Howdy, Friends: Earlier this year I posted a fully differential amplifier using sweep tubes and meant to develop 60 or more watts. I had promised a follow up post for the power supply which might wind up being of more interest to some of y'all than even the amplifier. Well, since then, the good news is that my consulting work has picked up considerably. 
The bad news is that my consulting work has picked up considerably I have not had time to refine the power supply design, much less begin the work on the amp itself. But I continue to think about things and read those wonderful papers from the 30s through the early 60s, before "guruship" came into vogue. One approach that keeps on catching my eye is one proposed by Peterson and Sinclair, around 1951. They called it a "single ended push pull". While I normally don't mess with whatever title an author gives their idea, I would argue that the name is a misnomer. It's more of a stacked push pull.
But no matter, what is really cool about their idea is that the transformer primary halves have no dynamic voltage differential between them in operation. This means that the transformer is way easier to design and one concentrates strictly on super tight coupling with no need to be concerned about capacitance. As I said, cool. As a professionally experienced magnetics designer, I really appreciate this feature. On the other hand, driving this thing, while not nearly as brutal as driving a Mac style stage or a G2 drive stage, is still problematic, as the upper tube has its cathode floating up and down to the tune of plus and minus the B+ minus whatever the minimum plate voltage is for the tube. While the actual G1 drive voltage is totally the same as it would be for a normal P-P, the reference for this voltage is moving up and down. Doable, just adds some complexity.
A perhaps more serious consideration is the cathode to heater rating. With a sweep tube running at 600 volts peak to peak swing, the cathode is moving +/-300V to the heater. It will take a separate heater winding for each upper tube (stereo) and connect its CT to the cathode or some other appropriate node. Puts a burden on that transformer as it just became a path for the audio signal, especially higher frequencies, through any leakage capacitance, etc, it may have. (and it will!).
Another thing is my original thought of using "schade" type feedback for the finals needs to be examined for execution. Adding a bit of CFB should still be possible. Back to benefits, the output transformer ratio is lowered, always a good thing for better primary to secondary coupling. I would urge y'all to download the article, General Radio Experimenter, October 1951, "A New Push-Pull Amplifier Circuit". Anyway, strictly in the thought stage now, totally open to ideas. I will still need to complete the High Conduction Angle, Low Peak Current power supply. Fully regulated, will be very stiff and very quiet. Y'all have a good one, talk to all soon Rene
The bad news is that my consulting work has picked up considerably
I have not had time to refine the power supply design, much less begin the work on the amp itself. But I continue to think about things and read those wonderful papers from the 30s through the early 60s, before "guruship" came into vogue. One approach that keeps on catching my eye is one proposed by Peterson and Sinclair, around 1951. They called it a "single ended push pull". While I normally don't mess with whatever title an author gives their idea, I would argue that the name is a misnomer. It's more of a stacked push pull. But no matter, what is really cool about their idea is that the transformer primary halves have no dynamic voltage differential between them in operation. This means that the transformer is way easier to design and one concentrates strictly on super tight coupling with no need to be concerned about capacitance. As I said, cool. As a professionally experienced magnetics designer, I really appreciate this feature. On the other hand, driving this thing, while not nearly as brutal as driving a Mac style stage or a G2 drive stage, is still problematic, as the upper tube has its cathode floating up and down to the tune of plus and minus the B+ minus whatever the minimum plate voltage is for the tube. While the actual G1 drive voltage is totally the same as it would be for a normal P-P, the reference for this voltage is moving up and down. Doable, just adds some complexity.
A perhaps more serious consideration is the cathode to heater rating. With a sweep tube running at 600 volts peak to peak swing, the cathode is moving +/-300V to the heater. It will take a separate heater winding for each upper tube (stereo) and connect its CT to the cathode or some other appropriate node. Puts a burden on that transformer as it just became a path for the audio signal, especially higher frequencies, through any leakage capacitance, etc, it may have. (and it will!).
Another thing is my original thought of using "schade" type feedback for the finals needs to be examined for execution. Adding a bit of CFB should still be possible. Back to benefits, the output transformer ratio is lowered, always a good thing for better primary to secondary coupling. I would urge y'all to download the article, General Radio Experimenter, October 1951, "A New Push-Pull Amplifier Circuit". Anyway, strictly in the thought stage now, totally open to ideas. I will still need to complete the High Conduction Angle, Low Peak Current power supply. Fully regulated, will be very stiff and very quiet. Y'all have a good one, talk to all soon Rene
SRPP with one more tube added.
Philips' version was more elegant.
http://www.audioxpress.com/article/The-SPP-Amplifier
Philips' version was more elegant.
http://www.audioxpress.com/article/The-SPP-Amplifier
Last edited:
Figure 1 looked promising, except for the floating tube heater and 2X B+, and B-. Half the usual primary turns without switching transients is a BIG plus. Worth the trouble, and 2X B+ is easy to get with a full Bridge rect. off most "tube" power xfmrs.
Then figure 2 requires twice as many primary turns in two separate windings, plus electrolytic caps in the power signal path. The two primaries need to be wound bifilar to avoid the switching transients they mention being solved by figure 1. Needs 500V driver tubes. Still 2X B+, B-, and a floating tube heater. This is a step backwards.
Check out Circlotrons. Much better (50% CFB).
(or stick with figure 1 using a pentode LTP driver/splitter, only a single floating heater winding needed versus usual P-P amplifiers)
Or Norman Crowhurst's "Twin Coupled" amplifier. (just one B+ supply)
..
Then figure 2 requires twice as many primary turns in two separate windings, plus electrolytic caps in the power signal path. The two primaries need to be wound bifilar to avoid the switching transients they mention being solved by figure 1. Needs 500V driver tubes. Still 2X B+, B-, and a floating tube heater. This is a step backwards.
Check out Circlotrons. Much better (50% CFB).
(or stick with figure 1 using a pentode LTP driver/splitter, only a single floating heater winding needed versus usual P-P amplifiers)
Or Norman Crowhurst's "Twin Coupled" amplifier. (just one B+ supply)
..
Last edited:
The figure 1 design, with the triode Concertina splitter and B- supply, eliminates output stage grid coupling caps. Adjusting tube bias level for equal idle current needs some work though. (some DC servo or bias adjusts likely needed)
Using a 1000V Mosfet splitter (IXTP01N100) would work well for the splitter and that eliminates one floating heater supply.
This should be the GO-TO design for all those cheap-skates out there trying to use power transformers for OTs. One can even series connect multiple xfmr primaries to get sufficient V rating (at 20 Hz) and primary inductance. (parallel connect the secondaries)
Using a 1000V Mosfet splitter (IXTP01N100) would work well for the splitter and that eliminates one floating heater supply.
This should be the GO-TO design for all those cheap-skates out there trying to use power transformers for OTs. One can even series connect multiple xfmr primaries to get sufficient V rating (at 20 Hz) and primary inductance. (parallel connect the secondaries)
Wow, lots of attention! Thanks, Y'all.
Now, to be sure, the Peterson/Sinclair version is a full blown, class AB or even B amplifier. It looked to me at first glance (all the time I have right now) that the Philips and other versions are all class A. I'm personally not the least concerned about electrolytic caps in the path, especially since in this circuit, they see little AC and are there to improve coupling between the primary halves. I want the power and efficiency of class AB or even B, practical with the glitch free topology.
As pointed out by Smoking Amp, the elimination, fundamentally, of switching transients by having zero dynamic differential potential change during any part of the cycle is super tremendous, and it's the ONLY reason I would consider going this route. I had already designed (but not wound) the transformer for the initial circuit and this one would be so, so much simpler and potentially better since no consideration need be given to interwinding capacitance and one goes for the best primary interwinding coupling possible (there is still intra winding capacitance and primary to secondary capacitance concerns but those are much more manageable than primary to primary considerations).
The power supply does not concern me much. Figure 6 of the original paper shows a way to use a single B+ supply and sweep tubes are happy with higher current and B+ in the 350V range. A concern is the care and feeding of the upper G2, already thinking about that, also doable.
The floating heater supply for the upper tube, while potentially inexpensive requiring only a simple transformer or separate winding, is still a potential liability since a leakage path to ground has thus been created for the audio signals, especially high frequency. But the node is relatively low impedance, less than 1K, the driving power is high, and NFB will help in making things flat. So maybe I'm putting too much weight on this issue.
I do like the big MOSFET driver idea if I go split load (concertina) driver way but I really want to maintain the original full differential thought so I'll look at how that kind of driver could be coupled. Already have some thoughts but I confess that so far, all this is just that, thoughts. No pencil has yet met paper on this subject, I have no calculations at all yet.
Thanks again, please keep it coming.
Rene
Now, to be sure, the Peterson/Sinclair version is a full blown, class AB or even B amplifier. It looked to me at first glance (all the time I have right now) that the Philips and other versions are all class A. I'm personally not the least concerned about electrolytic caps in the path, especially since in this circuit, they see little AC and are there to improve coupling between the primary halves. I want the power and efficiency of class AB or even B, practical with the glitch free topology.
As pointed out by Smoking Amp, the elimination, fundamentally, of switching transients by having zero dynamic differential potential change during any part of the cycle is super tremendous, and it's the ONLY reason I would consider going this route. I had already designed (but not wound) the transformer for the initial circuit and this one would be so, so much simpler and potentially better since no consideration need be given to interwinding capacitance and one goes for the best primary interwinding coupling possible (there is still intra winding capacitance and primary to secondary capacitance concerns but those are much more manageable than primary to primary considerations).
The power supply does not concern me much. Figure 6 of the original paper shows a way to use a single B+ supply and sweep tubes are happy with higher current and B+ in the 350V range. A concern is the care and feeding of the upper G2, already thinking about that, also doable.
The floating heater supply for the upper tube, while potentially inexpensive requiring only a simple transformer or separate winding, is still a potential liability since a leakage path to ground has thus been created for the audio signals, especially high frequency. But the node is relatively low impedance, less than 1K, the driving power is high, and NFB will help in making things flat. So maybe I'm putting too much weight on this issue.
I do like the big MOSFET driver idea if I go split load (concertina) driver way but I really want to maintain the original full differential thought so I'll look at how that kind of driver could be coupled. Already have some thoughts but I confess that so far, all this is just that, thoughts. No pencil has yet met paper on this subject, I have no calculations at all yet.
Thanks again, please keep it coming.
Rene
OK, folks, a few more thoughts on the application of this topology, at least for my specific goals.
The drive can indeed be obtained via an LTP pair, with the driver for the upper tube having its plate load resistor connected to the plate of the lower output tube (I'll call the upper output tube Vup and the lower Vlo to cut down on typing). Coupling does need to be either via capacitors or via a MOSFET follower. I'll look at each later as a detail.
Fixed biasing is straightforward for both.
Schade type feedback still works, as will CFB.
G2 supply for Vup is via a series regulator with a big enough output cap connected either from G2 to cathode or between G2 and bottom of the CFB winding if used (for a measure of UL operation). The regulator can be a simple device, a high enough voltage MOSFET with a zener reference returned to the cathode per above, in a simple Source follower configuration.
As to the cathode/heater consideration. The tube I want to use is the 6HJ5 sweep tube. It is hefty. If operated in class B1, which is not scary now with the good coupling available from the OT, with a B+ of 260 but limiting the swing to 200, the heater-cathode limit of 200 is respected. With a G2 voltage of 125, the power is now down to 40W, still good for me. Little in the way of curves is available for other G2 voltages but scaling is possible and could be I can get back to 60W with the lower B+ voltage by raising the G2 voltage.
Plate load impedance is only 520 Ohms. Cool! Looking forward to the better coupling between Pri and Sec possible with the lower turns ratio of only about 8:1, assuming an 8 Ohm speaker. Both the impedance and ratio will only get lower with a higher G2 voltage.
The above represent the only pencil to paper I have applied so far, need to do a lot more calculations. I'll try to do those in the next week or two and publish results and perhaps at least a scan of a paper sketch.
Rene
The drive can indeed be obtained via an LTP pair, with the driver for the upper tube having its plate load resistor connected to the plate of the lower output tube (I'll call the upper output tube Vup and the lower Vlo to cut down on typing). Coupling does need to be either via capacitors or via a MOSFET follower. I'll look at each later as a detail.
Fixed biasing is straightforward for both.
Schade type feedback still works, as will CFB.
G2 supply for Vup is via a series regulator with a big enough output cap connected either from G2 to cathode or between G2 and bottom of the CFB winding if used (for a measure of UL operation). The regulator can be a simple device, a high enough voltage MOSFET with a zener reference returned to the cathode per above, in a simple Source follower configuration.
As to the cathode/heater consideration. The tube I want to use is the 6HJ5 sweep tube. It is hefty. If operated in class B1, which is not scary now with the good coupling available from the OT, with a B+ of 260 but limiting the swing to 200, the heater-cathode limit of 200 is respected. With a G2 voltage of 125, the power is now down to 40W, still good for me. Little in the way of curves is available for other G2 voltages but scaling is possible and could be I can get back to 60W with the lower B+ voltage by raising the G2 voltage.
Plate load impedance is only 520 Ohms. Cool! Looking forward to the better coupling between Pri and Sec possible with the lower turns ratio of only about 8:1, assuming an 8 Ohm speaker. Both the impedance and ratio will only get lower with a higher G2 voltage.
The above represent the only pencil to paper I have applied so far, need to do a lot more calculations. I'll try to do those in the next week or two and publish results and perhaps at least a scan of a paper sketch.
Rene
6HD5 (same tube, different basing) has curves for Vg2 at 135 Volts. Can figure the scaling between them for grid2. (likely 1.5 power V to I at grid 2)
http://tubedata.milbert.com/sheets/084/6/6HD5.pdf
http://tubedata.milbert.com/sheets/084/6/6HJ5.pdf
and attached is, I think, 6HJ5 curves for 160Vg2
http://tubedata.milbert.com/sheets/084/6/6HD5.pdf
http://tubedata.milbert.com/sheets/084/6/6HJ5.pdf
and attached is, I think, 6HJ5 curves for 160Vg2
Yup, knew that, thanks! It appears that one is a somewhat of a pinout version of the other. Back when I was researching the older thread, I also found an octal, top cap version which is almost a match for either one of these two. Can't remember the type now, dropped it because the XYL said she thought no top caps look better. Bummer, I really wanted the top cap.
Back to the 6hd5/j5, the respective curves are at somewhat different G2 voltages which provides another data point. And, if the curves you provided are indeed for the 8HJ5, then 60W with 260V B+ is a sure bet.
Thanks!
Rene
Back to the 6hd5/j5, the respective curves are at somewhat different G2 voltages which provides another data point. And, if the curves you provided are indeed for the 8HJ5, then 60W with 260V B+ is a sure bet.
Thanks!
Rene
Edits:
Meant to say 6 and not 8.
Also meant to say 60W with the lower voltage provided the stage is run in Class B. Haven't calculated but intuitively I don't think AB will cut it, the tubes will overdissipate at idle.
An often missed advantage of Class B is the continuous load line with minimal curving when near zero, and a continuous impedance for the load line. May be why the Mighty Macs did so very well though operated in Class B.
Cheers
1
Meant to say 6 and not 8.
Also meant to say 60W with the lower voltage provided the stage is run in Class B. Haven't calculated but intuitively I don't think AB will cut it, the tubes will overdissipate at idle.
An often missed advantage of Class B is the continuous load line with minimal curving when near zero, and a continuous impedance for the load line. May be why the Mighty Macs did so very well though operated in Class B.
Cheers
1
Here is an interesting (definitely non-mainstream) idea by Goldenbeer for unusual "Schade"-like Fdbk. Its using Positive Fdbk "Schade" connections to remove 3rd harmonic of the LTP driver stage and to increase stage gain, for use by an outer N Fdbk loop. Not sure how one would merge this with a totem pole output stage and driver though. Some head scratching called for.
Interesting if applied to the RCA 50 Watt (tube manual) amplifier, since it would put hi-Z at the pentode driver plates instead of the usual low-Z of typical "Schade" designs. So the driver cathode N Fdbks will develop high loop gain beside linearizing the V to I function of the drivers.
http://www.diyaudio.com/forums/tube...ate-cathode-local-nfb-same-2.html#post5111755
The Positive "Schade" Fdbks at the driver plates could also assist the driver stage to develop large signal swings for a screen grid drive, or Crazy/Twin drive, output stage (Mosfet followers for grid 2 current demands). (BTW, none of this has been tested, simulate first!)
..
Interesting if applied to the RCA 50 Watt (tube manual) amplifier, since it would put hi-Z at the pentode driver plates instead of the usual low-Z of typical "Schade" designs. So the driver cathode N Fdbks will develop high loop gain beside linearizing the V to I function of the drivers.
http://www.diyaudio.com/forums/tube...ate-cathode-local-nfb-same-2.html#post5111755
The Positive "Schade" Fdbks at the driver plates could also assist the driver stage to develop large signal swings for a screen grid drive, or Crazy/Twin drive, output stage (Mosfet followers for grid 2 current demands). (BTW, none of this has been tested, simulate first!)
..
Last edited:
Intriguing ideas for sure! Up against the clock today and did not follow the given links yet, but I will first opportunity. The original Schade feedback scheme is, of course, NFB but of course this depends upon the phasing of the node to which it is introduced.
I would argue, however, that a proper plate to plate Schade FB scheme absolutely depends on the driver tube having a high plate resistance so it behaves as a constant current vs plate voltage node and thus make the operation effective. Anyway, let me shut up until I can follow the links
This is getting more and more exciting, folks.
Thanks for the feedback, please keep it coming.
Rene
I would argue, however, that a proper plate to plate Schade FB scheme absolutely depends on the driver tube having a high plate resistance so it behaves as a constant current vs plate voltage node and thus make the operation effective. Anyway, let me shut up until I can follow the links
This is getting more and more exciting, folks.
Thanks for the feedback, please keep it coming.
Rene
"that a proper plate to plate Schade FB scheme absolutely depends on the driver tube having a high plate resistance so it behaves as a constant current vs plate voltage node"
You are quite right. I had to go back and look at the Goldenbeer scheme again to refresh my memory, and it IS designed to work with Triode drivers, not pentodes.
Possibly a pentode driver with low Ri could work, unknown. One needs to check the polarity (expansive/compresive) of 3rd Harmonic for a pentode LTP as well.
Another possibility might be to put the positive feedbacks back to some pentode driver screen grids instead. (no longer crossed resistors then) Some simulation called for (assuming the tube models support accurate variable Vg2 effects, just 3/2 power law for Vg2 to Ip, but grid2 current modeling may be suspect)
You are quite right. I had to go back and look at the Goldenbeer scheme again to refresh my memory, and it IS designed to work with Triode drivers, not pentodes.
Possibly a pentode driver with low Ri could work, unknown. One needs to check the polarity (expansive/compresive) of 3rd Harmonic for a pentode LTP as well.
Another possibility might be to put the positive feedbacks back to some pentode driver screen grids instead. (no longer crossed resistors then) Some simulation called for (assuming the tube models support accurate variable Vg2 effects, just 3/2 power law for Vg2 to Ip, but grid2 current modeling may be suspect)
Last edited:
This is all turning into a very, very interesting intellectual challenge. Love it. Since it it still in the cards to use CFB via a coupled winding, the cathodes of Vup and Vlo become sources for a feedback signal which can go to the plates of respective drivers and become positive FB. A possibility anyway. Won't be current FB but voltage but, there and worthy of investigation.
Rene
Rene
OK, here is what I was able to do today on this project.
With a G2 voltage in the vicinity of 135-140, a B+ of 260 and swing restricted so 60V is minimum at the plate, 58 Watts of Class B1 can be obtained. Close enough.
Peak dissipation of 50 Watts for one half cycle occurs at Pout 25W. Average will be within (barely) the tube's rating. As I will likely be cranking more within the 10-15W range, this should be good. I'll buy a bunch of output tubes in case it proves they are not as sturdy as I think they really are. Maybe it will convince the XYL to let me rip out the duodecals and put in proper octals with cool plate caps!
Originally I thought I'd be able to feed the G2 of the lower tubes with a conventional regulated output and the upper with a floating series regulator. But if that is done, the transformer primary currents won't be balanced. So both upper and lower tubes will each need its own floating G2 supply. To be sure, what is meant by floating is not a scary, separate and isolated PT secondary for each. Rather, it's that each of them will be referenced to the cathode of the tube it serves. A big cap will be needed on the output, plus an isolation diode to prevent back flow through the MOSFET body diode, so the output can be held up for the portion of the cycle where the plate to cathode voltage goes below the drop out level of the regulator.
I've done a crude paper diagram so all the trouble spots can be identified. Maybe I'll have time next week to do a proper initial design and publish. But so far, driving the output stage with a pair of pentodes (one half of a 6JV8) in LTP fashion, returned to some adequate negative voltage, say about 100-150V, with the plate resistor of the upper driver to the plate of Vlo and the lower driver's plate resistor to B+ should do it.
I probably should tear into the OT next. If I can find the iron and bobbins I can actually wind my own. Either way, I'll have to design the thing all the way and then decide to build or buy. Can't think of any show stoppers but won't know for sure until pencil meets paper.
Thanks, Y'all!
Rene
With a G2 voltage in the vicinity of 135-140, a B+ of 260 and swing restricted so 60V is minimum at the plate, 58 Watts of Class B1 can be obtained. Close enough.
Peak dissipation of 50 Watts for one half cycle occurs at Pout 25W. Average will be within (barely) the tube's rating. As I will likely be cranking more within the 10-15W range, this should be good. I'll buy a bunch of output tubes in case it proves they are not as sturdy as I think they really are. Maybe it will convince the XYL to let me rip out the duodecals and put in proper octals with cool plate caps!
Originally I thought I'd be able to feed the G2 of the lower tubes with a conventional regulated output and the upper with a floating series regulator. But if that is done, the transformer primary currents won't be balanced. So both upper and lower tubes will each need its own floating G2 supply. To be sure, what is meant by floating is not a scary, separate and isolated PT secondary for each. Rather, it's that each of them will be referenced to the cathode of the tube it serves. A big cap will be needed on the output, plus an isolation diode to prevent back flow through the MOSFET body diode, so the output can be held up for the portion of the cycle where the plate to cathode voltage goes below the drop out level of the regulator.
I've done a crude paper diagram so all the trouble spots can be identified. Maybe I'll have time next week to do a proper initial design and publish. But so far, driving the output stage with a pair of pentodes (one half of a 6JV8) in LTP fashion, returned to some adequate negative voltage, say about 100-150V, with the plate resistor of the upper driver to the plate of Vlo and the lower driver's plate resistor to B+ should do it.
I probably should tear into the OT next. If I can find the iron and bobbins I can actually wind my own. Either way, I'll have to design the thing all the way and then decide to build or buy. Can't think of any show stoppers but won't know for sure until pencil meets paper.
Thanks, Y'all!
Rene
The older octal type with plate cap which is very similar to the 6HD5 and 6HJ5 is the 6DQ5. It has a bit higher cathode peak current and I imagine can cool the plate a bit better due to the plate cap structure.
It does have published curves for a bunch of different G2 voltages.
Back to work!
Rene
It does have published curves for a bunch of different G2 voltages.
Back to work!
Rene
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.