DIY RF Condenser Mics

[rogs]

Great stuff - and thanks for the Gerber zip files ..... I shall order 5 from JLC ( I won't try and 'cheat' -- yet!  )

Thank you so much for your efforts ... If this board works as well as it should, it'll be a huge step forward in making  this simple (but hopefully not bad!  ) 'hobby' RF biased mic into a cost effective buildable  reality....

 

[Gerard]

Can I suggest you add Rogs' circuit designation (RF.AMX8) to the silkscreen layer, to we can distinguish between different versions of the PCB.

 

[Khron]

As i mentioned, with the added groundplane on the top layer, i fully expect some capacitances to possibly need tweaking, but i guess time will tell

Compared to the tacked-on mod-board, i'd expect a bit less of the HF / RF to end up in the JFET (or the output PNPs), so at least in theory, the output should(?) end up a bit "cleaner". Whether that's audible or not, is a totally different discussion ;D

Can I suggest you add Rogs' circuit designation (RF.AMX8) to the silkscreen layer, to we can distinguish between different versions of the PCB.

That's actually a good point Zip file updated. I went "fancy" and added that on the soldermask layer ;D

 

[RuudNL]

I was awake last night and thought about the 'new' circuit. (Sometimes that happens...)
The part from the FET to the output looks very much like the classical Schoeps circuit.
Now, in most Schoeps circuits, there is a trimmer potentiometer to set the FET bias.
In the 'new' circuit the bias of the FET is the result of the FET parameters and the current through R4.
Wouldn't it be better to add a trimpot to set the working point of the FET, so that it can work in the most linear region?
Just a thought...

 

[rogs]

I agree that the new infinite impedance detector  FET - especially as it's  being used as a phase splitter as well - looks very much like  the Schoeps configuration..
However in this case, there is no DC bias of the FET gate.
The gate is grounded - from a DC point of view - by the secondary winding of T2. 

The 10MHz modulated carrier signal created by the bridge imbalance is amplified by the turns ratio (and tuning ) of T2.
This signal is presented to the gate and will swing + and - to ground.

It's possible to measure what appear to be a static DC bias across the 4k7 drain and source resistors, which suggests that a 'DC' current of around 1mA is flowing in the FET.
In reality this is actually using a DMM to look at the 'rectified' modulated carrier, which is effectively turned on and off  every half cycle.

So in effect, the optimum setting is determined by tuning for the maximum amplitude of the waveform presented at the gate...

The actual THD appears at present to be actually slightly better than a conventional Schoeps equivalent  circuit (using the same type of  capsule, FET and transistors) - even when that circuit has been optimised with a gate bias adjustment resistor.

The detailed maths of the infinite impedance detector are way above my 'pay grade'  ....but there are a number of articles online which describe the concept. I first got the idea from reading Rod Elliott's audio website (see here:  http://sound.whsites.net/articles/am-radio.htm  - about half way down the page ) which is quite readable!

 

[RuudNL]

I would like to try this. (Decouple the gate from ground for DC)
And see if there could be any improvement. Or not...

 

[rogs]

Hmmm..I'm not sure that would do very much?.......

Biasing the FET accurately in the 'standard' Schoeps mode will allow the (relatively) tiny voltages created by the capsule capacitive changes to be applied accurately to the most  linear part of the FET curve. 
I've found a 1MOhm multi turn pot across the source resistor can be adjusted to  help reduce second harmonic distortion.....

In this AM modulation case we have severall volts of AC being applied to the gate.... The challenge  seems to be to pitch the peaks of that voltage into the 'right' part of the FET  curve......  That 'sweet spot' is simple enough to find using your 'tune by tone' technique....

EDIT -  Not quite true ...... if the 'sweet spot' was not at  maximum amplitude (and therefore maximum Q), there would be 2 'sweet spots' - one each side of the max amplitude setting - and there isn't! 
It is easier to see that T2 is actually tuned  for maximum 'Q' - and therefore maximum  amplitude  -  by applying the 'scope to a lower impedance measuring point .. namely the bridge junction. 
You can then see there that the maximum voltage across T2 primary does coincide with maximum sensitivity.  So the voltage at the FET gate is quite a lot higher - and is essentially a linear sine.
The  drain and source resistors  provide negative feedback to help keep things under control.

The point is, there is no actual DC gate bias to adjust  in this configuration - as far as I can see... 

THD is measured at around 0.02%  at a nominal  -10dB output from my SD preamp. How much of that is distortion from my Sennheiser headphone transducer, and how much from the mic I can't measure....


 

 

[ricardo]

...  fit a JFET 'infinite impedance' detector in place of the diode rectifier.....

Advantages include:
• Much higher Q from T2  - load is now only a JFET gate
• Balanced output by using the JFET as an audio phase splitter as well.

Fitting that idea to the existing circuit , and the voltage swing now available at T2 secondary when the system is tuned can be way too high for the FET (15 -20V p-p) .

One obvious solution- increase R6 to reduce the oscillator DC supply.

Result  is  the oscillator current is now cut by around 50% -- and the sensitivity is increased by about another 6dB

Provisional schematic  is attached... (It may be that some of the new values still need to be 'tweaked'...)

The only disadvantage so far is that it's now a bit more tricky to get it all to fit onto Ruud's original prototype PCB ....

rogs, this is indeed a major development.  But I can't help pointing out that you could replace all the RF stuff with a single 1G resistor  8)

Jokes aside, you need to ask what you get for this complexity.

• 'immunity' to damp
• potentially much better noise

You should certainly get the first but you must be careful not to lose the second.

Making the Schoeps output 'balanced' introduces R10 4k7 and its noise. (The other Schoeps noise sources are much less)

Whether this is significant depends on the sensitivity of the source and its inherent noise output.  A louder source makes this noise less important as  its noise output drowns the extra resistor noise.  But if the source is very low noise .. like a good LDC .. the resistor noise can become dominant.

Zephyr.pdf in https://groups.yahoo.com/neo/groups/micbuilders/files/Mic%20Measurements/
has measurements of a very low noise variable pattern CK12 using Zapnspark's SP2PCM circuit.  You have to join.

You clearly see the effect of the extra resistor as patterns are switched because the LDC has low inherent noise.  The omni has the most noise cos the 2k2 noise is not diluted.  Cardioid doesn't have this resistor at all and Fig8 has it diluted.

A single noise measurement doesn't show this cos this is a HiZ mike and the '1G' noise dominates A-weighted measurements.  But the hissy resistor noise may be the most irritating noise cos the '1G' is very similar to ambient noise so less noticeable and this is in the critical 4kHz area where our hearing is most acute.

My guess is that RF mikes don't have '1G' noise .. ie quieter .. but the remaining noise is white so may be more noticeable.

rogs, is AMX7 louder than a HiZ Schoeps using the same capsule?

Is anyone in Oz going to make this with a CK12 capsule?  I'm happy to do detailed measurements of response & noise if I can borrow the mike for some weeks.

 

[rogs]

rogs, is AMX7 louder than a HiZ Schoeps using the same capsule?

Yes - by around 6dB.

The lowest noise commercial mic I currently have is a Rode NT1 (latest 2014 incarnation) and that  has a published Equivalent  Noise Level (A weighted) of 4dB ...The prototype RF.AMX7 has a noise floor  - measured, sadly, without the benefit of an anechoic chamber - as approximately  6dB worse than that.

But.... the RF.AMX7 is around 15dB more sensitive than the Rode - so it already comes over as a pretty quiet mic!

The most impressive improvement I've noticed from trying out the infinite impedance detector comes from  effectively releasing the full 'Q' potential of the tuned  T2 secondary, by removing the diode detector load.

The next 'tweaks' to be tried are  to find the optimum balance between the oscillator output level - the FET load resistors - and the available phantom power.

The inductors are specified as having a Q of 85 at 7MHz . We are using them here at 10MHz, so it will probably be a bit less.

We're also using them 'backwards' --( i.e. it's the secondary we need centre- tapped, not the primary.) 

T2 primary has a Z of around 750R, and with the impedance of the 2 bridge caps  at around 250R each, T2 doesn't sem to load the bridge too much.
The turns ratio of these transformers - in the mode we are using them here - is around 1:5, so there's quite lot of 'noise free' gain available .....certainly way more than I managed to get from my first prototype, using TOKO cans...

It seems to be turning out to be quite a useful little mic, for such a simple concept . (Thanks, Mr. Baxandall!  )

 

[abbey road d enfer]

I was awake last night and thought about the 'new' circuit. (Sometimes that happens...)
The part from the FET to the output looks very much like the classical Schoeps circuit.
Now, in most Schoeps circuits, there is a trimmer potentiometer to set the FET bias.
In the 'new' circuit the bias of the FET is the result of the FET parameters and the current through R4.
Wouldn't it be better to add a trimpot to set the working point of the FET, so that it can work in the most linear region?
Just a thought...

In the Schoeps circuit, bias has to be trimmed for putting the FET in its most linear region.
In the RF circuit, demodulation is a consequence of the non-linear characteristic of the FET; an extreme case would be a perfectly linear FET, where no demod would result.
I think this has to be investigated; I guess the signal at the source of the FET is not the same as that at the drain (source output is under heavy NFB when signal at the drain follows the x power 3/2 equation).

 

[rogs]

I think this has to be investigated; I guess the signal at the source of the FET is not the same as that at the drain (source output is under heavy NFB when signal at the drain follows the x power 3/2 equation).

The drain and source measure the same ...... using a DMM  one can measure around 4.5V 'DC' across both source and drain resistors...I appreciate it's not actually 'DC'  but the 'half' wave' rectified carrier which appears across both 1nF capacitors.
The audio appears to be the same  amplitude at each output  (inverted with respect to each other of course)

The drain and source of the J113 FET are interchangeable - according to the spec -I must try one with the drain and source connected the other way round!

I'm no theoretical engineer -- I leave the maths to the clever guys !  --  but combining this infinite impedance JFET detector with a phase splitting role appears to generate a symmetrical output from both drain and source?

It seems to me as if this is more like a switching 'sample and hold' technique being employed here... but I may well be wrong of course! 

 

[abbey road d enfer]

The audio appears to be the same  amplitude at each output  (inverted with respect to each other of course)

Indeed; that's the way it must be, since there is no current in the gate. I got somewhat carried away. :-[

It seems to me as if this is more like a switching 'sample and hold' technique being employed here... but I may well be wrong of course! 

I would really think it's the non-linearity of the transfer function that demodulates. Investigating that takes thorough experimentation that is not easy to carry out in DIY conditions.
A first approach would be to evaluate the audio output voltage as a function of RF level at the gate (for a constant modulation index, of course).

 

[ricardo]

is AMX7 louder than a HiZ Schoeps using the same capsule?

Yes - by around 6dB.

The lowest noise commercial mic I currently have is a Rode NT1 (latest 2014 incarnation) and that  has a published Equivalent  Noise Level (A weighted) of 4dB ...The prototype RF.AMX7 has a noise floor  - measured, sadly, without the benefit of an anechoic chamber - as approximately  6dB worse than that.

But.... the RF.AMX7 is around 15dB more sensitive than the Rode - so it already comes over as a pretty quiet mic!

We can argue about measurements & noise weightings but this make your mike

4dBA + 6 - 15 =sorta  -5dBA spl

assuming 2014 NT1 has a 'similar' (made in Sydney) capsule  to your CK12 clone.  So yours is one of a literal handful of mikes with sub 0dBA spl noise 

The most impressive improvement I've noticed from trying out the infinite impedance detector comes from  effectively releasing the full 'Q' potential of the tuned  T2 secondary, by removing the diode detector load.

The next 'tweaks' to be tried are  to find the optimum balance between the oscillator output level - the FET load resistors - and the available phantom power.

7mA is the 'optimum' power to take from P48V ...  with 5mA probably a safe maximum to cater for old N*** ****  8)

Your noise results show more S/N improvement won't be needed.

However, "15dB louder than NT1" means a 10-20dB pad would certainly be useful at times to prevent the mike overloading.  Maybe just a resistor across T2 secondary.

You should also check that the increased sensitivity doesn't kill your HF response too much from the high Q

My guess is that RF mikes don't have '1G' noise .. ie quieter .. but the remaining noise is white so may be more noticeable.

How does the noise of the NT1 sound compared to AMX7?  1G noise is 'red' ... so sounds like ambient noise in a quiet place but 'muffled'.

Can you post 15s of each as a WAV file with the same preamp gains?

For noise measurements, I wrap the mike in towels, put them in a stout wooden box and bury the box under a pile of blankets on a bed with a heavy weight on top.  I record 15s late at night and listen to each recording to make sure there are no insects & distant cars.  Fridge, fans & clocks disabled.

At Calrec, I did the same except the box sat on top of our anechoic wedges.

 

[Gerard]

• It would be nice if you could add an extra capacitor (2.5mm spacing)  across T1 secondary?
There's  a schematic showing the extra cap ( C3 - marked 'SOT') - here:  www.jp137.com/lis/RF.AMX8.jpg

Forgive a simplistic question from a mere 'mechanical', but what is this C3 ('SOT') for, and what value had you in mind for it?

Also, following Ricardo's suggestion earlier today, that there could be an option to add a resistor across T2 secondary to form a pad, what values would be appropriate for a 10 dB pad and for a 20 dB pad?

 

[rogs]

Forgive a simplistic question from a mere 'mechanical', but what is this C3 ('SOT') for, and what value had you in mind for it?

Also, following Ricardo's suggestion earlier today, that there could be an option to add a resistor across T2 secondary to form a pad, what values would be appropriate for a 10 dB pad and for a 20 dB pad?

C3 is included to allow for the  tuning of T1 to be maintained within the recommended inductor parameters. It is only likely to be required where the capacitive value of the capsule to be used is less than c.60pF.
Alternatively, one could try  using a higher frequncy for the oscillator.  That might reduce the 'Q' overall though ... so this is still subject to being tested.  (One of the reasons I included the word  'experiment' in the original project title!  )

There are a few notes on this in the pdf I linked to in post #150    -- but this is now quite a long thread, so they were easily missed! 
...(see here: http://www.jp137.com/lts/RF.AMX5.inductor.assembly.pdf )

Re: pads .....  I think I would be inclined to reduce the sensitivity  - if necessary - by increasing the value of R2  - or possibly R3  -(http://www.jp137.com/lis/RF.AMX8.jpg) , rather than by de-tuning T2.?.. but I'm sure someone will correct me if that's not a good idea!

I don't currently have any values for the figures you've suggested ... again, something to be tested out as the experiments progress.. 

EDIT: On reflection, I think Ricardo's idea to load T2 is the right one....

As a rough guide (within 1 dB or so)  fitting a 3k3 resistor across T2 secondary will attenuate the signal by c. -10dB.
Fitting a 1k resistor will  attenuate by c. -20dB.

Reducing  the sensitivity by modifying the oscillator amplitude is not as straightforward as it first seemed, with a relatively high impedance power supply (i.e. phantom power) ... too many variables.

 

[Khron]

Plenty of room to hang a resistor from the JFET gate to ground (and the associated pin-header to connect a switch, if needed / desired)

EDIT: On reflection, I think Ricardo's idea to load T2 is the right one....

As a rough guide (within 1 dB or so)  fitting a 3k3 resistor across T2 secondary will attenuate the signal by c. -10dB.
Fitting a 1k resistor will  attenuate by c. -20dB.

Reducing  the sensitivity by modifying the oscillator amplitude is not as straightforward as it first seemed, with a relatively high impedance power supply (i.e. phantom power) ... too many variables.

 

[rogs]

We can argue about measurements & noise weightings but this make your mike
4dBA + 6 - 15 =sorta  -5dBA spl
.....assuming 2014 NT1 has a 'similar' (made in Sydney) capsule  to your CK12 clone.  So yours is one of a literal handful of mikes with sub 0dBA spl noise 


.....Can you post 15s of each as a WAV file with the same preamp gains?

........For noise measurements, I wrap the mike in towels, put them in a stout wooden box and bury the box under a pile of blankets on a bed with a heavy weight on top.  I record 15s late at night and listen to each recording to make sure there are no insects & distant cars.  Fridge, fans & clocks disabled.

I think I may have been a fraction too optimistic with my noise figures  -- at least at 1KHz!.....  I had recorded the 2 mics by equalising the gain in the preamp before recording..... not really a very good idea of course....

I have re-done the recordings - taking on board your advice on how to create a home made 'anechoic chamber'!! 

The Rode NT1 is an example of the new 2014 'black' version, and is fitted with Rode's HF6 edge terminated capsule.

The RF.AMX8 prototype uses a Chinese 'C12' type edge terminated capsule.
(No where near as flat as the Rode - but not as bad as most cheap Chinese K.67 types)

The sound source is a single Sennheiser headphone earpiece, mounted within 30mm of both capsules (as close as possible to being equidistant).

The recording was made on my Tascam DR100Mk3 recorder, with an 80Hz HPF applied, to keep out the worst of the 'room rumble'.....


I have attached 4 'raw' 15 second waveforms - exactly as recorded - no adjustments made at all.

www.jp137.com/las/NT1.left.RF.AMX8.right.1KHz.wav
www.jp137.com/las/NT1.left.RF.AMX8.right.5KHz.wav
www.jp137.com/las/NT1.left.RF.AMX8.right.10KHz.wav
www.jp137.com/las/NT1.left.RF.AMX8.right.pink.noise.wav

As the file titles suggest, the Rode NT1 is recorded onto the left channel, the RF.AMX8 prototype onto the right.

As you can see, the 1KHz file shows roughly equal noise floor on both channels, once the signal gain has been increased on the Rode left channel recording to match the signal level on the right.

On the 5KHz, 10KHz  and pink noise recordings the RF.AMX8 noise floor comes out slightly lower than the Rode 

I think the reason for that is probably the non-linear frequency response of the 'Chinese' C12 type capsule....

Nevertheless, I think the RF prototype come out pretty well  for such a simple circuit.....

 

[rogs]

Plenty of room to hang a resistor from the JFET gate to ground (and the associated pin-header to connect a switch, if needed / desired)

I think an  SPDT 'On - Off - On' with the common connected to ground  should do the trick......

 

[Khron]

Is that for a two-step pad (-10dB and -20dB)?

In the mean time, i added a single 2k2 (split the difference) to the board.

I think an  SPDT 'On - Off - On' with the common connected to ground  should do the trick......

 

[rogs]

Is that for a two-step pad (-10dB and -20dB)?

......In the mean time, i added a single 2k2 (split the difference) to the board.

Most of my devices have just a single pad... Usually -10dB - (although my Sound Devices unit has a -15dB pad which is quite useful)... I must check, but I'm guessing your 2k2 resistor will probably be about -15dB?..

I'm wondering if it would be a good idea to change the board ident to 'RF.AMX9'  as well ...... otherwise we shall have 2 versions of RF.AMX8 --- and that's probably not a good idea?....