It may be more helpful to explain the problems at the filament level, rather than try to 'debug' various proposed solutions.
With directly-heated filaments, the fact that the filament serves both to carry the heating power, while at the same time conducting the signal current, imposes limitations:
The presence of the heating-VOLTAGE across the filament affects the triode's grid-cathode voltage, and skews it across the length of the filament. Wavebourn mentions the modelling image of the DHT as a number of triodes across the length of the filament, whose signal-current contributions are added up inside the filament wire. This model is helpful, because now we can see that each little 'sub-triode' has a different bias voltage, according to its position relative to the Heating + terminal.
From this observation, we must now accept that with triodes, different bias voltage also means different gm (mA/V slope) in each position. At the heating + terminal, the bias is closer to the grid (more dc anode current, higher gm), and more gm means a larger signal current, too. And since the filament has resistance, there is going to be a larger signal VOLTAGE at the heating + terminal. IOW, there is a differential signal across the filament.
Now consider what happens when something else is connected across the filament.
-- Another filament, carrying a different signal (PP opposite halves, or L vs. R channels) :
these differing signals will corrupt each other, since they are directly connected.
-- a capacitor:
This will short-circuit the differential signal, if the capacitance is large enough (39000-56000µF). Cap-Values lower than this will present a crossover frequency in the audio band, because the resistance of the filament is only about 3.2Ω for a 4P1L 2.1V mode. So now we are applying an unwanted EQ to the signal! Shorting the signal is also undesirable, both from the fact of running audio through a huge electrolytic, and from dissipating the audio signal.
-- a voltage regulator:
any regulator that senses the filament voltage, and applies feedback to set the output voltage (example LM317, LT1084 etc) will sense the differntial signal, too, and treat it as an error. These chips can muster an output impedance of 10mΩ or less at 650mA.
Now calculate how much current will be pumped into the filament, if the differential signal is 5mV - and next, consider the distortion profile of the IC regulators.
In the end, connecting any of these things across the filament will degrade the signal, by one or more of the processes above.
Naturally, the outcome of ignoring all of this and building it anyway, the amplifier will still amplify. Whether you can live with the degradations is a subjective question...
But if you are commited to a physical layout, or you are constrained by cost, or the amount of work involved in creating a heating solution without any of the above compromises, there is a simple alternative. Just replace the 4P1L with an indirectly heated tube, like a 6V6. These may not give the same level of performance or sound that a really-well implemented 4P1L amp can do, but it will be much better than 4P1Ls paralled filaments and with power supplies with large amounts of noise, and voltage-control loops.
With directly-heated filaments, the fact that the filament serves both to carry the heating power, while at the same time conducting the signal current, imposes limitations:
The presence of the heating-VOLTAGE across the filament affects the triode's grid-cathode voltage, and skews it across the length of the filament. Wavebourn mentions the modelling image of the DHT as a number of triodes across the length of the filament, whose signal-current contributions are added up inside the filament wire. This model is helpful, because now we can see that each little 'sub-triode' has a different bias voltage, according to its position relative to the Heating + terminal.
From this observation, we must now accept that with triodes, different bias voltage also means different gm (mA/V slope) in each position. At the heating + terminal, the bias is closer to the grid (more dc anode current, higher gm), and more gm means a larger signal current, too. And since the filament has resistance, there is going to be a larger signal VOLTAGE at the heating + terminal. IOW, there is a differential signal across the filament.
Now consider what happens when something else is connected across the filament.
-- Another filament, carrying a different signal (PP opposite halves, or L vs. R channels) :
these differing signals will corrupt each other, since they are directly connected.
-- a capacitor:
This will short-circuit the differential signal, if the capacitance is large enough (39000-56000µF). Cap-Values lower than this will present a crossover frequency in the audio band, because the resistance of the filament is only about 3.2Ω for a 4P1L 2.1V mode. So now we are applying an unwanted EQ to the signal! Shorting the signal is also undesirable, both from the fact of running audio through a huge electrolytic, and from dissipating the audio signal.
-- a voltage regulator:
any regulator that senses the filament voltage, and applies feedback to set the output voltage (example LM317, LT1084 etc) will sense the differntial signal, too, and treat it as an error. These chips can muster an output impedance of 10mΩ or less at 650mA.
Now calculate how much current will be pumped into the filament, if the differential signal is 5mV - and next, consider the distortion profile of the IC regulators.
In the end, connecting any of these things across the filament will degrade the signal, by one or more of the processes above.
Naturally, the outcome of ignoring all of this and building it anyway, the amplifier will still amplify. Whether you can live with the degradations is a subjective question...
But if you are commited to a physical layout, or you are constrained by cost, or the amount of work involved in creating a heating solution without any of the above compromises, there is a simple alternative. Just replace the 4P1L with an indirectly heated tube, like a 6V6. These may not give the same level of performance or sound that a really-well implemented 4P1L amp can do, but it will be much better than 4P1Ls paralled filaments and with power supplies with large amounts of noise, and voltage-control loops.
Thanks for the detailed response Rod. I'm not well versed in theory and technical knowledge but I'm able to grasp some of it anyway.
In fact, I am committed to a physical layout and constrained by cost. I'm modding a Magnavox amp that was pulled from an old console stereo. It actually uses 6V6s in stock form.
But I bought a bunch of 4P1Ls years ago and figured it's time to do something with them and I'm not up for a "from scratch" build right now.
And I often like to try doing things a little differently. Lots of folks here have built SE and PSE amps with them but nobody seems to have tried PP. And everybody runs them in triode so, of course, I want to at least try them in pentode. Yes, I know they're much more linear in triode and I may end up there. And I'll also be going with fixed bias, using a 9v Lithium battery.
So I will at least try using the Meanwell supply and parallel heating. I have no way of measuring the result other than with my ears. If I can't "live with the degradations", as you say, I'll forget about using the 4P1L and all the attendant direct heating issues.
Plan B would be to go with the 12P17L, which I also have in stock. According to the data sheet it's the exactly the same tube as the 4P1L except for the heater voltage and the fact that it's indirectly heated. The Magnavox even has two 6.3v windings so I can use them in series to heat them.
I wanted to at least try the 4P1Ls first though, partly because they're glass bottles. If I go with the 12P17Ls I guess I could just cut the metal casing off. I did that with some 2P29Ls that I tried when I was breadboarding preamp tubes last year.
I have a few of these Magnavox chassis, including a stock one that's been restored, so it should be apparent if the version with the Russian tubes doesn't measure up. Not that the stock Maggie is anything particularly great.
Plan C, if neither of the Russian tubes sounds good, would be to use 7C5s or 14C5s, loctal versions of the 6V6 and 12V6. Or put the octal sockets back in and try some other tube.
I would appreciate it if you - or anyone, actually - could comment on my (perhaps crazy and ignorant) question from my previous post. Would this even work and, if so, what effect would it have on the "anti-phase conflict" issue that you raised?
Last edited:
You better cut the upper part of 12P17L casing and glue the bottom part, so won't insert the tube the wrong way. You can remove the tension ring spring from the socket, so it won't lock. It is not a military transceiver that vibrates in the tank, right?
Good idea. When I cut the 2P29Ls I just marked the bottom of the tube but gluing the bottom back on would certainly be better. Hopefully I can get the 4P1Ls to work, though. No, it won't be used in a tank.
Could you comment on my alternate method of connecting the heaters? Would it work? Would there be any benefit in it?
I don't doubt that "anti-phase conflict" and some of Rod's other concerns are measurable but I'm wondering if the negative effects will be audible.
Hmmm. If you connect 2 tubes in parallel, would it change shomehow their internal construction? I don't think so.
Obviously not. I'm wondering if using the alternate wiring scheme for the two tubes in each channel would have any positive or negative effects.
Tube A positive to pins 1&7 and negative to pin 8 (center tap) and Tube B wired with positive to pin 8 and negative to pins 1&7. And the same for the other channel.
Would this work to counteract the "anti-phase conflict" issue that Rod raised? Or, in your opinion, is "anti-phase conflict" even an issue?
I sense some confusion between parallel-connected each 4P1L half-filament, and parallel connecting multiple 4P1L filaments....
Parallel connecting the individual 4P1L filaments (example: 1 and 7 to +, 8 to -) does not cause problems - and they work better like this.
Changing the electrical heating-polarity of pins 1/7 and 8 does not affect the problem of parallel connecting multiple 4P1L heating circuits. (The electrons don't know or care about pin numbers). The problem is the external circuit, not the tubes.
From what you say about your build, I would suggest making it compatible with the indirectly heated tube, and compare the 4P1L if you build with that first.
I've never seen any large transmitting tube that isn't a AC (!) directly heated one. The trick is to get heater voltage low also with these and to feed from a CT heater winding. Hence, the impact of the AC heater voltage can be ignored due to the large signal voltage levels at the control grid and due to the large mass of those filaments.
AC heating from a CT winding is preferable over DC heating; 'cause it can be shown that the complete cathode current always leaves the filament's negative terminal, regardless where it is grounded. I.e., the heater PSU becomes part of the signal circuit if a DHT or DHP is grounded at it's positive filament end or via a pair of »balancing resistors«.
Best regards!
> I've never seen any large transmitting tube that isn't a AC (!) directly heated one. The trick is to get heater voltage low also with these and to feed from a CT heater winding. Hence, the impact of the AC heater voltage can be ignored due to the large signal voltage levels at the control grid and due to the large mass of those filaments.
The intermodulation distortion from ac heating is the same with trafo-CT or with a pair of resistors. With 10V or 20V transmitter, the IMD is substantial, and plenty audible - especially at high grid voltage swings. No, thank you!
Also, the mass of the filament has no influence on hum, as Dmitri Nizhgorodov demonstrated many years ago:
On Correlation Between Residual DHT Filament Hum and AC Frequency
The intermodulation distortion from ac heating is the same with trafo-CT or with a pair of resistors. With 10V or 20V transmitter, the IMD is substantial, and plenty audible - especially at high grid voltage swings. No, thank you!
Also, the mass of the filament has no influence on hum, as Dmitri Nizhgorodov demonstrated many years ago:
On Correlation Between Residual DHT Filament Hum and AC Frequency
> AC heating from a CT winding is preferable over DC heating; 'cause it can be shown that the complete cathode current always leaves the filament's negative terminal, regardless where it is grounded. I.e., the heater PSU becomes part of the signal circuit if a DHT or DHP is grounded at it's positive filament end or via a pair of »balancing resistors«.
Not for my heating solution: the anode current cannot run back through the filament supply at all. If the F+ terminal is returned to ground, all the anode current runs through that terminal.
But ac-heating with CT causes the anode current (DC and signal) to run all around the transformer wiring and windings, and picking up even more noise as it goes.
Not for my heating solution: the anode current cannot run back through the filament supply at all. If the F+ terminal is returned to ground, all the anode current runs through that terminal.
But ac-heating with CT causes the anode current (DC and signal) to run all around the transformer wiring and windings, and picking up even more noise as it goes.
So, from a technical standpoint, the ideal arrangement would be to wire the tubes in parallel and heat each one individually. I understood that before starting this project but space and budget constraints prevent me from going that route. Despite their technical shortcomings, the Meanwell supplies provide a cheap, ready made, alternative.
So mixing the polarity on alternate tubes would have no effect on the "anti-phase conflict", which is the result of heating more than one tube from the same supply, which is what I need to do.
I think I got confused because some previous comments seemed to suggest that using a single supply to heat all 4 tubes was OK since I'm planning on using fixed bias.
Apparently, the use of fixed bias does not eliminate the "anti-phase conflict" issue.
I had originally planned on using one Meanwell supply for each channel. Would that offer any benefits over using a single supply for both channels? It sounds like it would not affect the "anti-phase conflict" issue, but would there be any other benefits?
I know most folks here are used to building state-of-the-art designs whose development relies heavily on the use of scopes and other sophisticated equipment to measure every detail. My builds are not so sophisticated and all I have to measure with is my ears. So while I recognize the theoretical and technical shortcomings, I'm mostly just concerned about what is practical and, above all, what is audible.
It's probably difficult for many of you to relate to a build like this. I do appreciate your responses, though, as well as your patient attempts to explain the technical details.
Certainly, if I detect any audible problems I can try the 12P17Ls.
Can any of your comments here be applied to my PP 4P1L project or are they only applicable to large transmitting tubes?
Just trying to limit any further confusion on my part.
Has anyone tried using AC to heat the 4P1L? I've seen suggestions that high frequency AC can be used with DH tubes.
I have measured it to my perfect satisfaction already.
In any event, my filament supply does not permit reverse current flow, by design.
As a consequence. I can build DHT stages with 100% of the anode current emerging from the Filament negative or positive, as I will. Yes even against the flow of heating current, if desired.
In any event, my filament supply does not permit reverse current flow, by design.
As a consequence. I can build DHT stages with 100% of the anode current emerging from the Filament negative or positive, as I will. Yes even against the flow of heating current, if desired.