Regulated screens

[Zozobra]

Hi all,

I've been pondering the screen supply recently. RC filters will sag a lot to get any meaningful amount of filtering and I don't know if I like/want that. LC filters are expensive and have their own inherent flaws so I thought why not regulate them? I'm certainly thinking its not necessarily a good idea to run the screens near the plate voltage anyway and the screens should have a happier life if I drop them from 450-500VDC to 400VDC and of course the supply for the preamp will be much cleaner too and there shouldn't by any significant loss of power. So how is the best way to achieve this with silicooties? How much voltage drop is feasible without needing to use a huge heatsink and then run into limitation there?

My first project to tinker with this will be an 8-knob AOR100 that I can gut as I'm not feeling it tonally and it was cheap (OT primary is 2K). It produces about 460VDC at the rectifier. So what do I need to figure out? I suppose the first thing is how to drop and regulate the voltage which I guess is ideally suited to MOSFETs and then I need to worry about the current limitation of the device which shouldn't be too extreme for a quad of EL34s and the preamp if I'm only regulating from the screens downwards. Hints and tips much appreciated!

[Steve Conner]

An oldie but still a goodie (been working great for 14 years)
Stabilised power supply

This produced two regulated voltages, one for the plate supply and another for the screens and preamp. The output is current limited to help it survive tube shorts. To keep the plate regulator MOSFET cool, a big fan-cooled heatsink was hidden inside the amp. But you could easily just leave out the circuitry for the second voltage output (M3 etc) and use the main output for your screens.

Regulating the screen voltage gives most of the benefit in terms of hum reduction and dynamics. Likewise, current limiting the screen supply gives most of the protection benefits. The regulator won't die if the screens short, and the screens won't melt if the speaker goes open. If I was doing this again, I wouldn't bother to regulate the plate supply, but I can't be bothered modifying it now.

The BFC62 was a cheap 800V MOSFET at the time. There are many better choices now.

[Zozobra]

Thanks! I shall have a read to work out which bits I need and then try to work out how to implement it on a turret board

[Steve Conner]

Well, I built it on a tagboard, but it wasn't a pretty sight

[Mick Bailey]

Perhaps not quite what you're after, but it works for me; after repairing a few VHT 12/20 MOSFET variable B+ supplies, I thought it may be a neat and cheap way of obtaining a variable screen supply on its own. It works really well, though you still need current limiting resistors for your screens. See the schematic in the manual. Very much of the 'power scaling' type of approach and simple to build. You need to add a 200R resistor right on the gate pin, though.

I have a lot of old FQA10HN80C left over from repairing industrial SMPS units so used one of those. They don't even get warm when bolted to the chassis. When picking MOSFETS the 'on' resistance is important in reducing heat. The lower the better.

VHT Amplification - Special 12/20 RT Combo

BTW, On conventionally-powered Hi-Fi amps I connect the screen to ground via an electrolytic on each tube - usually about 47uf or 100uf. Much better than without, but you lose some of the protection of just using resistors. Some also use a string of zeners to ground to give a regulated screen supply.

[R.G.]

What Steve said, but some additional thoughts.

why not regulate them?

The screen supply can be thought of as a kind of gain control for the tube. Higher voltage means higher gain-ish things. With that as the case,
- drooping screen voltage droops the tube gain, leading to one form of compression or sag
- ripple leads to a form of cross modulation
Eliminating these may make the amp sound much better to you, or they may be part of what you or someone else looks for as part of an indefinable incredible tone signature. No good way to tell ahead of time.

I'm certainly thinking its not necessarily a good idea to run the screens near the plate voltage anyway and the screens should have a happier life if I drop them from 450-500VDC to 400VDC and of course the supply for the preamp will be much cleaner too and there shouldn't by any significant loss of power.

If the preamp supplies are daisy-chained from the screen supply, and the screen supply is not T'd off the main supply chain. It can be done either way, and will likely sound different. Again, by regulating the preamp supply, you're removing another source of sag and "compression" that may be something that you or someone else likes.

So how is the best way to achieve this with silicooties?

"Best" is a loaded term. It is meaningless until you can define some way to measure what direction "better" is. This sounds like a bit of sophistry, but its not.

How much voltage drop is feasible without needing to use a huge heatsink and then run into limitation there?

You can fairly easily calculate this by using the heat analogy to Ohm's law. Temperature is analogous to voltage, heat to current. Thermal resistance is generally expressed as a "voltage" drop per "current" flow, as degrees-C per watt. To get what size heat sink you need, you find out what maximum screen current you're willing to allow your regulator do provide to the tube(s), and then what voltage will be across the regulator devices. Power in watts of heat generated is then P =V*I. The semiconductor device you use for a regulator will need to keep its junction below a value specified on its datasheet, generally 125-200C. You derate that some to help with longer life, perhaps to 100C absolutely maximum for the design, and then divide by the power to get degrees C per watt (C/W).

The device has some internal thermal resistance, which is spelled out for you in the datasheet, generally 0.5 to 2-3 C/W. You subtract that form the total C/W you calculated from heat generation, and if the number comes out below about 1.0C/W, you have a HUGE heatsink. If it comes out negative, even an infinite heat sink will not protect the device from burning up. If the remainder is in the 3-10 C/W range, you have a smallish, easy to find and use heat sink.

The heat you have to remove is tied to the PRODUCT of the voltage you're trying to drop and the maximum screen current your regulator will allow. There are some subtleties to calculating the performance of the regulator under current-limit conditions.

My first project to tinker with this will be an 8-knob AOR100 that I can gut as I'm not feeling it tonally and it was cheap (OT primary is 2K). It produces about 460VDC at the rectifier. So what do I need to figure out? I suppose the first thing is how to drop and regulate the voltage which I guess is ideally suited to MOSFETs

Good start. MOSFETs are easy to work with and heat-hardy.

and then I need to worry about the current limitation of the device which shouldn't be too extreme for a quad of EL34s and the preamp if I'm only regulating from the screens downwards.

One decision to make I've already mentioned - regulate the preamp from the screen supply in series, or from the main B+, tee-ing off the screens on their own. The amp may sound different in normal operation, and there will be issues with what happens when a screen shorts, screen current under massive overdrive, etc.

Another decision to make is whether you regulate screen voltage to all tubes in parallel or one regulator per tube.

Finally, regulating screen voltage can be a crude affair compared to what modern practice says is "regulation". MOSFETs are so easy to drive that you can actually just set up a resistor/zener to ground as a reference, and tie the MOSFET gate to that. The source is then regulated as well as the zener voltage is. A series resistor to the MOSFET gate and a current clamp using one bipolar device gives you current limiting. The resulting regulator can be as crude as one resistor, a string of zeners (to get to multi-hundred volts), three more resistors, the main MOSFET, and a small NPN. Or you can go wild with either discrete devices or ICs.

The LM317, by the way, can regulate hundreds of volts when buttressed with a MOSFET pre-regulator set to hold its incoming voltage to no more than 30V or so above the output. It works fine floating this way.

[Zozobra]

Thanks for the input guys. My preference is to try something simple initially to see how I like the sound so maybe the basic vvr approach will serve me well for testing purposes and if I like it I could then try more advanced circuits with current limiting and what not