Good point about biasing. I wasn't very clear. I was talking more generally about controlling the bias current, Id, through the 2sk2087c. I am seeing that it tends to run away, slowly but surely, at least on one of the devices I am testing (the other device is stable). Talking strictly about a single stage source follower with a choke. In this case, the gate bias voltage only controls the current through the device. I know that Ben faced something similar with his 2sk180 amp. So far, I've not found a reliable AND simple solution that would let me set a bias point and move away and not worry about it. Experiments continue. One sure way to control the current is to put additional resistance on the source pin beyond what is contributed by the choke. This results in more or less stable bias, but it wastes some voltage and power.
Yep, ZM has already done it. Bias the signal in front of an input buffer with no coupling cap between the buffer and the SIT, and no gate leak resistor at the SIT.
The regular circuit with a bias voltage at the SIT gate running through a gate leak resistor, the gate leak resistor is the issue. Leakage current from the SIT causes a voltage drop across the resistor which reduces the bias voltage at the gate. As the SIT heats up, the leakage current increases, the bias voltage at the gate drops, the drain current increases some more, and it snow balls. What I have done for devices that had higher leakage was to decrease the value of the gate leak resistor, say from 10k to 5k or 6k. If the leakage current is too high, then it might not work, but it is worth a try.
The regular circuit with a bias voltage at the SIT gate running through a gate leak resistor, the gate leak resistor is the issue. Leakage current from the SIT causes a voltage drop across the resistor which reduces the bias voltage at the gate. As the SIT heats up, the leakage current increases, the bias voltage at the gate drops, the drain current increases some more, and it snow balls. What I have done for devices that had higher leakage was to decrease the value of the gate leak resistor, say from 10k to 5k or 6k. If the leakage current is too high, then it might not work, but it is worth a try.
Yup, using a buffer without cap to the gate is a good way to do it, but I don't want a buffer. Want something simpler than that.
I started with just a voltage divider and that works about as good as a CCS with an LND150 to develop a stable voltage reference. The problem in both cases is the gate current, exactly as you described Ben. I had tried with a grid leak of 10k and that helps matters. In fact, it is totally stable for one of the 2sk2087s, but the other one creeps up slowly, and though it eventually stabilizes, it is not a repeatable solution.
So, we are back to the drawing board. I am considering using an optocoupler like Papa's designs. Stick a 0.47 ohm resistor on the drain (doesn't impact the SF, except a bit of voltage loss), use that to sample the current, and then connect the other side to the gate. Some sort of feedback mechanism is needed. This will count as "complex" but at least it won't count as another stage. I know what ZM is going say, just use the buffer. But hey, it is fun to explore.
edit: it is pretty solid with added resistance to the source and nothing on the gate. But then it is not easy to try various operating points.
I started with just a voltage divider and that works about as good as a CCS with an LND150 to develop a stable voltage reference. The problem in both cases is the gate current, exactly as you described Ben. I had tried with a grid leak of 10k and that helps matters. In fact, it is totally stable for one of the 2sk2087s, but the other one creeps up slowly, and though it eventually stabilizes, it is not a repeatable solution.
So, we are back to the drawing board. I am considering using an optocoupler like Papa's designs. Stick a 0.47 ohm resistor on the drain (doesn't impact the SF, except a bit of voltage loss), use that to sample the current, and then connect the other side to the gate. Some sort of feedback mechanism is needed. This will count as "complex" but at least it won't count as another stage. I know what ZM is going say, just use the buffer. But hey, it is fun to explore.
edit: it is pretty solid with added resistance to the source and nothing on the gate. But then it is not easy to try various operating points.
Last edited:
well, whatever you do, no way that you're getting even close in number to my Audiophoolery endeavors
so, feel free to cut short whatever you want, be it worthwhile or not
me, being a cheapskate, always trying that each stage or module is serving more than one duty .... this buffer not being any exception
-small input cap, due to high Rin of buffer
-stiff current path for SIT gate
-there is certainly more, but sometimes I'm clever, sometimes not in a mood to be the same
I assume this could be a good candidate for a THF-51 SE amp that follows NP's paper from BAF2022?
It has nearly only 2nd and 3rd harmonic at 1W out it seems.
Any listening experience?
I am on the list for 2 x VFET chassis for a THF-51 mono block build.
This design uses 2 x PCB. One for the "bias circuit" and one for the SIT as far I could see. Having a stereo chassis, a PCB on each heatsink could be a possibility?
It has nearly only 2nd and 3rd harmonic at 1W out it seems.
Any listening experience?
I am on the list for 2 x VFET chassis for a THF-51 mono block build.
This design uses 2 x PCB. One for the "bias circuit" and one for the SIT as far I could see. Having a stereo chassis, a PCB on each heatsink could be a possibility?
Yea, i noticed that me too. 😐
Zen Mod - Would it be reasonably doable to substitute the IXTN with THF51 or 2SK182E?
Since i have got 2 of each of them mighty fine Tokins 🙂👍
And. The amps would most certain look extra beautiful with twin bonus Tokins😎👍👍👍
There are a lot of IXTN N-channel devices for sale, but we need a P-channel for this design?
To use a THF-51, 2SK etc. the current design needs a bit of change, I think?
NP has in his PDF made also a design where a N-channel mosfet is used in the bias circuit.
Let us see if ZM finds a substitute else I may go with Mouser and wait for next spring
To use a THF-51, 2SK etc. the current design needs a bit of change, I think?
NP has in his PDF made also a design where a N-channel mosfet is used in the bias circuit.
Let us see if ZM finds a substitute else I may go with Mouser and wait for next spring