I think a tube front end is a great match for a VFET source follower.
If you prefer to keep things in the VFET family, a 2SK79 would be great for the front end / voltage gain (about 30x). These parts are very linear, and you can still find them without too much trouble.
If you prefer to keep things in the VFET family, a 2SK79 would be great for the front end / voltage gain (about 30x). These parts are very linear, and you can still find them without too much trouble.
I saw your curves for 2SK79 @mbrennwa , thanks for posting them, played with it for a bit in simulation (was able to find a Spice model). That's a pretty interesting part, look very linear. But I am more comfortable with designing tube gain stages, I have a good feel for what I think makes good sound with them, so I think that's the approach I will take. I might have the tube stage and VFET follower in separate chassis in case I ever want to change the front end rather than rolling the whole thing into an integrated design.
Has anyone tried a diode in the source circuit?
- My thoughts went there because I have a SIT with gate current and everytime I start it up without Vb (24V) then the gate current is higher afterwards. So there are some 'holes' being made because of current flowing from source to the V- gate bias (-7.5V in my case).
- Soooo, I thought, try a diode there to stop that flow. In simulation there is no impact (in some a slight increase in 3rd ahrmonic, up from none . . .). I have 8 Amp Cree SIC or 10 amp traditional. Both have non-neglicient resistance; the Cree is very very lineair, so favored.
- it will give a nice bias reference, two might even be good enough for some SITs, so there you have it: CREE BIAS, from triodes to SITs... Very handsome for the THF-51S: two diodes needed.
Last edited:
if gate demands current, without even analyzing how you wanna achieve that - if you stop current, you can't control it anymore
point is - in case of leaking current - to arrange circuit being able to give it, while staying firm voltage wise
so, source follower in front of hungry gate is perfect, in most cases
point is - in case of leaking current - to arrange circuit being able to give it, while staying firm voltage wise
so, source follower in front of hungry gate is perfect, in most cases
Here is an interesting follower with proper biasing:
https://www.diyaudio.com/community/...loaded-follower-amplifier.377830/post-6806645
https://www.diyaudio.com/community/...loaded-follower-amplifier.377830/post-6806645
I should test it with a very sensitive ammeter (I dont have that) but if the drain has zero voltage with Vg=-7V, there is a strange behaviour. The SIT after such a state will have more gate current! The bias will increase, I have to dial in a slightly larger Vg. Where does the new current settling come frome? I call it a hole. Stupid of course, but in the die, there is a physical complex behaviour.I've used LEDs for cathode bias of small signal triodes. However I may be missing something but I don't see how a diode at the source of a SIT would stop gate leakage current as the leakage is internal.
Ben, you have a SIT with gate current too, I remember. Have you seen the behaviour of that 2SK182 SIT change? Any ideas where that comes from?
Yes, one of my 2SK180 had very high gate leakage current. That was what drove me to try it in common drain. By putting the choke at the source, its 1 Ohm resistance provided bias in addition to the bias supply at the gate. It helped a bit but it did not alleviate the issue.
The biggest issue with high gate leakage was that it increased as the device got hot. That then caused the bias to drop and further increased the current and heat, and caused the gate leakage to increase more, and it would just keep on snowballing.
The gate leakage current affected the bias voltage through the 10K resistor from SIT gate to bias supply and ground. The gate leakage current goes to ground through the 10K resistor, causing a voltage drop across it, and drops the bias voltage by that amount. So one way to reduce the effect of the gate leakage current is to reduce the value of the resistor. I did that - I reduced the 10K to 6K - and that helped. Ultimately I got tired of constantly monitoring Vgs and Iq for runaway that I replaced the 2SK180.
The good fix is what Zen Mod did (in his Singing Bush amp). He got rid of the resistor from gate to ground that combined with the gate leakage current to reduce bias. He added a buffer in front of the SIT, and moved the bias supply to the input of the buffer to bias the input signal. The coupling cap between the buffer and SIT is also eliminated so that the bias voltage of the signal then biases the SIT.
Now with the SIT as common drain, a large voltage swing is required for the signal so the standard lower voltage JFET buffer that would work for common source would not work for common drain. A higher voltage buffer would be required. If I were to incorporate Zen Mod's scheme for common drain, I would try the DN2540 for the buffer. That is the buffer that Michael Rothacher used in his Luminaria preamp design that I have built.
The biggest issue with high gate leakage was that it increased as the device got hot. That then caused the bias to drop and further increased the current and heat, and caused the gate leakage to increase more, and it would just keep on snowballing.
The gate leakage current affected the bias voltage through the 10K resistor from SIT gate to bias supply and ground. The gate leakage current goes to ground through the 10K resistor, causing a voltage drop across it, and drops the bias voltage by that amount. So one way to reduce the effect of the gate leakage current is to reduce the value of the resistor. I did that - I reduced the 10K to 6K - and that helped. Ultimately I got tired of constantly monitoring Vgs and Iq for runaway that I replaced the 2SK180.
The good fix is what Zen Mod did (in his Singing Bush amp). He got rid of the resistor from gate to ground that combined with the gate leakage current to reduce bias. He added a buffer in front of the SIT, and moved the bias supply to the input of the buffer to bias the input signal. The coupling cap between the buffer and SIT is also eliminated so that the bias voltage of the signal then biases the SIT.
Now with the SIT as common drain, a large voltage swing is required for the signal so the standard lower voltage JFET buffer that would work for common source would not work for common drain. A higher voltage buffer would be required. If I were to incorporate Zen Mod's scheme for common drain, I would try the DN2540 for the buffer. That is the buffer that Michael Rothacher used in his Luminaria preamp design that I have built.
Last edited:
The gate current depends on temperature. If the SIT heats up, the gate current will be higher. Is this what you are seeing?I should test it with a very sensitive ammeter (I dont have that) but if the drain has zero voltage with Vg=-7V, there is a strange behaviour. The SIT after such a state will have more gate current! The bias will increase, I have to dial in a slightly larger Vg. Where does the new current settling come frome? I call it a hole. Stupid of course, but in the die, there is a physical complex behaviour.
I wouldn't call this a cascode. Rather, the buffer is loaded by a constant current sink (CCS), which helps with the linearity of the buffer. I did pretty much the same the DN2540 parts in the SuSy T. PuckFo, except that I used cascoded CCS (no pun intended, but...).Little steps for me. I understand your Singing Bush arrangement. I need to study your cascoded buffer and follower output.
I did a quick LTSpice model with the Singing Bush type of input bias: View attachment 1025881
I propose that you stop messing around and get that follower tested and get something done in lifeI drew ( purely) cascoded JFet buffer in my follower proposal, but still didn't found time to populate already present pcbs and do all necessary measurements
- Home
- Amplifiers
- Pass Labs
- 25W Single Ended Hammond 193V Choke Loaded 2SK180 L'Amp