Designing a fixed B+ regulator

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

electronaut

Well-known member
Joined
Sep 20, 2004
Messages
210
Location
Chicago
Hmmmmmm.......

I've been trying to work out a solution to this regulated B+ circuit and every direction I turn I seem to bump into a different wall.

Here's the goal:

Design a (reasonably simple) regulated high voltage (325V) power supply circuit capable of providing up to 40 mA, and as little as 12 mA, while being as clean and ripple-free as possible. Soft-start or timed delay required.

In order to keep with the "reasonably simple" requirement, I've decided against tube regulation, favoring a floating LM317 instead -- a modified and tweaked version of the circuits in Morgan Jones' book.

For the soft-start, I initially chose a 6X4 since those tubes are indirectly heated and have a natural delay, but I need about 50V drop-out across the regulator which means I need a minimum 375V at the second filter cap, and about 395 at the first filter cap. The regulator eats a few mA so round up to 50 mA, which at 395V is just outside the permissible range of the 6X4.

So I looked into the 6X5, but it is directly heated which means another transformer winding, plus I assume it will not have the same soft-start characteristics as an indirectly heated tube.

The EZ80 and EZ81 seem like possibilities, but from I can tell by looking at the datasheets, they would produce nearly 500 volts DC at input to filter when loaded lightly (12 mA), which means the regulator would have to drop out like 155 volts after the second filter which seems kind of insane.

So I could go with a solid-state rectifier and some kind of timed relay thing, but admittedly that kind of bugs me -- seems like just another gadget to break, plus I like the coincidental advantages of the indirectly heated tubes... but I am open to such a thing if there are no other practical options.

Lastly, I'm stumped by this whole idea that the resistance in series with the rectifier plates must be greater than Rs + N2 * Rp, where:

Rs = secondary resistance
N2 = square of the turns ratio
Rp = primary resistance

What the heck are they talking about? What resistance in series with the plates? I've tried to work this out but all my numbers seem very high, which I guess is a good thing but I may be doing it wrong.

Here's the current, not-quite happening plan:

HT_Reg.gif


Anyone have any suggestions?

Thanks, as always.

E.
 
I did something like this a while ago for a tube EQ that required about the same voltage. I think I wound up using a STP2N80 N-channel FET as a pass element after I wasn't happy with the survival prospects of a similar scheme with an LM317. Also, when you multiply up to that kind of output voltage the noise gets pretty obnoxious---we're taking 1.25V from a bandgap reference and getting hundreds of volts. Depending on the app that may not be an issue, and there is always passive filtering.

But, the thing did run to a fair number of parts. There was a low voltage bipolar pair as an error amp, cascoded with some high voltage parts. I think the voltage ref was a TL431 set to about 15V and R-C filtered since it is none too quiet either.
 
> Design a (reasonably simple) regulated high voltage (325V) power supply circuit capable of providing up to 40 mA, and as little as 12 mA, while being as clean and ripple-free as possible. Soft-start or timed delay required.

Why?

There are very few reasons tubes NEED regulated power. Clean, yes; regulated, no.

And why soft-start? Hot-starting radio tubes won't hurt them, and you turn-on all your tube gear and let it stabilize before you drink your morning coffee, right? That is, if you ever turn it off: better tube gear, I only turned-off in summer. I've had Macs and even Bogens that stayed on 24/7 10 months a year for years. (But those hot Dynas and some Fishers always made me nervous.)

Use a 240VAC winding, a couple 470uFd 400V caps (look inside the power supply of a dead server PC), and a choke with the most Henries you can find with less than 500 ohms resistance. And use sand-state diodes! 240VAC into silicon and a hefty C-L-C filter will get the ripple way down, and much less noise than the average simple regulator.

> as clean and ripple-free as possible.

No, that is a pie-in-sky wish, not a design goal. "As possible" does not limit the amount of time and money you will spend. If you really have too much time and money, people here can help. But more reasonably: figure-out and write-down your actual needs. 1mV ripple? 1uV ripple? Or (as is often the case) 1mV on the output stage, and 1uV on a low-current input stage that can be fed through additional filtering much cheaper than steamrollering the whole supply to the low-level stage's needs.

> Rs = secondary resistance
N2 = square of the turns ratio
Rp = primary resistance
What the heck are they talking about? What resistance in series with the plates?


Put an ohm meter on the high-voltage winding. Probably about 100 ohms, more or less. They want you to add the line-volt winding transformed by the impedance ratio (square of the voltage ratio) but if the HV is the only winding, you can just double the secondary resistance and be in a ballpark. (If the heater winding power rating approaches the plate-winding power rating, which may be true, then you should do both measurements and the square-job, or conservatively assume the secondary resistance is near-enough to "all" the resistance.)

> the regulator would have to drop out like 155 volts after the second filter which seems kind of insane.

Not at all! Regulation generally starts from a voltage nominally 1.5 times higher than the regulated voltage. After all, you "need" regulation because your raw power source varies, either in actual voltage or in drop under load. If it were stable to 5%, you might not bother regulating. But more likely the utility company sags 10%, the transformer sags 20%. And now that total variation is high, you stick in a pass-element, which has its own load-sag. You can cancel all that with an amplifier, but you have to start from someplace far above all the sags added together, which usually winds up 1.5 times higher than the regulated output. So a 325V supply starts from 487(!) volts and drops 162V at nominal line and light load. 487-10%= 440V when utility sags, -20%= 360V for transformer and C/L filter sag, leaving only 360V-325V= 35V across the pass element. OK for sand-state, but you won't find a pass-tube (or hollow rectifier!) with only 35V drop at 40mA unless it eats BIG heater power. (6L6 will be on the verge of drop-out.)

What do you REALLY need for regulation and ripple? And why does everybody today wanna soft-start little tubes, when Back In The Days we only pre-heated BIG 10KW transmitter tubes?

FWIW: I have built something like that (without the IMHO pointless soft-start). I needed to drop the screen-grid voltage on a pair of 6L6 (because I had replaced the obsolete inefficient 5U4 rectifier with modern silicon: the plates could take the higher voltage but the screens wanted to stay near where they were). Screen current on 6L6 can be high or low (or even negative): voltage divider resistors would have had to waste a LOT of bleeder current (I hate heat). I sorted "60V" 1N914 diodes for reverse-breakdown: most were around 110V at 1mA. I picked three that came to about 350V, and they biased the base of a high-voltage transistor on a heatsink. While 6L6 screens "shouldn't" suck big screen current, I put a resistor in the collector to catch things if a screen shorted. That (and my high-voltage diff-pair neon-lamp balance indicator) worked fine for years. True, P-P 6L6 isn't too fussy how clean its screen supply is.
 
[quote author="PRR"]There are very few reasons tubes NEED regulated power. Clean, yes; regulated, no.[/quote]
I might have one of those reasons. I want to be able to make different amplifier circuits (with perhaps wildly different current demands) , stuff them in a two-channel box, and know ahead of time that I will have 325 V to work with, and that having one heavily loaded left channel won't interfere with a lightly loaded right channel.

[quote author="PRR"]And why soft-start?[/quote]
Well I would certainly NOT trade noise for soft-start. I suppose I always liked watching the voltage rise slowly when I used 6X4 tubes. :grin:

In reality the noise performance is a much more important spec.

[quote author="PRR"]If you really have too much time and money...[/quote]
Sore subject... I still owe my girlfriend for last month's rent! :oops:

[quote author="PRR"]write-down your actual needs. 1mV ripple? 1uV ripple?[/quote]
Let's say 1uV! :wink:

Seriously, I do intend this to be used in low-level input stages, but undoubtedly those will be followed by output stages which can provide the extra filtering. 1mV should suffice.

By the way, (begin laughing here) how do I even MEASURE <1mV ripple?? My scope gets all blurry and my meter reads .6mV with the leads shorted.


I happen to have a 40H choke with about 495 ohms. I'll monkey with that a while and see what I come up with...

Thanks PRR for the very informative post, as usual.

And thanks bcarso for the tip on the STP2N80. If you ever feel like sharing your circuit, I'd love to try & make sense of it.

Cheers...

-E.
 
If I can dig the circuit up I'll share it. In defense of the regulation, this project was an extravagant, mostly d.c. coupled, mostly tube circuit with a lot of gain, done for an audiophile speaker designer, a confirmed bottlehead, who wanted a huge amount of bass boost to compensate for the rolloff intrinsic to his woofers.

The application cried out for sand state but I thought I would see just how far I could get. In the end the flicker noise (old term for 1/f noise) made the low freq channel distortion performance essentially impossible to measure---good thing the ear isn't too sensitive down there. Meanwhile the high frequency channel worked pretty well except the 6C45 supertriodes had a tendency to ring at 10kHz with a mechanical Q of ~3000. Otherwise they were amazingly low noise.

I'm now doing a solid-state version (whenever I get to it...).
 
Bulletproofing a SS regulator for the tube environment is tough, see

http://www.tubecad.com/february2000/page2.html
http://www.tubecad.com/march2000/page2.html


Good link for tube regulator design

http://members.aol.com/sbench/reg1.html

Check out schematics of Tektronix scopes like the 530 or 561A, or HP gear like the 400D voltmeter for schems of simple tube regulators that work extemely well. No need to reinvent the wheel.
 
Heh. I see 6AS7's in your future...

There is a pretty good all-tube regulator schematic in the Radiotron Designer's Handbook 4th Ed. My father built it and remember him grieving over what he had to pay for the special "voltage reference" tube.
 
Don't ever use solid state with B+ unless it's a 1N4004. Sooner or later, it will get nuked. Been there, done that. Plus it's just not politically correct to junk up a classic tube circuit with stupid chips.

That's the Bunny:

b+_reg.jpg
 
while i agree with PRR that many voltage regulator designs solve unimportant problems, one factor that makes me want to shy away from choke-based supplies is the ever dreaded COST factor. perhaps i've been looking in the wrong places, but most of the "right" chokes i've found cost nearly as much as the transformer.

by comparison, the design in this article (as referenced by scott s) isn't so overcomplicated, and the additional parts cost only a fraction of what a choke would cost.

http://www.tubecad.com/march2000/page2.html

img9.gif


my question is, could the choke be removed if the regulator above were used? and are there any drawbacks to the circuit above?

thanks,

ed
 
Not ignoring you Ed, but it's a chore to see how well that circuit works. Note that the "practical" one on the second page of the ref has certain essential components like protection diodes for the op amp inputs.

Chokes are indeed expensive but they are sure bulletproof by comparison to sand state.
 
thnaks for responding. i thought the tubecad article mentioned by scott was one i had read before, but i was actually thinking of this one:

http://www.glass-ware.com/tubecircuits/High_Voltage_Regulator.html

this article goes into a bit more detail. i'm going to be building the circuit sometime, and i'll let everyone know how it turns out for me. other than ripple, what should i test for? any particularly important failure modes i should check for?

ed
 
quote: "any particularly important failure modes i should check for? "

Well, don't test for short circuit protection ;-) since the circuit doesn't have any.

The author fails to mention the opamp he/she is using, almost as if it were some generic item like a resistor. It is not! But then this is from a presumably antisand site. From the 1M resistor (mislabeled 1m, which means milliohm, or when applied to a capacitor in tubeland, micro-), it had better be a low input current amp at least.

No spec on the zener across gate-source of the pass transistor. A ten-volt 500mW part ( like a 1N5240B) would work. With the low floating rail of the voltage-doubled lightly-loaded 6.3VAC trafo you couldn't easily do damage to the FET gate anyway.

The problem with circuits like this is that they often destroy themselves, or the operator, before issues like stability against oscillation can be explored.

There must be additional material about this design?

Having gone on now, the basic topology looks like it could work. If the advice is followed and the thing be made to be stable with the 5V out configuration, there's a good chance it will be ok with high voltage.
 
Couple other thoughts: the diodes, spec'd as 1N4007, may have enough leakage current to cause a significant error at the noninverting input terminal. If they are matched this will be less of a problem because the currents will tend to cancel.

The bootstrapped nature of the beast throws my simulator into conniption fits, especially with a capacitive load on the output, but when I can get it to run suggests that a LF411 or other bifet amp for the op amp, and an IRF840 for the pass transistor, should work well. Happily, the topology doesn't throw a lot of additional voltage gain into the feedback loop so the stability margin issues are reduced.

The exact value of the output voltage is probably not gong to be an issue, but if it were, a better choice using 1% standard values for the lower-right-hand R's would be 154k and 140k. The 5k above those should be a 4.99k.
 
Back
Top