Yet another question about HV Maida Regs

[jgrg1]

Hi,

I'm in the design phase of building a new guitar valve amp, and I'm looking at regulating the HT using a Maida regulator. My requirements are 400V DC @ 200mA as I plan to use it to supply HT to the entire amp (yes, I know the PP rejects common-mode noise, but if I have a well-regulated HT, why not use it). Having read the Morgan Jones book, I figured I'd try to design one for my higher voltage and current requirements. The amp will be an 2xEL34 PP with 5xECC83s. I already built 2 other amps (a 2xEL34 and a 4xEL34 PP).

See attached image for my PSU design (which has been ripped off of various posts here and from Jones' book).

1. Can anyone confirm my reasoning:
I have a 310Vac secondary, which I reckon will give me around 437V out (including diode drops). Given my C1 filter, the Vripple is 9Vpp at 200mA draw, which means the DC Vin will be around 432.5V +/- 4.5V.

Now, I'm having a little trouble fully understanding the operation of the maida regulator as there are some conflicting equations on the web. In particular, the exact formulae used to calculate the values of R4 and R5 in my schematic. My understanding is that (please confirm?):

Vout = (1.25V / R4) * (R4 + R5)

Following Jones' logic, he's used a high value R5 of 220k to reduce the power dissipation through it, and he then sets R4 to the appropriate value based on the equations I've shown in the diagram. So for my 400V out, R1 should be around 709 ohms.

2. The calculation of R3 (the resistor that sets the current for the D1 12V zener) I've also lifted from Jones' book. In that, the current should be above the zener current of the diode (5mA). So I make that a Vdrop of 432.5V in and 400V out, less the 12V zener voltage of 20.5V. If I want 6.5mA through the zener, V=IR gives R3=3076 ohms. So I'd use R3=3k ohm for 6.7mA. Is this correct?

3. The only other thing that has confused me is the presence and value of R2. From what I read, this seems to act as a current limiter for the IRF840 pass device. Is that correct? And if so, what value should I use for my current requirements? I couldn't find any description in Jones' book of what value to use. Also, what power rating should it be.

4. Should C2 be dropped?

5. Does D4 have any protection effects at all (I've seen it in various maida schematics)?

6. Also, is the IRF840 a suitable pass device?

Cheers all
J

[tomchr]

Hi,

I've looked quite a bit at the Maida regulator. It works very well as long as the cascode device (the IRF840 in your circuit) is kept within its SOA limits. Otherwise, it goes poof in a hurry.

Check out this thread: High Voltage Regulators (Maida or zener)

In my circuit, I needed 475 V, 200 mA but I also needed it to be able to start up into a 47 uF capacitive load. The capacitor represents a dead short at start-up, and the cascode device did not appreciate having 600 V across it while supplying about 1 A (peak) to charge the cap. Even with a 10A rated device, it still went poof. You won't have as much trouble as you're "only" supplying 450-ish V to your regulator. But it's something to be aware of. I ended up designing a floating regulator that I sell circuit boards for (see my website). It works quite well.

That said, the Maida regulator is pretty solid as long as you observe the SOA limit on the cascode. I was quite happy with it as a supply for my 300B until I converted to a Loftin-White configuration where the supply "sees" a 47 uF cap.

To work well, the Maida regulator needs to supply minimum 5~10 mA. So the resistor between the OUT and ADJ pins (R4 in your case) need to be no larger than 250 ohm. Most people use 120 ohm, but for high-voltage circuits this results in lots of power dissipated in the "lower" resistor (R5 in your case). So many people go with 220 ohm.

The equation for Vout is: Vout = 1.25*(1+R5/R4), which I think is what your math reduces to. I'm getting Vout = 390 V with your values.

It looks like you're on the right track. Your calculation for the zener resistor makes sense. Beware of the power dissipated here...

R3 will present a "ghetto" current limiter. I say "ghetto" as it really doesn't work that well. Basically, it works by reducing the available Vgs on the IRF840 and input voltage on the LM317. The current limit will be very soft and its trip point depend heavily on part-to-part variation and temperature. The LM317 needs about 2.5 V drop-out to be happy. So figure R3 < (Vzener - Vgs_max - 2.5)/Iout_max.

C2 will cause the LM317 to die a sudden and painful death if there's ever a momentary short circuit on the output. Remove it.

D4 isn't strictly needed, but depending on how your bias circuits shut down, you may end up with a negative voltage on the output of the Maida imposed from elsewhere in the circuit. I'd leave it in. Diodes are cheap.

Is IRF840 suitable? Don't know.... Look at your start-up requirements for current and compare with the SOA graph in the datasheet. Basically, the cascode should be selected to survive a DC condition where it supplies the start-up current into a short circuit.

Also note that the MOSFETs have the drain connected to the case/tab. So you'll need good isolation here. Beware of burrs and such on the heat sink.

Oh... Another commonly overlooked thing is to ensure that your regulator has sufficient drop-out voltage available when the mains input voltage falls to its lowest value (I generally design for 10 % variation in mains voltage).

~Tom

[jgrg1]

Yeah, I read that thread. Interesting stuff, but the end circuit looks far too complicated for me (on top of everything else that I have to build). So presumably if the maida is feeding a "standard" preamp supply chain of resistor/cap droppers, the pass device won't "see" a purely capacitive load, or at least be limited in terms of current draw by the dropper resistors?

With regards to the minimum current, presumably it's not the divider chain that has to draw that current, it's the entire circuit? So as long as I have a load drawing more than 10mA, the LM317 will regulate?

Regarding power dissipated in the zener resistor. I make that 6.7mA with 20.5V, which is 0.14W? Or is there more at work here?

"So figure R3 < (Vzener - Vgs_max - 2.5)/Iout_max.". So by this I reckon R3 < (12 - 2 - 2.5) / 0.2 = 37.5R? So a 33 or 27 ohm resistor should be ok? With a power dissipation of 1.5W?!

Will do, consider C2 ditched.

Hmm... so the AC in could drop by as much as 10%?! In practice, does this ever happen?

Another question, in various schematics I've seen around maida's, some also define a 100R between the zener reg and the gate of the pass device. Is this sensible?

[tomchr]

Originally Posted by jgrg1

So presumably if the maida is feeding a "standard" preamp supply chain of resistor/cap droppers, the pass device won't "see" a purely capacitive load, or at least be limited in terms of current draw by the dropper resistors?

In the traditional RC chain, the R's tend to be sizable - 10's of kOhm. So the inrush current will be limited and the Maida regulator will likely survive.

In my circuit, the regulator 'sees' a large cap in series with 750~900 ohm. When fed 480 V, that's well over 600 mA peak on power-up. I apply roughly 600 V to the input of the regulator, so on start-up, that's 360 W peak dissipation. 720 W for a stereo amp. The regulator pass device turns into silicon slag instantly unless I have a slow turn-on and current limiter.

Originally Posted by jgrg1

With regards to the minimum current, presumably it's not the divider chain that has to draw that current, it's the entire circuit? So as long as I have a load drawing more than 10mA, the LM317 will regulate?

Correct. And if you can guarantee that your circuit always draws more than 10 mA, you're good. This either means that you have a bleeder resistor somewhere or that your circuit is designed to have the heaters come up well before the B+. Recall, tubes don't draw current with cold heaters...

Originally Posted by jgrg1

Regarding power dissipated in the zener resistor. I make that 6.7mA with 20.5V, which is 0.14W? Or is there more at work here?

P = I * E --> P = 0.0067 * 20.5 = 0.14-ish W. That sounds right. That's all there is to it.

Originally Posted by jgrg1

"So figure R3 < (Vzener - Vgs_max - 2.5)/Iout_max.". So by this I reckon R3 < (12 - 2 - 2.5) / 0.2 = 37.5R? So a 33 or 27 ohm resistor should be ok? With a power dissipation of 1.5W?!

I don't know off the top of my head what the worst-case Vgs for the IRF840 is at the current you're running, but 27 ohm doesn't sound too far off. 1.5 W dissipated sounds right so use a 5 W type. Yes. Really... Welcome to high-current Maida regulators.

Originally Posted by jgrg1

Hmm... so the AC in could drop by as much as 10%?! In practice, does this ever happen?

All the time. The spec is +/- 5 % in the US, but I think that's for the steady-state value. I know the voltage at my house dips well below 10 % of the nominal value when a heavy load turns on (think laser printer, dryer, etc). even though the distribution/pole transformer sits in my front yard.

Originally Posted by jgrg1

Another question, in various schematics I've seen around maida's, some also define a 100R between the zener reg and the gate of the pass device. Is this sensible?

Mike Maida's original regulator had a bipolar darlington transistor for the cascode device. In those, a base resistor is not used. For a MOSFET version like you're building, I'd put in the 100 Ohm "gate stopper" resistor. The traces/leads to it need to be as short as possible. Its purpose in life is to prevent the MOSFET from oscillating.

I forget if you showed in your original schematic, but you need an input cap (Vin to GND) as close to the MOSFET as possible. Something like 1~10 uF polyprop is suitable. This prevents oscillations as well (don't ask how I know... )

~Tom

[tomchr]

Here's Mike Maida's original app note by the way. It has a few good nuggets in it... National Semiconductor Linear Brief #47

~Tom

[tomchr]

Another way to accomplish a soft start-up is to use a vacuum tube rectifier with a slow warm-up. I've used a Maida regulator with a 5AR4 to supply 380 V, 100 mA. That worked pretty well. Just mind the voltage drop across the rectifier. You'll also have higher ripple voltage as the tube rectifiers can't handle as much cap as a solid state rectifier.

~Tom

[jgrg1]

Hi,

thanks for the help. My old uni supervisor (electronics genius) independently corroborated your advice on the R3, and suggested adding the 100R resistor to reduce oscillation.

Just for reference, the Vgs threshold for the IRF840 is quoted as 2-4V. The dropout for the LM317 is 2.5V. So according to my calculations, if Vbs=2V, R3=37R and if Vbs=4V, R3=27R. So if I go with R3=27R, worst case if Vbs=2V, Imax=277mA. Which should be ok.

As an aside, what actual effect on tone does regulating the entire PSU have on an amp. I read AC Ripple in a Class AB Power Amp and they came to the slightly odd conclusion that:

"In spite of the practically identical voltage curves (with the exception of the 100Hz ripple), in terms of sound the amplifier is no longer recognizable. Its tonal characteristics are completely different."

I was just wondering if anyone else has experienced this effect at all. Where does PSU sag have the most effect; presumably the power amp?

I'm going to stick with solid state regulation for the moment, but may leave the power amp "unregulated".

Cheers

[tomchr]

Originally Posted by jgrg1

Just for reference, the Vgs threshold for the IRF840 is quoted as 2-4V.

Yeah, that's the threshold for Id = 250 uA. You're running a lot more than that. Look further into the data sheet for the Id vs Vgs curve and read the worst case Vgs for the drain current you're running.

Originally Posted by jgrg1

The dropout for the LM317 is 2.5V. So according to my calculations, if Vbs=2V, R3=37R and if Vbs=4V, R3=27R. So if I go with R3=27R, worst case if Vbs=2V, Imax=277mA. Which should be ok.

I take you mean Vgs not Vbs. Vb = Vs.... Vb = Vbulk. Vgs = 2 V is NOT the worst case. Either set up the math symbolically and deduce the worst case condition or do the math for the two extremes. Recall that the goal here is to ensure that the current limiter does NOT trip during normal operating conditions.

Originally Posted by jgrg1

As an aside, what actual effect on tone does regulating the entire PSU have on an amp.

Try it and find out... I'm a measurements kinda guy. I'd rather not get into discussions about how many veils have been lifted by making certain modifications. But if I can measure the 50/60 Hz "grass" change on my spec an, I will consider a lowering of the 50/60 Hz IMD to be an improvement.

Originally Posted by jgrg1

I'm going to stick with solid state regulation for the moment, but may leave the power amp "unregulated".

Regulating the input stage and letting the output stage run on an unregulated supply is certainly a valid option. That's done all the time in solid state amps -- and is actually also effectively what the classic RC decoupling of the input stages does. You could go fancy and use a capacitor multiplier (like is done for solid state). Options. Options.

You're on the right track with the Maida regulator, though. You're almost done with the design. Now build it!

~Tom