Here is a dirt simple PP 2a3 amp. I've built something very similar, it works fine. You could increase the B+ to 400 volts without changing much, a few changes in the power supply is all, and try the 807s in triode mode, see how it sounds! Scroll down a bit to see the schematic.
https://audiokarma.org/forums/index...ilt-a-je-labs-push-pull-2a3-amplifier.929923/
https://audiokarma.org/forums/index...ilt-a-je-labs-push-pull-2a3-amplifier.929923/
Plate resistance of a triode is not is not an absolute figure.
rp (internal plate resistance) of 1700 ohms for 807 (or 6L6) is as measured at vintage operating points of Vp = 250V, Ip = 40mA. That's only 10W plate dissipation.
Using 6L6GC, you can easily increase the Ip (plate current) to 80mA with 250V Vp and get much lower rp. Probably down to 1000 ohms.
I'm assuming your 12" full-range Philips speakers are 8 ohms (or are they 16 ohms?). If 8 ohms, then each 807-triode in class A will 'see' a load of 4k ohms.
These days, when fashioning a push-pull triode amp design, many constructors will optimize the design with no global NFB. Then apply just enough NFB to 'take the edge off'. This often winds up being about 6dB NFB applied (lower the total amplifier gain by half). So if you design your amp to have 25X gain, apply NFB to end up with 12.5X gain and that means you've applied 6dB NFB.
Generally speaking, people who build push-pull triode-wired beam tube amps use too low a plate-plate primary impedance OPT, which gives more power but higher THD. In these cases, the better sound comes from applying a little NFB.
Typical OPT primary impedance for 6L6/807 is 6.6k ohms. Sometimes 5k ohms is used to get more power output (due to earlier onset of class AB operation) but that yields higher THD.
DTN Wiliamson specified a 10k OPT primary for better damping and more class A operation (both tubes in the push-pull pair stay conducting plate current in the face of higher input voltage swings). The penalty there is lower power output, but the benefit is lower THD.
You want lower THD? You get lower power out, but less NFB needed for pleasant sound.
You want more power output? You get more THD, so more NFB will be needed for pleasant sound.
Everything is a compromise.
_______________________________________
You have a quad of KT88s. Why not use them? Triode operation looks very promising...
With Vp = 250V and Ip = 160mA (!!), the rp is only 670 ohms. Used at Vp = 400V and Ip of perhaps 75mA, rp should still be no more than about 1000 ohms. Damping should be good, even without NFB, and you may get your 20W.
The GEC data sheet says this about KT88 push-pull triode class A operation:
With Vp-k (plate to cathode voltage) of 425V, Ip of 85mA (Pdiss = 36W), it says you can get 25W out from a 4k ohm PP load. The drive voltage needed is 35V peak (70V peak to peak).
If you want to be able to use the amp without global NFB, you may want to minimize THD in the first stage.
The following are just illustrations of basic push-pull tube amp topologies, not finished, tested designs.
I've found the best sound without global NFB came from a topology using a twin-triode Long-Tailed Pair (LTP) for the first stage phase splitter, DC-coupled to a push-pull differential driver, which drives the push-pull output stage. The cascade of differential stages helps AC balance, which helps cancel H2 and H4 distortion. Push-pull KT88s are going to need a fair amount of drive voltage, so you will likely need those three stages (Input LTP, PP driver, PP outputs). You will need at least one negative voltage supply for the tail of the input LTP.
The 'Mullard' style topology uses a single triode voltage amplifier stage, DC-coupled to a LTP phase splitter-driver stage, driving the output tubes. Since the phase splitting stage is also driving the output tubes, that stage has a difficult job. I think that's why 'Mullard' style designs have a bit higher THD open loop than the triple-PP stage topology. BUT... It's a lot easier to put together a 'Mullard' style amp. No negative voltage supply is needed since the LTP-driver stage sits on top of a large cathode resistor (R13 in the schematic below). Any remaining imbalance can be trimmed out by adjusting the value of the plate resistor of the non-inverting side of the LTP-driver stage (R3 in the schematic below).
The above could be improved by replacing R13 with a constant-current sink, which would enforce balance and remove the need for R3.
My experience from building a few amps of both the above topologies is that the 'triple-differential' (first schematic) style sounds generally cleaner, without that 'romantic' tubey warmth some crave. The 'Mullard' style (second schematic) sounds warmer, because of some H2 from that first single-ended voltage amplifier stage, and from more likely phase splitter imbalance.
The third standard topology is the 'Williamson' style, which you've been looking at. That one can be made to be very clean, but since there are four stages (input voltage amp > split load inverter > push-pull driver > push-pull outputs) there is likely to be low frequency instability, and if you apply a lot of NFB, high frequency instability too. You can make a 'Williamson' style amp without NFB, but then it will be very sensitive and noise becomes an issue. Perhaps a Williamson-style amp made with all low-mu triodes would be a worthwhile experiment.
In simulation, the above has open loop sensitivity of 132mV to 1W out into a 4 ohm load. It's shown with 8dB NFB applied (R4 and R20).
I suppose you could use a pentode or 12AX7 for the first stage into a split-load inverter, straight into the KT88 outputs, but then the first stage will be working very hard to provide the drive voltage necessary for the output stage. That single voltage amp will need to swing 50V or more into the split-load inverter grid. That's at the edge of what a 12AX7 can do, and I don't know what the linearity of a pentode would be swinging that much voltage. It might be okay, I don't know... I imagine you'd want to apply NFB to reduce THD, so you'd be better off with a pentode or UL output stage for more open loop gain with which to drive the NFB loop. And around and around we go...
rp (internal plate resistance) of 1700 ohms for 807 (or 6L6) is as measured at vintage operating points of Vp = 250V, Ip = 40mA. That's only 10W plate dissipation.
Using 6L6GC, you can easily increase the Ip (plate current) to 80mA with 250V Vp and get much lower rp. Probably down to 1000 ohms.
I'm assuming your 12" full-range Philips speakers are 8 ohms (or are they 16 ohms?). If 8 ohms, then each 807-triode in class A will 'see' a load of 4k ohms.
These days, when fashioning a push-pull triode amp design, many constructors will optimize the design with no global NFB. Then apply just enough NFB to 'take the edge off'. This often winds up being about 6dB NFB applied (lower the total amplifier gain by half). So if you design your amp to have 25X gain, apply NFB to end up with 12.5X gain and that means you've applied 6dB NFB.
Generally speaking, people who build push-pull triode-wired beam tube amps use too low a plate-plate primary impedance OPT, which gives more power but higher THD. In these cases, the better sound comes from applying a little NFB.
Typical OPT primary impedance for 6L6/807 is 6.6k ohms. Sometimes 5k ohms is used to get more power output (due to earlier onset of class AB operation) but that yields higher THD.
DTN Wiliamson specified a 10k OPT primary for better damping and more class A operation (both tubes in the push-pull pair stay conducting plate current in the face of higher input voltage swings). The penalty there is lower power output, but the benefit is lower THD.
You want lower THD? You get lower power out, but less NFB needed for pleasant sound.
You want more power output? You get more THD, so more NFB will be needed for pleasant sound.
Everything is a compromise.
_______________________________________
You have a quad of KT88s. Why not use them? Triode operation looks very promising...
With Vp = 250V and Ip = 160mA (!!), the rp is only 670 ohms. Used at Vp = 400V and Ip of perhaps 75mA, rp should still be no more than about 1000 ohms. Damping should be good, even without NFB, and you may get your 20W.
The GEC data sheet says this about KT88 push-pull triode class A operation:
With Vp-k (plate to cathode voltage) of 425V, Ip of 85mA (Pdiss = 36W), it says you can get 25W out from a 4k ohm PP load. The drive voltage needed is 35V peak (70V peak to peak).
- You have 8k primary OPTs. Using PP KT88-triodes and an 8 ohm speaker, you'll see less power, better damping, lower THD than above.
- Your OPTs are pretty big, so should be okay with the higher plate currents of KT88.
- If you have a 4 ohm speaker, your "8k ohm" OPT will present a load of 4k ohms to the KT88-triodes. Performance should be something like the above. Maybe include a switchable NFB loop, to enable better damping if/when you need/want it.
- You have a quad of KT88s already.
- A B+ of 425V is not too difficult to achieve. 450V and 500V rated passive parts are not difficult to get.
If you want to be able to use the amp without global NFB, you may want to minimize THD in the first stage.
The following are just illustrations of basic push-pull tube amp topologies, not finished, tested designs.
I've found the best sound without global NFB came from a topology using a twin-triode Long-Tailed Pair (LTP) for the first stage phase splitter, DC-coupled to a push-pull differential driver, which drives the push-pull output stage. The cascade of differential stages helps AC balance, which helps cancel H2 and H4 distortion. Push-pull KT88s are going to need a fair amount of drive voltage, so you will likely need those three stages (Input LTP, PP driver, PP outputs). You will need at least one negative voltage supply for the tail of the input LTP.
The 'Mullard' style topology uses a single triode voltage amplifier stage, DC-coupled to a LTP phase splitter-driver stage, driving the output tubes. Since the phase splitting stage is also driving the output tubes, that stage has a difficult job. I think that's why 'Mullard' style designs have a bit higher THD open loop than the triple-PP stage topology. BUT... It's a lot easier to put together a 'Mullard' style amp. No negative voltage supply is needed since the LTP-driver stage sits on top of a large cathode resistor (R13 in the schematic below). Any remaining imbalance can be trimmed out by adjusting the value of the plate resistor of the non-inverting side of the LTP-driver stage (R3 in the schematic below).
The above could be improved by replacing R13 with a constant-current sink, which would enforce balance and remove the need for R3.
My experience from building a few amps of both the above topologies is that the 'triple-differential' (first schematic) style sounds generally cleaner, without that 'romantic' tubey warmth some crave. The 'Mullard' style (second schematic) sounds warmer, because of some H2 from that first single-ended voltage amplifier stage, and from more likely phase splitter imbalance.
The third standard topology is the 'Williamson' style, which you've been looking at. That one can be made to be very clean, but since there are four stages (input voltage amp > split load inverter > push-pull driver > push-pull outputs) there is likely to be low frequency instability, and if you apply a lot of NFB, high frequency instability too. You can make a 'Williamson' style amp without NFB, but then it will be very sensitive and noise becomes an issue. Perhaps a Williamson-style amp made with all low-mu triodes would be a worthwhile experiment.
In simulation, the above has open loop sensitivity of 132mV to 1W out into a 4 ohm load. It's shown with 8dB NFB applied (R4 and R20).
I suppose you could use a pentode or 12AX7 for the first stage into a split-load inverter, straight into the KT88 outputs, but then the first stage will be working very hard to provide the drive voltage necessary for the output stage. That single voltage amp will need to swing 50V or more into the split-load inverter grid. That's at the edge of what a 12AX7 can do, and I don't know what the linearity of a pentode would be swinging that much voltage. It might be okay, I don't know... I imagine you'd want to apply NFB to reduce THD, so you'd be better off with a pentode or UL output stage for more open loop gain with which to drive the NFB loop. And around and around we go...
Last edited:
That's the 'Mullard' style topology done very simply, with no global NFB, and no attempt to enforce balance. I remember a long time ago, I heard an amp basically like that, same exact tubes used, built with Tango OPTs, at the old Fi shop in NYC. It sounded warm and sweet. Too syrupy and 'tubey' for my taste, but a nice sound.
This I can probably wire up as soon as I can get the 807 tube bases. Speaking of that what would be a good EU based source of 807 sockets and anode caps? I will prefer the NOS bakelite types.
Yes these are rated at 8 ohm. I believe these are mid 70's vintage. Both the drivers are in fine fettle but they do get shouty with my SE EL34 triode strapped.
I unfortunately do not have KT88's. Can your first simulation use 6SN7's instead of 6j6 and 12BH7?
I believe for the present I have sufficient pointers on how to take this forward. Lets first try the simple schematics suggestions and then try the more elaborate ones. Thanks everyone for the excellent set of advices. I humbly express my appreciations.
PS: I checked up the PT with me. The exact spec is as below:
Sec 1 : 370v - CT - 60v - 370v (180mA) and not 320 - 0 - 320 as mentioned by me in error.
Sec 2 : 5v (3A)
Sec 3 : 3.15 - 0 - 3.15 (2A)
Sec 4 : 6.3v (4A)
I also have a pair of chokes rated at 8H 150mA with a DCR of 90 ohms. Probably this would scrape by for mono block builds.
Let me sink in all the info and come up for any further guidance.
Chinese do not seem to offer Bakelite type sockets for 807. But as a last resort I will plonk for them.
Yes, it could be converted to 6SN7 for both the input LTP and the push-pull driver stage. I actually had a circuit like that running, years ago, to drive push-pull 300B tubes. It had a 6FQ7 in the first stage and 6SN7 in the second stage. It worked fine, but my 6SN7s kept crackling, going noisy. I didn't know about exceeding the heater to cathode limit back then! I gave up on that amp, unfortunately...
Not Built
Not Tested
Experimental
Example Only
? ? ? ?
They look like nice ideas.
Trust but Verify
No Verification?
You can try it if you want, and then report to us, or if needed ask for help.
I see one of those schematics does have a dominant pole series RC from the input tube plate to ground (often used to help tame global negative feedback). It can not be applied/used on some of the other schematics because of the topology; or unless it is a push and a pull matched dominant pole, from the plates to ground.
And regarding the word Simple, using global negative feedback is not always simple.
Especially the amplifier that has 2 sets of push and pull Coupling Cap RC networks (4 RC; 2 for one stage to stage; and another 2 for another stage to stage). That makes at least three low frequency poles one after the other; first RC pair, second RC pair, output transformer inductance (and Loudspeaker load as it reflects to the output transformer primary; might be considered to be 4 sequential low frequency poles).
Just a lot more complex of amplifiers than I ever expect to build.
Happy designing, happy building, happy listening.
Not Tested
Experimental
Example Only
? ? ? ?
They look like nice ideas.
Trust but Verify
No Verification?
You can try it if you want, and then report to us, or if needed ask for help.
I see one of those schematics does have a dominant pole series RC from the input tube plate to ground (often used to help tame global negative feedback). It can not be applied/used on some of the other schematics because of the topology; or unless it is a push and a pull matched dominant pole, from the plates to ground.
And regarding the word Simple, using global negative feedback is not always simple.
Especially the amplifier that has 2 sets of push and pull Coupling Cap RC networks (4 RC; 2 for one stage to stage; and another 2 for another stage to stage). That makes at least three low frequency poles one after the other; first RC pair, second RC pair, output transformer inductance (and Loudspeaker load as it reflects to the output transformer primary; might be considered to be 4 sequential low frequency poles).
Just a lot more complex of amplifiers than I ever expect to build.
Happy designing, happy building, happy listening.
Last edited:
True!
The more stages, the harder it gets to keep the amp stable with global NFB around it.
The 'Williamson' style would be about the toughest, as is often pointed out.
That 'Mullard' style circuit example I posted is actually a published design and was built (not by me, but by a well-respected member of this forum) something like 20 years ago.
The others are just untested, unbuilt examples, to show what the various basic topologies for push-pull 6L6-oids look like.
Then there's your basic Dynaco MkIII or Stereo 70 style. That one looks simple, but requires negative feedback to sound good, so it's not so simple after all.
If this circuit were to be used with a different OPT, the 750mmf (750pF) shunt cap around the 1000 1W feedback resistor would need to be adjusted on test for best looking square waves on a 'scope. There's also that 12pF cap from the plate supply feed to the grid of the 6AN8A triode (the split load inverter section), and the 390pF cap going from one of the KT88 screen grids to the cathode of the input stage pentode.
That's a fair amount of high frequency compensation going on for the feedback loop.
Looks simple at first, but there's some subtle stuff going on there...
Please forgive me for adding another inflection. With these transformers and chokes another option is available, a multiple B+ power supply. Using HV rectifiers in the 5U4, GZ34, etc. class and choke input, you'll get a B+ of about 300VDC at a comfortable 150mA. This value would allow you to optimize the operating point of your 807s in the sweetest, most class-A region, at the expense of peak power output, down to maybe +8dBW in full triode connection, but with best available linearity and best speaker driving characteristics.
The problem with 300VDC is that it's marginal for driver stages, so a secondary B+ needs to be generated. This can be done with a 1N4007 (or modern equivalent) diode connected at the junction of the HV rectifier's cathodes and the input choke, feeding some CRCRC filtering (or include another smaller-current choke - go crazy). This should give you somewhere North of 400VDC, enough to allow hi-mu triode direct coupled to medium-mu triode split-load inverter (the "7247 input stage") generous headroom.
A little extra effort, beyond what commercial designs could afford, but allows the signal path to be kept as clean as possible, so maybe worthwhile.
Always good fortune,
Chris
