• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

If UNSET and the RCA50W Had a Baby

What's UNSET? See here. It's a clever way of wrapping series-applied voltage feedback around an output stage and sharing some of the idle dissipation in the output stage between a mosfet follower and the output tube (which can get you more power if you increase B+ accordingly), apparently recently discovered by Mr. Tubelab and Mr. Smoking-Amp nearly simultaneously.

What's the RCA 50W amp? See attached schematic.

It's a 50W push-pull amp with three nested feedback loops that apparently can do 50W@0.1% THD, which ain't too shabby. The innermost loop is parallel-applied voltage feedback from plate to grid of the output tube. Surrounding that is series-applied voltage feedback from plate of output tube to driver cathode. And then there is the global loop that goes from amplifier output to input tube.

I've always had mixed feelings about this amp. It has a lot done right, but my back-of-the-envelope calculations show that the ouput tube plate to driver cathode feedback can't be very effective due to the fact that the low impedance load at the plate of the driver spoils a lot of the potential driver gain. Driver cathode degeneration lowers gain further. I'm not even totally sure that it has more gain than the gain that is trying to be set with the resistor ratio that is there.

Anyway, I've been playing with output tube plate to driver cathode feedback and have been having exceptional results with high gain drivers. It makes me wonder what would happen if we fixed the RCA amp a bit and used series-applied voltage feedback around the output tube instead of parallel-applied feedback, so I decided to run with that idea and I have attached a simplified conceptual schematic.

I made the output tubes two KT88 UNSETs (UNPPTs?), which frees up the driver to use a very high impedance load and develop some serious gain. Oh yeah.

This is pretty similar to my driver for my recent 826 SE amp experiments, only I reversed the input and feedback connections to make the phasing correct for negative feedback. Bias on the 6BN11 stage is not very stable over time due to absurdly-high gain, so a bias servo is probably mandatory on that stage. Open-loop gain of that stage is ~2600. This provides a lot of feedback and I expect resulting driving impedance on the primary of the transformer will be somewhere between 10 and 20 Ohms. Using a Hammond 1650R (or something with similar low copper losses) will result in a Zout of ~0.5 Ohms or so. Distortion will be extremely low.

In my mind, at that point, global feedback is optional, so I have omitted it. It could be added back in with another gain stage. I also like input transformers for the immunity to ground loops that they offer so I included one. Obviously, I have omitted some necessary components such as stoppers, protection diodes, and something to tie input transformer secondary to some level other than what the leakage currents pull it to.

The UNSET output configuration offers the opportunity to share some of the idle dissipation between the mosfet and the power tube, so I have increased B+ to 530V. This still puts 450V across the output tubes like in the RCA amp, but now we can hit over 75W with the same class of output tube, and have plenty of idle current to keep idle output tube transconductance high.

I decided to pull the output tube plate to driver grid feedback network down near GND with a CCS. My intuition tells me that having a high AC impedance at this node would have a balancing effect on the amplifier. I'm not sure how well this would work (haven't simmed it or anything). The other option would be to ground the center point (maybe with a trim pot in the center to adjust out AC imbalances in the two halves).

Anyway, I think it would make a good amp that could make your ears bleed with low distortion.

As always, comments/criticisms welcome.
 

Attachments

  • RCA Original.jpg
    RCA Original.jpg
    160.9 KB · Views: 1,646
  • RCA_50W_CCS.jpg
    RCA_50W_CCS.jpg
    115.8 KB · Views: 777
apparently recently discovered by Mr. Tubelab and Mr. Smoking-Amp nearly simultaneously.

I have been working on this off and on for nearly 10 years. There have been several versions, but the UNSET is the first to become a Tubelab board. The final production version is being laid out now.

Smoking-Amps version does essentially the same thing through two different paths. His version wraps the feedback around both the driver and output stage via a mosfet in the cathode of the driver stage. This version can offer a lower overall THD and output impedance than the UNSET which has a local plate to grid feedback loop in each stage.

If UNSET and the RCA50W Had a Baby

The UNSET is a rather burly amp with big bulging output tubes, the biggest seen in the USA, the 6LW6. It didn't show much interest in a 60 year old RCA amp, but tends to prefer pretty redheads. It chose a younger mate to have a baby, a direct descendant of that old RCA, a much younger "Engineers Amp."

The baby is still with some rather childish looking output tubes, actually the smallest known in the USA, the 6GF5. Still developing, the input stage is rather immature, but the childish output stage can crank out 80 WPC for hours without whining......I can't wait to see how this child grows up.

Actually the "child" is in surgery now. I'm performing an input stage transplant. The version currently implemented on the board also suffers from operating point shift particularly when the line voltage changes. I'm now implementing a separate phase splitter, and I may try some of the tricks shown here.

The original breadboard of this design is the combination of the driver PCB and the green perf board output stage on the right of a Tubelab SPP board. The first proto of a complete amp board is above the SPP board. These were built in mid 2018, before the UNSET. I have never been happy with the driver / phase splitter in this design, so it got shelved for the UNSET, which does not need a phase splitter.
 

Attachments

  • P1850371_x.jpg
    P1850371_x.jpg
    451.1 KB · Views: 179
  • P1850356_x.jpg
    P1850356_x.jpg
    391.1 KB · Views: 124
  • P1850355_x.jpg
    P1850355_x.jpg
    358.9 KB · Views: 177
  • P1850379_x.jpg
    P1850379_x.jpg
    548.8 KB · Views: 185
  • P3730285_x.jpg
    P3730285_x.jpg
    761.5 KB · Views: 673
  • P3730281_x.jpg
    P3730281_x.jpg
    839.1 KB · Views: 685
  • P3730279_x.jpg
    P3730279_x.jpg
    713.9 KB · Views: 690
  • P1850350_x.jpg
    P1850350_x.jpg
    649.2 KB · Views: 193
Last edited:
Smoking-Amps version does essentially the same thing through two different paths. His version wraps the feedback around both the driver and output stage via a mosfet in the cathode of the driver stage. This version can offer a lower overall THD and output impedance than the UNSET which has a local plate to grid feedback loop in each stage.

Did I miss something? I thought he was describing essentially what you did, and I was the one who thought of the output to driver cathode mosfet. There's too much thinking going on here!

I love watching your projects unfold.

I'd like to build this amp here but realistically it will be years before I can get around to it. The problem is I build a breadboard version and I can't stop trying new things! Next thing I know, another year has gone by.
 
I thought he was describing essentially what you did, and I was the one who thought of the output to driver cathode mosfet. There's too much thinking going on here!

Sorry, you are correct. That's what I get for trying to think late at night after spending the day outside in the serious heat. I got confused as to who did what.

After a night's sleep and some searching through old threads I know that smoking-amp and I came up with the idea of driving both G1 and G2 at roughly the same time. I called it dual drive, he called it crazy drive.

The circuit of a crude but working test amp is in post #30 here:

G1=G2/mu Scaled Drive Strawman Design

I tried this as a possible cure to the issue of melting screen grids in a high powered screen driven amp. It helped, but did not cure the problem.

The p-fet in the driver cathode was indeed your idea, and it relieves the burden of having a big p-fet in the cathode of the output tube where a failure can cause big fireworks when the fet shorts. My experiments with a mosfet in the cathode of the output tube go back more than 10 years, but usually ended in exploding fets and blown tubes when pushing into the 200 watts per pair area. Today's fets are finally good enough to live here.

After the many successes with screen drive over a 10 year period kept being clouded with a few spectacular failures involving tube arcs and exploding mosfets, I decided to back up and do some experiments. I created the little test board seen here.

I called it the Grand Unified Theory, or "GUT" board. I simply put a mosfet on every electrode of the tube except the plate. G1 and G2 got complementary pairs so that current could be sourced, or sinked. Each fet had its own bias and drive circuits so that each electrode could be driven in phase, or out of phase with the others. I spent nearly a year turning knobs and taking data, before coming up with the circuit seen in the UNSET.

I had tried many of these experiments individually at various times over a 10 year period, but never collated all of the results in one place to see what worked and didn't.

The results of smoking-amps curve tracer photos kept pulling me back to dual - drive, but my experiments kept going off in different directions.

I stated some of what I found in a screen drive thread back in 2013 here:

swing the cathode and maybe G1 with a PNP or somesuch.

I went down this road maybe 10 years ago by standing a P channel mosfet follower on its head in the cathode circuit of a conventional pentode while grounding the grid. It seemed to work, but I lost interest for some reason, maybe blown mosfets.

The thought was boosting the efficiency due to the "power pass through" seen in grounded grid RF amplifiers. If you are applying 50 volts of signal swing to the cathode as drive, that 50 volts is added to the total output swing seen by the OPT.

The problem back in 2013 was that the P-channel fets of the day did not have the SOA needed to live in the cathode circuit of a high power sweep tube.

The UNSET concept allows for the elimination of the negative voltage power supply allowing for some real cheap HiFi grade amps running off a single cheap power transformer.

Here is a single channel amp running a pair of 50C5 radio tubes. It could be built for under $50 and makes 20 watts. It did so for 14 hours straight. Only the three tubes on the right are being used. The other two would be the guitar preamp, but are needed to complete the series heater string. It runs from a Traid N-68X isolation transformer.

I had plans for making a ultra cheap series string guitar amp using 50B5's since ESRC had over 10,000 of them and I could get them cheap. That's not going to happen now.
 

Attachments

  • P3730287_x.jpg
    P3730287_x.jpg
    675.7 KB · Views: 199
  • P3730057_x.jpg
    P3730057_x.jpg
    737.3 KB · Views: 145
  • P3730052_x.jpg
    P3730052_x.jpg
    976.5 KB · Views: 167
Ah, okay, I just wanted to make sure there wasn't an interesting discussion that I missed at some point.:D

Well, I certainly love how in the UNSET output stage, part of the B+ is eaten by the fet at idle but is given back to the tube as it drives toward saturation. It's a beautiful thing.

Otherwise, this approach is very similar in effect to the KT88 amp I built for my brother with the shunt network driven from a p-channel FET follower. It just never occurred to me to drive the cathode in a series feedback arrangement when I was building that.

I put -50V on the mosfet drain to keep Crss low. Do you find that isn't necessary? I haven't reviewed the datasheet for the part you used (I forgot what part it is).

I'm really loving the results I am getting with series feedback to the driver stage, but the problem for many would be that it works so well that the amp has kind of lost its "tube character". Damping factor is high and distortion very low.

As I ran through experiments that led me here, I went from ~1% THD at 1W down to 0.027% and I definitely could hear the timbre of plucked strings change. The high distortion, high Zout early experiments sounded really cool with Norah Jones-type music, but they fell apart playing loud "workout music," which is what this amp mostly plays right now since it is in the garage with all of the weightlifting equipment. It really rocks in that application.
 
The high distortion, high Zout early experiments sounded really cool with Norah Jones-type music, but they fell apart playing loud "workout music,"

Exactly my observation. The typical 300B SE with no feedback has a DF of somewhere between 2 and 5 on a good day. Plenty of 2H, and some 3H. Norah, "chick singer with a guitar or a violin" type music sounds best on this system. Complex music gets eaten up by IMD.

Someone called "Tubelab" does not want to admit what's currently feeding the thrift store Cerwin Vegas in my gym, but it doesn't glow and runs on a laptop brick......It's only temporary because the only working UNSET board is currently sitting tubeless on the desk here next to the PC as I work on the layout for the next, and hopefully final iteration before ordering boards.

The "offspring" board as I may call it really slapped those little Cerwin Vegas around, in far better style than a $20 class D amp, but it's on the operating table getting a driver stage transplant.

I put -50V on the mosfet drain to keep Crss low. Do you find that isn't necessary?

There are several small fets with low Crss that work in the driver. None of them have models in LT spice, so what's in the spice file is not what I used in the amp. The board currently has a VP0106 in it. Crss is about 5 pF with anything over 5 volts across it. I set the bias on the grid at about 25 volts putting about 12 volts across the fet at idle.

The output fet is either a FQPF5P20 or a FQPF9P25 depending on ho big a tube I'm running and how hard I'm pushing it. I currently have the 9P25's in this board because I have seen it run at 30+ watts output for an hour or more. The fets get pretty hot at that level and the tubes are rather red. No failures were seen. I have turned the B+ up as high as 550 volts on 500 volt caps, with idle current in the 150 mA range. I don't want to push it farther. Exploding electrolytics are messy!
 
Exactly my observation. The typical 300B SE with no feedback has a DF of somewhere between 2 and 5 on a good day. Plenty of 2H, and some 3H. Norah, "chick singer with a guitar or a violin" type music sounds best on this system. Complex music gets eaten up by IMD.

There are several small fets with low Crss that work in the driver. None of them have models in LT spice, so what's in the spice file is not what I used in the amp. The board currently has a VP0106 in it. Crss is about 5 pF with anything over 5 volts across it. I set the bias on the grid at about 25 volts putting about 12 volts across the fet at idle.

The output fet is either a FQPF5P20 or a FQPF9P25 depending on ho big a tube I'm running and how hard I'm pushing it. I currently have the 9P25's in this board because I have seen it run at 30+ watts output for an hour or more. The fets get pretty hot at that level and the tubes are rather red. No failures were seen. I have turned the B+ up as high as 550 volts on 500 volt caps, with idle current in the 150 mA range. I don't want to push it farther. Exploding electrolytics are messy!

I just had another idea. I'm thinking there could be a switch-selectable plate load for the driver tube in my 826 amp. CCS load for when I want low distortion, resistor for when I want an effects box. Keeping the feedback divider constant would keep overall gain similar in the two positions, but gain inside the loop would vary. I don't think I'd hot switch it or anything but it could be like having two kinds of amps in one box.

I compared the FQPF5P20 and the FQPF9P25 to the fet I normally use (FQP3P50) and while it doesn't have the isolated case it does have significantly more SOA margin over the other two and the capacitance stays low to ~15V, whereas the other two start rising steeply at ~20V.

It seems like with any of these, though, a -15V supply for the drain could be beneficial for keeping out of the high capacitance zone at all signal levels, and wouldn't add too much to the dissipation. I'd probably use a cheap AC-DC brick to do it.
 
The RCA 50 Watt-er has been a mystery to me too for some time. As you said, the driver gain gets eaten up by the "shunt Schade" (low impedance) around the output tube grid. Spoiling the middle driver cathode loop effectiveness. I can only guess they wanted something fool-proof (no instability) for at-home constructions.

Still, to get 0.1% distortion, there must be some kind of curious synergy going on there. Maybe one feedback loop left a slight transfer curvature one way and the other loop left some curvature the other way, and they balanced the two against each other? The Citation II has got something similar going on too, I suspect.

One further twist could be to use "Crazy drive" configured grids for the output N Fdbks, with the new cathode driving Mosfet for input drive. I need to set this up on the curve tracer to see if it works well, and to see if an extra Mosfet follower is needed to isolate the screen grid current.

Theoretically, with linear operation, the screen grid would be operating at some constant V fraction of the plate V, due to this Crazy N Fdbk, so would be drawing a proportional current. That would make it intercept a constant fraction of plate current, so would look like a resistive load to the feedback divider (so could be integrated into the R divider then without a Mosfet follower, or wishful thinking anyway) Load current may mess this up though, not sure yet.

A curve trace plot should illuminate the operating range. Hopefully I can run a plot off tomorrow. "Crazy feedback" should then give "triode" curves with tilted straight lines, if that can still be called a triode.

Maybe "Straight Wire And Gain" (a SWAG stage) would be a better name. Gee, I think I've heard that name before. Something to do with Scientific... :D
 
Last edited:
I normally use (FQP3P50) and while it doesn't have the isolated case it does have significantly more SOA margin

Not really. Fairchild / ON puts SOA curves in their data sheets, but the DC curve is just a graph of it's dissipation rating, 85 watts in the case of the FQP3P50.

Working as a transmitter designer at Motorola led our group to the early discovery that fets do exhibit a secondary breakdown effect similar to BJT's, but it has been given a different name that I don't remember now. We found this the hard way in GaN parts, but it exists in Si and SiC parts too. Some of the mosfet manufacturers have papers discussing the issue, and have proper SOA curves for their parts in the data sheets. IXYS makes a line of fets for linear operation and they are far more expensive that the typical switchmode parts we tend to use. Some 100+ watt switchmode mosfets are only good for a few watts in the upper end of the linear region, but not all data sheets reflect this.

Those of us trying to build linear voltage regulators have dealt with this issue, and that has been the delay with the UNSET board. The screen regulator is dissipating about 5 watts when the amp is run at the edge of clipping. I have blown several fets on "full power for several minutes" tests, including some rated for 90 watts. Of course a shorted fet stuffing 500 volts into the screen grid of a sweep tube blows stuff up, so I need to redo the layout in that part of the board to put a 5 watt dropping resistor in series with the fet.

Here is a picture of the test setup that led to the "offspring" board. There is a dead fet in the picture. I don't know the number because it went into the trash when I cleaned up the bench. It has an isolated case, so it is not a FQP3P50. The picture was taken in June of 2018. My box of fets has a bag of 10 FQP3P50's with 4 parts missing. Obviously I moved on for a reason. Much of my testing was done with 6HJ5's which are capable of a lot of plate current. After one cracked and another exploded due to shorted fets, I switched to the 6GF5's they are smaller and cost $1 each. The offspring board is actually laid out to accept either tube (hence the funny schematic around the output tube), but I have not tried 6HJ5's in it yet.

I had the offspring board on my bench to take some pictures a couple days ago before ripping up the driver. I noticed a bunch of circuit modifications done some time ago. The third picture in post# 2 was taken Thursday. The fourth picture was taken about 2 years ago. The schematic of the board as originally built is included here. I took the feedback from the plate of the output tube to the screen grid of the driver instead of the control grid as you have done. I am using driver tubes from the $1 list which are TV IF amp tubes, and some have undocumented semi-remote cutoff characteristics (verified with a hammer). I am also using the driver as a LTP phase inverter, so there is a CCS in the tail, with input to the control grid. This works great for a narrow range of operating points, and I find myself turning the CCS pot every time I fire this up, or whenever the heat pump kicks on (line voltage variations).

The little 20 watt 50C5 board used two copies of the UNSET in push pull driven by a split load phase inverter made with a mosfet, so that's where I'm going here.

Once that's working, I'll try moving the feedback to G1 and adding a CCS in the plate. I have a small guitar amp with a CCS in the plate load of a pentode for stupid amounts of gain in a single stage. Running a $1 pentode at a stage gain well over 1000 makes for a very microphonic amp.
 

Attachments

  • P1390483_x.jpg
    P1390483_x.jpg
    689 KB · Views: 172
  • C-Drive_12pin_Amp.pdf
    126.3 KB · Views: 156
Last edited:
The RCA 50 Watt-er has been a mystery to me too for some time. As you said, the driver gain gets eaten up by the "shunt Schade" (low impedance) around the output tube grid. Spoiling the middle driver cathode loop effectiveness. I can only guess they wanted something fool-proof (no instability) for at-home constructions.

Still, to get 0.1% distortion, there must be some kind of curious synergy going on there. Maybe one feedback loop left a slight transfer curvature one way and the other loop left some curvature the other way, and they balanced the two against each other? The Citation II has got something similar going on too, I suspect.

One further twist could be to use "Crazy drive" configured grids for the output N Fdbks, with the new cathode driving Mosfet for input drive. I need to set this up on the curve tracer to see if it works well, and to see if an extra Mosfet follower is needed to isolate the screen grid current.

Theoretically, with linear operation, the screen grid would be operating at some constant V fraction of the plate V, due to this Crazy N Fdbk, so would be drawing a proportional current. That would make it intercept a constant fraction of plate current, so would look like a resistive load to the feedback divider (so could be integrated into the R divider then without a Mosfet follower, or wishful thinking anyway) Load current may mess this up though, not sure yet.

A curve trace plot should illuminate the operating range. Hopefully I can run a plot off tomorrow. "Crazy feedback" should then give "triode" curves with tilted straight lines, if that can still be called a triode.

Maybe "Straight Wire And Gain" (a SWAG stage) would be a better name. Gee, I think I've heard that name before. Something to do with Scientific... :D

Well, I can't wait to see the results.

I did some experimentation with CCS-loaded pentodes a while back with plate-grid feedback for a high swing driver. At the same time, I was reading a book by Douglas Self where he went over resistor distortion. He showed results that thick film resistors generated significant 2nd harmonic distortion (their resistance varied with applied voltage). He also showed that wirewound resistors had no such effect.

I was using TO-220 power thick film resistors for my feedback resistor so I decided to try a wirewound and see if I got measurably better results. Distortion increased, and I can only surmise that the 2nd harmonic from the feedback resistor was cancelling with some distortion in the circuit. Sometimes making a circuit worse in just the right way makes it better.
 
Not really. Fairchild / ON puts SOA curves in their data sheets, but the DC curve is just a graph of it's dissipation rating, 85 watts in the case of the FQP3P50.

Ah, so they just flat-out lie in the datasheet. I was wondering why so many more expensive IXYS parts had such worse SOA...

I built a bias servo board some time ago and I'm currently putting together a parts order to build that up so I don't have to keep tweaking the bias on the input stage of my breadboard amp. I'll set up a channel for the output stage as well while I'm at it.
 
Maybe they lie, or maybe the datasheet was printed before this phenomenon was fully understood.

Some parts do not have a DC specification. Just sticking the word mosfet into Mouser, narrowing the results to through hole and in-stock, then choosing the first part got me this data sheet. It is a huge 450 watt part, with no DC spec, but the 1 mS spec had the SOA reduced to 150 watts. DC is probably in the 50 watt range. Crss for such a big part is real low at 3 pF, but it needs 100 volts to get there.

We used to think that mosfets were immune from secondary breakdown. The work I did on GaN RF power fets led to the purchase of about $10K worth of Flir video camera equipment capable of looking at the individual fet cells on a die. We were using a pulse rated 50 watt fet for CW power in the 5 watt range. The parts could run forever into all sorts of abuse like open or shorted load at 100C, then just randomly die under minimally stressful conditions.

I set up a transmitter running about twice the normal power and it ran for nearly a month before it blew. We had captured the failure on video, and it looked like one cell started heating up, then another....then an avalanche effect in under one second.

This was late 2013, and I took the buyout 3 months later. I know that the part vendor was in and out of our lab daily during those 3 months, but since I had accepted early retirement, I was removed from the project, and the lab. Sometime around 2018 that product started shipping, so the issue was resolved.

It seems that this issue had been lying dormant since the dawn of the mosfet, but todays smaller geometry parts are far more susceptible. If the part only shows switchmode applications, it may be susceptible to second breakdown. Some switchmode parts actually show this in the data sheet, some don't.

ON / Fairchild doesn't seem to show this fact, but I have blown up several ON parts in simple screen regulator circuits. My UNSET board has been working fine with an STmicro part in it since January, but I can't control what get's put into my boards by the builder, so I'm sticking a 5 watt resistor in the drain lead to reduce the risk.

My first attempt at a linear 600 volt 1 amp power supply ended with a lot of blown fets. No further work has been done, but I did buy some of those expensive IXYS parts. It also appears that depletion mode fets are somewhat immune from this effect, but I have not tested them under near spec limit conditions.
 

Attachments

  • STWA70N65DM6.pdf
    930.2 KB · Views: 95
Lateral mosfets didn’t seem to have that second breakdown issue. But they’re hard to come by these days, and not really what you need for a screen regulator. When contemplating the next really big amp, I’ll only consider running the screens from a lower voltage auxiliary supply. Even then with peak g2 currents up to am amp maybe an amp and a half I’m not sure I’d trust it to any switching mosfet. If the voltage differential is 100 volts or less you can get *biploars* that will do the job. I’m thinking about a darlington using a mosfet and an MJL4281 for the screens on the big sucker.
 
When contemplating the next really big amp....a darlington using a mosfet and an MJL4281 for the screens on the big sucker.... with peak g2 currents up to am amp maybe an amp and a half

My "last big tube amp" has been in the planning / dreaming stage for far too long. It's time to actually build something before I get too old to move it. I'm looking at a vacuum tube kilowatt (500 WPC) using big TV sweep tubes, probably 6 per channel. Plate voltages will be in the 650 to 700 volt range with screen voltage at 150 volts. The screen current will be low (10 to 25 mA per tube) except when the plate voltage gets pulled down below 150 volts. This should occur infrequently with music, but peaks over 100 mA per tube are possible. The amp should be capable of continuous full power operation until the line breaker trips (several minutes).

I will definitely not try to drop 500+ volts in a mosfet in an amp of this magnitude.

Meanwhile, I have an evil experiment in mind. I'm trying to get my first prototype UNSET working (some parts were removed). Then I will make a few modifications to the feedback circuitry, add a CCS in the plate of the drivers, and get it all working as a two channel SE amp. Once that is working to my satisfaction, I will drive the two channels out of phase with a push pull OPT across both outputs. I would like to see something in the 200+ watt range.

It's safe to assume that something will go BOOM.
 
I don’t see any reason why you couldn’t drive the cathodes with a big rugged bipolar. Doing a darlington or CFP with a smaller low capacitance mosfet would certainly beat the high input and reverse capacitances you’d have to deal with to get a mosfet that can take as much abuse in the linear region as an MJL 4281. The first time I saw anything like this it was in a Peavey guitar amp - they used TIP3055’s to drive the cathodes oh 6L6’s. G1’s were grounded, so no local NFB - which meant a Hiiiiiiigh output impedance. They did clamp the output with diodes, to keep from arcing over or ruining output transformers.
 
I've sketched up an attempt at this approach (output plate to driver cathode feedback) on a Unity-Coupled amp and I have to say it works out beautifully. Instead of having the feedback resistor attached to output tube plate, it can be attached to output tube cathode and there is no huge DC offset heating the feedback resistor all the time.

The feedback network can also be set up like the input stage of an instrumentation amp, like I tried to do up above, but there is no CCS required to pull the network down to the level of the input tubes. It is already sitting at just over a volt, which works fine.

Very simple and elegant.

I'm almost tempted to do surgery on my Unity-Coupled amp right now, but I'll enjoy the glow of the 841s a year or two more while I build my SE amp.
 

Attachments

  • UC_6BN11.jpg
    UC_6BN11.jpg
    118 KB · Views: 259
The first time I saw anything like this it was in a Peavey guitar amp - they used TIP3055’s to drive the cathodes oh 6L6’s. G1’s were grounded

Music Man had some guitar amps in the early 80's that put a common emitter BJT in the cathodes of 6L6GC's or EL34's. All the preamp stuff used opamps. They had a unique sound that I was not fond of. At least in the humid south Florida area there were some flaming failures of the OPT's, tubes and sockets. This might have been due to the 700 or so volts of B+ that the big versions ran.

Doing a darlington or CFP with a smaller low capacitance mosfet would certainly beat the high input and reverse capacitances

I am using two different fets now, one is 13 pF and the other is 27 pF so for now the OPT is the main bandwidth bottleneck.

If I stick any silicon in the signal path of a tube amp it is a follower, either mosfet or BJT. I have some older MJW21195's left over from some CFP experiments involving a tube / BJT pair for output stage duty. I could try some of them fed by a small signal mosfet.

It's Eagle. It does circuit board layouts as well.

I got started with Eagle in the early 1990's when their US office was in Boca Raton Florida about a mile from where I worked and went to school. It was version 2.6 and ran on DOS. There was a guy who worked there that literally taught me how to use it. I would drive over with my laptop and have him help me with my screw ups.

I am now stuck at version 5.11 since Eagle was sold to Farnell, then to Autodesk. Neither of them would offer me an upgrade to their current version for a price that I could afford. Autodesk now wants a $600 / year lease for their pro version which is required to do the large PC boards needed for tube amps. They do have a free version which does single page schematics and small PC boards.