Replacing GOhm resistors with diodes, the theory behind it

[abbey road d enfer]

Just to make things a tad clearer, what do you mean by "bias resistor"? A resistor inserted between source and ground? Or one connected to a negative source?

 

[MicUlli]

In this case the bias resistor would be inserted between JFET source and the node where the anode of the diode is connected. It should provide Vgs going down, but it does NOT because the diode is reverse biased.

 

[Admiral-dk]

The reverse Biased Diode is kind of the secret here, as the very tiny Leakage current through it, equals a Resistor in the Giga-Tera -Ohm area.
That said - I'm not sure that ALL JFets are equaly good in this application .... especially if their own Gate-Source-Diode has too Big a leakage current ...!

Per

 

[thor.zmt]

Adding a bias resistor requires an external diode biased in flow direction

Maybe we use the term bias resistor differently. I should have used source resistor.

And no, the source resistor does not need a forward biased diode external to the Fet. It works by raising the voltage on the source terminal above the reference point of the gate resistor (here a diode) and thus making the gate more negative, pinching off the current through the J-Fet.

The reverse biased diode is best understood as a very high value, but finite resistance resistor that sinks the gate current, which is not quite zero, just very low. Well, as long as it is reverse biased.

To forward bias it needs the gate around 0.5V lower than the reference point for the diode.

Thor

 

[abbey road d enfer]

In this case the bias resistor would be inserted between JFET source and the node where the anode of the diode is connected. It should provide Vgs going down, but it does NOT because the diode is reverse biased.

Very confusing. The "node where the anode of the diode is connected is the FET's source, so the resistor would be in parallels with the diode...? In that case the resistor should be about a Gohm, then what the diode would be good for?

 

[abbey road d enfer]

Maybe we use the term bias resistor differently. I should have used source resistor.

That makes sense.

And no, the source resistor does not need a forward biased diode external to the Fet. It works by raising the voltage on the source terminal above the reference point of the gate resistor (here a diode) and thus making the gate more negative, pinching off the current through the J-Fet.

This is the basic principle of source bias, isn't it? Otherwise known as automatic bias by the ancients.

 

[abbey road d enfer]

Lots of confusion here by using words where pictures (schemo) would speak more eloquently.

 

[thor.zmt]

Lots of confusion here by using words where pictures (schemo) would speak more eloquently.

Note, the J113 suggests a lower SNR by a few dB.

Much will depend on the specific circuit intended. What Ulli shows in his "best J-Fet" thread is a rather different (and arguably better) circuit than what I have in mind.

One last point. J-Fet's inherently designed for microphones have actuallynthis schematic:



As noted before, only the reverse diode is needed.

Thor

 

[rankot]

As i mentioned in the paper: BAV199 from Infineon or NXP. BAV199 from other manufacturers have a much higher reverse current. Maybe there are alternatives, not tested yet. Small parallel capacitance is also important, BAV199 shows approx. 4 pF.

Then BAT41 seems to be a good alternative for TTH, with only 2pF capacitance? Or BAV17-BAV21, they are listed as 1.5pF, BAV70 as well.

 

[abbey road d enfer]

View attachment 105554

That is much clearer, and probably very different than what Ulli had in mind.

One last point. J-Fet's inherently designed for microphones have actuallynthis schematic:

View attachment 105555

Indeed, that's the diagram published by e.g. Sanyo for 2SK1578. But are all the other usual suspects similar?
My understanding is that the resistor in this equivalent diagram is actually a reverse-polarized diode, which is represented as a resistor, because that's what it is functionally, and the other is represented as a diode because it is there for protection.
I have serious doubts about Gohm resistors in IC fab.

As noted before, only the reverse diode is needed.

Agreed.

 

[terry setter]

I'm not a transistor guy, I only design and build tube mics, but the idea of using a diode pair to achieve extremely high resistance with lower noise is very interesting to me. I know that the diode pair must be "bootstrapped" in order to work as a high value resistor. Can anyone describe a non SMT circuit where a bootstrapped diode pair could be used to replace the grid resistor in a typical plate-loaded circuit (given the availability of 120-200V B+ and 5-6.3V heater supply)?

It will be interesting to see if the lower noise of a diode pair is, overall, quieter than what the increase in shot noise would add. Usually, increasing the value of grid reistors in tube mic circuits reaches its maximum benefit somewhere around 3-4 Gigs, then the shot noise becomes problematic.

By the way, I did extensive tests to locate diodes to use in biasing the tubes and found that the UF4007 and Cree CSD01060A Schottky diodes were extremely quiet in a signal path, far more so than most others and way beyond LEDs. The UF4007 is rated at 10 micro-amps and the CSD01060A is rated at 100 micro-amps.

Thanks in advance for any help you can provide.

 

[thor.zmt]

My understanding is that the resistor in this equivalent diagram is actually a reverse-polarized diode, which is represented as a resistor, because that's what it is functionally,

Incorrect. What is drawn is a reverse biased diode including a resistor representing the very high but finite parasitic parallel resistance of the diode.

If wanted to be complete, we'd add a capacitor in parallel as well and a series resistor.

I am also aware of examples that show no resistor and antiparallel diodes. This represents the J-Fets parasitic forward PN junction as separate diode that in reality doesn't exist as a separate part.

and the other is represented as a diode because it is there for protection.

Nope, that is not a protection diode, it is our "Gigaohm" resistor.

I have serious doubts about Gohm resistors in IC fab.

Why? You use a reverse biased diode, voila, you got one, which suffices for the job.

Thor

 

[thor.zmt]

Can anyone describe a non SMT circuit where a bootstrapped diode pair could be used to replace the grid resistor in a typical plate-loaded circuit (given the availability of 120-200V B+ and 5-6.3V heater supply)?

Use a very low leakage current diode, go for something in a glass case. You should only use one reverse biased diode, not an antiparallel pair.

The 1N4148 is rated as 4pF reverse capacitance and 20nA reverse current at 20V reverse voltage. That is 1GOhm.

So we are barely ok. Look for less reverse current (and less reverse capacitance if possible.

A J113 J-Fet is rated for 1nA gate current (and 5pF). Diode connected it would give around 15GOhm. Lower current and Transconductance J-Fets will likely be higher effective resistance.

J201 is rated at 100pA/20V, so that is 200GOhm. It starts to look interesting me say. The datasheet actually shows 4pA @ 5V for 25 degrees, which is 1.25TOhm and 1pF.

Thor

 

[abbey road d enfer]

Incorrect. What is drawn is a reverse biased diode including a resistor representing the very high but finite parasitic parallel resistance of the diode.

OK, thanks for correcting me.

Why? You use a reverse biased diode, voila, you got one, which suffices for the job.

OK, it's adequate for some jobs but its significant non-linearity may preclude its use for other jobs. The reverse saturation current being essentially constant vs. voltage, the equivalent resistance is variable.

 

[abbey road d enfer]

I know that the diode pair must be "bootstrapped" in order to work as a high value resistor.

There is no need for bootstrapping. High ohmic value of reverse bias is an intrinsic characteristic of a PN junction.

It will be interesting to see if the lower noise of a diode pair is, overall, quieter than what the increase in shot noise would add.

Can you be more specific here?
"lower noise of a diode pair" compared to what?
"increase in shot noise" resulting from what?

Usually, increasing the value of grid reistors in tube mic circuits reaches its maximum benefit somewhere around 3-4 Gigs, then the shot noise becomes problematic.

What type of tubes can operate with such high valu grid resistors? You typical 12AX7 or EF86 gets crazy with these values. Not to mention leakage from coupling caps.

By the way, I did extensive tests to locate diodes to use in biasing the tubes

Biasing how? You mean as a replacement for the grid resistor?

 

[thor.zmt]

OK, thanks for correcting me.

OK, it's adequate for some jobs but its significant non-linearity may preclude its use for other jobs. The reverse saturation current being essentially constant vs. voltage, the equivalent resistance is variable.

Absolutely, but in a Capacitor Microphone meant to be semi-mass produced?

As long as we are mainly dealing with signals in the mV range, dirty silicone makes great high value resistors.

Bootstrapping the dirty silicone deals with the level limitations.

Thor

 

[abbey road d enfer]

Absolutely, but in a Capacitor Microphone meant to be semi-mass produced?

Indeed, when quality can be compromised.

As long as we are mainly dealing with signals in the mV range, dirty silicone makes great high value resistors.

I have no doubt they are useful in the particular case of loading capacitive sources.
Now an hypothetic (but believable) rev biased diode resulting in 1Gohm and 1pF has a -3dB point at 160Hz, which precludes it being used e.g. as the feedback resistor of an opamp.

Bootstrapping the dirty silicone deals with the level limitations.

Boostrapping is not always an option. Some (many) designs result (and require) full swing across a resistor. Now, do they need Gohm resistors to start with?
Bootstrapping allows creating enormous virual impedance, even with standard monolithic resistors.

 

[thor.zmt]

I have no doubt they are useful in the particular case of loading capacitive sources.

Yup.

Now an hypothetic (but believable) rev biased diode resulting in 1Gohm and 1pF has a -3dB point at 160Hz, which precludes it being used e.g. as the feedback resistor of an opamp.

Depends.

May I offer a J-Fet Op-Amp, using the dirty silicone to make the DC connection and using coupling capacitors to the AC loop using conventional value resistors?

In this case the dirty silicone is the (DC) feedback resistor.

Boostrapping is not always an option.

All engineering is a compromise.

Some (many) designs result (and require) full swing across a resistor. Now, do they need Gohm resistors to start with?

Sure. Hey, I'm Kool with anything that works.

Bootstrapping allows creating enormous virual impedance, even with standard monolithic resistors.

Indeed. Virtual components. I find few people think this out to the logical conclusion... But that, as they say, is another debate for another time.

Thor

 

[abbey road d enfer]

May I offer a J-Fet Op-Amp, using the dirty silicone to make the DC connection and using coupling capacitors to the AC loop using conventional value resistors?

In this case the dirty silicone is the (DC) feedback resistor.

Of course. Distortion at DC is not a real subject.

 

[ccaudle]

The reverse saturation current being essentially constant vs. voltage, the equivalent resistance is variable.

Would simulation likely do a good job analyzing the in circuit behavior?

I have not quite figured out how to simulate the capsule behavior (changing capacitance with amplitude) from the previous THD discussion, but it seems like these two phenomena of changing capsule capacitance and changing effective resistance could interact. I am curious to know if the magnitude approaches anything like audible. If the resulting distortion is 130dB below signal levels it becomes just an academic curiosity.