Giaime,
The FET follower has certain technical advantages in a general sense, but my thought is, why go to all the trouble of building a tube amp when you would then use a FET in the signal path? You'd be able to hear a difference between the FET follower and a CF, I'd wager. I'm not saying which one you might prefer. It's one thing to use solid state CCSs and SS power supply support, but this FET is in the direct "line-of-fire". You certainly don't need amps of output current (you'd melt the grid of the 6L6s). You also don't need one ohm of driving impedance. You should be able to get a low enough output Z with the right tube choice for a CF and proper set up. And whatever output Z you get with a CF would be in the range of a grid stopper value for the 6L6 anyway. Yes, the PS voltage may have to change, but you've already got 350, 400, and volts 450 available. Save the IRF820s; you might be able to use them to make CCSs in the CF cathodes circuits!
I seem to be almost alone in blowing this trumpet: with the FET follower, you're loading the plates of the 6SN7 with the FET’s variable capacitance from drain to gate (Cdg), which creates non-linearities that are “out-of-character” for an otherwise promising tube amp. See the excerpt from the IRF820 data sheet. Even though the Cgs is bootstrapped, it still may be a factor too.
Just one man’s opinion, of course
The FET follower has certain technical advantages in a general sense, but my thought is, why go to all the trouble of building a tube amp when you would then use a FET in the signal path? You'd be able to hear a difference between the FET follower and a CF, I'd wager. I'm not saying which one you might prefer. It's one thing to use solid state CCSs and SS power supply support, but this FET is in the direct "line-of-fire". You certainly don't need amps of output current (you'd melt the grid of the 6L6s). You also don't need one ohm of driving impedance. You should be able to get a low enough output Z with the right tube choice for a CF and proper set up. And whatever output Z you get with a CF would be in the range of a grid stopper value for the 6L6 anyway. Yes, the PS voltage may have to change, but you've already got 350, 400, and volts 450 available. Save the IRF820s; you might be able to use them to make CCSs in the CF cathodes circuits!
I seem to be almost alone in blowing this trumpet: with the FET follower, you're loading the plates of the 6SN7 with the FET’s variable capacitance from drain to gate (Cdg), which creates non-linearities that are “out-of-character” for an otherwise promising tube amp. See the excerpt from the IRF820 data sheet. Even though the Cgs is bootstrapped, it still may be a factor too.
Just one man’s opinion, of course
Attachments
Nothing seems to provoke strong opinions as much as sticking a piece of sand directly in a critical part of the signal path.
I agree that a mosfet is overkill to drive a 6L6 type tube. I experimented with mosfet drive while developing my SimpleSE amp. The additional complexity required for a PowerDrive circuit did not buy enough sound improvement to justify its use.
It is a different story when you are driving a large DHT (845 211 833A) well into A2. When I first developed this circuit I was experimenting with an 845 in A2. I tried driver transformers, cathode followers made with triodes and pentodes, bipolar transistors, IGBT's and mosfets. The mosfet wins both measurement wise and sonically.
Not all mosfets are created equally. I tested all of the high voltage mosfets that were available from DigiKey at the time. The Toshiba 2SK2700 was the best. It also has the lowest transfer capacitance and total gate charge. It outperformed all of the IR fets in HF response and sound quality. There may be some new fets that are beter, but I haven't tested any recently.
The secret to avoiding voltage related non linearities is to run a lot of voltage across the fet. This is where the capacitance curves are almost flat. I use 400 volts in my 845 amp. I also noticed that the sound quality improves with fet current up to aboud 10 mA, so I run the fet at 20 mA. This obviously requires a big heat sink.
When I first posted the PowerDrive circuit on my web site, I was criticized by a lot of people (who never tried it) even to the point of being called transistorlab! Then the e-mail started arriving from people who tried PowerDrive. In the time since I published the circuit ALL of the e-mail from people who tried it was positive.
Like any circuit, it has its place. I wouldn't use it just anywhere, although I have never seen it cause sonic harm. It works well with most DHT's. I saw no improvement with 45's and 2A3's, but works well with anything bigger (including 300B's). Where it really shines is in applications where grid current is drawn. The grid impedance of a DHT will drop severely when it goes positive. Many DHT's remain linear in this region if the driver does not distort in this region. I measured grid current over 100 mA on signal peaks in an 833A tube. The mosfet is also indespendible in screen drive circuits.
I agree that a mosfet is overkill to drive a 6L6 type tube. I experimented with mosfet drive while developing my SimpleSE amp. The additional complexity required for a PowerDrive circuit did not buy enough sound improvement to justify its use.
It is a different story when you are driving a large DHT (845 211 833A) well into A2. When I first developed this circuit I was experimenting with an 845 in A2. I tried driver transformers, cathode followers made with triodes and pentodes, bipolar transistors, IGBT's and mosfets. The mosfet wins both measurement wise and sonically.
Not all mosfets are created equally. I tested all of the high voltage mosfets that were available from DigiKey at the time. The Toshiba 2SK2700 was the best. It also has the lowest transfer capacitance and total gate charge. It outperformed all of the IR fets in HF response and sound quality. There may be some new fets that are beter, but I haven't tested any recently.
The secret to avoiding voltage related non linearities is to run a lot of voltage across the fet. This is where the capacitance curves are almost flat. I use 400 volts in my 845 amp. I also noticed that the sound quality improves with fet current up to aboud 10 mA, so I run the fet at 20 mA. This obviously requires a big heat sink.
When I first posted the PowerDrive circuit on my web site, I was criticized by a lot of people (who never tried it) even to the point of being called transistorlab! Then the e-mail started arriving from people who tried PowerDrive. In the time since I published the circuit ALL of the e-mail from people who tried it was positive.
Like any circuit, it has its place. I wouldn't use it just anywhere, although I have never seen it cause sonic harm. It works well with most DHT's. I saw no improvement with 45's and 2A3's, but works well with anything bigger (including 300B's). Where it really shines is in applications where grid current is drawn. The grid impedance of a DHT will drop severely when it goes positive. Many DHT's remain linear in this region if the driver does not distort in this region. I measured grid current over 100 mA on signal peaks in an 833A tube. The mosfet is also indespendible in screen drive circuits.
George,
You may have found a part that works well, as long as it is biased with a very high Vds. I took a snapshot of the capacitance curves from the 2SK2700 data sheet (a part I haven't used). They're typically hideous below 100 volts, but data for above 100 volts are MIA. Yes, as a general rule FET capacitance curves do flatten at higher Vds values. Do you have data for Vds > 100 volts? You say that at 400 volts "the capacitance curves are almost flat". In my experience, even minute amounts of capacitance modulation can give a character to the sound. Capacitance modulation gives rise to phase intermodulation, which can be thought of as jitter in the time domain. We now appreciate the extreme sensitivity to the ear to digital jitter - even 100s of picoseconds matter. Testing for minute levels of analog phase intermodulation is very difficult when you also have amplitude intermodulation present simultaneously. So conventional measurements are out (I’m working on a PM measuring scheme, time permitting). Of course, even a tube CF can have RC time “constant” modulation, not so much from capacitance change (very little change, but not zero), but more from the variable R of the driving plate. But the tube’s capacitance is also typically smaller in absolute magnitude, reducing absolute and variable phase shift in the audio band with the same driving resistance. Personally, I would not be inclined to add a FET to a tube circuit in the absence of a strong reason to avoid a tube. But I respect your experience here, and would like to know more.
You may have found a part that works well, as long as it is biased with a very high Vds. I took a snapshot of the capacitance curves from the 2SK2700 data sheet (a part I haven't used). They're typically hideous below 100 volts, but data for above 100 volts are MIA. Yes, as a general rule FET capacitance curves do flatten at higher Vds values. Do you have data for Vds > 100 volts? You say that at 400 volts "the capacitance curves are almost flat". In my experience, even minute amounts of capacitance modulation can give a character to the sound. Capacitance modulation gives rise to phase intermodulation, which can be thought of as jitter in the time domain. We now appreciate the extreme sensitivity to the ear to digital jitter - even 100s of picoseconds matter. Testing for minute levels of analog phase intermodulation is very difficult when you also have amplitude intermodulation present simultaneously. So conventional measurements are out (I’m working on a PM measuring scheme, time permitting). Of course, even a tube CF can have RC time “constant” modulation, not so much from capacitance change (very little change, but not zero), but more from the variable R of the driving plate. But the tube’s capacitance is also typically smaller in absolute magnitude, reducing absolute and variable phase shift in the audio band with the same driving resistance. Personally, I would not be inclined to add a FET to a tube circuit in the absence of a strong reason to avoid a tube. But I respect your experience here, and would like to know more.
Attachments
I have no data on the 2Sk2700 beyond what is in the data sheet. When I was designing the PowerDrive circuit for the 845SE I noticed that the capacitance curves for most fets flatten out as the Vds is increased. This was confirmed by discussions with a few of the CMOS IC designers at work.
When I had the circuit breadboarded on the bench I was using variable power supplies for the positive and negative fet voltages. I noticed that as the negative supply was increased (more neative) the sound quality (and distortion spectra) did not change after the voltage was sufficient to assure current flow through the fet at the most negative grid voltage excursion.
I did notice a change in sound quality and a slight improvement in the higher order harmonics (already at nearly the noise level) by turning up the positive supply voltage. An 845 is probably worse case due to the large voltage swing required at the grid. I tried bipolar transistors in this application but they sounded noticibly worse even though the capacitance issue should not be as much of a problem. It is also hard to find BJT's with good SOA capability at this voltage level. The driver tube was a 45 running at 25 mA (CCS loaded) so there was sufficient current and slew rate capability to drive the fet or the bipolar transistor.
I do agree that the variable capacitance can cause phase distortion and this could be audible. This may have been what I was hearing when I was playing with the power supply knob. The effect was most noticible on HF sounds. like cymbals (especially ones with loose rivets in them). The kinds of sounds that are absoulutely destroyed by MP3's.
I find that the abrupt change in grid impedance caused by entering the positive grid region will cause a much more objectional distortion (sounds like crossover) that requires a very low impedance driver, at least while sourcing the grid current. If you think about it, the mosfet driver (or any follower) circuit is asymetrical. The current that it can source is limited by the ON resistance of the active device and its power supply. The current it can sink is controlled by the source (cathode) resistor.
In this case we are trading one distortion for another, which is usually the case in any amplifier design. I believe that a mosfet can be a useful driver in places where grid current is drawn, and yes you have to use a very small grid stopper with ferrite beads to avoid yet another nonlinearity.
At the time that I discovered PowerDrive I had tried most of the common types of driver circuits. When I found the PowerDrive circuit I stopped looking at the driver and went on to build the amp. I have never looked back, and I am probably guilty of overusing the circuit, but the amps that I have put it in have been well received as "punchy and dynamic". It could be that they are just colored in a slighthy different way when compared to the average 300B amp.
I tried the PowerDrive circuit on 6L6 type tubes, and I didn't find much change in a normal music situation. It did improve the overload characteristics when I ran my guitar through the amp and pushed it into hard clipping. I did however get an e-mail from a user who said that it made his KT-88 (HiFi) amp come to life. I have no idea what the operating conditions were. Most happy users were running transmitting tubes, particularly the 811A.
If I ever find the time I could think of several other driver topologies to try:
Cathode follower using a high perveance sweep tube that is capable of sourcing high current peaks.
Emitter folower using a HV darlington pair.
Mosfet follower with newer low capactance mosfets.
Some hybrid combination of the above.
If you ever work out a way to measure Phase Intermodulation Distortion, it could be another tool to further refine the driver (and other) circuit for better performance.
When I had the circuit breadboarded on the bench I was using variable power supplies for the positive and negative fet voltages. I noticed that as the negative supply was increased (more neative) the sound quality (and distortion spectra) did not change after the voltage was sufficient to assure current flow through the fet at the most negative grid voltage excursion.
I did notice a change in sound quality and a slight improvement in the higher order harmonics (already at nearly the noise level) by turning up the positive supply voltage. An 845 is probably worse case due to the large voltage swing required at the grid. I tried bipolar transistors in this application but they sounded noticibly worse even though the capacitance issue should not be as much of a problem. It is also hard to find BJT's with good SOA capability at this voltage level. The driver tube was a 45 running at 25 mA (CCS loaded) so there was sufficient current and slew rate capability to drive the fet or the bipolar transistor.
I do agree that the variable capacitance can cause phase distortion and this could be audible. This may have been what I was hearing when I was playing with the power supply knob. The effect was most noticible on HF sounds. like cymbals (especially ones with loose rivets in them). The kinds of sounds that are absoulutely destroyed by MP3's.
I find that the abrupt change in grid impedance caused by entering the positive grid region will cause a much more objectional distortion (sounds like crossover) that requires a very low impedance driver, at least while sourcing the grid current. If you think about it, the mosfet driver (or any follower) circuit is asymetrical. The current that it can source is limited by the ON resistance of the active device and its power supply. The current it can sink is controlled by the source (cathode) resistor.
In this case we are trading one distortion for another, which is usually the case in any amplifier design. I believe that a mosfet can be a useful driver in places where grid current is drawn, and yes you have to use a very small grid stopper with ferrite beads to avoid yet another nonlinearity.
At the time that I discovered PowerDrive I had tried most of the common types of driver circuits. When I found the PowerDrive circuit I stopped looking at the driver and went on to build the amp. I have never looked back, and I am probably guilty of overusing the circuit, but the amps that I have put it in have been well received as "punchy and dynamic". It could be that they are just colored in a slighthy different way when compared to the average 300B amp.
I tried the PowerDrive circuit on 6L6 type tubes, and I didn't find much change in a normal music situation. It did improve the overload characteristics when I ran my guitar through the amp and pushed it into hard clipping. I did however get an e-mail from a user who said that it made his KT-88 (HiFi) amp come to life. I have no idea what the operating conditions were. Most happy users were running transmitting tubes, particularly the 811A.
If I ever find the time I could think of several other driver topologies to try:
Cathode follower using a high perveance sweep tube that is capable of sourcing high current peaks.
Emitter folower using a HV darlington pair.
Mosfet follower with newer low capactance mosfets.
Some hybrid combination of the above.
If you ever work out a way to measure Phase Intermodulation Distortion, it could be another tool to further refine the driver (and other) circuit for better performance.
tubelab.com said:
BJTs won't work right in such an application since the 845 grid presents a high impedance when it is not drawing current, but a much lower impedance when it does. The input impedance of an Emitter Follower is dependent on the load impedance. Having that widely varying load impedance magnified by a factor of (1 + h(fe) ) isn't a good thing.
Parametric frequency multiplication is a bytch, and makes for problems in RF circuits as well. Any solid state RF oscillator requires a bandpass filter to clean up the mess. If you don't use one in xcvr circuits, that's your problem; in xmtr circuits, it becomes everyone's problem. Some solid state designs are so horrible that the resulting low level noise can raise the noise floor over entire ham bands.
An externally hosted image should be here but it was not working when we last tested it.
Now, I don't always agree with everything Doug Self says, but this one is not one of those times:
For audio useage, "one ill-concieved package" is certainly an accurate description. It is very bad, and if used as a follower will magnify the varying impedance of grid current/no grid current by an even bigger factor. I wouldn't waste my time fooling with one of those for any audio project where fidelity was a design criterion.
Going back to my amp, the schematic remained the same, but I modified the first stage employing an ECC88 as a LTP, thus reducing gain and gNFB (now 6dB), because it's just enought, and less gNFB = less problems.
Now about 40W with 1% distortion, without stressing the 6L6 (I go in AB2 but limiting only +10Vgk).
Now about 40W with 1% distortion, without stressing the 6L6 (I go in AB2 but limiting only +10Vgk).
40 watts in AB2 is not a problem for a good pair of 6L6GC's. 1% might be, unless good (accurately balanced, symmetrical) transformers are used.
The current flowing from C to E is the base current times beta, so the input (base) impedance change is the change at the emitter reduced (divided) by beta. That is the reason for trying a darlington connected pair (beta in the 10000 range). I don't like the "ill conceived package" parts either.
I better stop talking like this before I get called transistorlab again!
I design solid state transmitters for a living, so I know about this. All I can say is remember the old illegal CB amplifiers the tube and solid state versions created enough spectrum polution to render TV sets unwatchable for miles. We now call that spectrum polution BPL.
tubelab.com said:
Keep talking like that, and you won't have to worry about that ever again. That is just plain wrong. If you have an emitter current of 1.0mA, then your r(e)= 26.0R. Let's say that the emitter resistor (unbypassed) is: 1K0. So what's the input impedance? It's: (1K0 + 26)(1 + h(fe) ) assuming h(fe)= 100, then Zi= (1026)(101)= 103K63 (of course, this doesn't take into consideration the value of the base bias resistor(s) which appear in parallel with that base impedance). It most certainly is not: 1026/101= 10R16! If that were the case, BJTs would be all but unusable. Those low input impedances are what you see when using grounded base amps, and you can get away with it since you can impedance match when making RF voltage amps. Of course, impedance step up means voltage step down, so you lose a lot of potential stage gain there.
I hope they let you do it right, and not use bad PLLs. Even some pricey rigs have PLLs with way too much phase noise.
I work on TX's that operate in the public safety bands. Good phase noise is required by law (and enforced). The limiting factor is the phase noise requirements needed for 70 db of adjacent channel rejection on the RX. The TX (same PLL) is by design very good.
Now that I reread what I wrote I realize how stupid it sounds. I should have used the word current instead of impedance. Other than that, I believe we are saying the same thing except I am thinking about it from a current perspective. I am also ignoring the lesser terms for simplicity.
The base current is beta times smaller than the emitter current, so the (tube) driver sees a smaller current variation due to the abrupt increase in grid current. If beta was infinite the base would draw no current (neglecting the bias resistors) and the driver tube would never see the grid current change.
If the output tube grid goes from zero (no current) to 100 mA (measured with an 833A) the base current will go from zero to 1 mA (neglecting bias and load resistors and assuming a beta of 100). If beta is increased to 10000 (darlington connected pair of hfe =100), the resulting base current will go from zero to 10 uA. If the driver tube is cranking along at 25 mA, the 10 uA will not be noticed.
Yes, the impedance changes over a large ohmic range, but if the absolute input impedance of the darlington is high enough it will be swamped by the plate resistor of the driver tube.
In spite of this confusion, the real issue is finding BJT's that have a suitable SOA to live long in this application. Since the capacitance in no longer an issue, I may be able to turn down the voltage to allow some transistors to work. This is a moot point, since I don't have the time to try any of this any time soon.
George and Miles,
As a former RF design electrical engineer, and a ham (although not a very active one), my background sounds similar to both of you. My feeling is that audio can benefit from RF thinking and practices. In particular, noise floor modulation mechanisms, and phase and frequency modulation considerations can cross-over into audio design. I believe we could increase the correlation between listening results and testing results by considering these and other “RF” issues, not to mention designing better sounding amps. I wish I had more time to contribute in this area.
One more note: the BJT, while not having the dramatic capacitance problems of MOSFETS, do have variable capacitances that are pertinent to audio. The base-to-emitter capacitance can be enormous as well as emitter-current-variable, although it is rarely considered because it is shunted by the relatively small base-emitter resistance (a short TC) and it is bootstrapped by the emitter which follows the base to a great (although non-linear) degree. The real culprit typically is the capacitance of the reverse-biased collector-base diode which significantly modulates with Vcb changes - think of it as a varactor. It may have a smaller magnitude compared to a comparably rated MOSFET, but it’s no less ugly. The Miller effect amplifies it, and its effect on the amp is dependent upon driving resistance, of course.
I have said before that I think the whole area of capacitance modulation is the biggest unheralded difference between the sound of tubes and transistors. If I can find the time to finish, I plan to publish on these topics, along with test results and measurement techniques. We own several businesses that I thought would be fun early retirement diversions, but they have turned out to be 24/7 time consumers (I keep telling myself that “busy” is what I wanted for these businesses, but now I have to live with that).
Good discussion guys. Giaime, are you convinced yet to try a tube buffer instead?
As a former RF design electrical engineer, and a ham (although not a very active one), my background sounds similar to both of you. My feeling is that audio can benefit from RF thinking and practices. In particular, noise floor modulation mechanisms, and phase and frequency modulation considerations can cross-over into audio design. I believe we could increase the correlation between listening results and testing results by considering these and other “RF” issues, not to mention designing better sounding amps. I wish I had more time to contribute in this area.
One more note: the BJT, while not having the dramatic capacitance problems of MOSFETS, do have variable capacitances that are pertinent to audio. The base-to-emitter capacitance can be enormous as well as emitter-current-variable, although it is rarely considered because it is shunted by the relatively small base-emitter resistance (a short TC) and it is bootstrapped by the emitter which follows the base to a great (although non-linear) degree. The real culprit typically is the capacitance of the reverse-biased collector-base diode which significantly modulates with Vcb changes - think of it as a varactor. It may have a smaller magnitude compared to a comparably rated MOSFET, but it’s no less ugly. The Miller effect amplifies it, and its effect on the amp is dependent upon driving resistance, of course.
I have said before that I think the whole area of capacitance modulation is the biggest unheralded difference between the sound of tubes and transistors. If I can find the time to finish, I plan to publish on these topics, along with test results and measurement techniques. We own several businesses that I thought would be fun early retirement diversions, but they have turned out to be 24/7 time consumers (I keep telling myself that “busy” is what I wanted for these businesses, but now I have to live with that).
Good discussion guys. Giaime, are you convinced yet to try a tube buffer instead?
I am still an RF design engineer (although I don't think it is going to last much longer) and a ham (but I no longer own a radio). I know about the C-B capacitance modulation. That was exploited heavily in the MDS receivers that were popular about 25 years ago. The receiver LO was tuned over a 200 MHz range (at 2.2 GHz) by varying the supply voltage. There again the key to limiting this effect is running the BJT at the highest possible voltage, but then you get into SOA problems. When I tried this I remember that I could never get good sound out of a BJT.
I have been working on a simple SE amp that uses 6L6's, KT-88's, or EL-34's. I tried mosfet drive to see if it made any difference, but if there was any improvement (or degradation) I (and my friends) couldn't here it so the mosfet was eliminated. I also tried both a resistor and a CCS chip as the plate load for the 12AT7 (a tube not known for linearity). It tested and sounded much better with the CCS load. The 12AT7 is just one of those triodes that likes a CCS load.
I am not afraid of using sand (followers and CCS's not gain stages) where it demonstrates some benefit. I will also eliminate it if it does not show obvious improvement.
As I said before, I have no time for experiments right now. In the environment where I currently work, those of us who have been there a long time have to work extra hard to keep our jobs.
I have been working on a simple SE amp that uses 6L6's, KT-88's, or EL-34's. I tried mosfet drive to see if it made any difference, but if there was any improvement (or degradation) I (and my friends) couldn't here it so the mosfet was eliminated. I also tried both a resistor and a CCS chip as the plate load for the 12AT7 (a tube not known for linearity). It tested and sounded much better with the CCS load. The 12AT7 is just one of those triodes that likes a CCS load.
I am not afraid of using sand (followers and CCS's not gain stages) where it demonstrates some benefit. I will also eliminate it if it does not show obvious improvement.
As I said before, I have no time for experiments right now. In the environment where I currently work, those of us who have been there a long time have to work extra hard to keep our jobs.
It is instructive to compare the distortion spectra of a common cathode triode with a common cathode triode plus MOSFET source follower. I've done it (and done listening comparisons); the cathode followers have not replaced the IRF820s I use.
The CFs are much too wimpy to punch the screens of the output tubes without severe compromises in distortion. Perhaps a stronger and lower source Z tube than a 6SN7 would give better results, but the FET works so well, I'm not tempted to start punching holes in the chassis for 6528s.
The CFs are much too wimpy to punch the screens of the output tubes without severe compromises in distortion. Perhaps a stronger and lower source Z tube than a 6SN7 would give better results, but the FET works so well, I'm not tempted to start punching holes in the chassis for 6528s.
The Crss is the capacitance from gate to drain. Since the drain is AC ground in a follower, this is indeed the important spec. For this fet the capacitance is not only low, the curve goes relatively flat above 20 volts. Like the guy in the movie says "I gotta git me one of these". I will get some of these next time I order parts from DigiKey.
Well, it is either fets or sweep tubes driving sweep tubes. My amp has fets (screen drive 6AV5) but I could be tempted to try a sweep tube next time I work on this amp. It is still on the breadboard, so no holes are needed.
In my setup, I just use small screw-on heatsinks, the ones that are about 2x3cm with fingers. Alternatively, you can screw then down to the chassis (using an insulator!); the extra drain-to-ground capacitance is harmless.
The heatsinking requirements are very moderate- you don't have to run them in Class A. Mine are set at +200V on the drain, 4mA idle current. At full tilt, the Vds drops to about 80V with the current going up to about 20mA. The heatsinks are almost unneccessary.
The heatsinking requirements are very moderate- you don't have to run them in Class A. Mine are set at +200V on the drain, 4mA idle current. At full tilt, the Vds drops to about 80V with the current going up to about 20mA. The heatsinks are almost unneccessary.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.