null' point these tiny (a couple of millivolts) residue signals appear to be at their maximum amplitude .

Seems to be consistent with "maximum tuned setting"

It would be nice to try and discover how the inductor tuning settings are  affecting the FET noise output though.....
[/quote] I believe you have everything to figure this out. Plot noise vs. RF level.

 

[rogs]

Many thanks for your comments and suggestions..... some more testing needed, I think!....

 

[cyrano]

Thanks, rogs, for the link and the insight. Food for thought!

 

[gyraf]

I'm not sure I understand what you mean by 'dead capacitance'? ....

It's that the "active", i.e. moving part of the diaphragm is a relatively smaller portion of the total capsule capacitance - mainly because of the membrane retaining ring being at front potential. Thought this might have upset the system.

But I see your explanation.

btw. I find this thread to be some of the most exiting things that has happened around here for a long time - following closely, but I have nothing to add at this time..

Jakob E.

 

[rogs]

...."It would be nice to try and discover how the inductor tuning settings are  affecting the FET noise output though".....
I believe you have everything to figure this out. Plot noise vs. RF level.

You're right of course --  I should have given it a little more thought before I posted....

The 'no noise' null point  occurs simply because - at maximum tuned RF level - the FET is in saturation.
With no volts across the FET  the capacitors remain essentially fully charged during the whole cycle.
And that leaves only the noise from the 'Schoeps' output stage....

The oscillator output voltage needs to be as large as possible so that - in conjunction with T1 being tuned to maximum - the voltage across T1 secondary presents as large an AC voltage as possible across the bridge.
T2 is then tuned to allow the FET gate to be biased to the optimum setting.

This will not be at maximum of course - for the reason mentioned above - and will vary for each bridge and FET combination.
But there is an optimum setting which can be easily derived, by tuning for maximum audio from a source presented near to the capsule. (This approach was suggested by Ruud very early on in the project).

There is some noise introduced from the oscillator inductor assembly  - although I suspect it's constitution is complex, as suggested by Baxandall on page 595 of his paper.

What does make a huge difference to the overall performance is the sensitivity of the capsule. Using one of my cheap K67 Chinese capsules I can get a signal/ noise ratio similar to my Rode NT1, even though the RF circuitry  introduces more than 10dB of noise into the system.
My K47 - and some C12 capsules - appear to  have a much lower sensitivity, and thus a much reduced overall noise performance.

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.

In any event, using a high sensitivity capsule renders to noise floor low enough not to worry too much about!