Replacing GOhm resistors with diodes, the theory behind it

[k brown]

I can't speak for others, but the reason I do it is go get higher PV on the capsule, for more output. Also, more predictable results across various tubes.

I believe the simple C-37A circuit put over 100v on the capsule. Some resistor values changed over the course of production, which might have resulted in differing capsule voltages.
https://groupdiy.com/threads/sony-c-37a-questions.153/page-2

 

[abbey road d enfer]

I can't speak for others, but the reason I do it is go get higher PV on the capsule, for more output. Also, more predictable results across various tubes.

Increasing the grid resistor does not change the level; it improves noise, until a certan limit.

 

[abbey road d enfer]

That was my whole point - the Royer/Altec/C-37A doesn't use any gig-value resistors, and is waaay simpler.

Indeed it is simpler.
However the C37A does not have the noise benefit of a very high value resistor. The determiing element is the 100meg grid resistor, which makes it intrinsically 10dB noisier than a 1Gohm.
The Altec and Royer are odd, because they don't have a grid resistor.
The grid is floating DC-wise, which is definitely bad practice.
It may be acceptable with nuvistors, though.
David Royer says he relies on the various leakages to replace the grid resistor, I think he's lucky it works. That would certainly not be accepted for any life-sustaining equipment!

 

[k brown]

Indeed it is simpler.
However the C37A does not have the noise benefit of a very high value resistor. The determiing element is the 100meg grid resistor, which makes it intrinsically 10dB noisier than a 1Gohm.
The Altec and Royer are odd, because they don't have a grid resistor.
The grid is floating DC-wise, which is definitely bad practice.
It may be acceptable with nuvistors, though.
David Royer says he relies on the various leakages to replace the grid resistor, I think he's lucky it works. That would certainly not be accepted for any life-sustaining equipment!

I don't think it's only 'Royer's Luck' at work here; worked pretty well for Altec on the M20 'Lipstick' and M11/21B 'Coke-bottle' mics, as well, using 5840 and 6AU6.
https://www.coutant.org/altec21b/index.html
Altec described it: "The backplate of the microphone gets it's polarization through the elevation of the cathode voltage ground potential".

The method may require careful selection of tubes, however.

My Oktava MK-012s (especially with the hypercard capsules) that I gave the 5840 Royer mod, are some of the best-sounding mics I have.

 

[abbey road d enfer]

I don't think it's only 'Royer's Luck' at work here; worked pretty well for Altec on the M20 'Lipstick' and M11/21B 'Coke-bottle' mics, as well, using 5840 and 6AU6.

The problem I have with it is that it is in complete contradiction with a fundamental in electronics, which stipulates that no active element in a circuit must be left without a galvanic connection to ground.
And I have never seen this mode of operation described in reference books.
I'll check RDH when I have time, maybe there is something in it.

Altec described it: "The backplate of the microphone gets it's polarization through the elevation of the cathode voltage ground potential".

I don't question this at all. What I question is how the operating point of the tube is predictable and stable.

The method may require careful selection of tubes, however.

Which I translate as "proper operation relies on unpublished and parameters that are not factory tested."

 

[k brown]

The problem I have with it is that it is in complete contradiction with a fundamental in electronics, which stipulates that no active element in a circuit must be left without a galvanic connection to ground.
And I have never seen this mode of operation described in reference books.
I'll check RDH when I have time, maybe there is something in it.

I don't question this at all. What I question is how the operating point of the tube is predictable and stable.

Which I translate as "proper operation relies on unpublished and parameters that are not factory tested."

Royer also says in the article that the idea is probably best suited to SDCs.

 

[thor.zmt]

This wouldn't work for simulating the distortion due to capacitance variation. You need a non-linear cap, expressing the fact that C is in the form 1/(1+x), x representing the instantaneous sound pressure.

Correct. But as said, knowing other distortion sources in the system, am I worried about 0.1% distortion on a 0.1% capacitance change for a microphone to record music with?

A similar reason is often given for neglecting head amp noise.

I am not NEGLECTING anything, I am merely putting things into context so we can focus on what matters.

Actually, a lot of discussions are about the frequency response of microphones, particularly how to tame the HF rise due to diaphragm resonance and diffraction, which is not such a big issue because it can easily be corrected electronically.

True. A while back I proposed a microphone design that was universally rejected. It would have been based on massive numbers of "digital" mems microphones (ones that have a basically flat response), suitable DSP that allows almost any pickup pattern to be created and wireless 32Bit/192kHz digital Output.

Naturally the DSP can simulate also the coarser frequency response variations and distortion of common recording Mic's.

The fact that a nuisance can be deemed negligible is not a good reason for not investigating it.

We should investigate enough to see where in the system overall set of sound quality impairments this effect ends up.

Thor

 

[thor.zmt]

Altec described it: "The backplate of the microphone gets it's polarization through the elevation of the cathode voltage ground potential".

This works because the grid of the Tube also includes an implicit diode. In the golden age of tubes, variations of grid leak biasing abounded. These days people go on endlessly about how it cannot work.

The understanding of how tubes work certainly has declined a lot.

Thor

 

[rock soderstrom]

That was my whole point - the Royer/Altec/C-37A doesn't use any gig-value resistors, and is waaay simpler.

I've built it with 5840, 6AU6 and 6AK5.

Just to add to the list, the Swedish SELA T12 does not have a grid resistor either. I just set it up and it works flawlessly with various 5654/6AK5 tubes.

This has also been discussed here on the forum, member ln76d could not tell any difference between with and without high impedance grid resistor.



Post in thread 'ST12 - 6AK5 - SELA T12 rip-off - SIMPLEST TUBE MICROPHONE IN THE WORLD, FULL PRO' ST12 - 6AK5 - SELA T12 rip-off - SIMPLEST TUBE MICROPHONE IN THE WORLD, FULL PRO

 

[thor.zmt]

The problem I have with it is that it is in complete contradiction with a fundamental in electronics, which stipulates that no active element in a circuit must be left without a galvanic connection to ground.

But it has, cathode to ground.

The grid actually emits electrons, the more the closer the grid-cathode voltage gets to 0V.

In effect this inherent mechanism keeps the grid negative relative to the cathode in the circuit where the grid terminates in a capacitor. The capacitance of the capsule acts to store the charge and provides a stable bias.

The cathode will operate with very high degeneration and will float up to a voltage that can be determined by tube curves, with ~ 0V G-K.

The circuit is essentially self regulating with the capsule providing storage for the tube bias voltage.

For oversimplification, imagine the tube has a GOhm resistor inherent between grid & cathode.

And I have never seen this mode of operation described in reference books.
I'll check RDH when I have time, maybe there is something in it.

Look under "contact bias".

I don't question this at all. What I question is how the operating point of the tube is predictable and stable.

As said, operation is based on ~0V G-K. Details in each and every tube datasheet.

Which I translate as "proper operation relies on unpublished and parameters that are not factory tested."

No it doesn't.

Thor

 

[abbey road d enfer]

Correct. But as said, knowing other distortion sources in the system, am I worried about 0.1% distortion on a 0.1% capacitance change for a microphone to record music with?

My point is that, in order to consider something is negligible, it must be quantified. It's the only way for giving an objective answer to a question.

A while back I proposed a microphone design that was universally rejected. It would have been based on massive numbers of "digital" mems microphones (ones that have a basically flat response), suitable DSP that allows almost any pickup pattern to be created and wireless 32Bit/192kHz digital Output.

Microphone arrays are not new, they are just getting easier to put in practice.
There are a few companies working on them, mainly for conference and communications applications, with much larger budgets than any audio specialist company. Apple, Microsoft, Samsung are in the game.

We should investigate enough to see where in the system overall set of sound quality impairments this effect ends up.

It's exactly why I suggest using a mathematical/physical model, in order to set the record straight, which is what I did when the question was asked to me a long time ago. I didn't want to believe blindly the Neumann litterature.
Now, figures don't tell the whole story. There are forms of distortion that are more pernicious than others. That's what subjective tests are for.

 

[abbey road d enfer]

This works because the grid of the Tube also includes an implicit diode. In the golden age of tubes, variations of grid leak biasing abounded. These days people go on endlessly about how it cannot work.

The understanding of how tubes work certainly has declined a lot.

Thor

I'm not privvy to the spellbooks and grimoires of the ancients.

 

[abbey road d enfer]

But it has, cathode to ground.

? K is grounded, for sure, but grid is not.

The grid actually emits electrons, the more the closer the grid-cathode voltage gets to 0V.

My basic and naive knowledge tells me that a continuous flow of electrons reaching an electrode would result in ever increasing negative voltage if there is no discharge path.
This type of stable operation indicates there is actually a discharge path.
Actually, my view on it is that once the grid has reached a negative potential high enough to reduce the flow of electrons to a trickle, grid current decreases to a point where leakage becomes sufficient to compensate the remaining flow of electrons hitting the grid. However I fail to model and quantize this discharge path. It is very hard for me to admit taht a whole industry relied on uncontrolled parameters.
Of course when the grid becomes negative enough it ceases to attract electrons. We have two concurrent phenomenons that stabilize the operation point.

In effect this inherent mechanism keeps the grid negative relative to the cathode

Of course, it's the basic principle of contact potential bias.

in the circuit where the grid terminates in a capacitor. The capacitance of the capsule acts to store the charge and provides a stable bias.

Without a discharge path, the voltage is continuous increasing. The Van de Graaf generator voltage would increase infinitely if there was no air ionization for discharge.

The cathode will operate with very high degeneration and will float up to a voltage that can be determined by tube curves, with ~ 0V G-K.

Curves are the results of experiments, they show observations, they don't explain why.
Considering the number of circuts that operate this way, I can just acknowledge it works.
I fail to see how the operation near zero Vgk helps understanding how grid voltage is regulated.

 

[abbey road d enfer]

Post in thread 'ST12 - 6AK5 - SELA T12 rip-off - SIMPLEST TUBE MICROPHONE IN THE WORLD, FULL PRO' ST12 - 6AK5 - SELA T12 rip-off - SIMPLEST TUBE MICROPHONE IN THE WORLD, FULL PRO

Starting at post #19, member ricardo asks a series of interesting questions..

 

[rock soderstrom]

Starting at post #19, member ricardo asks a series of interesting questions..

Yes, that's right. That's why I linked to it.

"It would be good if ln76d, or anyone with these mikes w/o '1G' grid resistors on a Tube mike, could test for 'pop overload recovery'."

No unusual behaviour so far.

"Now that us theoretical nerds have trashed it all out, 8) lets hear from those who have tried it in practice."

It works flawlessly so far.

But I understand very well that many very experienced engineers have problems with this kind of grid biasing. If I were doing this for a commercial project, I would install a high impedance grid resistor to ensure stability under all circumstances.

Interestingly, SELA has also done this with the successor microphone T24.

 

[thor.zmt]

? K is grounded, for sure, but grid is not.

Well, it is mechanically related with a fixed distance to cathode and there is a flow of electrons...

If the Grid is left open the Tube becomes a Diode and the grid potential will "float" to a voltage between cathode and Anode determined by the geometry.

As the grid is typically close to the electron emitting cathode and far from the electron attracting anode, it would be a few Volt above the cathode.

Now, let's connect the grid to a capacitor.

The grid potential will initially cause current to flow, the grid emits negatively charged big fat lazy green electrons.

These will charge the capacitor NEGATIVELY, with respect to the Grid's "open circuit" potential.

Eventually the grid will become sufficiently negative to pinch off the current in the Tube.

This means less electrons reach the grid and the charge in the capacitor falls allowing more current to flow and the capacitor potential will rise allowing more current.

Basically the circuit works as a result of the parasitic grid current and given non-gassy tubes will reach an equilibrium with the grid a few 100mV lower than the cathode and will remain there stable.

My basic and naive knowledge tells me that a continuous flow of electrons reaching an electrode would result in ever increasing negative voltage if there is no discharge path.

But there is a discharge path. Through the dielectrics.

Yes, this circuit fundamentally requires parasitics to work, it actually only works on parasitics, but it works fine and stable with good quality tubes (which makes it less ideal in the 21st century).

This type of stable operation indicates there is actually a discharge path.

Of course is. Air has finite resistance.

Actually, my view on it is that once the grid has reached a negative potential high enough to reduce the flow of electrons to a trickle, grid current decreases to a point where leakage becomes sufficient to compensate the remaining flow of electrons hitting the grid. However I fail to model and quantize this discharge path. It is very hard for me to admit taht a whole industry relied on uncontrolled parameters.

The parameter is NOT uncontrolled. It's quite simple, the tube works close to Vgc = 0V. In this region the grid emission will provide a small negative bias when the grid is connected to a capacitor. Note that the negative will be relative to the cathode, NOT ground.

Thus the cathode is pulled up by appx. the diode operation current and the grid is stabilised by the grid-emission at a few 100mV below the cathode voltage.

Without a discharge path, the voltage is continuous increasing. The Van de Graaf generator voltage would increase infinitely if there was no air ionization for discharge.

Incorrect, eventually the grid stops emitting enough electrons to overcome the losses in the dielectric and this happens before the Anode Current is pinched off completely.

Curves are the results of experiments, they show observations, they don't explain why.
Considering the number of circuts that operate this way, I can just acknowledge it works.

Of course it does.

I fail to see how the operation near zero Vgk helps understanding how grid voltage is regulated.

Because only near 0Vgc is there enough grid emission for the whole thing to work... Otherwise the Grid emits insufficient electrons to have this mechanism function.

And that is main parameter here, which IS CONTROLLED (by inherent design).

It does mean that circuit must be designed explicitly for a SPECIFIC tube and needs reasonably narrow tolerance, so it is not something to randomly "roll any tube that fits the base" into.

Thor

 

[thor.zmt]

Now back to my own line of reasoning.
As remarked, what interested me overall was to make a bunch of decent, cheap and easy recording Mic's both large and small diaphragm using the usual cheap chinese Mic's as Body donor and the readily available capsules, as alternative to buying the overpriced China made "recording microphones" available generally. And yes, easy obtainability and low cost of all parts is paramount to me.
The emphasis is on decent, cheap & easy. P48 only use was a given as well as the option (not shown) to use non-standard P68 to power the circuit giving 60V Capsule Bias.

Here something I would try out with SMD parts on veroboard:



Fairchild J201 remains in production as MMBFJ201 in SOT-23 SMT. Cascoding lowers HD. Using the J201 for the reverse biased diode should give some of the beast results.

As we need to buy reasonable quantities to get sensible pricing, just use it, rather than buying extra diodes or 1GOhm resistors.

The rest is pretty much the old Schoeps design, except optimised to using P48 as bias source.

For J201 with a very abnormal Idss R1 & R2 may need adjusting. They are meant to have ~ 4V across them.

Using the Phantom Power source as bias, the Capsule gets 40V, not 60V which reduces output levels from the capsule by around 3.5dB. Yes, it will also change the sound, but we are not using genuine AKG or Neuman capsules and we are not trying to make accurate copies of 2,500 USD Mic's, just something cheap, easy & functional.

With a 1G resistor we get 64dB unweighted SNR @ ~94dB SPL while with J201 as diode the sim predicts 74dB. Note that is unweighted. A-Weighting will give a lot lower noise. So using the Diode gives us more SNR back than the lower bias loses and we have a VERY SIMPLE microphone.

No oscillators etc. to make 60V (and track noise into the Audio circuit).

For ~134dB SPL (without the pad enabled) THD will be < 0.5% almost pure H2. With the pad enabled HD is reduced by ~ 12dB and of course, so is SNR.

The point here is that if want soft mic overload with loud instruments or vocals, leave the Pad off, otherwise turn loud down a trifle with the Pad to get a relatively clean signal.

Now I need to get all my boxes shipped here.

Thor

 

[abbey road d enfer]

Capacitance change is modest. If we have 60V bias, a 10% change in capacitance will give 6V out.

For 1% it would be 0.6V. And for 60mV peak (~40mV RMS) we get 0.1% capacity change.

Abstract from the Radiomuseum archive of the Hiller M59. (translated from german by Google trans).
"The measurement of the capsule capacitance (35pF) and thus the distance between the membrane and the counter-electrode is carried out in a high-frequency circuit according to Riegger. The deflection of the membrane at normal sound pressure is about 0.001 u, the resulting change in capacitance delta C is about 1.4pF."
4% !
I must say, I'm inclined to think you are correct. Whatever means "at normal sound pressure"?
Just shows there is a lot of info and counter-info. hard to sort the wheat from the chaff.
The M59 is interesting because its grid leak resistor is very uncommonly high 7Gohm.
Abstract from the BBC report/survey of M59 (again from teh Radiomuseum archive).:
"It will be noted that a 7000 megohm grid leak is used; this high value of resistance imposes very severe requirements on the capsule insulation and on the permissible grid current. The valve is of special design. Its electrodes are brought out to wire ends, with which contact is made by very small screw connectors; valve replacement therefore requires some skill."

 

[thor.zmt]

Abstract from the Radiomuseum archive of the Hiller M59. (translated from german by Google trans).
"The measurement of the capsule capacitance (35pF) and thus the distance between the membrane and the counter-electrode is carried out in a high-frequency circuit according to Riegger. The deflection of the membrane at normal sound pressure is about 0.001 u, the resulting change in capacitance delta C is about 1.4pF."
4% !
I must say, I'm inclined to think you are correct. Whatever means "at normal sound pressure"?
Just shows there is a lot of info and counter-info. hard to sort the wheat from the chaff.
The M59 is interesting because its grid leak resistor is very uncommonly high 7Gohm.
Abstract from the BBC report/survey of M59 (again from teh Radiomuseum archive).:
"It will be noted that a 7000 megohm grid leak is used; this high value of resistance imposes very severe requirements on the capsule insulation and on the permissible grid current. The valve is of special design. Its electrodes are brought out to wire ends, with which contact is made by very small screw connectors; valve replacement therefore requires some skill."

OK, logic dictates that capacitance change is more or less linear to SPL.

A 1% change of capacitance forces 1% of the stored charge or requires similar extra charge to be drawn.

If we have 60V bias voltage, when coupling into an infinite impedance the output should chabe by 1% of the bias voltage or 0.6V.

With finite grid resistors and capacitive loading this changes a little, but generally speaking, we will be on the reservation.

Thor

 

[abbey road d enfer]

With a 1G resistor we get 64dB unweighted SNR @ ~94dB SPL while with J201 as diode the sim predicts 74dB. Note that is unweighted. A-Weighting will give a lot lower noise.

The effects of weighting on white noise are documented and generally tend to produce a +/-10dB difference between CCIR (essentially unweighted) and "A" weighted.
The noise levels calculated or simulated are nothing without convoluting them with the head amp's response and the audition spectrum.
A LDC capsule has a higher KTC noise than a SDC, however, once noise is passed through the CR filter constituted by the capsule and grid/gate resistor/diode/whatever, the perceived noise level is lower, if only because the LF cut-off decreases at 6dB/octave and the KTC noise voltage at 3dB/octave.