Regulated Voltage Supplies

[bob p]

I'm thinking about using a regulated B+ rail for a HiFi amplifier. I've never designed one from scratch, and all of the reference material I've found so far in my vintage books has been more theoretical than practical.

The amp I'd like to build is a 100-W fixed bias KT88. The B+ rail needs to be 550V. The monoblock version of this amp should draw about 100 mA at idle, and maybe 375 mA under full load. (the numbers would double for a stereo amp.) Needless to say, that kind of current demand requires an extremely low impedance power supply. Even with a choke input design the models suggest that there will be a fair amount of sag. I've been thinking that it might worthwhile to look into a regulated B+ rail.

One of the problems I've encountered is that I haven't been able to find good examples of a supply that could handle this kind of B+ level and load. If anyone could point me in the right direction, I'd appreciate it.

[R.G.]

I'm thinking about using a regulated B+ rail for a HiFi amplifier. I've never designed one from scratch, and all of the reference material I've found so far in my vintage books has been more theoretical than practical.

The amp I'd like to build is a 100-W fixed bias KT88. The B+ rail needs to be 550V. The monoblock version of this amp should draw about 100 mA at idle, and maybe 375 mA under full load. (the numbers would double for a stereo amp.) Needless to say, that kind of current demand requires an extremely low impedance power supply. Even with a choke input design the models suggest that there will be a fair amount of sag. I've been thinking that it might worthwhile to look into a regulated B+ rail.

One of the problems I've encountered is that I haven't been able to find good examples of a supply that could handle this kind of B+ level and load. If anyone could point me in the right direction, I'd appreciate it.

It's really not that difficult, other than needing good high voltage layout. Think MOSFET. You can get 1kV MOSFETs which have Idmax of 1A to 3A pretty easily and cheaply. You can roll your own regulator from any of a number of discrete circuits, or there are a few examples of using an LM317 floated on the HV to drive a series pass device.

You might be able to get by with a simple source follower supply since the current demands are so modest ( relative to what power MOSFETs usually do).

Just set up an N-channel with drain to the raw supply, source to the output, and put a stack of 75V or 100V zeners between gate and ground, and a high value resistor from drain to gate. The resistor pulls up the zener voltages, and the gate holds the source no more than a few volts lower than the gate.

The current gain in this setup is so high that you get much better response than you'd expect.

[Steve Conner]

Well, this is the regulated supply I've been using for the last 8 years or so, with no failures.

http://www.scopeboy.com/psu.html

It is basically R.G's source follower supply, but the "zener" is adjustable.

There is one mistake that I just noticed, R13 should be connected to the other side of R11, otherwise the foldback current limiting is non-functional.

To be honest, I am thinking of junking the regulator or relegating it to the 350V rail for the screen and preamp supplies. This might make the amp sound more interesting when cranked, not to mention reducing the silicootie count... Since I changed from EL34s to 6550s, and started using 8 ohm speakers with my 16 ohm-only OT, the regulator runs out of headroom and drops out when I crank it anyway.

PS: The unregulated B+ was 550V at idle, the regulated rail is 460V, and the average current draw can be over 300mA when playing loud.

[Amp Kat]

Both of those circuits are good and even a N-Channel mosfet like the PE350 would work. I use KOC's power scaling circuit as a regulator for high voltage and it works very well. You could always series some OA2 regulators but the drop in voltage could be a problem but they are rock solid.

[bob p]

when i started the thread i anticipated that the answer that would come from the EEs would be based on MOSFETs and Zeners.

i have to admit that i am operating with a bit of a handicap, as my knowledge of circuits precedes the advent of SS semiconductors. on the bench, it would be hard for me to tell a MOSFET from a Zener, and i have to confess that I'm really not up to the task of rolling my own regulated SS power supply without first doing a fair amount of reading. unless i had a chance to copy a suitable circuit from an existing schematic, i doubt that i could come up with a SS regulated supply without doing a LOT of homework.

in contrast, i'm very much up to speed on why VR tubes like the OA3, OB3, OC3, and OD3 aren't up to the task: they have a maximum operating current of 40, 30, 40 and 40 mA, respectively. Although its easy enough to "stack" three OD3 on top of an OC3 in the classic shunt-regulator configuration to get 550 regulated volts, the problem is that the stack won't handle the 800mA of current required by the application.

The next logical step was to look at vacuum tube based solutions. There are triode/pentode mode regulators, error amplifiers, etc. The error amplifier circuit is as old as dirt and its probably the right tool for the job, but I haven't been able to find any good examples of circuits that can handle the voltage and the current rating that i'm looking for. I'm thinking that the solution will either require tubes that aren't in production any more, or a modified circuit that would require the use of multiple control tubes (types that are currently in production).

setting the aesthetics of a tube vs. SS solution aside, i'm on the fence as to whether it would be better to approach the problem with multiple control vacuum tubes in an error amplifier type of circuit, or whether it would be better to approach the problem from a SS perspective. either way would involve a lot of homework learning about individual devices that i'm not familiar with. rather than re-inventing the wheel i'm hoping that i can find good example circuits, and maybe some online reading that addresses the design issues.

[R.G.]

i have to admit that i am operating with a bit of a handicap, as my knowledge of circuits precedes the advent of SS semiconductors. on the bench, it would be hard for me to tell a MOSFET from a Zener, and i have to confess that I'm really not up to the task of rolling my own regulated SS power supply without first doing a fair amount of reading. unless i had a chance to copy a suitable circuit from an existing schematic, i doubt that i could come up with a SS regulated supply without doing a LOT of homework.
...
setting the aesthetics of a tube vs. SS solution aside, i'm on the fence as to whether it would be better to approach the problem with multiple control vacuum tubes in an error amplifier type of circuit, or whether it would be better to approach the problem from a SS perspective. either way would involve a lot of homework learning about individual devices that i'm not familiar with. rather than re-inventing the wheel i'm hoping that i can find good example circuits, and maybe some online reading that addresses the design issues.

Imagine you could get a pentode that had a 1kV plate voltage rating, a gm of 1000ma/V, could dissipate between 50W and 100W, and was biased on by forward grid voltage, not a negative one.

That, in a nutshell, is a high voltage power MOSFET. The sole and only (for this discussion at least) difference in a pentode pass element and a power MOSFET is that the power MOSFET does not need an "off" bias. Other than that, it's a "pentode" that people in the Golden Age would have killed one another for.

There really is no issue in my mind. Items like power supplies, where the effects on audio quality are tenuous at best, are better done with the roughest, toughest, highest power rating devices that can reasonably be bought. Tubes are a vanishing breed. It's only fair that semiconductors should give their lives to protect tubes, isn't it?

I'll tell you what - clearly state the design objectives and I'll modify Steven's power supply to do it...

Seriously. About all I need to know from here is what the minimum load/maximum voltage of your raw supply is and what it sags to under full load of 375/750ma. It's possible that if that's a big number that more power MOSFETs will be needed to spread the power dissipation out. But that factor itself says "do it with MOSFETs" because the dissipation in a tube will be lower per device and you'll need more tubes than MOSFETs.

I really like that variable zener you did, Steven. Did you come up with that yourself?

[Steve Conner]

Hi guys,

R.G., as far as I know, I did invent that variable zener circuit

Bob, MOSFETs may be theoretically better, but why on earth shouldn't you use tubes in your power supply if you prefer them for non-technical reasons. I've talked to plenty of EEs who say I am crazy for using tubes at all. I just answer that I know fine well they aren't as good as semiconductors, but I just have an irrational liking for them.

You could check out tubes designed for series-pass regulator duty like the 6C33 and the 6080/6AS7. The 6C33 is still made in Russia. Unfortunately audiophools recently started using big banks of these tube types in their OTL stereo amps, which isn't doing the prices any good

At the risk of wandering off-topic: OTL! Now there's a waste of tubes if I ever saw one

One possible "gotcha" with MOSFETs is that devices designed for switching use in power supplies won't necessarily be able to dissipate their rated power in linear mode with a high voltage across them. There are all sorts of factors that can lead to uneven heating in the device, and the new VMOS devices don't necessarily even have a negative temperature coefficient at all voltages and drain currents. (The old stuff about MOSFETs sharing current in parallel was from back in the days of lateral MOSFETs.)

They may work just fine in linear use, but the manufacturer doesn't guarantee it. I was pretty surprised that Ampeg used IRFP260s and 9260s in their new solid-state amps, for this reason.

Steve

[Rob Mercure]

Bob,

Much as I love tubes - even to have coined the term "silicooties" that we now use about 8 years ago - I have to agree that using tubes for regulators may be too much "trouble." At that I refer to the space/weight requirements. With that said TV horrible output tubes can pass lotsa current and have high plate voltage specs. Steve mentioned the venerable 6AS7/6080 - I've got a few of these - but HO tubes are cheap and I'd be willing to sell you a few if you really wanted to use them as "pass" tubes. I might even have a few compactron V output/oscillator triodes in a common envelope but I suspect you'd have to parallel the pass sections.

Also if you can find a copy of "NBS Handbook of Preferred Circuits Navy Aeronautical Electronics Equipment," Vol. 1, Electron Tube Circuits, 1963 there are scads of regulated tube power supply circuits to pick and choose from. Unfortunately I don't have this book but it is referred to endlessly in one of my "cook books" - "Sourcebook of Electronic Circuit," John Markus (what didn't he write/edit?) 1968 McGraw-Hill - with many circuits included in this book.

Rob

[bob p]

thanks everyone for your help.

I'll tell you what - clearly state the design objectives and I'll modify Steven's power supply to do it...

Seriously. About all I need to know from here is what the minimum load/maximum voltage of your raw supply is and what it sags to under full load of 375/750ma. It's possible that if that's a big number that more power MOSFETs will be needed to spread the power dissipation out. But that factor itself says "do it with MOSFETs" because the dissipation in a tube will be lower per device and you'll need more tubes than MOSFETs.

thanks for the generous offer to help. right now I'm still working with Duncan PSUD models. that's good in that i'm completely flexible at this point as I haven't committed to the iron yet and I don't have any parts that i'm forced to work around, but that's bad in that the best figures that I can come up with right now are modeled guesstimates, which may or not be accurate in the real world.

The PSU designs I've been working on so far have been old-fashioned brute force deisgns: high voltage PT using 720 series Hammonds, HUGE 12-lb chokes, HUGE snap-in caps, etc. Such brute force designs might not even be necessary once a regulation circuit enters the picture, so its likely that the iron requirements would be dramatically revised.

The best firm data that I have then, are the design criteria in the Genalex KT88 data sheet for a 100-W, Class AB1, fixed bias, UL application. At 560 plate volts, each pair of KT88 will draw 100mA at idle and 325 mA at full load. The remainder of the amp should only add about 50 mA of current; each channel only uses a single 12AU7, 12AX7 and 12BH7 with each pair of KT88. The total peak current demand for all of the tubes should be 375mA/750 mA mono/stereo.

Instead of starting with one of my power supplies at one and, and the tubes at the other, and working toward the regulation circuit in the middle, would it make more sense to just scrap my existing PSU designs, and just work backward from the tubes? If we can do that, its easy enough to say that I need a 560 VDC regulated B+ rail that can handle from 120mA to 375 mA per channel, and work back from there to determine what I need for the regulation circuit, and then select a transformer and input filter that's suitable for the job.

By relieving the iron-based constraints, does this make the task easier?

[Amp Kat]

I don't know about the other guys but I thought you were going to use the regulator from the tubes back rather than in the middle from the get go. You also didn't mention a tube or SS rectifier but would think since you want regulation a SS would be better for several obvious reasons. Rob is correct of the circuitry of an all tube regulator and the size and complexity of the circuit. IMO the way to go is to use the MOSFET Regulator right off the bridge rectified out to a pot with the wiper and around 1k to the gate. Pin 3 of the pot to the drain and an 8 volt zener from gate to source.Pin 1 to a 33k to ground. Then you can add a choke pi filter for the screen filtering and vary your plate voltage to desired voltage. A STW9NB90 N-Channel Power Mosfet has a Vds of 900 V and ID of 9.7 Amps. Small and compact IMO is the way to go and I'm sure RG and Steves circuits are very close and compact and would work great.

[R.G.]

R.G., as far as I know, I did invent that variable zener circuit

Very nice work. My compliments.
(a) may I use it for Bob's power supply?
(b) what kind of regulation do you get from it? I'm guessing it's sub-single-volt regulation based on the gm of a power MOSFET.

Bob, MOSFETs may be theoretically better, but why on earth shouldn't you use tubes in your power supply if you prefer them for non-technical reasons. I've talked to plenty of EEs who say I am crazy for using tubes at all. I just answer that I know fine well they aren't as good as semiconductors, but I just have an irrational liking for them.

You're right about this - if what you're doing is for aesthetic reasons, there is no reason not to use whatever you like; hence my comment about artistic purposes aside.

I do however have a irrational liking for using silicon to enhance and extend the lives of hollow-state component. It seems like appropriate justice to me.

... the new VMOS devices don't necessarily even have a negative temperature coefficient at all voltages and drain currents. (The old stuff about MOSFETs sharing current in parallel was from back in the days of lateral MOSFETs.)

They may work just fine in linear use, but the manufacturer doesn't guarantee it. I was pretty surprised that Ampeg used IRFP260s and 9260s in their new solid-state amps, for this reason.

Yes. D-MOS devices have an initial positive tempco, which drops to 0 and reverses to negative tempco at 2-3A depending on the device. So where you're running them at high voltage/low current, you still have to ensure sharing by using ballasting or other tricks. That would be the case here. But in high current/low voltage operation, they do share by the negative tempco region. You have to know where the power stress is and design accordingly.

... Such brute force designs might not even be necessary once a regulation circuit enters the picture, so its likely that the iron requirements would be dramatically revised.

I believe that is correct.

Instead of starting with one of my power supplies at one and, and the tubes at the other, and working toward the regulation circuit in the middle, would it make more sense to just scrap my existing PSU designs, and just work backward from the tubes? If we can do that, its easy enough to say that I need a 560 VDC regulated B+ rail that can handle from 120mA to 375 mA per channel, and work back from there to determine what I need for the regulation circuit, and then select a transformer and input filter that's suitable for the job.

By relieving the iron-based constraints, does this make the task easier?

It does make things much easier. About the only guess work is to pre-estimate the regulation of the transformer so you don't get the power supply sagging out of regulation at full current.

Did you want a single power transformer or dual power transformers, completely separate per channel?

[Steve Conner]

Hi R.G.,

Sure you can use it, I'd be flattered ;-) Even if I tried to stop you, it's been in the public domain for a good 7-8 years now. About the silicon vs. tubes thing, I totally agree. Tubes in all the places where they actually improve the tone, silicooties everywhere else you can get away with it.

The regulation was about 0.5V from no load to full load, IIRC. Plenty good enough for a tube circuit, I think.

[bob p]

It does make things much easier. About the only guess work is to pre-estimate the regulation of the transformer so you don't get the power supply sagging out of regulation at full current.

Did you want a single power transformer or dual power transformers, completely separate per channel?

do i have a choice? i had expected that my choice would be limited by the available iron. when you get to these kinds of voltage and current ratings, there just aren't many options for transformers. a stereo amp with a single PSU would be great (it would also allow the option of building either a 100-W stereo amp or the "behemoth" 200-W monoblock). but i'm not sure if its actually do-able because of the iron limitations. for that reason i had been expecting that high power, completely separate monoblocks would be the only available option.

i don't have a really good handle on what sort of excess (current vs. voltage) is required for the regulation circuit. there are plenty of options with respect to voltage in the Hammond line, but the current rating on all of them seems to be pushing the limits. i don't know if its even possible to actually get 750 mA of regulated power out of the Hammond 700 series. will that limit our options? are there other iron options that i'm not aware of?

on a related note, one thing that i had thought would be an interesting project would be a general purpose variable voltage regulated power supply. I've seen adjustable tube circuits that do this in the voltage range of 250-400V at 200mA, but I've never seen anything that would do 400-600V at anywhere from 100-750mA. THAT would be a really cool generic circuit for experimenters. i was surprised that nobody had done it already. is it even feasible?

[R.G.]

You are of course limited by the iron; that's why I asked that one up front.

The simplest thing is to grab one transformer per channel and build it. The other options are to haunt the surplus forums and otherwise go find a transformer that will put out the required power for a 200W.

If you a single supply the Hammond 717 seems like the best choice. It's simpler - you don't have to find a 1A choke for the input to a choke filter like the 722 would need.
Let's sketch out the numbers.

The 717's 510-0-510 makes for about 720V nominal, and 770 at high line conditions. At .75A out, that's 540W nominal, 578 high line power that's going somewhere. The difference between the 400-600V into the amplifier has to be dissipated by the regulator. If you want 400V out and 750ma, that's 240W of dissipation in the regulators that comes out as heat for linear regulators.

But yes, there are two Hammonds that will produce the necessary power.

It gets hot in there. What would be ideal is a transformer that put out a nominal of about 620V or 680V high line.That's about 439-0-439 or a 440-0-440 transformer at 750ma.

These are big trannies, and probably expensive. As well as a ticklish thermal design. In fact, the thermal design might dominate the rest of the design.

Divide and conquer might work better.
Hammond provides a transformer that might do a single channel pretty well. It's the 712, which provides 425-0-425 or 600Vdc at nominal line, probably 650 at high line. The thing will sag some, but I haven't found the regulation specs. This one can probably do regulated 550 at 375ma, one per channel. With tubes, that's not enough headroom, but with MOSFETs you could probably do it.

I didn't see any integrated ones with plate and filament, so there's a filament transformer or two to put in there as well.

The generic regulated supply is possible for experimenters, but it's going to be EXPENSIVE which is probably why it's not done. Actually, HP built and sold one similar to that; the B+ wasn't as high, nor the current, but it was fully regulated, provided filament, screen and bias outputs. I have one... but it would be very tough to replicate.

In regulated power supplies this big, it usually makes sense to preregulate the HV at the power line by phase control, then do a secondary regulator to smooth out any phase control funnies.

But as a brute force thing, no, conceptually it's not hard to do.

[Steve Conner]

Why are you using transformers meant for FWCT rectifiers? They don't make efficient use of the copper in the secondary. A tranny for a bridge rectifier will be considerably smaller for the same rating.

Once you go to a bridge rectifier, you may be able to find an industrial control panel transformer to suit. You know the ones for running 120 or 240V equipment when you only have three phase supply available with no neutral.

I don't know what they are like in the states, but over here, you can get one that transforms between 240V and 480V with tappings at 380, 400, 415, in various sizes up to 1kVA. I'd be very surprised if that wouldn't power your proposed amp, even if it might look and sound like a small substation

I have experimented with phase control in the past, and I made a voltage-controlled phase angle firing circuit for SCRs and triacs, that would just need a PID controller added to run it closed-loop. But I'd not relish the thought of getting phase control to work without introducing buzz into the audio, and the transient response is lousy too.

I have one of those regulated power supplies too, but sadly it's pretty old and in poor condition, and not an HP It uses large KT88-like beam tetrodes for the pass tubes, and has no current limiting except for fuses.

I also know a guy who used to do a lot of power electronics work. He picked up a high voltage switched mode bench power supply that puts out 1.5A at 600V. It's only about half the size of an old VCR and weighs under 20lbs. So switchers can be really good for this kind of thing, the only downside is the price (that model sells used for about $500 on ebay!)

PS. I found a guy actually selling them targeted at tube audio people:fleabay link