DIY RF Condenser Mics

[abbey road d enfer]

Attached graph of RF signal at FET's gate vs capsule's variations

 

[abbey road d enfer]

Attached graph showing FET current vs. RF gate voltage

 

[ricardo]

Not sure I quite understand what you are saying;  What is the "desired bias" value. Could you please elaborate a bit.

I discuss FET bias in my MicBuilders FETbias.doc

Abbey has in fact answered your question but in a rather roundabout way.

It's about setting good working conditions for the FET.  The voltage across R4 / 10 gives the current through the FET.  The voltage at Drain & Source gives the voltage across the FET.

In FETbias.doc, which describes simple FET circuits up to & including the Schoeps, this is set by the Source resistor & sometimes a potential divider for the Gate.

Here, the Gate voltage is set by the level of the RF presented to it.  Changing the level of the RF by tweaking the bridge will change the bias of the FET.

I don't know at present what the 'desired bias' is as the desired operating conditions are changing with each iteration.  But FETbias.doc tells you the sort of stuff you need to consider.

Which leads me into an important point which we've overlooked ...  In AMX8, R10, the Drain resistor is fed directly from the O/P BJTs.  But that means it is wide open to sh*t on the P48 supply.  The only 'smoothing' is C7 47u.

This may be OK with rogs Sound Devices but there are many prosumer recorders/soundcards (and some 'pro' ones) with terrible noise on their P48.  An important example is early TASCAM DR680 but I think the Mk2 has sorted this out.

Moving R10 & C13 to the other side of R2 and another 47u added across C6 which decouples the oscillator will help this.

Then you need to watch oscillator noise doesn't get into R10.  Also as the voltage available to the FET is now less, you need to be more careful with your bias.

 

[Gerard]

I discuss FET bias in my MicBuilders FETbias.doc

Thanks, Ricardo. I'll check it out and refresh my memory; I last read that doc several years ago.

Which leads me into an important point which we've overlooked ...  In AMX8, R10, the Drain resistor is fed directly from the O/P BJTs.  But that means it is wide open to sh*t on the P48 supply.  The only 'smoothing' is C7 47u.

This may be OK with rogs Sound Devices but there are many prosumer recorders/soundcards (and some 'pro' ones) with terrible noise on their P48.  An important example is early TASCAM DR680 but I think the Mk2 has sorted this out.

That is a very important point for me. I use an early model TASCAM DR-680 and, though I have added extra smoothing to the P48 supplies on channels 1-4, I have not been able to do so for channels 5 & 6. My Brahma and my Oktava MK-012 4D are both quite sensitive to noise on the P48 supplies, the Brahma especially so. I will probably be using the RF condenser mics on channels 5/6 with the sh*tty P48 supply smoothing, so they'll need to be resistant to noise on the phantom power supply.

 

[Khron]

How about shoehorning in some capacitor-multipliers? In the simplest form, it's just an NPN, a resistor and a capacitor, and less than 1V dropped.

Sometimes, straight up adding capacitance just doesn't cut it.

Thanks, Ricardo. I'll check it out and refresh my memory; I last read that doc several years ago.
That is a very important point for me. I use an early model TASCAM DR-680 and, though I have added extra smoothing to the P48 supplies on channels 1-4, I have not been able to do so for channels 5 & 6. My Brahma and my Oktava MK-012 4D are both quite sensitive to noise on the P48 supplies, the Brahma especially so. I will probably be using the RF condenser mics on channels 5/6 with the sh*tty P48 supply smoothing, so they'll need to be resistant to noise on the phantom power supply.

 

[Gerard]

Do you mean in the mic or in the TASCAM? In the TASCAM DR-680 I have now got the P48 smooth enough on channels 1-4, where they are located on an auxiliary PCB that is relatively easily accessed; but I can't get adequate access to the main PCB to do the same for for channels 5-6 (or even to find the location of the P48 supplies for those channels.

 

[Khron]

In the Tascam. Is the layout really that convoluted, that the two pairs of 6.8k resistors can't be found (or accessed)? Cutting the trace and tacking on a cap multiplier should be relatively easy to achieve, i would imagine (or at least hope).

Do you mean in the mic or in the TASCAM? In the TASCAM DR-680 I have now got the P48 smooth enough on channels 1-4, where they are located on an auxiliary PCB that is relatively easily accessed; but I can't get adequate access to the main PCB to do the same for for channels 5-6 (or even to find the location of the P48 supplies for those channels.

 

[Gerard]

Unfortunately, yes it really is. Nearly a quarter of the TASCAM DR-680 is taken up by the battery tray (8 x AA batteries). It's small! Yet it packs six analogue inputs with phantom power, six analogue outputs, and a stereo S/PDIF in and out. It's possible, if not especially easy, to record all eight input channels if you put two inputs on the S/PDIF in. What with the physical size of the XLR/TRS sockets on channels 1-4, the TRS sockets for channels 5-6, six RCA connectors for the line outs and two more for the S/PDIF in and out, as well as a memory card slot, a 24 V power socket and that battery tray, a lot of space is used inside for connectors; it's very tightly packed inside.

The electronics is spread over a main logic board and three or four auxiliary boards, all tightly packed with SMD components. Apparently the switches on the top cover use pads on the main logic board as switching contacts; there doesn't appear to be room for conventional switches. To get at the main board requires removing some dozens of screws, cables, the auxiliary PCBs, external sockets and other overlaid parts.

I managed to get the main analogue input board out; it's the first one that comes out; that allowed me to add eight extra smoothing capacitors on the phantom power distribution to those sockets. But it has only the combined XLR/TRS connectors and P48 supplies for channels 1-4. The next two boards to come out were the DI/DO and AO boards; they did not have anything for analogue inputs 5/6. At that point I damaged the locking catch for a ribbon connector on one of the boards and am reluctant to dig further; but I did get far enough to see that the TRS connectors for channels 5/6 seemed to be connected to the main logic board, not to any more accessible auxiliary PCB.

But having got that far, it was apparent that removing the main board would likely be a significantly bigger task than all that had gone before as all the top panel switches would no longer be held in place, and possibly would no longer be held together. My courage failed at that point  :-\

Someone has reported that they managed to get TASCAM in Germany to accept to modify their DR-680, to make the phantom power quieter. But it seems to me, having read his posting about it, TASCAM only modified channels 1-4, i.e. added extra capacitors on that auxiliary board. I may be mistaken; I read it some years ago, when I had just got my DR-680 and discovered the problem. TASCAM would not even respond to my request for the same modification.

This is now the only reference I can find to the modification: https://www.gearslutz.com/board/showpost.php?p=9675848&postcount=29. Unfortunately the original was on Audiomasters Forum, which seems to have disappeared; I can't find that thread even on the WayBackMachine (I can see it in the index, but not bring it up).

 

[rogs]

I suspect the capsule and C4 are just acting as a potential divider for the RF signal.  That would explain the single twiddle peak.  ie AM modulation is NOT due to the LC frequency changing.

If so, T2 is unnecessary.  It ups the voltage but this is already too high for comfort.  It also raises the impedance which raises the 'noise resistance'.  Removing T2 would also remove the noise contribution of the transformer and its associated bits....

I tried out a sample with just T1 fitted, and with a nominal FET gate  resistor (c.100K?) to ground ...
Works OK, but it does reduce the sensitivity by around 24dB... In addition, there was a significant increase in the noise floor (not from the extra preamp gain required). I can only assume that - as the noise level varies with T1 tuning - it comes from the oscillator?.....
So I fitted T2 again! 
One other observation I've made...As discussed above, the bridge needs to be unbalanced slightly....
One would assume that dC would be a slightly larger percentage of the C4/ capsule series capacitors if C4 was slightly lower in value than the capsule. In my experiments, the reverse seems to be true.
With a C12 style capsule that measures around  100pF, making C4  115pF rather than 85pF actually increases the sensitivity by  around 6dB, which was a surprise.

One other thing I've noticed ... when tuning the coils for maximum  audio level it's interesting to note that when running the audio output through a spectrum analyser it's possible to 'tweak' the max setting  for minimum 2nd harmonic distortion, without affecting the out put level noticeably.....
Much along the same lines as 'tweaking' the bias on the 1G gate resistor with a multi-turn pot across the source of a Schoeps style impedance converter FET ....

I've been very  interested to follow the comments and observations you and Abbey have been making...
I look forward to see what results come from the mods and changes that are derived from your notes by the more experienced experimenters here on the board, as the project 'moves up a gear' as it were....

 

[abbey road d enfer]

I tried out a sample with just T1 fitted, and with a nominal FET gate  resistor (c.100K?) to ground ...

I would think it has no effect on bias because the resistance of the xfmr's secondary is humpteen times smaller. However it may damp the tank a little and reduce sensitivity ...?

With a C12 style capsule that measures around  100pF, making C4  115pF rather than 85pF actually increases the sensitivity by  around 6dB, which was a surprise.

It seems normal to me, since the tank is then tuned with 115pf and 100pf in series, which is close enough to the nominal 47pF.
The main issue with this latest arrangement is that the tank is damped by the 100R from the oscillator so the response is quite broad and does not show the resonance that increases the RF level.

 

[rogs]

I would think it has no effect on bias because the resistance of the xfmr's secondary is humpteen times smaller. However it may damp the tank a little and reduce sensitivity ...?

With T1 only fitted, the input to the gate is taken from the junction of C4 and the capsule -- which has no DC component to gnd.

Although I now know that some of the FET bias is a result of the rectified RF, I had assumed the FET gate would still need a nominal DC ground reference?...

 

[abbey road d enfer]

With T1 only fitted, the input to the gate is taken from the junction of C4 and the capsule -- which has no DC component to gnd.

Aah...OK, then.

 

[ricardo]

I use an early model TASCAM DR-680 and, though I have added extra smoothing to the P48 supplies on channels 1-4, I have not been able to do so for channels 5 & 6. My Brahma and my Oktava MK-012 4D are both quite sensitive to noise on the P48 supplies, the Brahma especially so.

I thought Brahma did a mod which sorted this out and issued a recall for early mike users with the problem like yourself.

Did this not happen?

 

[Gerard]

I thought Brahma did a mod which sorted this out and issued a recall for early mike users with the problem like yourself.

Did this not happen?

I never saw any mention of such a recall.

 

[ricardo]

I need to be very careful what I say.  I've a good grasp of noise in HiZ mikes but I'm a novice at RF stuff.  The Baxandall mike was my one serious RF project apart from a SMPS which worked well but was far too complicated.  40 yrs ago, I had the Great Guru to hold my hand too.

I never played with the 'simple rectifier' version .. going straight to his PSD.

There's a real danger I'm pontificating from the wrong orifice on RF stuff. 

I tried out a sample with just T1 fitted, and with a nominal FET gate  resistor (c.100K?) to ground ...

Works OK, but it does reduce the sensitivity by around 24dB... In addition, there was a significant increase in the noise floor (not from the extra preamp gain required).

I don't think the 100k is needed with a full wave rectifier or the PSD. Maybe with our half wave rectifier.  Half wave rectifier noise is complicated.  When conducting, the 'next' stage sees the tuned circuit which is low noise (?).  But when not conducting (most of the time), the O/P stage sees R7 4k7 which is noisy.

But IS the tuned circuit LoZ and hence low noise?  Or are we just benefiting from higher output?  How does your 100k affect this?

Perhaps its best to Series Tune the output of the bridge like Baxandall does, even with the simple 1/2 wave rectifier.

I can only assume that - as the noise level varies with T1 tuning - it comes from the oscillator?.....

Carefully re-reading Baxandall shows he thinks a Balanced Bridge & PSD will better deal with oscillator noise ... though I can't see how this has more than a limited effect.  And what is the noise of our Crystal controlled Colpitts compared to Baxandall's Hartley?

It may be having the bridge slightly wonky reduces oscillator noise cos less carrier wave.  In that case, my suggestion of using the 'full' RF would be noisier

... when tuning the coils for maximum  audio level it's interesting to note that when running the audio output through a spectrum analyser it's possible to 'tweak' the max setting  for minimum 2nd harmonic distortion, without affecting the out put level noticeably.....
Much along the same lines as 'tweaking' the bias on the 1G gate resistor with a multi-turn pot across the source of a Schoeps style impedance converter FET ....

The FET bias is almost wholly set by the rectified RF so this isn't surprising.  You ARE in fact tweaking the bias of a Schoeps mike.

Much of this is a beach bum talking to himself in an attempt to understand some of the black magic.  What we DO have is a benchmark which is AMX8 and we should not make any moves without looking back at this.

About all I'm sure of is that making C13 10n will reduce noise & flatten its response (with rogs capsule) compared to NT1  8)

But we should look at the effect of noisy P48 too as this & the very high output are IMHO, the 2 most important points to address.

 

[abbey road d enfer]

The FET bias is almost wholly set by the rectified RF so this isn't surprising.

Source voltage without RF carrier is 1.1V. Compare with 2.7V for extreme RF level. I would think the FET's threshold voltage to be a significant factor, as it is in a Schoeps-type circuit.

  You ARE in fact tweaking the bias of a Schoeps mike.

Regarding the role of impedance converter and phase-splitter, yes, but as a detector, I'm not so sure. Detection is a non-linear behaviour. I would think setting the op point is a balance between linearity as a basic audio amp and efficiency as a detector.

 

[rogs]

....But we should look at the effect of noisy P48 too as this & the very high output are IMHO, the 2 most important points to address.

I have 6 mic input sources to hand  that include phantom power....  I had tried out the various versions of this project with 5 of them and had not experienced any problems, noise wise.
Having now tried the 6th  - a cheap Behringer mixer - and I can confirm that it does indeed introduce some hum  into the AMX8 prototype as currently presented.
I have no problem with my own Hi-Z Schoeps type circuit connected to the same input....so maybe time to copy the Hi-Z Schoeps voltage regulation? ....

Adding the components highlighted in blue on the attached schematic seems to solve the P48 noise problem.
In addition, it also makes the mic much more predictable  ---( I'd been having problems matching sensitivities between prototypes for example, with the constantly variable voltages playing havoc with calibration). 
Using a regulated supply (much like standard Schoeps of course) makes life a lot easier.

I've managed to add the necessary components to a small 'piggy back' strip board assembly added to Khron's AMX8 PCB - fortunately without any track cuts! 
No where near as elegant of course, but it does seem to prove that controlling the P48 variables - as suggested by Ricardo - does work well.

As to reducing sensitivity? --- Some of my early prototypes were indeed over sensitive - and not always under effective control! This new regulated version seems much more predictable.

There are a number of other component changes which  should help improve things a bit...
•C2 is now 47pF (improves 10MHz oscillator symmetry)
• R10 and R4 are now 2k2 which  - together with a new rail voltages of 9V -  sets the FET  bias such that there is now c. 3 volts across both R4 and R10, and 3V across Q4 D-S.  There should also be a little less noise by using a lower value resistor for drain and source resistors.

If it is still necessary to reduce the sensitivity further, then adding  pad resistors across T2 secondary  - suggested by Ricardo and implemented by Khron in his AMX9 layout - should do the job. 
Sadly, they will  not reduce the noise floor by the same degree of attenuation.

A way of reducing the noise floor - as well as the signal level - using internal attenuation, rather than an external pad remains one of the outstanding project objectives......

 

[Khron]

AMX9 layout updated w/ zener regulation

Not sure much else can be shoehorned into this BM800-sized board, though ;D
Maaaaaybe if i scoot the cans closer to the "top" edge of the board, and the output pads & filtering elements closer to the "bottom" edge...

PS: For cost-cutting reasons, the 12v zener could be made up of two 6.2v zeners, like the ones across the output PNP's.
One less BOM item to worry about, plus better chance of bulk discounts for multiples

 

[rogs]

AMX9 layout updated w/ zener regulation

Not sure much else can be shoehorned into this BM800-sized board, though ;D
Maaaaaybe if i scoot the cans closer to the "top" edge of the board, and the output pads & filtering elements closer to the "bottom" edge...

PS: For cost-cutting reasons, the 12v zener could be made up of two 6.2v zeners, like the ones across the output PNP's.
One less BOM item to worry about, plus better chance of bulk discounts for multiples

Wow! -- I really couldn't see how that was going to be achievable -- great job! 

Maybe the ident should now be RF.AMX10  ?..... and if you could post the Gerber files that would be great.....

Re. the zener....... I'm currently paying 0.04p (including VAT) for a 500mW 12V Zener -- so I thought - hang the extra expense!   

 

[Khron]

Here ya go.

https://www.dropbox.com/s/sc1y3m0l4uvwxs8/RFmic_gerber.zip?dl=0

And the zener substitution was just a thought