Got this idea when discussing Shoogs EL86. MJK showed the embryo of this one with his 417A Spud.
I get indications of 5-6W with quite low THD, this with SE signature. The input triode does need 400V so its working point is not optimal, but this is as good as it gets. Distortion-cancellation helps a bit
.
Balance is set with the "bias" voltage at the MOSFETs gate. There is a problem with current balance though, as screen current varies.
Any suggestions to make it even simpler?
I have the chassis, tubes and OPTs just have to order the PT.
Feel free to use the Spice-file at the bottom. Just change *.text to *.asc and open in LTSpice, the tube-models are included
http://www.eflatjump.se/SchadeECL86PPMOS.txt
I get indications of 5-6W with quite low THD, this with SE signature. The input triode does need 400V so its working point is not optimal, but this is as good as it gets. Distortion-cancellation helps a bit
.Balance is set with the "bias" voltage at the MOSFETs gate. There is a problem with current balance though, as screen current varies.
Any suggestions to make it even simpler?
I have the chassis, tubes and OPTs just have to order the PT.
Feel free to use the Spice-file at the bottom. Just change *.text to *.asc and open in LTSpice, the tube-models are included
An externally hosted image should be here but it was not working when we last tested it.
http://www.eflatjump.se/SchadeECL86PPMOS.txt
Last edited:
Got this idea when discussing Shoogs EL86. MJK showed the embryo of this one with his 417A Spud.
I get indications of 5-6W with quite low THD, this with SE signature. The input triode does need 400V so its working point is not optimal, but this is as good as it gets. Distortion-cancellation helps a bit
Balance is set with the "bias" voltage at the MOSFETs gate. There is a problem with current balance though, as screen current varies.
Any suggestions to make it even simpler?
Replace the tail CCS with a resistor?:
http://www.diyaudio.com/forums/showpost.php?p=1599359&postcount=7
A balance servo is probably needed with the toroid. Maybe regulate the tube current using the screen voltage? Not very simple though. Probably easiest to make the MOSFET follow the tube DC.
Startup is strange with the MOSFET hogging all the current until the tube conducts, but nothing burned up.
It would be cool to make this into a complete amp. A little re-arrangement of feedback and coupling and this could drive class A2
Cheers,
Michael
Last edited:
Maybe the DC xfmr balance could be tweaked by adjusting R29 (lower probably) so the increasing screen current roughly balances out (subtracts out) the non-linear increasing component of the U1 plate current at large signal excursion.
Or use a sense resistor in the U1 screen current to pull the M1 gate voltage slightly upward (another P type transistor needed) to control DC balance.
Don
Or use a sense resistor in the U1 screen current to pull the M1 gate voltage slightly upward (another P type transistor needed) to control DC balance.
Don
Last edited:
SE <---> P-P control idea:
If M1 were another identical pentode to U1, call it U2, a SE <---> P-P control can be implemented by controlling the effective gm of U2:
A differential pair or Op Amp would compare an adjustable DC reference voltage (set at the nominal common cathode voltage) with either a fast or slow reference ("A") derived from the common cathodes. It would control the grid of U2 to keep these matched.
A pot would span between the direct common cathode voltage (fast) to a LP filtered common cathode voltage (slow). The pot's wiper would then provide the reference "A" above.
When the pot is at the "slow" end, U2 grid 1 is controlled slowly to keep DC balance in the xfmr. So giving a P-P signature with essentially matched U1 U2 gms.
When the pot is at the "fast" end, U2 grid 1 is controlled quickly to keep the common cathode voltage rock stable. U2 looks like a very high gm tube in this case. So giving a SE signature from the very unmatched U1 U2 gms.
The adjustable DC reference could still have some DC balancing signal added to it for DC servo balancing.
Don
If M1 were another identical pentode to U1, call it U2, a SE <---> P-P control can be implemented by controlling the effective gm of U2:
A differential pair or Op Amp would compare an adjustable DC reference voltage (set at the nominal common cathode voltage) with either a fast or slow reference ("A") derived from the common cathodes. It would control the grid of U2 to keep these matched.
A pot would span between the direct common cathode voltage (fast) to a LP filtered common cathode voltage (slow). The pot's wiper would then provide the reference "A" above.
When the pot is at the "slow" end, U2 grid 1 is controlled slowly to keep DC balance in the xfmr. So giving a P-P signature with essentially matched U1 U2 gms.
When the pot is at the "fast" end, U2 grid 1 is controlled quickly to keep the common cathode voltage rock stable. U2 looks like a very high gm tube in this case. So giving a SE signature from the very unmatched U1 U2 gms.
The adjustable DC reference could still have some DC balancing signal added to it for DC servo balancing.
Don
Hey Don,
Good idea. For listening evaluation this must be great to be able to dial in the correct amount.
Solving the dc-balance is my main concern at the moment. Probably a servo is needed. It should, as I see it, take the pentodes anode as reference comparing, with the MOSFETs drain and with the correction voltage applied into MOSFETs gate. It is beyound my knowledge though to solve it.
Good idea. For listening evaluation this must be great to be able to dial in the correct amount.
Solving the dc-balance is my main concern at the moment. Probably a servo is needed. It should, as I see it, take the pentodes anode as reference comparing, with the MOSFETs drain and with the correction voltage applied into MOSFETs gate. It is beyound my knowledge though to solve it.
Most DC servos work to keep cathode current constant - which would be no use to you here.
If you used a conventional EI OT and an 807 as the output pentode the problem would be a lot less. 807's were designed to draw much less screen current than most pentodes. I believe that most pentodes draw about 20% of the anode current through the screen - but the 807 is less than 5%. Coupled with the lower plate current/higher plate voltage of a 807 this makes the screen imbalance much less of an issue.
Shoog
If you used a conventional EI OT and an 807 as the output pentode the problem would be a lot less. 807's were designed to draw much less screen current than most pentodes. I believe that most pentodes draw about 20% of the anode current through the screen - but the 807 is less than 5%. Coupled with the lower plate current/higher plate voltage of a 807 this makes the screen imbalance much less of an issue.
Shoog
Hey Shoog,
I suggested a servo to compensate for current-differences in the OPT only. The servo has to be referenced to B+. Don´t know how to design it though.
This one was ment to be a Spud(single tube amp). Have to breadboard this one to see what happens IRL with respect to current imbalance. Also have a pair of Nuovotem 6V-toroids waiting. Might try a simple currentmirror first.
As currentmirrors/CCS/LM317 use some of the necessary cathode-swing there will be a tiny problem with them together with low-Ug tubes like ECL86. Maybe one should add some positive gridbias to make more room and compensate? Or go for cathode-resistor and accept higher THD?
I have a 6L6(807 w/o top cap)+6E5P(trioded) on the drawingboard, but that one will be a Schaded-SE. Will use a rewired 125ESE at 3,4kohm for that one.
I suggested a servo to compensate for current-differences in the OPT only. The servo has to be referenced to B+. Don´t know how to design it though.
This one was ment to be a Spud(single tube amp). Have to breadboard this one to see what happens IRL with respect to current imbalance. Also have a pair of Nuovotem 6V-toroids waiting. Might try a simple currentmirror first.
As currentmirrors/CCS/LM317 use some of the necessary cathode-swing there will be a tiny problem with them together with low-Ug tubes like ECL86. Maybe one should add some positive gridbias to make more room and compensate? Or go for cathode-resistor and accept higher THD?
I have a 6L6(807 w/o top cap)+6E5P(trioded) on the drawingboard, but that one will be a Schaded-SE. Will use a rewired 125ESE at 3,4kohm for that one.
Last edited:
I had a similar problem with designing a 6080 amp with the OT in the cathodes. My problem was that I needed perfect current balance which could be achieved with a CCS, but that would have created different cathode voltages - which would have created a DC current through the OT. I am still thinking about that one. I think Broskie has a DC servo which would address my issue. He revived the notion of the Garter Bias, and has recently produced a transistorized version. Interesting musing - but not a solution to your particular problem.
I am not certain that there is a solution to your particular problem. It seems to me that a DC servo on the anode side is likely to inject large amounts of distortion (I could be wrong though).
Have you considered going triode or pentode and supplying the DC to each winding side with a CCS (ie not CT'd). The cathodes would have to be none CCS which would make the self splitting very poor.
I really don't think there is a good solution to this issue.
Shoog
I am not certain that there is a solution to your particular problem. It seems to me that a DC servo on the anode side is likely to inject large amounts of distortion (I could be wrong though).
Have you considered going triode or pentode and supplying the DC to each winding side with a CCS (ie not CT'd). The cathodes would have to be none CCS which would make the self splitting very poor.
I really don't think there is a good solution to this issue.
Shoog
Ahh, I hadn't caught on to the screen current imbalance with a pentode on one side and a Mosfet on the other. I see the problem now.
Some ideas:
1) The dual pentode idea I mentioned above, with a gm booster on the slave, that at least matches the two sides current draw to 1st order.
2) The screen current drawn by the pentode side could be returned to the plate circuit by using a Mosfet follower (with g2 stopper resistor) to power the screen grid. The follower's drain then gets a big cap to the plate for AC current and a CCS to a somewhat higher B+ than the plate for DC current. (that way the plate voltage can drop below the screen voltage and the Mosfet still works.) The follower's gate just gets a stopper resistor to a DC reference. (the reference could be a Zener to ground, with a pull-up resistor to B+. Putting a big cap on the Zener will provide for slow ramp up of the screen voltage at power-up)
The Mosfet follower gate, being at HV already, could provide a convenient control point for a DC servo operating off the xfmr primary voltages. (A P type SS diffl. pair comparing the primary voltages and then controlling the reference V on the gate. Or this could control the gate V of the P-P slave side Mosfet, but that's near ground potential, so HV P type diffl parts needed then.)
3) Power the screen grid from a CCS to B+, and put a Zener diode from the screen grid down to the cathode. This subtracts screen current from the cathode, leaving just plate current plus a constant CCS current. Now the cathode and Mosfet source currents can be controlled by various schemes at ground potential.
Easiest would be just two CCS's with a big cap between them at the bottom (Ken's idea, like Shoog's design, separate CCS on left and right sides). The cathode CCS would need to include the top side CCS current plus the plate current.
Oh, if measuring the xfmr primary voltages for servo-ing, keep in mind that the two primary resistances may not be exactly the same. Hmm, the Shade feedback resistor DC current needs considering too. Could just put a dummy resistor on the other side I guess. Or use the xfmr V servo approach.
Don
Some ideas:
1) The dual pentode idea I mentioned above, with a gm booster on the slave, that at least matches the two sides current draw to 1st order.
2) The screen current drawn by the pentode side could be returned to the plate circuit by using a Mosfet follower (with g2 stopper resistor) to power the screen grid. The follower's drain then gets a big cap to the plate for AC current and a CCS to a somewhat higher B+ than the plate for DC current. (that way the plate voltage can drop below the screen voltage and the Mosfet still works.) The follower's gate just gets a stopper resistor to a DC reference. (the reference could be a Zener to ground, with a pull-up resistor to B+. Putting a big cap on the Zener will provide for slow ramp up of the screen voltage at power-up)
The Mosfet follower gate, being at HV already, could provide a convenient control point for a DC servo operating off the xfmr primary voltages. (A P type SS diffl. pair comparing the primary voltages and then controlling the reference V on the gate. Or this could control the gate V of the P-P slave side Mosfet, but that's near ground potential, so HV P type diffl parts needed then.)
3) Power the screen grid from a CCS to B+, and put a Zener diode from the screen grid down to the cathode. This subtracts screen current from the cathode, leaving just plate current plus a constant CCS current. Now the cathode and Mosfet source currents can be controlled by various schemes at ground potential.
Easiest would be just two CCS's with a big cap between them at the bottom (Ken's idea, like Shoog's design, separate CCS on left and right sides). The cathode CCS would need to include the top side CCS current plus the plate current.
Oh, if measuring the xfmr primary voltages for servo-ing, keep in mind that the two primary resistances may not be exactly the same. Hmm, the Shade feedback resistor DC current needs considering too. Could just put a dummy resistor on the other side I guess. Or use the xfmr V servo approach.
Don
Last edited:
re: Ken
"One could boost the effective gm of a pentode on anti-duty by
injecting some see-saw paraphase to either the grid or screen."
Hmm, it does boost the gm, but I think this approach will null out the SE effect. Notice that extreme slave gm would cause the common cathodes to freeze up voltage-wise.
Putting equal complementary drive voltages on both grid sides (aka normal P-P) does nearly the same thing except for some residual odd order harmonic wobble left on the cathodes.
I think to get SE effect, you will need some local neg. feedback loop, with gain, monitoring the common cathode voltage (to freeze it solid), like I mentioned above for the dual identical pentode scheme.
Don
"One could boost the effective gm of a pentode on anti-duty by
injecting some see-saw paraphase to either the grid or screen."
Hmm, it does boost the gm, but I think this approach will null out the SE effect. Notice that extreme slave gm would cause the common cathodes to freeze up voltage-wise.
Putting equal complementary drive voltages on both grid sides (aka normal P-P) does nearly the same thing except for some residual odd order harmonic wobble left on the cathodes.
I think to get SE effect, you will need some local neg. feedback loop, with gain, monitoring the common cathode voltage (to freeze it solid), like I mentioned above for the dual identical pentode scheme.
Don
If you get the plates doing the equal and opposite, I think you find the cathode voltage
between them will freeze, for the full SE effect. Same as the folded cascode model with
the much higher Gm on the right.
I second your suggestion for screen CCS, with shunt regulation (zenier dump to cathode).
This gives the stable answer for cathode bias in almost every Pentode circuit.
between them will freeze, for the full SE effect. Same as the folded cascode model with
the much higher Gm on the right.
I second your suggestion for screen CCS, with shunt regulation (zenier dump to cathode).
This gives the stable answer for cathode bias in almost every Pentode circuit.
Last edited:
There will be a difference in sonics between the Zener-shunted screen scheme and the Mosfet follower-ed screen scheme.
The Zener-shunteded scheme throws out the screen current, giving the usual sloping/curved pentode plate curves normally seen on datasheets. This reduces some 2nd harmonic by compressing the curve set on the left side. Some tubes with "controlled knee" characteristics take this to extremes to get rid of 2nd harmonic (single ended video amp stages, 6GN8, 6JG5, 6JE8 for example).
The Mosfet follower-ed screen scheme adds the screen current back into the plate output, giving nice flat top pentode curve characteristics (not like seen on the data sheets). More parts though. Still gives 3/2 power versus the g1 voltage though, but more 2nd harmonic.
The Zener-shunteded scheme throws out the screen current, giving the usual sloping/curved pentode plate curves normally seen on datasheets. This reduces some 2nd harmonic by compressing the curve set on the left side. Some tubes with "controlled knee" characteristics take this to extremes to get rid of 2nd harmonic (single ended video amp stages, 6GN8, 6JG5, 6JE8 for example).
The Mosfet follower-ed screen scheme adds the screen current back into the plate output, giving nice flat top pentode curve characteristics (not like seen on the data sheets). More parts though. Still gives 3/2 power versus the g1 voltage though, but more 2nd harmonic.
Last edited:
It works for plate V lower than screen V. The CCS goes to a higher B+ than the plates use, so the Mosfet drain can swing further. (actually, the "CCS" has to be a DC stabilized gyrator so as to set the drain idle voltage in the middle of the swing range, and the higher CCS B+ has to be quite a bit higher for a center tapped xfmr load that swings above B+) The drain is cap coupled to the plate. The 2nd B+ is a pain though.
Alternatively, could just put a 100 V Zener and paralled cap between drain and plate, so the drain stays 100 V above the plate. Then the CCS gets a reverse protection diode in series with it and can then go to the normal B+. The CCS keeps the cap charged during half of the sine wave. Sort of a 50% duty bootstrap arrangement.
Or just use a resistor in place of the CCS. Skip the 100V Zener too. Drain will average at B+. But will bootstrap to 2xB+. (to clarify: drain has cap to plate and resistor to B+, source drives screen, gate goes to reference DC)
Alternatively, could just put a 100 V Zener and paralled cap between drain and plate, so the drain stays 100 V above the plate. Then the CCS gets a reverse protection diode in series with it and can then go to the normal B+. The CCS keeps the cap charged during half of the sine wave. Sort of a 50% duty bootstrap arrangement.
Or just use a resistor in place of the CCS. Skip the 100V Zener too. Drain will average at B+. But will bootstrap to 2xB+. (to clarify: drain has cap to plate and resistor to B+, source drives screen, gate goes to reference DC)
Last edited:
DN2540 datasheet:
http://www.supertex.com/pdf/datasheets/DN2540.pdf
They claim it is free from thermal runaway. BUT, I see two opposite variations on the graphs. The transconductance goes down with increasing temp. but turn off gate threshold increases with temp. too. Looks like it is probably safe if you keep the temp. below 100 degrees C.
http://www.supertex.com/pdf/datasheets/DN2540.pdf
They claim it is free from thermal runaway. BUT, I see two opposite variations on the graphs. The transconductance goes down with increasing temp. but turn off gate threshold increases with temp. too. Looks like it is probably safe if you keep the temp. below 100 degrees C.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.