Built a JFET Source follower with JFET current sink load microphone

[pasarski]

I recently stumbled on this article.

https://sound-au.com/articles/jfet-design.htm#s7

I decided to try the JFET Source follower with JFET current sink load (Fig. 7.4) in a microphone as the author says it's a good choice for a high impedance source. I don't know if this has ever been used in a commercial microphone or not.

(Following is based on simulations and I'm by no means a expert in that, actually i just learned how to use LTspice, so take it with a grain of salt)

If you are wondering about the low value output coupling cap, it's just an experiment. It makes a quite substantial bump at about 100hz and rolls out the lows below that. The idea was to reduce plosives while not making the mic sound thin. The bump is dependent of the inductance of the trafo primary (I don't know the spec for a trafo I used so I just listened and chose a value that sounded best to me). You could use larger cap (also for C2) for a full bandwidth mic. How large, is again dependent of the primary inductance. Larger value actually gives a slight low end boost.

The microphone has crazy amount of headroom for a simple circuit. The downside is that it's quite low gain, in a SM7 ballpark. I used a Chinese T-8 trafo which is 7:1 ratio. I will change it to Neutrik NTE4 which is 4:1 (inverted) which would suffice in this circuit afaik (have to play again with the output cap after that). I think I gain about 5db more that way? I probably will also try it with a DC-DC converter at some point but it's really not needed for close miked vocals with a decent low noise preamp.

The other downside is that the JFET's and the source resistors have to be closely matched (how closely I don't know). You could see it as an upside also, it's not easy to mass produce.

I used a B1 body and capsule. Some could consider it too bright without de-emphasis, but it's ok for me for now. You can add an de-emphasis network or use a flatter capsule.

It's an ugly b*****d inside. I should learn to make neat layouts and keep the component leads straight, so sorry for the graphic material

 

[k brown]

I've used this in microphones with great success; simply replacing R183/283 with 1G (and appropriately smaller cap in front of it):

 

[pasarski]

I've used this in microphones with great success; simply replacing R183/283 with 1G (and appropriately smaller cap in front of it):

Looks similar but without the source resistors, biasing done differently if I understand it correctly. You also probably have changed the 25k pot for a high value resistor?

 

[kingkorg]

I recently stumbled on this article.

https://sound-au.com/articles/jfet-design.htm#s7

I decided to try the JFET Source follower with JFET current sink load (Fig. 7.4) in a microphone as the author says it's a good choice for a high impedance source. I don't know if this has ever been used in a commercial microphone or not.

(Following is based on simulations and I'm by no means a expert in that, actually i just learned how to use LTspice, so take it with a grain of salt)

If you are wondering about the low value output coupling cap, it's just an experiment. It makes a quite substantial bump at about 100hz and rolls out the lows below that. The idea was to reduce plosives while not making the mic sound thin. The bump is dependent of the inductance of the trafo primary (I don't know the spec for a trafo I used so I just listened and chose a value that sounded best to me). You could use larger cap (also for C2) for a full bandwidth mic. How large, is again dependent of the primary inductance. Larger value actually gives a slight low end boost.

The microphone has crazy amount of headroom for a simple circuit. The downside is that it's quite low gain, in a SM7 ballpark. I used a Chinese T-8 trafo which is 7:1 ratio. I will change it to Neutrik NTE4 which is 4:1 (inverted) which would suffice in this circuit afaik (have to play again with the output cap after that). I think I gain about 5db more that way? I probably will also try it with a DC-DC converter at some point but it's really not needed for close miked vocals with a decent low noise preamp.

The other downside is that the JFET's and the source resistors have to be closely matched (how closely I don't know). You could see it as an upside also, it's not easy to mass produce.

I used a B1 body and capsule. Some could consider it too bright without de-emphasis, but it's ok for me for now. You can add an de-emphasis network or use a flatter capsule.

It's an ugly b*****d inside. I should learn to make neat layouts and keep the component leads straight, so sorry for the graphic material

Very interesting! Since it's source follower i presume you can go with 1:1 transformer.

 

[Voyager10]

The DC operating point of the circuit depends critically on the JFET matching; unfortunately LTSpice will give you identical FETs so it's much better in theory than practice.

You can avoid this by connecting gate resistor R2 to a fixed voltage (say 2-3V) from a simple resistor voltage divider, instead of to R1. You then don't need R1 (connect J1 source directly to J2 drain).

The FET drain current is then set by J2/R3 alone, and the DC level at the output will be your 2-3V gate voltage, plus J1's Vgs value for that drain current.

(The "2-3V" value is arbitrary, it's a balance between having enough output headroom, and reducing the capsule polarization voltage).

 

[pasarski]

Very interesting! Since it's source follower i presume you can go with 1:1 transformer.

I tried to simulate the output impedance and came to a conclusion that you need about 4:1 ratio transformer for about 200ohm output impedance. Not sure if that's correct. The impedance rises towards lower frequencies drastically btw but that doesn't seem to be a problem.

 

[pasarski]

The DC operating point of the circuit depends critically on the JFET matching; unfortunately LTSpice will give you identical FETs so it's much better in theory than practice.

You can avoid this by connecting gate resistor R2 to a fixed voltage (say 2-3V) from a simple resistor voltage divider, instead of to R1. You then don't need R1 (connect J1 source directly to J2 drain).

The FET drain current is then set by J2/R3 alone, and the DC level at the output will be your 2-3V gate voltage, plus J1's Vgs value for that drain current.

(The "2-3V" value is arbitrary, it's a balance between having enough output headroom, and reducing the capsule polarization voltage).

Great info. Will definitely try it.

I did match the JFETs by IDSS as close as I could from the ones I had. What is the main symptom if they are not matched well enough?

 

[Voyager10]

At DC, the R1 and R3 currents are identical, so the Vgs for J1 and J2 are the same. If the FETs are identical, they have the same Id, and the same Vgs, and all is well. If not ...?

The only other variable we have in play here is the variance of Id with Vds. J1 and J2 will settle on (different) Vds values so that their Id/Vgs's match. The Vds's determine the output DC level, and also the capsule voltage.

 

[k brown]

Looks similar but without the source resistors, biasing done differently if I understand it correctly. You also probably have changed the 25k pot for a high value resistor?

Well, since it's adapted as a mic FET, none of the parts before the input caps (which are changed to 1000pf) are used, and the outputs parts are changed to those typical of impedence-balanced mics: 68 uF > 47 ohm > pin 2; same parts between ground and pin 3. Plus RF shunt caps between pins 2,3 and ground.

The 18v is supplied by batteries; can be up to 24v.

 

[k brown]

The DC operating point of the circuit depends critically on the JFET matching; unfortunately LTSpice will give you identical FETs so it's much better in theory than practice.

You can avoid this by connecting gate resistor R2 to a fixed voltage (say 2-3V) from a simple resistor voltage divider, instead of to R1. You then don't need R1 (connect J1 source directly to J2 drain).

The FET drain current is then set by J2/R3 alone, and the DC level at the output will be your 2-3V gate voltage, plus J1's Vgs value for that drain current.

(The "2-3V" value is arbitrary, it's a balance between having enough output headroom, and reducing the capsule polarization voltage).

You just basically described the circuit in post #2.

 

[pasarski]

You just basically described the circuit in post #2.

Ok I try to understand this with my parcellar knowledge of electricity basics..

I understand the Voyagers circuit and it's easy to modify my circuit to it. I don't understand your version so easily. The other circuit leaves the R3 in place for biasing but your circuit doesn't. So the stuff at the bottom right must be there to generate a negative voltage to the gate? Also, what do you mean by "appropriately smaller cap in front of it"?

 

[pasarski]

The DC operating point of the circuit depends critically on the JFET matching; unfortunately LTSpice will give you identical FETs so it's much better in theory than practice.

You can avoid this by connecting gate resistor R2 to a fixed voltage (say 2-3V) from a simple resistor voltage divider, instead of to R1. You then don't need R1 (connect J1 source directly to J2 drain).

The FET drain current is then set by J2/R3 alone, and the DC level at the output will be your 2-3V gate voltage, plus J1's Vgs value for that drain current.

(The "2-3V" value is arbitrary, it's a balance between having enough output headroom, and reducing the capsule polarization voltage).

Got it working in a simulation. The headroom is not as crazy (still huge). No matter how I try to bias it it seems to clip the positive halve of the signal much earlier than the other circuit starts to clip (which it does quite symmetrically). There's still plenty enough headroom for a condenser mic though, and it actually seems to be more linear at the levels a condenser capsule is capable of delivering.

 

[k brown]

Ok I try to understand this with my parcellar knowledge of electricity basics..

I understand the Voyagers circuit and it's easy to modify my circuit to it. I don't understand your version so easily. The other circuit leaves the R3 in place for biasing but your circuit doesn't. So the stuff at the bottom right must be there to generate a negative voltage to the gate? Also, what do you mean by "appropriately smaller cap in front of it"?

Provides voltage to the gate, yes - but it's not negative.

Circuit was designed by Nelson Pass for use as a HiFi 'buffered passive preamp'. Here's the original article:
https://www.passdiy.com/pdf/B1 Buffer Preamp.pdf

I used it to make extremely compact 'remote head' mics, with captive cables and separate battery box. All that's in the mics heads are the two FETs, 820-1000pf cap, 1G gate resistor, and the voltage divider parts..

 

[Voyager10]

If you omit R3 (i.e. set it to zero), the drain current for both FETs will be the Idss value of J2. The 2N3819 will be in the 10mA range, which is going to be too much for a phantom-powered circuit.

Regarding clipping, looks like we could do with a capacitor from J1's drain to ground, otherwise the voltage drop across R4 will increase dramatically for positive output current swings.

 

[k brown]

If you omit R3 (i.e. set it to zero), the drain current for both FETs will be the Idss value of J2. The 2N3819 will be in the 10mA range, which is going to be too much for a phantom-powered circuit.

Regarding clipping, looks like we could do with a capacitor from J1's drain to ground, otherwise the voltage drop across R4 will increase dramatically for positive output current swings.

That seems odd - the 2SK170BL circuit I posted only draws 0.01 amp; 3819 really that diferent?

 

[Khron]

the 2SK170BL circuit I posted only draws 0.01 amp

Isn't that 10mA?

 

[k brown]

Isn't that 10mA?

Right you are!

Egg firmly on face.

_____

Since I always run these mics on rechargable batteries, it never occurred to me that phantom power might not supply sufficient current.

 

[k brown]

Incidentally, it was only after using the B1 Buffer circuit with the Primo EM23 that I realized just how amazing they are. Putting them on a Crown CM-700 was also an eye-opener.

Considering that all my examples are pushing 50 years old, and that they are alum-sputter electrets, they hold their own with far more pricey mics. I bought about half a dozen taken off Teac and Tascam
mics and cherry picked the two best for a matched pair.

I never had any of the EM21 cardioids that I cared much for; I imagine that the foam rings that provide to proper rear venting do not age well, and the plastic retaining rings that apply the correct pressure on them are no longer doing so. Very curious if current production are using a different, longer-lasting material.

But those omnis punch way above their weight.

 

[hifizen]

Reminds me a little of the Audio-Technica mic pre front end, except they add a 3rd JFET, "Erno Borbely" style, as a self-bootstrapped cascode on the follower FET to neutralize it's gate-drain capacitance...

I don't know if Mr. Borbely came up with the arrangement (probably not), he's just stuck in my mind as one of the most prolific advocates of it.

From Mr. Borbely's famous article "JFETs The New Frontiers Part 2" (a balanced pair version of this cascode arrangement) ... 2SK389 is basically a dual packaged 2SK170. Slice this circuit in half and put that B1 buffer's JFET current source in place of the J511 CRD shown, and you have something pretty close to the A-T front end:

 

[hifizen]

..oh, except the A-T circuit just uses the source follower output, not a gain resistor at the drain. And back-to-back diodes to bias up the gate from a separate voltage divider with AC bootstrap from a push-pull bipolar output buffer. I should just open up one of my mics and trace out the circuit.