Bipolar Cap Mult DC Heaters

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

atavacron

Well-known member
Joined
Jan 28, 2009
Messages
358
I’m trying to work out the best way to supply heater voltage to eight JJ E88CC or EH 6922 via PCB. AC is right out. Regulation is unnecessary (and wasteful). There are totem poles in the circuit, and while I don’t necessarily have to elevate the heater supply, I hear it’s good practice with an eye towards cathode lifespan. Doing it bipolar seems to make more sense than stacking the heater voltage atop a reference voltage. Luckily, two diode drops and two Vbe drops get me from rectified potential back down to filament spec. My lack of a center tap does complicate things.

IMG_7478.jpeg

R and C values are generic for scale. That’s millifarads, not micro.

Do I need loads from my new rails to the reference voltage? At the 7.5V mark, the 6.3V mark, or from the bases? Could I nix the center reference entirely after the rectifier and just cap across the rails? Any suggestions for the right NPN / PNP pair in through hole? Are there significant advantages in moving to a Sziklai approach in this application?
 
Last edited:
(PS - I didn’t draw the RFI stuff and inrush limiting on the primary, but it’ll be there, scout’s honor. And I think I need beefier diodes.)
 
Last edited:
A bump for V2, less ripple / more headroom / better "regulation." LMK any thoughts, cheers!

heater psu.jpg
 
Interesting project, I don't think I can really help you, I guess you know more about this topic.

I just have questions:
Doing it bipolar seems to make more sense than stacking the heater voltage atop a reference voltage.
What is the advantage of making the power supply bipolar compared to the B+ voltage divider to elevate the heater voltage? The component complexity is greater and therefore more can break in the long term.

Which Schottky diodes are you using for the rectifier?

Please keep us up to date on how things are going... (y)
 
Last edited:
Eight x 6922 needs 2.4A heater current at 6.3V. I would wire the heaters in series so you need 1.2A at 12V then drop in a small MeanWell 12V switcher. Rate it at twice the nominal current to handle inrush so a 25W or 30W one would do it:

https://www.farnell.com/datasheets/2547985.pdf

Cheers

Ian
Thanks Ian. I have noticed that you’ve mentioned moving to switch mode in your builds in the last few years. Sure would be easier than what I’m investigating here. I’ll be using a trafo with a 6.3V winding, and this isn’t a one-off project, so I’m trying to utilize it in a repeatable way without…uh…learning to design a switcher. Have you considered doing so? And what level of ripple have you found tolerable when using the Meanwells? The 12V/2.1A has 120mV.
 
I guess you know more about this topic.

Barely.

What is the advantage of making the power supply bipolar compared to the B+ voltage divider to elevate the heater voltage?

The power supply is remote and the circuit is entirely on a PCB with PCB mounted tube sockets. I can’t run AC in parallel with DC down the PSU cable, and I don’t really want it dirtying up the PCB either. It’s not strictly necessary to elevate the heaters with this tube type, but since I’m committed to DC, I might as well make life easier for the cathodes.

Which Schottky diodes are you using for the rectifier?

Dunno. Ones rated for 1.5A, or a little 3A chassis mount bridge if that exists. And I have to research whether putting just 10n, or 10n +10R, in parallel with each is as beneficial as with a conventional 1N4007 bridge.

Please keep us up to date on how things are going... (y)

Surely
 
The power supply is remote and the circuit is entirely on a PCB with PCB mounted tube sockets. I can’t run AC in parallel with DC down the PSU cable, and I don’t really want it dirtying up the PCB either. It’s not strictly necessary to elevate the heaters with this tube type, but since I’m committed to DC, I might as well make life easier for the cathodes.
But why use two 'bipolar' cap multipliers? Why not a conventional single cap multiplier, so the other wire is a simple common return?
 
But why use two 'bipolar' cap multipliers? Why not a conventional single cap multiplier, so the other wire is a simple common return?
Pass element dissipation cut in half, lower heat regardless of transistor chosen, right? Or could I accomplish the same thing by paralleling a coupla TO-220s fed by the same TO-92? More bulk capacitance up front too I suppose.

Also, of all people, you must know the answer to this: Is there any uneven wear effect upon the heater when using a single supply? I saw something on DIYAudio about that but didn’t investigate enough to tell if it was making a mountain out of a molehill.
 
Last edited:
Thanks Ian. I have noticed that you’ve mentioned moving to switch mode in your builds in the last few years. Sure would be easier than what I’m investigating here. I’ll be using a trafo with a 6.3V winding, and this isn’t a one-off project, so I’m trying to utilize it in a repeatable way without…uh…learning to design a switcher. Have you considered doing so? And what level of ripple have you found tolerable when using the Meanwells? The 12V/2.1A has 120mV.
I have had excellent results using MeanWell switchers for heaters. Noise has never been a problem - quite the opposite. Everything I have built using them has been quieter than I ever got with linear heater supplies. As long as you run the switcher comfortably inside its rating you should be fine. I use nothing else now.

Cheers
 
Pass element dissipation cut in half, lower heat regardless of transistor chosen, right? Or could I accomplish the same thing by paralleling a coupla TO-220s fed by the same TO-92? More bulk capacitance up front too I suppose.

Also, of all people, you must know the answer to this: Is there any uneven wear effect upon the heater when using a single supply? I saw something on DIYAudio about that but didn’t investigate enough to tell if it was making a mountain out of a molehill.
Dissipation-wise it doesn't matter which way you do it. Two in parallel or two bipolar. Or just a big series resistor! If you find the symmetry appealing, who am I to argue.

There is no uneven wear, whichever way you do it, since we're talking about indirectly heated valves.
 
Eight x 6922 needs 2.4A heater current at 6.3V. I would wire the heaters in series so you need 1.2A at 12V

Wahrschau ...

ECC series tubes and the US equivalents (e.g. ECC8/6DJ8/6922) are not designed for series connection. It is not possible to guarantee correct sharing of the voltage. There were tubes specifically designed series connection of heaters, for the ECC88/6DJ8 the series heater connection version is PCC88/7DJ8.

Thor
 
Wahrschau ...
Ha ha, I haven't heard that for a long time! 😂
ECC series tubes and the US equivalents (e.g. ECC8/6DJ8/6922) are not designed for series connection. It is not possible to guarantee correct sharing of the voltage.
In my experience, however, they do it accurately enough (at least if they are from the same manufacturer). But essentially you're right, that's what the PCC88 was developed for.
 
Last edited:
Also, of all people, you must know the answer to this: Is there any uneven wear effect upon the heater when using a single supply? I saw something on DIYAudio about that but didn’t investigate enough to tell if it was making a mountain out of a molehill.
Uneven cath wear may be an issue with a direct-heated power triode, but not on a low-po indirect-heated tube.
Anyway, making the heater voltage balanced (referenced to what?) does not change anything to the supposed cathode wear, since the voltage gradient from one end to the other is the same.
 
It is not possible to guarantee correct sharing of the voltage. There were tubes specifically designed series connection of heaters, for the ECC88/6DJ8 the series heater connection version is PCC88/7DJ8.
Ah, but if one wanted to throw away some tens of mA and blow a bunch of PCB area on wirewound pots across the pairs with wipers at the midpoint… 😉 [EDIT to add: Does it matter if one heater in a string is 6.1V and one is 6.5V?]

Unfortunately we’re stuck with either the JJ E88CC or the EH 6922 as the only modern production options for short run / boutique but reproduceable product. So there will be some imbalance vs PCC88/7DJ8. Additionally the JJs average mu is 25-28 (according to Eurotubes) while the EH average is 30 (according to Ian), so one can no longer design for 32-34. And the JJs are only viable in bulk, or if you buy pre-selected.

@ruffrecords are you consistently measuring 300ish mA on the EH heaters? the JJs are consistently around 320mA I hear.

In my experience, however, they do it accurately enough (at least if they are from the same manufacturer).
Got any data points for me on balance in series?


Dissipation-wise it doesn't matter which way you do it.
Concern #1 alleviated.
If you find the symmetry appealing, who am I to argue.
I do, but not 5-more-capacitors appealing.
There is no uneven wear,
Concern #2 alleviated.


(referenced to what?)
1/4 of HT, which is midway between the cathodes in a SRPP or WCF. That’s for minimizing stress between heater and cathode, rather than the (alleviated) concern about uneven heater wear.

So anyways:

I am going to redraw this as a single supply, still using the secondaries in series to minimize current draw and thus ripple. There will be more headroom and I’ll be able to attenuate voltage a little bit at the base. Q’s:

— Is a full wave bridge still the best option if I want to have some spare current available for 12V relays?

— Am I correct to continue to leave the windings floating and to only reference my elevating voltage at the output of the supply?

— Should I continue to Sziklai or should I move to a depletion mode mosfet? I don’t quite understand what kind of Vgs will automatically occur (if any). It is my understanding that enhancement mode mosfets need not apply in a low voltage supply with these kind of margins.

— If I can get away with a mosfet I’d really like to take full advantage of the high input impedance and use a RCRC string more on the magnitude of 10K/10uF or 47K/2.2uF, so I can use el cheapo PET caps. If someone wants to post the appropriate way to get a beta of 1000+ out of a compound pair with modern production through hole mosfets, it sure would save me a lot of googling. Only the pass one needs to be TO-220.
 
Last edited:
I would just add that since there is no any voltage reference in the design, it cannot be expected that the two voltages will add up to 12.6V.
 
I would just add that since there is no any voltage reference in the design, it cannot be expected that the two voltages will add up to 12.6V.
HT/4 at the output. I still can’t quite grok why that doesn’t continue to the CT of the windings, but I’ve been told (elsewhere) to float ‘em.
 
HT/4 at the output. I still can’t quite grok why that doesn’t continue to the CT of the windings, but I’ve been told (elsewhere) to float ‘em.
moapms comment refer to the fact that the circuit is not a regulator, it's only a capacitance multiplier, so the voltage is bound to fluctuate with variations of mains voltage.
You could connect the CT to your 1/4HT reference; it wouldn't change much since currents in both outputs are identical, so the CT is roughly at the same potential than the common point.
 
Last edited:
IMG_7508.jpeg

The limits of my understanding on display here. I’m not sure what compound mosfet current amp will net me the least voltage drop, but the point would be maximum headroom, and you adjust output voltage with the 1M trimmer (which will change to a fixed resistor and a smaller trimmer in series once the drop question gets hashed out). Also I’m unsure whether the drains need their own current isolation resistors, and what an appropriate value for the output cap might be. Depends on the source impedances I suppose.
 
Back
Top