a year ago I have come across a stash of GU50s. The human brain thinks simple in those situations, see shiny, want shiny, even if it serves no purpose. SO I took em.
Today I have tought to myself, HMMMMM these things are rate 40W anode dissapation Ua1kV and Ik 350mA
I think I can easily pull out 100W RMS out of a single pair. Well, would be easier if I didnt want a UL output transformer with cathode feedback windings too.
Here is the setup:
Yes youre seeing it right, 1kV anode voltage. With this voltage I could probably get into the 130W RMS territory (as the cathode current in this setup is peaking 250mA nowhere near the Ik limit of the tubes)
At this voltage getting acceptable class AB behaviour is not that easy as it requires to drive tubes into 30-35W dissapation but according to the simulator it yealds 2.7%THD at 100W RMS
The parameters for the transformer are
10k Ra-a
30%UL
10%CFB
The one issue I am seeing here is the quite large negative G1 voltage. At very negative voltages grids start to behave like electron deflectors..something like a beamforming electrode but in an undesireable place. I might reconsider what to do about it, either lower Ug2 but then having to drive grids positive. If anyone has experience with such negative bias voltages with these tubes or tubes of any kind, precise values or whereabouts your contribution would be really appreciated.
The exact reason I went with 600V Ug2 was because I didnt want to push grid current into them. The GU50s do however have the magical capabillity of withstanding up to 1W ! of grid 1 dissapation. That is quite insane If you ask me, however G2 dissapation kinda sucks a little. But that is not an issue as my G2 voltages are largely below the anode ones so not much G2 current flows until the anode voltage falls below G2 voltage and then larger current starts to flow which is sorta limited by the 100ohm resistors (while still not impeding gain)
The cathode feedback windings hugely increase the drive requirements right up to 600V p-p. This is certainly a problem to overcome with both a tube or transistor driver. But I have a solution.
This abominaton:
C1 and C2 replace grid cappacitance of GU50s (datasheet maximum value)
At first I did not plan it to come out like this, but I thought "what the hell lets give it a chance anyways"
Interstage transformer with cathode feedback to prevent phase shifting.
Everything is DC coupled, no bass frequency compromises HOWEVER I see a slight issue with this. While the voltage on the anode resistors are going to be fairly predictable thanks to the current sources, if the tubes are mismatched that will cause a current flow difference between one anode to another one and consequently DC current will flow trough the interstage transformer.
I dont know of a small power still manufactured tube that could give me a 600V p-p swing without great distortion, melting it or straight up ripping the emissive coating off the cathodes. So it was either a cascode of I dont know what tube that could also give reasonable drive power without melting, or an interstage transformer. So I thought if I have to use an interstage transformer might aswell overdo it. And that was the result of such thinking.
I dont know the winding ratios yet because I just threw down the priary inductance by gut feeling and then adjusted everyting arround until I got desireable results and gains.
Any feedback on this insanity would be appreciated. Please do leave feedback as I mostly draw this sort of stuff up trought the night half asleep and I might miss some fundamental errors why this cant work or wont work the way I want. Suggestions on how to make it better or fix it if broken are greatly appreciated.
Today I have tought to myself, HMMMMM these things are rate 40W anode dissapation Ua1kV and Ik 350mA
I think I can easily pull out 100W RMS out of a single pair. Well, would be easier if I didnt want a UL output transformer with cathode feedback windings too.
Here is the setup:
Yes youre seeing it right, 1kV anode voltage. With this voltage I could probably get into the 130W RMS territory (as the cathode current in this setup is peaking 250mA nowhere near the Ik limit of the tubes)
At this voltage getting acceptable class AB behaviour is not that easy as it requires to drive tubes into 30-35W dissapation but according to the simulator it yealds 2.7%THD at 100W RMS
The parameters for the transformer are
10k Ra-a
30%UL
10%CFB
The one issue I am seeing here is the quite large negative G1 voltage. At very negative voltages grids start to behave like electron deflectors..something like a beamforming electrode but in an undesireable place. I might reconsider what to do about it, either lower Ug2 but then having to drive grids positive. If anyone has experience with such negative bias voltages with these tubes or tubes of any kind, precise values or whereabouts your contribution would be really appreciated.
The exact reason I went with 600V Ug2 was because I didnt want to push grid current into them. The GU50s do however have the magical capabillity of withstanding up to 1W ! of grid 1 dissapation. That is quite insane If you ask me, however G2 dissapation kinda sucks a little. But that is not an issue as my G2 voltages are largely below the anode ones so not much G2 current flows until the anode voltage falls below G2 voltage and then larger current starts to flow which is sorta limited by the 100ohm resistors (while still not impeding gain)
The cathode feedback windings hugely increase the drive requirements right up to 600V p-p. This is certainly a problem to overcome with both a tube or transistor driver. But I have a solution.
This abominaton:
C1 and C2 replace grid cappacitance of GU50s (datasheet maximum value)
At first I did not plan it to come out like this, but I thought "what the hell lets give it a chance anyways"
Interstage transformer with cathode feedback to prevent phase shifting.
Everything is DC coupled, no bass frequency compromises HOWEVER I see a slight issue with this. While the voltage on the anode resistors are going to be fairly predictable thanks to the current sources, if the tubes are mismatched that will cause a current flow difference between one anode to another one and consequently DC current will flow trough the interstage transformer.
I dont know of a small power still manufactured tube that could give me a 600V p-p swing without great distortion, melting it or straight up ripping the emissive coating off the cathodes. So it was either a cascode of I dont know what tube that could also give reasonable drive power without melting, or an interstage transformer. So I thought if I have to use an interstage transformer might aswell overdo it. And that was the result of such thinking.
I dont know the winding ratios yet because I just threw down the priary inductance by gut feeling and then adjusted everyting arround until I got desireable results and gains.
Any feedback on this insanity would be appreciated. Please do leave feedback as I mostly draw this sort of stuff up trought the night half asleep and I might miss some fundamental errors why this cant work or wont work the way I want. Suggestions on how to make it better or fix it if broken are greatly appreciated.
It's easy to make a circuit that works with perfect parts.
Engineering is making it actually work.
Engineering is making it actually work.
Being that the GU50 is cheap, might it not be easier to run parallel push pull at lower voltages and easier drive requirements? You wont need the high voltage wire, caps, and output transformers that will cost a fortune. Driving them will be easier too without an interstage transformer. It’s fun to think about doing what you’re proposing but hard and expensive to really do it right.
CaterpilarSK
Ambitious design, an issue with it is the 600V on G2, the GU50 data sheet recommends a maximum of 300V. in my experience 350V works but 600V is asking for trouble. Wavebourn, a fellow contributor has much experience with the GU50 operating his Pyramid design at Anode 800V, G2 270V and Ip 30mA into an 8 - 10K load for 100W thereabouts. Regarding your driver, Zintolo has recently posted a novel simpler design that can be adapted, also Gary Pimm's CCS loaded pentode. Anyway 600V PP is 212V RMS, very tall order even for a tube driver stage. Good luck.
Ambitious design, an issue with it is the 600V on G2, the GU50 data sheet recommends a maximum of 300V. in my experience 350V works but 600V is asking for trouble. Wavebourn, a fellow contributor has much experience with the GU50 operating his Pyramid design at Anode 800V, G2 270V and Ip 30mA into an 8 - 10K load for 100W thereabouts. Regarding your driver, Zintolo has recently posted a novel simpler design that can be adapted, also Gary Pimm's CCS loaded pentode. Anyway 600V PP is 212V RMS, very tall order even for a tube driver stage. Good luck.
Yeah I have guessed the 600V on G2 might be an issue..simulator says its not going over Pg2 but reality will tell. I wanted to go UL so this was my only option without having to run grid current into G1.
Problem is that for such high voltage and power I desire from the setup with relatively low distortion trough a battalion of local feedbacks there is no simple solution.
The driver would definitely work. It just looks to be too good to be true in the simulator. Because of all the DC couplings that is...every part is exactly identically matched. Thats an impossibillity in reality. So Ill try to build it and if it fails to go as planned Ill just add two coupling cappacitors before the 12AT7 triodes.
Problem is that for such high voltage and power I desire from the setup with relatively low distortion trough a battalion of local feedbacks there is no simple solution.
The driver would definitely work. It just looks to be too good to be true in the simulator. Because of all the DC couplings that is...every part is exactly identically matched. Thats an impossibillity in reality. So Ill try to build it and if it fails to go as planned Ill just add two coupling cappacitors before the 12AT7 triodes.
unused potential. If i can squeeze ALL the performance out of the tube I will. I have done this similarly with a single pair EL34 ran at 800V 10kOhm 75-80W RMS and the output stage alone had a THD of 2% at max power. Didnt have the Cathode feedback windings otherwise I am sure those would get me MUCH lower. ran 500V on G2 with 43%UL. -60V. And the rest was basically a Mullard 5-20 with a better driver able to provide a much larger voltage swing
You could swap a 12ax7 for a 12ay7 (U6) in your simulation to look at mismatch. I would not be coupling 3 DC stages in that way.
You can't exceed 300V g2-k with GU50. Working at those high B+ isn't even necessary.
Plan to dc couple the driver to g1 because they like A2/AB2 alot.
I'd skip all those feedbacks and use a simple pentode configuration with local feedback (current of voltage as you like).
You can get almost twice the power you are claiming.
Don't do it in real world.
"Don't do it in real world."
Why not thought ? What is going to happen? G2 shouldnt melt as the anode is still far higher potential than the grids (until it isnt and it pulls a bit of grid current into G2 but nothing dramatic. The amp works happily and delivers quite the punch.
Why not thought ? What is going to happen? G2 shouldnt melt as the anode is still far higher potential than the grids (until it isnt and it pulls a bit of grid current into G2 but nothing dramatic. The amp works happily and delivers quite the punch.
People here who have a lot of experience with GU50 say exactly that, do not go over 300V for G2.
@Wavebourn has a lot of experience and designs using GU-50, would be interesting to hear his opinion.
He says what I said: 900-1000V B+, local current feedback, 10-12k Raa and screens around 270V IIRC.
While its nice to know what, I would also like to know WHY? What happens to the tube or inside the tube that limits us to 300V Ug2?
As you increase Vg2, it takes more and more negative Vg1 to shut it off. At higher negative vg1, Grid leakage currents matter more and more, and it gets harder to maintain a given Iq. Eventually, it will get finicky and the current will simply run away.
The spec of Vg2 is 250V max. G2 captures some of the electrons like the plate so will dissipate some power. Ultimately the wire on G2 will melt. The spec does give 5W max on G2 ltspice will predict G2 dissipation.
Other option if you don't mind is run 2 tubes in parallel rather than pushing them out of spec. Bias the G1's separately.
As for UL you can drop the screen voltage from the UL taps with some 100V 5W zeners. This will increase your UL %. Or you can do a voltage combiner of HT 0V and plate followed by a follower (valve or MOSFET) to drive the screen. Just options...
Other option if you don't mind is run 2 tubes in parallel rather than pushing them out of spec. Bias the G1's separately.
As for UL you can drop the screen voltage from the UL taps with some 100V 5W zeners. This will increase your UL %. Or you can do a voltage combiner of HT 0V and plate followed by a follower (valve or MOSFET) to drive the screen. Just options...
Last edited:
grid leakage currents to where exactly?
The spec of Vg2 is 250V max. G2 captures some of the electrons like the plate so will dissipate some power. Ultimately the wire on G2 will melt. The spec does give 5W max on G2 ltspice will predict G2 dissipation.
Acorrding to LTSpice with a quite accurate model, G2 dissapation is nearly nonexisting in idle conditions. I have stated earlier that G2 dissapation is low until the anode voltage falls below G2 voltage. Even then its a pulsed 3W RMS load. I was worried that I might melt G2 but from what the simulator says its safe.
The entire reason for me going that high voltage on G2 was to avoid pushing G1 current even thought the GU50 can withstand 1W of G1 dissapation (certainly a useful quality for a class C RF amplifier)
It may be OK. One issue is the curves on the data sheet only go up to 250V so ltspice is extrapolating. If you have a real valve in circuit take a current measurement on the screen.
I tend to put two in parallel KT77/6550 etc as I know then its in spec and I don't need the high HT. Avoiding AB2 is good as its difficult to drive G1 without distortion or the bias slipping.
I tend to put two in parallel KT77/6550 etc as I know then its in spec and I don't need the high HT. Avoiding AB2 is good as its difficult to drive G1 without distortion or the bias slipping.
The problem with your #1 3 cascade LTP is that very small mismatches on the first stage will become very large to the point of the final stage having one tube hard on and one tube hard off. If you like the idea of a single high gain voltage stage, a better option is say a couple of pentodes EF86 as LTP followed by a cathode follower ECC88. Other method which I use a a JFET/12AX7 cascode (gain x400) followed by a ECC88 cathode follower. Both don't need the interstage transformer.
I do plan to measure real world values once the thing is built, I am more worried I will kill myself in the process of doing so (1kV anode and 600V G2 is NO JOKE)
Now to solve the problem of where to get 1kV DC
Also I like your idea of swapping in the 12AY7 model to test for
On another note, this circuit requires a 600V p-p signalto be driven into limitation. The output I mean. I need a circuit that will give me that and for sure if I dont want to melt tubes I cant do it without an interstage transformer.
The driver has lots of gain however it shouldnt be an issue. Because there is local feedback for the interstage driver (which also sorts out phase issues to a degree) and the first diff is basically the inpu of the amplifier.
The IN is basically my +4dBu audio signal and the feedback comes from the output transformer secondary. Much like a transistor amplifier with a differential amplifier would be.
And while I absolutely love EF86 tubes, they are SO EXPENSIVE that I can buy 3 12AX7s for the price of a single EF86 😕. So as much as I love those tubes (and fantasies of UL connecting EF86 tubes with interstage transformers and other fun stuff) is off limits for now.
I am worried only about DC mismatch in the entire long tail. AC mismatch should be handled by the current sources in the cathodes.
Now to solve the problem of where to get 1kV DC
Also I like your idea of swapping in the 12AY7 model to test for
On another note, this circuit requires a 600V p-p signalto be driven into limitation. The output I mean. I need a circuit that will give me that and for sure if I dont want to melt tubes I cant do it without an interstage transformer.
The driver has lots of gain however it shouldnt be an issue. Because there is local feedback for the interstage driver (which also sorts out phase issues to a degree) and the first diff is basically the inpu of the amplifier.
The IN is basically my +4dBu audio signal and the feedback comes from the output transformer secondary. Much like a transistor amplifier with a differential amplifier would be.
And while I absolutely love EF86 tubes, they are SO EXPENSIVE that I can buy 3 12AX7s for the price of a single EF86 😕. So as much as I love those tubes (and fantasies of UL connecting EF86 tubes with interstage transformers and other fun stuff) is off limits for now.
I am worried only about DC mismatch in the entire long tail. AC mismatch should be handled by the current sources in the cathodes.
Grid leakage currents develop a voltage across the bias resistor. Which raises the actual Vg1, which in turn increases plate current. Too much of this going on and g1 loses control of the valve. That’s the problem with requiring a large negative bias just to get the required Iq. Especially at a kV on the plate. That’s the real reason for g2 voltage limits. It’s the max g2 for reliable g1 control. The g2 dissipation is a separate issue entirely. That usually won’t shoot up till you’re clipping.
It’s not as big a deal in class C operation where it’s turned off HARD, and not expected to maintain a few tens of mA idle plate current. You apply a negative-enough Vg1 where it won’t run away if it develops a few uA grid leakage.
Well you could make R1 R8 a 220R pot but even so you seem to have a lot of gain. I would have thought 2 stages of LPT is enough considering most amps have two stages. With an interstage transformer and OPT you won't be able to apply a lot of NFB without stability issues.
Then lets assume direct coupling.
One end of the transformer winding gets attached to regulated fixed negative voltage supply and the other end trought a 1kOhm resistor gets attached to the grid.
An ABSOLUTE worst case scenario I could imagine 100uA flowing into the grid would give a 0.1V voltage change. thats, practically laughable in comparison to how much negative voltage you actually need to get idle current low enough.
Even trough a 100kOhm resistor 100uA is only a SINGULAR volt which makes very little difference in anode current and is very unlikely to runaway. Though I have not yet seen grid leakage currents on tubes in a condition when the grid is far lower than cathode voltage except I saw a PCL86 that was faulty and was pulling the grid too high of a voltage with a 470kOhm grid to ground resistor.
G2 dissapation only becomes a problem near clipping that is true. Well at least from my expectations and the simulator.
Changing R1/R8 for DC mismatch could be a way to go, however it will affect AC performance aswell (unless I put quite sizeable cappacity cappacitors in paralel...)
It is a 3 differencial design purely out of sheer fun and curiosity.
This part of the circuit has "only" 36dB gain, thats after the interstage transformer
The goal here was to use local feedback (L4 L5 are part of the interstage transformer) to reduce gain and also prevent phase shifting as much as possible while not having to overwork the first stage diff amp where it wouldnt have enough gain left to do its job. And while I agree U5/U6 configuration has copious amounths of gain on top of the 36dB of gain that the whole U3 U4 U2 U1 setup has, but we also have to keep in mind I need a 600V P-P drive signal. Thats no joke. So I think this is still reasonable for the application at hand. If it turns out to be a big problem I can still just inscrease the amounth of feedback on L4 L5 to reduce gain (hopefully increase stabillity) but will see if its necessary.