Well, one weird turn deserves another.
This amplifier output stage takes the Graeme Cohen design to upside down.
This "Twilight-Zone" amplifier takes a P-P output stage and splits it into two SE amplifier outputs. It even looks practical and economical. It may be pointless, but that remains to be seen. Ordinary P-P OTs are used ( no gaps ), so some cheap speaker 70V distributon Xfmrs could be used.
The batteries shown as B1 and B2, or floating supplies, are low voltage. They are between two points with the same nominal AC voltage, but they insert DC current to just null the tube DC currents in the primary windings ( DC current = Vbatt / Rwinding ). Since each tube will have some small opposite SE distortion components, the small inductors L1 and L2 are inserted to isolate the difference. These could be just a string of ferrite beads, and could be bifilar ( two wires thru the beads ).
If the design is stretched to amplify two -different- ( L and R ) SE signals, the L1 and L2 inductors could be something like a 1000:1000 Ohm bifilar interstage xfmer possibly. They have to handle the DC tube compensation currents, while isolating the AC signal differences.
This amplifier output stage takes the Graeme Cohen design to upside down.
This "Twilight-Zone" amplifier takes a P-P output stage and splits it into two SE amplifier outputs. It even looks practical and economical. It may be pointless, but that remains to be seen. Ordinary P-P OTs are used ( no gaps ), so some cheap speaker 70V distributon Xfmrs could be used.
The batteries shown as B1 and B2, or floating supplies, are low voltage. They are between two points with the same nominal AC voltage, but they insert DC current to just null the tube DC currents in the primary windings ( DC current = Vbatt / Rwinding ). Since each tube will have some small opposite SE distortion components, the small inductors L1 and L2 are inserted to isolate the difference. These could be just a string of ferrite beads, and could be bifilar ( two wires thru the beads ).
If the design is stretched to amplify two -different- ( L and R ) SE signals, the L1 and L2 inductors could be something like a 1000:1000 Ohm bifilar interstage xfmer possibly. They have to handle the DC tube compensation currents, while isolating the AC signal differences.
Last edited:
Looks like one can go even further. Use just one P-P OT with the two tubes as normal and put the L1, L2 inductors in series with the tube plates. No batteries this time. L1, L2 get changed to very small xfmers with the same turns ratio (to secondary) as the main OT. The L1 and L2 still absorb the distortion differences between the tube outputs (class A ), but then their low turn secondary is put in series (separately) with the main OT secondary to re-form the two SE outputs again. They can be low turns for the same stage input signal, since they only carry the SE distortion component voltages. Heavier wire for the secondary side wire.
For the case of two separate ( L and R ) input signals, one is just back to the old classic econo scheme of a small OT to carry the difference signal to the main OT channel, and L and R are reconstituted at the main OT secondary by adding/subtracting the difference channel secondaries. Usually found in some stereo consoles.
The difference channel was often done as a small SE Amp, and sometimes combined with the main OT using a gapped center leg E lamination. P-P around the outside path and SE thru the center leg. Not practical for DIY though.
Two birds with one stone.
For the case of two separate ( L and R ) input signals, one is just back to the old classic econo scheme of a small OT to carry the difference signal to the main OT channel, and L and R are reconstituted at the main OT secondary by adding/subtracting the difference channel secondaries. Usually found in some stereo consoles.
The difference channel was often done as a small SE Amp, and sometimes combined with the main OT using a gapped center leg E lamination. P-P around the outside path and SE thru the center leg. Not practical for DIY though.
Two birds with one stone.
Last edited:
Thanks fellow DIYers; I'm beginning to understand.
For the relatively small advantage of making something useful out of bits someone has lying around and might as well dispose of in something other than the trash can / metal recycle house, the disadvantages win.
1. The doubled up or at least higher power supply voltage value. However that comes about, it is an offsetting extra expense.
2. An inductor in the plate circuit. A good inductor. Or a SS constant current source, or large value resistor or tube current source with its own isolated heater power. If that part cant be made simply from another distribution transformer primary winding, its a significant offset. Plus that part burns a lot of energy; managing that heat offsets any hoped for "savings".
3. Mark mentioned that - in a guitar amp context - perhaps a semi-saturated iron core sounds way different than a core at zero on its BH curve. An "all for nought" possible outcome is a significant offset if building a guitar amp this way.
4. Thanks for the link to the Broskie article! Sounds like PSRR is an issue with the Parafeed and any expense in dealing with that, or implementing some feedback trick to get rid of it is an offset to any "economy" in making use of these particular parts. Although same is true using a proper SE OPT I believe.
For the relatively small advantage of making something useful out of bits someone has lying around and might as well dispose of in something other than the trash can / metal recycle house, the disadvantages win.
1. The doubled up or at least higher power supply voltage value. However that comes about, it is an offsetting extra expense.
2. An inductor in the plate circuit. A good inductor. Or a SS constant current source, or large value resistor or tube current source with its own isolated heater power. If that part cant be made simply from another distribution transformer primary winding, its a significant offset. Plus that part burns a lot of energy; managing that heat offsets any hoped for "savings".
3. Mark mentioned that - in a guitar amp context - perhaps a semi-saturated iron core sounds way different than a core at zero on its BH curve. An "all for nought" possible outcome is a significant offset if building a guitar amp this way.
4. Thanks for the link to the Broskie article! Sounds like PSRR is an issue with the Parafeed and any expense in dealing with that, or implementing some feedback trick to get rid of it is an offset to any "economy" in making use of these particular parts. Although same is true using a proper SE OPT I believe.
Read Broskie's article again:
quote:
A parafeed power amplifier sidesteps the problem of poor PSRR that most single-ended amplifiers exhibit.
:unquote
poor PSRR is what plagues conventional SE designs
while it is less of a problem with parafeed ...
quote:
A parafeed power amplifier sidesteps the problem of poor PSRR that most single-ended amplifiers exhibit.
:unquote
poor PSRR is what plagues conventional SE designs
while it is less of a problem with parafeed ...
But isnt that if you design it with the large plate inductor - and/or use his plate to cathode feedback? A resistor plate load still has the problem?
true, the bigger the resistor value, the less ripple appears at the plate,
but a resistive plate load not only burns an insane amount of power, it also gives you only 1/2 the voltage swing (1/4 output power) for the same supply voltage compared to choke load (parafeed) or transformer load (non parafeed), therefore
I don' t think its a viable option at all ...
but a resistive plate load not only burns an insane amount of power, it also gives you only 1/2 the voltage swing (1/4 output power) for the same supply voltage compared to choke load (parafeed) or transformer load (non parafeed), therefore
I don' t think its a viable option at all ...
Maybe I should just buy this https://decilliaamplification.com/single-25 Only $3k+ and a year to wait for one, but probably built with a lot less Mickey in the mouse...
Like these by Steve Bench? 👍
Steve Bench Link
http://diyaudioprojects.com/mirror/members.aol.com/sbench101/
http://diyaudioprojects.com/mirror/members.aol.com/sbench101/#THE MATRIX
Looks like DSB SC
🙂
Yep, you are entering the MATRIX. Not recommended for a hobbyist intro. though. Half the tubes, but till needs two OTs.
(DSB SC double sideband suppressed carrier )
For cheap OTs, the Circlotron allows one to series connect industrial control xfmrs to get enough primary turns. Then a split bobbin industrial xfmr for making the two floating power supplies.
For lowest cost, I favor a modest tube OTL with P and N type current mirrors with gain for final outputs, arranged to use low voltage power for the output stage. Low heat, no OT, cheap, and authentic tube sound preserved. Unfortunately, one needs to do some transistor matching to make the mirrors.
(DSB SC double sideband suppressed carrier )
For cheap OTs, the Circlotron allows one to series connect industrial control xfmrs to get enough primary turns. Then a split bobbin industrial xfmr for making the two floating power supplies.
For lowest cost, I favor a modest tube OTL with P and N type current mirrors with gain for final outputs, arranged to use low voltage power for the output stage. Low heat, no OT, cheap, and authentic tube sound preserved. Unfortunately, one needs to do some transistor matching to make the mirrors.
Last edited:
The N Crowhurst Twin Coupled OT Amp would allow cheapo OTs from industrial control xfmrs, since the primary turns can be effectively doubled. You have an example Amp even. Using cheap 6LU8 tubes as I recall. I would say this is the most practical and low cost Amp topology for cheap OTs. Need to find some industrial control xfmrs with dual 240V primaries and a 12 V secondary.
Last edited:
Circlotron, McIntosh and Crowhurst were the Gen2 designs of their era. All identically maximized OPT performance (possibilities of bifilar lower impedance windings) by stressing driver stages (relatively easy gigs, but not trivial) and one could easily argue that this is still the best strategy in 2025.
The single-ended argument that biasing the OPT away from the zero-crossing in the B/H curve remains. I built one in the mid 1990s, during the Sound Practices cult era, with the requisite type 845 outputs and octal drivers, oil capacitors, as Anthony Quinn said in Zorba the Greek "the whole catastrophy" (those too young to remember the reference will live to repeat it, but I digress). The kink in the iron core B/H curve of permeability through the zero crossing only appears as an inductive shunt across the output valves' load.
A certain gentleperson who contrributes here has informed us of both the importance of and a strategy to minimize distortion arising from B/H curve variation, emphasizing the AP patented combination of positive current sensing feedback and negative voltage sensing feedback. But he's some kind of wizz-bang and nobody's caught on yet.
Always the best fortune,
Chris
The single-ended argument that biasing the OPT away from the zero-crossing in the B/H curve remains. I built one in the mid 1990s, during the Sound Practices cult era, with the requisite type 845 outputs and octal drivers, oil capacitors, as Anthony Quinn said in Zorba the Greek "the whole catastrophy" (those too young to remember the reference will live to repeat it, but I digress). The kink in the iron core B/H curve of permeability through the zero crossing only appears as an inductive shunt across the output valves' load.
A certain gentleperson who contrributes here has informed us of both the importance of and a strategy to minimize distortion arising from B/H curve variation, emphasizing the AP patented combination of positive current sensing feedback and negative voltage sensing feedback. But he's some kind of wizz-bang and nobody's caught on yet.
Always the best fortune,
Chris
As one of them has DC in the primary and both transformers have to be identical, they both need to be gapped ... 🙁
I did a quick sim of Mr.Cohen's dual SE and the output stage alone did not suppress both odd and even harmonics as claimed in his paper. Just the even harmonics. Looks exactly like any conventional PP. It seems that the symmetric negative feedback across 3 stages actually does away with 3rd and 5th.