Replacing GOhm resistors with diodes, the theory behind it

[ccaudle]

That's true for the signal. But input noise created by the JFET will be amplified a little bit more. In reality it does not play an important role. Find attached a measurement for comparison...

That is interesting. Does the 33pF in the graph name refer to a 33pF capacitor in the circuit to simulate a small-ish microphone capsule?

Relevant to this topic, I have always had a difficult time building my intuition for how noise contributions work with a capacitive source. In the whitepaper there is the statement "Rg has to be as high as possible" but I do not know how to reconcile that with the fact that noise voltage in a resistor is proportional to the resistance. Doesn't making Rg as high as possible then make the thermal noise voltage of the resistor as high as possible?
This forum doesn't seem to support any math markup in BB Code, but something like En=sqrt(4*Boltzmann*tempK*bandwidth*resistance)

Is the case that the current noise dominates, so the lower current noise of a high value resistor is more important than the increased voltage noise?
Would this be the correct way to calculate total noise?
Vtotal=sqrt(Vresnoise^2 + (inoise*Xc)^2)

If that is correct, it would make sense why the noise spectral density rises at low frequencies, since Xc is increasing continuously with decreasing frequency.

 

[abbey road d enfer]

Remember that the calculated noise voltages (or currents) do not tell the whole story.
The noise energy is determined by the capsule's capacitance only. Google "KTC noise".
The perceived noise level results from the convolution of the noise spectrum with the audition spectrum.
The capsule's capacitance forms a low-pass filter with the gate (or grid) resistor. Increasing the resistor shifts the spectrum towards VLF, where most of it becomes inaudible.
Basically, doubling the gate resistor results in a 6dB improvement of perceived noise.
Bootstrapping doesn't change a thing in this respect, because the part that creates noise is the physical resistor, not a virtual or reflected one.

 

[abbey road d enfer]

I do not know how to reconcile that with the fact that noise voltage in a resistor is proportional to the resistance. Doesn't making Rg as high as possible then make the thermal noise voltage of the resistor as high as possible?

I hope the preceding post answers your question.

 

[Admiral-dk]

Thank you for setting the Record Straight Abbey Road
I reconized some of the formulaes from my youth (I haven't used them since). So the very impressive improvement is down to the two BAT199 in Anti-Parallel, actually works as a kind of Resistor two Magnitudes higher in resistance then the 1G Ohm ~ 100G Ohm ->
Very close to the Teoretical minimum Noise for this Circuit

 

[kingkorg]

Thank you for setting the Record Straight Abbey Road
I reconized some of the formulaes from my youth (I haven't used them since). So the very impressive improvement is down to the two BAT199 in Anti-Parallel, actually works as a kind of Resistor two Magnitudes higher in resistance then the 1G Ohm ~ 100G Ohm ->
Very close to the Teoretical minimum Noise for this Circuit

Polarizing capsule at higher voltage then traditional 40-60v can make significant improvement as well. However, there is a point where all of this stops making sense. I polarized once a capsule of NT1 (which is a dead quiet mic to start with) with 100v. The experiment stopped making sense, as the ambient noise even in a very quiet treated room was exceeding any electronic noise.

 

[ccaudle]

I hope the preceding post answers your question.

Thanks, it did, and also gave me a chance to review the wikipedia entry on Johnson-Nyquist noise. I had always skimmed over the case of capacitor dominated noise because I rarely needed to work with the details, but the article does give a nicely concise summary of how the kTC formula is derived.

 

[Admiral-dk]

Tim Campbell told me that increasing the Polarizing Voltage, also reduces the Frequency responce + changes the sensitivity of the capsule.
This makes sense, as the increased Voltage also increase the pull towards the opsite polarity (no matter what that is - it's the difference that counts) .... Actually bending the membrane.
We came across this when debating a FM Microphone - where it isn't the change in charge but in the capacitance, that gives the Modulation => signal out .... and why that is to be prefered to the traditional application.

 

[Mr. Moscode]

I would always recommend the connection between JFET gate and GOhm resistor (or diode) and capsule wire soldered in free air (no contact on pcb). A datasheet from TI (LMC6001) gives useful hints, see attachment

Not only in mics but in tube gear where you have megohm Rs. I had an amp in for service where the bias and offset was drifting around and took a lot of time to trace leakage to the board only. It was an amp owned by someone who lives by the ocean in a salt air environment and the salt deposits became resistors. You could measure leakage resistance values in less than megohms. A trace can provide a lot of contact area for the salt deposits. I cleaned the board in soap and water, dried with a heat gun or oven and sprayed the board with a polyurethane sealer coat. No more drift.

 

[abbey road d enfer]

The experiment stopped making sense, as the ambient noise even in a very quiet treated room was exceeding any electronic noise.

There is a minimum to the noise even in a super dead and super isolated room. It's the thermal noise of air particles hitting the diaphragm.
The only solution is to cool the ambient or decrease the atmospheric pressure.
Both options are usually not welcome by talents.
The equivalent noise level depends on the acoustic resistance of the diaphragm.
I asked one of my partners who's a Doctor in acoustics the formula for calculating it; he couldn't.
I know there's a paper from the Philips tech bulletin taht deals with it but could never find it.

 

[MicUlli]

There is a minimum to the noise even in a super dead and super isolated room. It's the thermal noise of air particles hitting the diaphragm.
The only solution is to cool the ambient or decrease the atmospheric pressure.
Both options are usually not welcome by talents.
The equivalent noise level depends on the acoustic resistance of the diaphragm.
I asked one of my partners who's a Doctor in acoustics the formula for calculating it; he couldn't.
I know there's a paper from the Philips tech bulletin taht deals with it but could never find it.

The min. acoustic noise in a super isolated room was measured by Microsoft (astonishing..) at a level of -20,6 dB SPL A-wheighted. The brownian motion of air particles at 20 °C and 1013 hPa was calculated to be -23,0 dB SPL A-wheighted.
Compare this to quoted noise levels of LDCs (+4 dB SPL A ..+20 dB SPL A) or SDCs (+10 dB SPL A .. +26 dB SPL A).
Obviously condenser mics are at least 27 dB noisier than they could be in theory
Yes, one main noise source in condenser mic capsules is the acoustic damping of the membrane and cavities. It can be calculated, but it is not easily done..
The other main noise source is electronics. It can be also calculated but measurements rule. In order to minimise self noise electronics should be tuned appropriately..

 

[abbey road d enfer]

The min. acoustic noise in a super isolated room was measured by Microsoft (astonishing..) at a level of -20,6 dB SPL A-wheighted. The brownian motion of air particles at 20 °C and 1013 hPa was calculated to be -23,0 dB SPL A-wheighted.

Thanks for the info. I'll try to find links to this.

Obviously condenser mics are at least 27 dB noisier than they could be in theory

Actually there are some mics that have a better noise performance, in the vicinity of -6dBspl. They use very little damping and electronic compensation.
Some people have tried to use them for music recording and reported bad results...

 

[rogs]

Tim Campbell told me that increasing the Polarizing Voltage, also reduces the Frequency responce + changes the sensitivity of the capsule.
This makes sense, as the increased Voltage also increase the pull towards the opsite polarity (no matter what that is - it's the difference that counts) .... Actually bending the membrane.
We came across this when debating a FM Microphone - where it isn't the change in charge but in the capacitance, that gives the Modulation => signal out .... and why that is to be prefered to the traditional application.

I am certainly of the opinion that using the same capsule with both a 60V DC potential applied (as in THIS application) and as part of a amplitude modulated RF bias mic ( as in THIS application ) it does seem to sound 'different', the latter approach apparently providing the 'better' results?
Sadly, I do not have the equipment - or probably the technical expertise! - to be able to measure this difference objectively, but it certainly seems to be audibly evident .
Understandably perhaps - the capsule is obviously less physically 'stressed' with only few volts of RF volts applied...

The relative noise levels created by the two variations is an entirely different story -- the theory behind the noise produced by the RF configuration is way above my 'pay grade' !

 

[KA-Electonics.com]

This forum doesn't seem to support any math markup in BB Code, but something like En=sqrt(4*Boltzmann*tempK*bandwidth*resistance)

Alt+251 = √
Alt+253 = ²

More codes here: Special Characters — Alt Keyboard Sequences
And here: Code Chart (ALT and HTML) for Greek Letters & Symbols
Some common ones we use:

 

[rock soderstrom]

Alt+251 = √
Alt+253 = ²

More codes here: Special Characters — Alt Keyboard Sequences
And here: Code Chart (ALT and HTML) for Greek Letters & Symbols
Some common ones we use:
View attachment 105455

Thanks, very useful!

 

[thor.zmt]

Hello All,
after finishing some experiments successfully i want to share the results. I replaced a 1GOhm resistor by a double diode and found the noise floor getting down considerably. Find attached a document i created to explain how it works..

May I suggest that you only need one Diode?

The J-Fet itself has a PN junction between Source & Gate that acts as the second diode.

It reduces capacitance and leakage and increases impedance more.


This shows a basic Shoeps style Frontend with Diode and bias for J-Fet via reverse biased diode and Mic Bias from Phantom power (40V instead of 60V).

Sim suggests 111fA and 111mV for 1TOhm which seems to match the Datasheet for BAV45.
SNR is unweighted and re 10mV (balanced out). With 79dB re. 10mV (1Pa 0r 94dB) the unweighted self noise would be 15dB, I suspect A-Weighting will take off at least 6dB.

It would be followed by now noise P-MOS Fets in my case, which allow very low value coupling capacitors, allowing film or C0G ceramic types to be used. Here 10Hz would be around -0.5dB.
My Idea was to make some good basic Mic's without excessive complications (hence bias from Phantom Power).

For improved dynamic range the whole circuit was designed to be able to use non-standard P68 to give 60V capsule Bias.

Sadly everything was shelved as I had to relocate rapidly and the parts etc. are in storage in another country, though they may eventually come here and let me continue.

Thor

 

[Gus]

J201?
Real ones often have an IDSS of about .6mA IIRC.
electret capsule JFET operation?

 

[thor.zmt]

J201?
Real ones often have an IDSS of about .6mA IIRC.
electret capsule JFET operation?

No, I have some of these "magic" J-Fet's the video mentions. But I have no spice model for them and the diode is internal.

Thor

 

[MicUlli]

May I suggest that you only need one Diode?

The J-Fet itself has a PN junction between Source & Gate that acts as the second diode.

It reduces capacitance and leakage and increases impedance more.

View attachment 105470
This shows a basic Shoeps style Frontend with Diode and bias for J-Fet via reverse biased diode and Mic Bias from Phantom power (40V instead of 60V).

Sim suggests 111fA and 111mV for 1TOhm which seems to match the Datasheet for BAV45.
SNR is unweighted and re 10mV (balanced out). With 79dB re. 10mV (1Pa 0r 94dB) the unweighted self noise would be 15dB, I suspect A-Weighting will take off at least 6dB.

It would be followed by now noise P-MOS Fets in my case, which allow very low value coupling capacitors, allowing film or C0G ceramic types to be used. Here 10Hz would be around -0.5dB.
My Idea was to make some good basic Mic's without excessive complications (hence bias from Phantom Power).

For improved dynamic range the whole circuit was designed to be able to use non-standard P68 to give 60V capsule Bias.

Sadly everything was shelved as I had to relocate rapidly and the parts etc. are in storage in another country, though they may eventually come here and let me continue.

Thor

Of course the JFET has an intrinsic diode, but if you make use of it you must live with VGS at several mV. This forces IDSS (f.e. 20mA for J113) flowing. Not a good idea.
I want to make a hint to a new thread called JFET shootout starting in 5 minutes...

 

[thor.zmt]

Of course the JFET has an intrinsic diode, but if you make use of it you must live with VGS at several mV. This forces IDSS (f.e. 20mA for J113) flowing. Not a good idea.
I want to make a hint to a new thread called JFET shootout starting in 5 minutes...

You can add a bias resistor to run at less than IDSS and still only use one diode.

So what is your beef?

Thor

 

[MicUlli]

"My beef is" that your circuit shows the external diode reverse biased. When i add a bias resistor it changes nothing..
Adding a bias resistor requires an external diode biased in flow direction