Discrete Heater Regulator?

[jhaible]

Here's a little tube heater PSU I've built (see page 2)

My design goal was a ultra-low-drop regulator, so I used a FET, and a voltage multiplier to get the higher Gate voltage.
The opamp is just in for current limiting. If you don't need that, the only active components are the FET and the tiny TL431.

JH.

 

[Matt Syson]

Hi
If you really want to drop that amount of voltage use a chunky, heatsinked resistor then your regulator, this will also add some short circuit protection although heater shorts would just tend to get brighter.....
Switchmodes are a bad idea as you would need so much filtering on the in and out to get rid of the noise. Valves were designed to 'get along' with AC heaters and produce reasonable performance we should just be 'icing the cake'. Lots of big, dumb, cheap power transistors (2N3055 or similar) are the answer to reliable supplies. These 'megawatt in a TO220 case' designs are all very well but you can't get the heat out especially if you use insulating washers.
Matt S

 

[ulysses]

[quote author="Matt Syson"]Switchmodes are a bad idea as you would need so much filtering on the in and out to get rid of the noise.[/quote]

I think it's time to kill off the myth that switching supplies are inherently noisy. They can be made to be very, very quiet. And a tube filament, which is intended to be run with 100% ripple at 60Hz, certainly won't even blink at a heater supply with a couple of microvolts of noise at around a couple of MHz.

 

[bcarso]

That's going to be a hard "myth" to kill off! :roll: Show me a switcher with a "couple of microvolts" of noise and I'll eat my hat (fortunately I don't have one, but you know what I mean). A few hundred perhaps, but usually peppered with nasty narrow spikes well above the rms voltages.

That's not to say you can't eventually get the radiated and conducted noise reduced to where it is not a problem, but it is a whole lot of work. Resonant mode switchers work pretty well, but the magnetics still all radiate to some extent.

In tube circuits the likelihood of local demodulation is pretty small, so a bit of HF crap is not likely to be creating an audible artifact. But it may well screw up the surrounding equipment.

Having said that, regular old chunky power diodes are a pain in and of themselves, and a lot of mains freq power supplies of old violate EMC just from the reverse recovery spikes. And the magnetic fields are often terrible even with toroids, and their emission when intercepted by loops in nearby circuitry is certainly audible.

So, if I were totally space-constrained a very-carefully designed switcher might well make sense. But it is not for the faint of heart IMO.

 

[Matt Syson]

Hi
Yes I agree that switchers CAN be made very quiet but considering the scale of the work proposed earlier which are simple homebrew projects the process of making a reliable super quiet switcher is in a different ballpark, involving filtering and extra metalwork to screen the active bits.
The circuits above can be knocked up with 'junkbox' parts and with a bit of fiddling can be made to work successfully.
Matt S

 

[CJ]

output device takes place of 100000 MFD.

 

[ulysses]

[quote author="bcarso"]That's going to be a hard "myth" to kill off![/quote]

The technology behind switching supplies has progressed by huge leaps and bounds over the past few years, driven largely by the cell phone market. National Semiconductor, and others, have come out with and continue to develop all kinds of very simple switching controllers and they keep getting better. Small battery-powered devices keep getting smaller, which means they have to be more efficient (smaller batteries) but the switching circuit has to get smaller too. The best way to do that is to increase the switching frequency. Do the math and see how much smaller your filter caps can be when your ripple is at 1.2MHz instead of 120Hz. The iron gets tremendously smaller too. In audio circuits that don't need to fit in your pocket, this affords a greater margin in filtering. Also, it's a whole lot easier to shield MHz noise than it is to shield 60Hz hum. We're talking about millimeters and copper as opposed to inches and iron. Yes, it's more work to design a switching supply (though that is less the case as more and more of the circuit is getting integrated into the controller chips). The audio crowd tends not to know enough about switchmode circuits to knock one out as an afterthought as people tend to do with linear supplies. But poorly designed linear supplies aren't any better than poorly designed switching supplies. With some effort, it can be very worthwhile. There is some very high-performance gear out there now that uses well-designed, very clean switching supplies. One example I'll give is the Great River equalizer. You can't get more high-performance than that.

 

[bcarso]

Quote: "Do the math and see how much smaller your filter caps can be when your ripple is at 1.2MHz instead of 120Hz."

Yeah---but do the math on the effect of parasitic inductances in your smaller caps and associated traces as well. And again---show me this switcher with "a few microvolts" of noise. Also, show me a switcher that has comparable load change transient response to a good linear, and show me one that works equally well at zero load current as it does at intermediate to maximum currents.

I am well aware of the advances in switchers (and for that matter, switching amplifiers). They are still not for the faint of heart, or those possessing less than really great intuition and powerful simulators that model distributed EM behavior, if you are truly trying to get to the same peformance levels as well-designed linear regulators.

If you are space-constrained and power-constrained, and are going to make a few million units of some product, I agree that switchers are the way to go.

 

[PRR]

> it's a whole lot easier to shield MHz noise than it is to shield 60Hz hum.

Not in this case. Heater wiring radiates noise largely through capacitance. The pFd of capacitance from heater to cathode won't pass much current at 60Hz, lots of current at 1MHz. This can be demonstrated at lower frequencies: a poorly filtered "DC" heater supply's 120Hz etc buzz is much more obnoxious than just 60Hz, due to 6dB/oct rise in capacitive coupling (and of course, the rise in Fletcher Munson).

> poorly designed linear supplies aren't any better than poorly designed switching supplies.

Gotta agree with that.

 

[Matt Syson]

Back to the original question, interesting that the discussion has become.
How about a constant current regulator since you are using valves which are a known load. About 70 years ago this problem had been sorted by using a baretter.
Matt S

 

[Larrchild]

bcarso mentions:

OTOH using a current source has the disadvantage of leading to thermal runaway and failure if the filament is already deteriorating---if the resistance goes up the power goes up proportionally.

 

[Svart]

how did I miss this thread?

I am well aware of the advances in switchers (and for that matter, switching amplifiers). They are still not for the faint of heart, or those possessing less than really great intuition and powerful simulators that model distributed EM behavior, if you are truly trying to get to the same peformance levels as well-designed linear regulators.

It won't happen. There will always be an inherent noise when moving back and forth between power levels at considerable speeds.

Anything we attempt to do to *lessen* the affect such as smoothing leading and trailing edges of our square waves only bring up back to pulsating positive going sines..

Switching at higher and higher speeds do save on cost and size but start to bring other problems, as Bcarso mentioned, with the parts, the designs, the PCBs, and everything else. As we move to higher and higher freqs with our transmissions of phones and so forth we will also start to see radiations plague the designs once again as we catch up to our band-aids that we slapped on the gear of last year.

reminds me of something that pops up here a lot that Bcarso, myself and others have discussed.. strapping caps in parallel with noisy rectifier diodes to squish the noise. we could discuss this again but the bottom line is we move the noise to another part of the spectrum instead of actually getting rid of it.

 

[Larrchild]

Maybe Tesla's concept of rectifying the waveform from a big shortwave transmitter to create energy was meant to be used across-town.

But it bears mentioning that half of the gear I use to measure noise floor has switchers in it. hmm.

Anyway, I was just proposing to vary the duty-cycle of analag's pass device at a not-too-rf rate, and keep the linear supply feeding it. Still needs a nice filter section on the tailpipe, but would it be any more manageable, noise wise?

 

[ulysses]

[quote author="Larrchild"]But it bears mentioning that half of the gear I use to measure noise floor has switchers in it. hmm.[/quote]

Spezackly. An awful lot of digital gear uses switching supplies now too, and there are converters out there getting suspiciously close to 24-bit performance. Where's the noise? Way the hell down there, that's where.

All the complaints you guys have raised are valid, but they also apply (usually more so) to LF supplies. Silicon turning itself on and off tends to generate noise. We do a pretty good job of sledgehammering that ripple out of there when it's 120Hz, and we have to worry about the harmonics extending up through the audio range. Sliding up to the RF range isn't conceptually dissimilar, except that the fundamental and the major harmonics aren't anywhere near the audible range.

That said, I'm not building switching supplies myself yet because design and layout are a lot more difficult. But I'm headed that direction.

 

[Larrchild]

and we have to worry about the harmonics extending up through the audio range. Sliding up to the RF range isn't conceptually dissimilar

Unless you listen to PRR's facts on 60hz vs RF induced in heaters. Then they are as dissimilar as possible.

 

[Svart]

well I think that the solution is not to try to make them noiseless because that isn't going to happen, I think the solution is to make the other circuits immune.