Regulated B+ supply board

[thorpej]

Hey folks -

I guess I'm easily distracted... I've been working on my own copy of the @kward bench power supply, and in the back of my mind is this EL84M amp based on Heathkit AA-100 iron that I've been nudging along ever so slowly.

I've done a few design iterations of the power supply for the amp in simulation, and I have been planning on doing a regulated supply. But the work on the bench supply made me think -- hey, maybe I should take a crack at a Maida-style power supply using a power MOSFET rather than the 6550 array... From Kevin's earlier bench power supply thread, I know he had trouble with solid state pass device, but it's not like this thing has to stand up to the rigors of a bench unit!

Now, having made the bench supply boards on thru-plated perf boards, I don't care to really care to do that again... especially since if I like it I'll stick with it... and with small runs of custom PCBs being so cheap these days, that's what I'm going to do.

This is basically the floating LM317 section from Kevin's bench supply, but controlling a 700V 8A IXYS power MOSFET rather than the 6550 array in Kevin's unit or the Darlington voltage follower in the original Maida design. Rather than the adjustment pots, there are 4 footprints for resistors for the adjustment divider -- an upper leg of 2 parallel resistors, and a lower leg of 2 parallel resistors. The parallel pairs provide some flexibility in fine-tuning the divider ... or you can use flying leads to an external adjuster network if you really need to. The lower leg of the divider has footprints for vertically-mounted wire-wound power resistors, because they'll have to dissipate some heat... in the case of my EL84M amp, 300 upper and 100K lower for 421V results in the lower leg dissipating a continuous 1.8W.

I put the 2 TO-220 packages on individual Aavid Thermalloy heat sinks that are flat on the back, and flush with the edge of the board. The idea is that you can mount the the board standalone or attach the heat sinks to your metal chassis if extra heat sinking is required.

This board doesn't do it all... in particular, you need to have a reservoir cap after the rectifier, and you can also use a small value power resistor (33 Ohms in my case study) to handle some of the workload at higher currents so that the MOSFET doesn't have to dissipate so much power. Those are left as an exercise for the reader, as are the upper and lower divider resistor values for your specific application.

As for capacitors, I'm using a 10,000 hour Nichicon electrolytic for the output reservoir, and Panasonic rectangular polypropylenes in the film cap positions. None of the resistors require anything better than 5% tolerance, except the adjustment divider, but only because you want predictable results. The only exotic components are the 9.1V Zener and the 22V TVS diode -- and they're not especially exotic.

I spec'd a 600V-rated TE AMP 3-position screw terminal block with a 10.16mm pitch, but you could skip that entirely and just solder flying leads if you like. Mouser doesn't sock these, but they're readily available at Digi-Key.

I'm about to submit the Gerbers for the initial test run of these boards, so the "completely untested" grain of salt applies... but I've already had one friend inquire about kits. I could certainly entertain that if there's sufficient interest.

Feedback welcome.

 

[s-petersen]

I built this one, it seems to work fine.
I

 

[kward]

Just a word of caution with any Maida-style supply: The voltage dropping mosfet (or BJT if you go that way) should not see more than say 100V drop max at sustained currents of say 100 mA (and even at that it will require decent heat sinking). It doesn't matter how beefy the mosfet is--in a linear supply with constant voltage and current demands, even the most massively built silicon transistor will still blow out if over driven in a linear supply. Even if you overrate the heat sink, the limit increases somewhat, but you will still run into the limits, and they are much much less than the stated datasheet values. (I believe this is because those datasheet values are stated assuming switching applications, where the device is "on" for only a fraction of the the total duty cycle).

So what that means is once you determine your desired regulated output voltage, you need to work backwards all the way back to the transformer to determine the needed AC secondary voltage, so that after DC conversion, no more than 100V (@ 100 mA) will be dropped across the device. Then find the closest off the shelf unit that meets the transformer specs you determined. If you have a lower current load, you can get away with dropping more voltage across the mosfet. If you have a higher current load, you need to derate even further. About the max current I was able to draw with a single dropping device was 300 mA @ 20V drop. But the thing got awfully hot also, so not an actual workable solution, but it more or less stated the bar from which you'd need to derate from.

 

[thorpej]

Yah, in my initial application, the MOSFET drain sees 440V (with about 4Vpp of ripple centered around that) at full power (which draws ~285mA), which is being dropped to 426V by the MOSFET, and is 422V on the other side of the LM317. The MOSFET I picked for the job is a fairly new IXYS part with a 150W-before-derating dissipation rating; at full power, simulation tells me the MOSFET is dissipating 4.2W. I'm sharing the load with a 33R 5W resistor in series with the drain (not shown on the schematic for the regulator board because it would be connected to the "IN" terminal) that, at full power, dissipates 2.8W and drops ~10V. Simulation tells me I could bump that to 47R and still have sufficient headroom for the LM317, which gets the MOSFET down to 2.5W dissipation at full power.

The MOSFET is actually dissipating more (5.5W) at idle (100mA) because not as much is dropped across that dropping resistor when it's 33R. A 47R resistor would reduce that a little, but only just (down to 5.3W). I may do a little more adjusting ... i.e. it's not really clear that the 10R I have in series with the LM317 is buying me much ... I put it there to take some of the load off the LM317 at higher currents, but the LM317 is barely even breathing hard in any case. E.g., if I eliminate the 10R resistor in series with the LM317, it looks like I can bump the dropping resistor to 51R, which gets the idle MOSFET dissipation down to 5.2W and only 1.8W at full power (the resistor is dissipating 4.5W and dropping 14V).

 

[thorpej]

...and a companion board, for the screen supply. This originated with @dcgillespie's universal EFB(tm), as I recall... But @kward has used this on several of this builds, as well. The idea is to provide a semi-regulated screen supply that's referenced to B+... as it sags, so does the screen voltage (and, in a full-blown EFB universe, so-inversely-does the bias voltage). It's comprised of a voltage divider and filter cap to smooth out any ripple, along with a power resistor to provide some voltage drop before the power MOSFET. The voltage divider sets the screen voltage (desired screen voltage + 4V, or whatever the threshold voltage for the MOSFET is). Because the MOSFET is a voltage-controlled device, barely any current needs to flow... so relatively high-value resistors can be used in the divider, meaning the ripple-smoothing filter cap can be fairly low-value. I spec'd 1uF for the board because it'll cover practically everything (in my initial application 0.47uF is plenty), and I wanted the board to be generic and configurable.

The upper leg of the divider is formed by R1, R4, R3. The lower leg is R2, R5, R6. R3 and R4 are in series, as are R5 and R6. R1 is parallel with R3+R4, and R2 and parallel with R5+R6. This gives a ton of flexibility in configuring the voltage divider to obtain the desired voltage.

R7 is a power resistor designed to provide some voltage drop to take some of the load off the MOSFET. Make sure to space it up away from the board when mounting the part. In my initial application, the resistor dissipates quite a lot more power than the MOSFET (which never dissipates more than about 2.5W).

The way I've laid these boards out, with the flat-backed Thermalloy ML33G heat sink, means you can smear some thermal compound on the back of the heat sink and mount these boards to the chassis using the head sink (make sure to insulate the TO-220 package if there's an exposed tab!). This will take up less space under the hood as the boards will be "vertical".

 

[thorpej]

Had the boards fabricated, and tonight I built up one of the Maida boards and one of the screen regulator boards. I also built up some double-snap-in-cap boards for cases like these where I want to stack two caps for voltage handling... the cap stack is center-tapped in case one wants to tap the 1/2 B+.

Encountered an unexpected issue, though! The IXYS power MOSFET on the Maida board would not take the insulating bushing from a standard TO-220 insulated mounting kit. I ended up having to drill out the hole in the tab. (I also didn't expect to have an exposed tab... the Mouser part number implied an insulated tab, but no big deal...)

The screen board worked perfectly when I tested it, providing precisely 75% of the input voltage on the output no matter what the load. I have some additional stuff to build before I test the Maida board.