(Updated schematic here)
I had the good fortune of scoring 10 of these devices in the initial feeding frenzy after someone kindly posted about the and their potential. I shared my bounty with others here on diyAudio and now I want to share an amp that I built with these devices.
I am calling it the Total Domination VFET! TDV for short. TDV means something in my line of work and IMO conveys the giant-killer properties of this device. It has the bass of a powerful solid state amp and the liquid highs that we come to expect from VFET (or DHTs).
Here's the schematic:
I had a lot of trouble controlling the bias on one of the two devices I started with. The VFET needs a Vgs of about -2.0V to -3.0V for between 1 to 2 amps of current. The inductor will provide some of it, which means the rest of it has to come from the bias supply. To provide -1.0V to -2.0V is very tricky. And it needs to be with a somewhat decent impedance. ZM has a very good solution for this problem but I wanted something simpler. The gate draws current and can get into a cycle of drawing current, increasing the bias, and thereby drawing more current.
I tried a LOT of things over the past couple of weeks and tried to make the bias control as simple as possible. This is the result of that exploration. A separate plus/minus supply is needed, but it can be very simple. Center-tapped trafo with a bridge rectifier and caps. From there, 7812 and 7912 regulators are used and they need small caps (10uF) on the input and output. The regulators are very important to maintain a stable voltage. The gate of this VFET draws a small amount of current and some of the parts can be very mischievous. The output of the regulators is connected via a 10k pot and you take the bias from the wiper. That's pretty much it. I ended up with something that is not as simple as I would've liked because it needs a separate supply and regulators, but it is not too bad IMO.
The rest of the amp is very simple. It is a source follower and needs voltage gain from the preamp. I found the sweet spot to be 1.4 amps. I'm sure there are other operating points, but I didn't want to go much past 1.75 amps given the inductor and heatsinking I had. Besides, there is plenty of power with 1.4 amps of bias.
Power up procedure: Bring the bias up very slowly. Use a Variac if you have one. First test the bias supply and set it up for an output of about -3.0V to the VFET gate. Then connect the VFET. With -3.0V on the gate, only a few milliamps will flow. As the gate bias is reduced (towards 0V), the current will increase. Aim for about 66% - 75% of the final bias current value. Then watch it carefully over half an hour or so. The current will drift upwards and should settle after about half an hour. That's the time to slowly increase it to the target. It might move around over the next hour. Check back and dial it in again. Then you should be set.
And that about does it. Note that this is my attempt at making the amp as simple as possible. You can make it even more simple by adding 1.0-1.5 ohms of resistance between the inductor and the source pin. Then the gate bias arrangement can be ditched altogether making it massively more simple. A 10k resistor from gate to ground is all that is needed (in my testing, the mischievous device ran away with anything higher than 10k, whereas the other one was fine even with 47k). The trade-off is that you cannot play around with the operating point and two devices will likely need different amounts of resistance to keep the bias current the same. I think it sounds better without the extra resistance.
I had the good fortune of scoring 10 of these devices in the initial feeding frenzy after someone kindly posted about the and their potential. I shared my bounty with others here on diyAudio and now I want to share an amp that I built with these devices.
I am calling it the Total Domination VFET! TDV for short. TDV means something in my line of work and IMO conveys the giant-killer properties of this device. It has the bass of a powerful solid state amp and the liquid highs that we come to expect from VFET (or DHTs).
Here's the schematic:
I had a lot of trouble controlling the bias on one of the two devices I started with. The VFET needs a Vgs of about -2.0V to -3.0V for between 1 to 2 amps of current. The inductor will provide some of it, which means the rest of it has to come from the bias supply. To provide -1.0V to -2.0V is very tricky. And it needs to be with a somewhat decent impedance. ZM has a very good solution for this problem but I wanted something simpler. The gate draws current and can get into a cycle of drawing current, increasing the bias, and thereby drawing more current.
I tried a LOT of things over the past couple of weeks and tried to make the bias control as simple as possible. This is the result of that exploration. A separate plus/minus supply is needed, but it can be very simple. Center-tapped trafo with a bridge rectifier and caps. From there, 7812 and 7912 regulators are used and they need small caps (10uF) on the input and output. The regulators are very important to maintain a stable voltage. The gate of this VFET draws a small amount of current and some of the parts can be very mischievous. The output of the regulators is connected via a 10k pot and you take the bias from the wiper. That's pretty much it. I ended up with something that is not as simple as I would've liked because it needs a separate supply and regulators, but it is not too bad IMO.
The rest of the amp is very simple. It is a source follower and needs voltage gain from the preamp. I found the sweet spot to be 1.4 amps. I'm sure there are other operating points, but I didn't want to go much past 1.75 amps given the inductor and heatsinking I had. Besides, there is plenty of power with 1.4 amps of bias.
Power up procedure: Bring the bias up very slowly. Use a Variac if you have one. First test the bias supply and set it up for an output of about -3.0V to the VFET gate. Then connect the VFET. With -3.0V on the gate, only a few milliamps will flow. As the gate bias is reduced (towards 0V), the current will increase. Aim for about 66% - 75% of the final bias current value. Then watch it carefully over half an hour or so. The current will drift upwards and should settle after about half an hour. That's the time to slowly increase it to the target. It might move around over the next hour. Check back and dial it in again. Then you should be set.
And that about does it. Note that this is my attempt at making the amp as simple as possible. You can make it even more simple by adding 1.0-1.5 ohms of resistance between the inductor and the source pin. Then the gate bias arrangement can be ditched altogether making it massively more simple. A 10k resistor from gate to ground is all that is needed (in my testing, the mischievous device ran away with anything higher than 10k, whereas the other one was fine even with 47k). The trade-off is that you cannot play around with the operating point and two devices will likely need different amounts of resistance to keep the bias current the same. I think it sounds better without the extra resistance.
Last edited:
Yes, it should work with other Tokin VFETs though I haven’t tried them yet.
I am exploring a version with negative supplies, which might simplify it further, we’ll see.
This effort is based on learning from Ben, ZM, and of course, Papa. I didn’t think the gate bias would be such a problem having worked with the Sony devices before. But it is quite a challenge and kind of fun to solve the puzzle! You want something that is reliable, repeatable, and something that sounds good.
I am exploring a version with negative supplies, which might simplify it further, we’ll see.
This effort is based on learning from Ben, ZM, and of course, Papa. I didn’t think the gate bias would be such a problem having worked with the Sony devices before. But it is quite a challenge and kind of fun to solve the puzzle! You want something that is reliable, repeatable, and something that sounds good.
Thanks for catching! Yes, there should be a big cap on the output. Will revise the schematic.
Good question. Tried that but there are two reasons it doesn’t work as well:
1. The lowest an LM337 will go is about -1.2V which will put an upper limit on the bias current. I never went below -1.2V on my devices but I got up to -1.3V at higher currents.
2. For the voltages we want on the gate (about -1.5V), we are operating very close to 0V and so the trimmer position is also closer to 0R with a single reg. The reg itself provides some resistance to ground but it is small. This sets the input impedance and I wanted at least close to 5k.
You could use a very big pot to get some resistance even at the -1.2V position, but then the current through the bias circuit becomes small and becomes comparable to the gate current, which is in the microamp region. Now what happens is when the gate current changes, it causes a change in the voltage across the trimmer causing the bias to slide. So, you also want a decent amount of current flowing through the bias circuit so that changes in gate current does not affect it.
Last edited:
well, I like things little more controlled
if nothing else, I would use buffered potential divider (pot), then properly dimensioned gate shunting resistor to bring biasing voltage to gate
small TO92 bjt is appropriate for buffer duty
attaching napkin sshot impossible, typing on dumbphone
if nothing else, I would use buffered potential divider (pot), then properly dimensioned gate shunting resistor to bring biasing voltage to gate
small TO92 bjt is appropriate for buffer duty
attaching napkin sshot impossible, typing on dumbphone
That should work. If you are going to make a Mouser/Digikey order, I'd also get some 0.47 ohm and 0.25 ohm 3W resistors. Just to try out the simple version.
I would like seeing that napkin drawing. I like the idea of a buffered bias supply.
Thanks for posting. Doesn’t look too complex. I tried something similar with FETs. Actually I tried many, many combinations with FETs but was not happy either with the complexity, stability, or sound.
I am experimenting with a scheme called back biasing using a single resistor between ground and the return leg of the last cap. Let’s see how it works.
I am experimenting with a scheme called back biasing using a single resistor between ground and the return leg of the last cap. Let’s see how it works.
Did anyone look at the input capacitance and gate current of the 2SK2087C? Would a buffer be a good idea to drive the 2SK2087C gate?
(for comparison, THF51 has approximately 1 nF and 10...100 nA into the gate)
(for comparison, THF51 has approximately 1 nF and 10...100 nA into the gate)
Back Biasing!
Below is a drawing of a couple of simple versions of the amp.
#1 is simply a bit of added resistance on the source pin to get the Vgs in the right zone. In the two devices I have, one needs about 1.0R whereas the other needs about 1.25R to get about 1.4 amps. This resistance is in addition to the choke, which has a DCR of about 0.68 ohms. The gate resistor is 10k or if that proves unstable, then go down until it stabilizes.
#2 is called "back biasing" where you obtain a negative voltage from the same supply and winding by inserting a resistance between the negative rectifier terminal and ground. I read about it in Blencow's book. The idea is as old as tubes and was used in old tube receivers to develop the negative bias needed for tubes. (ZM, you mentioned Papa did this at some point too. I have not seen it, can you share more?).
Both these arrangements are kind of self-regulating. More current leads to more negative Vgs, which reduces the current.
With this arrangement the choke is directly connected to the source. Not yet sure if that is better, but I find this very neat. I am listening to it now. No noise, no hum. Bias is very stable. It retains the thundering bass at least. More listening impressions later. Fun stuff!
Below is a drawing of a couple of simple versions of the amp.
#1 is simply a bit of added resistance on the source pin to get the Vgs in the right zone. In the two devices I have, one needs about 1.0R whereas the other needs about 1.25R to get about 1.4 amps. This resistance is in addition to the choke, which has a DCR of about 0.68 ohms. The gate resistor is 10k or if that proves unstable, then go down until it stabilizes.
#2 is called "back biasing" where you obtain a negative voltage from the same supply and winding by inserting a resistance between the negative rectifier terminal and ground. I read about it in Blencow's book. The idea is as old as tubes and was used in old tube receivers to develop the negative bias needed for tubes. (ZM, you mentioned Papa did this at some point too. I have not seen it, can you share more?).
Both these arrangements are kind of self-regulating. More current leads to more negative Vgs, which reduces the current.
With this arrangement the choke is directly connected to the source. Not yet sure if that is better, but I find this very neat. I am listening to it now. No noise, no hum. Bias is very stable. It retains the thundering bass at least. More listening impressions later. Fun stuff!
Yes, that works just the same, except now the resistance is in parallel with the 1/Gm of the VFET in terms of output impedance (depending on which one is the dominant one, it might be better this way). I did hear it briefly and it sounds perfectly good. In fact, I'm feeling like both simple #1 and #2 are better sounding than the one in the first post.
Edit: PRR, I see it now. Taking the output from the source directly might be better from an output impedance perspective and also getting some more power.
Edit: PRR, I see it now. Taking the output from the source directly might be better from an output impedance perspective and also getting some more power.
Last edited: