Bipolar Cap Mult DC Heaters

[atavacron]

[…] we engineers shouldn't encourage such ideas that are technically not correct. Although they actually work.

I think you just summed up the consensus. Nobody’s saying that series/parallel is ideal. Most are saying that you can get away with it within reason, and everyone knows that any tube or human can be an agent of chaos. I am not seeing an impediment (within my simple early simulation) to making all eight of these filaments a parallel affair, so I will proceed along those lines but with PSU elements that are equally applicable to series/parallel.

 

[atavacron]

The secondary windings are for full wava rectification, use them like that for HT.

Why? Assuming DCR/winding count is precise.

 

[atavacron]

Alrighty, here's where I'm at with this, below. Floor box on the left, one of four channels on the right. For people just tuning in: sorry to disappoint, but the heater supply is neither bipolar nor a cap mult. I'm leaving about 300-350mV on the table, BTW, by not springing for more spendy heater regulators. Should be OK since I only need 6V at the end of the day. Please do point out any errors, other than how I should be making it a SMPS.

 

[thor.zmt]

Please do point out any errors, other than how I should be making it a SMPS.

Do you have accounted for low mains (-10%)? Doesn't look like it to me.

Have you looked at the dissipation in the regulator IC with high line (+10%)? If what you show is nominal, then you might have 10VA dissipation.

This means heatsink plus isolator pad and TO-220 package thermal resistance should be 4K/W to keep the junction temperature below 70...80C.

Commonly TO-220 is 2K/W, Isolator 1K/W leaving 1K/W for heatsink.

If instead you used 3pcs of LM317 in parallel Each IC only handles 3W, so much higher thermal resistance will control temperature.

More modern solutions (that shall not be mentioned) can further increase tolerance to mains voltage variation and can cut hear losses even further.

The current limiter on the HT should be foldback type. Or it will not protect the transistor well.

Otherwise I would suggest that it does everything that is "not a great idea" in an audio power supply.

But it will produce DC out.

Why? Assuming DCR/winding count is precise.

Because if there is a difference, it will not matter. In parallel it does....

Does the manufacturer recommend parallel operation?

Thor

 

[emrr]

Don't want to hear it, but again, especially if for production, I don't see how you beat $12-14 in single piece price for switchers over a transformer and PCB. Transformer alone is higher cost. Maybe you have a pile of surplus to burn.

 

[Cqwet Dbdfte]

Will do. If I have inrush current limiting (it’s a toroid), can I add a 100n X7R in parallel to bring that down even further? I see that people tend to not do that on the reservoir cap but i don’t quite know why.

I thought the dropout for a depletion mosfet used as a source follower was just RdsOn * current. No? That part is like $8 in 10 qty btw, you really pay for the low resistance.

Tightly twisted AC heater wires may reduce stray fields to insignificance in your bundle.
Center tapping the PSU is a waste.
The dropout of a NPN series pass is VceSAT, not Vbe.
LDO typically use PNP pass device.
An LM317 needs at least 2.5V to regulate effectively.
A N-MOS requires 10V higher gate than source voltage to drive it into low resistance, there are IC chips that will do that, for + rail regulators, or why not use a P-MOS, or put the N-MOS in the negative rail?
Tube data depends on heater temps, which is voltage related, using series may or may not work, great, adding some uncertainty.
Try the tightly twisted heater cable first, and measure any induced noise in the output signal.
Try your circuits in the free LTspice and save time and money. Super easy to use.

 

[thor.zmt]

Tightly twisted AC heater wires may reduce stray fields to insignificance in your bundle.
Center tapping the PSU is a waste.
The dropout of a NPN series pass is VceSAT, not Vbe.
LDO typically use PNP pass device.
An LM317 needs at least 2.5V to regulate effectively.
A N-MOS requires 10V higher gate than source voltage to drive it into low resistance, there are IC chips that will do that, for + rail regulators, or why not use a P-MOS, or put the N-MOS in the negative rail?
Tube data depends on heater temps, which is voltage related, using series may or may not work, great, adding some uncertainty.
Try the tightly twisted heater cable first, and measure any induced noise in the output signal.
Try your circuits in the free LTspice and save time and money. Super easy to use.

Seconded on all counts. Except:

"Center tapping the PSU is a waste."

That depends on the mains transformer design.

"An LM317 needs at least 2.5V to regulate effectively."



At around 1A it is closer to 2V. MIC29512 arguably needs much less:


Multiple MIC29512 in parallel will also reduce thermal challenges.

Thor

 

[Cqwet Dbdfte]

In the circumstances of thread, DC heater looks like unmotivated complications. DC heat will reduce AC fields in a wire bundle, but how much?
Realistically some load side HV filtering is done.
In my experience 2.5V is the number to remember, and that dropout is subtracted from the regulated DC. Some margin is needed for the ripple Vpp. The lowest point of the Vpp has to be higher than the Vout+dropout+some margin, if clean DC is really needed. An unregulated filtered DC with a series resistor may suffice.
There is very little AC AF coupling between the heater and cathode in an indirectly heated tube.

 

[thor.zmt]

There is very little AC AF coupling between the heater and cathode in an indirectly heated tube.

There is a parasitic rectifier diode between cathode (which becomes the anode of the parasitic diode) and heater (which becomes the Cathode).

The AC heater then creates a voltage difference that causes the AC voltage to be rectified.

Special low noise audio tubes as a result had twisted filaments (which are much harder to make than normal ones) to cancel the fields. Remember the filaments are oxide coated (insulation) tungsten alloys glowing orange in operation.

I have also found many NOS tubes with notable heater-cathode leakage.

For reliable low noise at 21st century standards DC heaters are often needed, IME.

Thor

 

[ruffrecords]

There is a parasitic rectifier diode between cathode (which becomes the anode of the parasitic diode) and heater (which becomes the Cathode).

The AC heater then creates a voltage difference that causes the AC voltage to be rectified.

Heater elevation will reverse bias this diode.

Special low noise audio tubes as a result had twisted filaments (which are much harder to make than normal ones) to cancel the fields. Remember the filaments are oxide coated (insulation) tungsten alloys glowing orange in operation.

It is an unfortunate fact of life that every single spiral filament tube I have tested has been hopelessly microphonic.

I have also found many NOS tubes with notable heater-cathode leakage.

For reliable low noise at 21st century standards DC heaters are often needed, IME.

Thor

My experience exactly.

Cheers

Ian

 

[abbey road d enfer]

Why? Assuming DCR/winding count is precise.

Only one diode voltage drop instead of two. No issues with voltage balance.

 

[abbey road d enfer]

Center tapping the PSU is a waste.

In what way?

A N-MOS requires 10V higher gate than source voltage to drive it into low resistance,

IINM, the OP wants to use a depletion MOSFET.

 

[merlin]

In what way?

Two-diode rectification is less efficienct than bridge rectifier. The ripple current in each half winding is 1/sqrt2 times the total RMS. For the same windings (two in parallel with a bridge, versus the same two in series with 2 diodes), you can only pull about 70% as much DC out compared with a bridge.

 

[atavacron]

Two-diode rectification is less efficienct than bridge rectifier. The ripple current in each half winding is 1/sqrt2 times the total RMS. For the same windings (two in parallel with a bridge, versus the same two in series with 2 diodes), you can only pull about 70% as much DC out compared with a bridge.

Right. For the required pull here at 6V downstream, I have to use a bridge. I suppose the windings-in-parallel issue could be addressed by rectifying each winding separately with a quad of those super low drop schottkys and joining the outputs. Same for HT but with UF4004s. Side benefit: keeps the diodes cool. Doesn’t provide the lower drop benefit of the diodes being warm though.

to respond to some other concerns…

The base premise of the power supply is to use the lowest possible HT secondary that will still allow adequate headroom for that godawful LR8 to work under low line conditions. I’ve been doing most calcs at 117V and accounting for 112V and 124V. If someone in Japan buys a unit I will specify a step down mains trafo. I could redesign some things to use a lower HT if necessary/recommended, say 190V instead of 205V. I did spend an inordinate amount of time planning this out with a 160V trafo previously.

I did more work on the circuit and total draw at maximum test output whilst burdened with 600 ohm loads and phantom on will be ~32.5mA per channel @ 205V, or 130mA total. The HT windings on this trafo are spec’d at 280mA each @ 175V. I will look at feedback current limiters but if anyone wants to post examples I’d be much obliged. I don’t like that I have to current limit twice, on both HT and heater lines, but the outboard PSU cable is of course a major liability. That’s much more of a concern than someone jamming a tube into a socket in incorrect orientation. At least the heater regs do the limiting for me, I wish the same could be said for HT.

I have no problem paralleling the HT pass elements or the heater supply regulators; that’s why I’m specifying $3-4 parts instead of $8-10 parts to begin with.

The main reason to attempt 6V heaters here is because it will allow people to replace V1 themselves when it gets tired. The design uses a lowest-common-denominator approach to gain, so that even a JJ E88CC will work if it comes down to it. I’ve built in a mu compensation approach to the circuit, whereby you turn up a trimmer to calibrate level as the tube ages. But after a certain point there will be no more trim to turn, and it would be nice to not have to send out two new production 6922s to people at that time. Turns tube rolling from a liability into a possibility, and the only thing I have to point out is that you gotta use a matched example for V2, the SRPP tube. There’s very little closed loop gain in the output stage (it’s a trim), so it is entirely likely that no one will need to replace V2 for a coupla decades in normal periodic use.

 

[Cqwet Dbdfte]

In what way?

Going back thru a regulator to a CT transformer (a low impedance path) when a multi kohm resistive divider would do, as leakage current is very small.


IINM, the OP wants to use a depletion MOSFET.

Selection is very thin, and biasing is no fun unless you drop out some volts, where in this case voltage headroom is meager, or make a complicated bias circuit. The effort does not match the need. Depletion mode FETs for constant current sources and loads, yes, but not this.
The GaN SiC FETs has had slow acceptance until compounded with MOS FETs, making them enhancement mode.
From OP's circuit examples I can see he is not ready for this pain.

 

[atavacron]

.
From OP's circuit examples I can see he is not ready for this pain.

My knowledge extends to enhancement mode; the three HT-related mosfets on the channel card itself can be enhancement, everything else needs to be depletion in the current architecture. Actually two (HT and phantom) because I got rid of the 105V bias thing. That ST 3A guy seems to be the 2024 preferred part. Loadwise my quiescent is approx 84mA sans phantom, approx 98mA with 4 channels of phantom, or 114mA/128mA respectively in ye olde big-square-waves-into-all-600-ohm-loads test scenario. Much easier to say “about 100-110mA in normal use.”

 

[Cqwet Dbdfte]

Which 3A ST part?

 

[thor.zmt]

Right. For the required pull here at 6V downstream, I have to use a bridge.

I doubt it. Put it into PSUD as CT with one of the big TO-220 or larger case dual schottky diodes.

Remember current draw per winding is RMS. Given each winding only sees 50% Duty Cycle the peak current in the winding can be pretty much double. PSUD shows RMS current.

Using TINA-TI (free) gives more accurate results, PSUD is good for first order approximations.

The base premise of the power supply is to use the lowest possible HT secondary that will still allow adequate headroom for that godawful LR8 to work under low line conditions.

Why use that "godawful LR8" then?

Instead you can make a 2-pin high voltage CCS from a pair of Mosfet's or BJT's with sufficient voltage rating that has less less dropout and then use that to power up a Zener diode string?

I will look at feedback current limiters but if anyone wants to post examples I’d be much obliged.

NOT FEEDBACK, Foldback.

https://www.sciencedirect.com/topics/engineering/foldback-current-limiting#:~:text=%E2%80%9CFoldback%E2%80%9D%20current%20limiting%20reduces%20the,of%20the%20pass%20element's%20SOA.&text=and%20so%20foldback%20ratios%20of,impractical%20for%20low%2Dvoltage%20regulators.

It is usually build into IC regulators. An alternative regulator would be this BTW (not mine):



No "godawful LR8", lower dropout as well.

The schematic misses the part that can make this circuit short circuit proof, move the 10V Zener at the gate to the other side of the 18 Ohm resistor, current limiting is then at around 0.39A and if mounting Mosfet and LM317 on the same heatsink (insulated) the 317's thermal limiting will protect the MOSFET during sustained shorts.


Thor

 

[C.B. - Boudio]

Why do you want to do it that way? In most applications, a hum-balance pot or two 100 ohm resistors to earth wil do, I think. Or just use a single diode rectification and something like a 680mF capacitor and maybe a small resistor. If your line voltage is not all over the place, and your wire dress is decent, why worrying?
But, I may be wrong of course (I probably am ;-) )

 

[atavacron]

But, I may be wrong of course (I probably am ;-) )

Heater DC enters the unit directly on to the channel PCBs. There is no wire dress. AC is not an option.