Power Supply Design Strategy, From Morgan Jones

[GeorgeF]

Hi All,

This is a spinoff from the PSUDII power supply modeling thread, that I decided to give its own place.

I've been studying the power supply design chapter in Morgan Jones' Valve Amplifiers. It's quite mathematical, and I certainly don't understand it all, but I hope I've managed to derive a useful design strategy from it. (Note I didn't say the *best* design strategy.) I'm post it here, for your consideration and comments.

Here's what I've come up with--anything good should be credited to Jones. Regarding anything stupid, well, the source is obvious:

0. Create a filter consisting of either an initial LC filter, or a CLC filter, followed by as many RC stages as necessary. In other words LCRCRC...
or CLCRCRC... (this is not Jone's prescription, but my interpretation of it for this discussion--of course there are myriad ways to make a filter--this just seems a reasonable pair of strategies)

1. Select transformer VAC and current rating

Current rating should be at least as high as the anticipated load. VAC should be such that the DC component after the first filter element (either L or C) is a bit higher than the desired B+, to allow for additional voltage loss after filtering. I decided on 20V over B+

Initial inductors supply DC of approximately 0.9VAC. Large initial capacitors supply DC of approximately 1.4VAC, down to 0.9VAC for small caps--this allows tuning of the DC voltage.

2. Select either an inductor or capacitor as the initial filter element.

As seen above, inductors supply less voltage, but are less subject to voltage changes in response to load change. They also are more finicky, and require a minimum load (current flow) in order to function correctly. With too little load the voltage also drifts up towards 1.4VAC.

Capacitors supply more voltage and do not require a minimum current draw to filter correctly, but are more prone to changing voltage in response to changes from the nominal load. I selected a capacitor input, and the following steps follow that path.

3. If going capacitor input, size the initial cap so ripple is 5% of DC.

I'm guessing/hoping this keeps inrush current to a reasonable level , along with the corresponding RFI. My manipulation of Jones' equations for this results in:

C1 = I_Load/(5 * V_B+)

4. To stabilize the filter, add a resistance in series with the inductance in the the LC filter to give it a Q of 0.5

My manipulations give the equation:

R_additional = 2*Sqrt(L/C) - R_inductor

5. Determine the R necessary to drop the voltage at the LC filter to the desired B+

6. Decide how many RC stages are necessary to drop the ripple after the LC filter to an acceptable level, divide Step 5's R by that number, and calculate the value of the C's in each of these RC stages.

Unfortunately you'll have to buy the book to do this. It's easy once understood, but it involves consulting a table. It wouldn't be fair to Jones to reproduce it here. Jone's strategy seems elegant, for it allows iterative RC filters to do the work of a single RC filter, with smaller individual capacitors and less total capacitance.

Example:

I'm building an Aikido Headphone Amp based on John Broskie's Aikido Headphone Amplifier Recipe on TubeCad.com. It' requires a B+ of 250V, and I'm guessing that it has constant average current draw of 50mA, so I went with a capacitor input filter. I also want to incorporate a 10H 270R inductor I have, along with a 100uF ASC oil filled capacitor for the final cap.

Using the above strategy I came up with this as the filter sequence:

C(L+R)CRCRCRC.

Specifically 225VAC, 40uF, 10H/270R + 424R (for Q = 0.5), 83uF, 147R, 90uF, 147R, 90uF, 147R, 90uF.

PSUDII gives me a B+ of 250V and ripple of 13uV, and I do not see any ringing in response to stepped load changes, just a smooth rise/fall to a new voltage.

Whew! I hope folks find this useful in some fashion, and invite your comments.

Best Regards,

George

[NickG]

Sounds good, nice informative post.

BUT...

I wonder how it will sound. You will only get ringing from a LC stage, so the distributed RC stage will further ensure that you get no overshoot.

It will be interesting once its built, to compare how it sounds with different power supply designs.

I bet (IMHO) that a simple CLC or LC followed by a regulator will sound better.

[Ianm2]

nice post, do you need so much filtering tho, seems large overkill?

if you carefully design a clc filter, it shouldn't oscillate.

[GeorgeF]

Originally Posted by Ianm2

nice post, do you need so much filtering tho, seems large overkill?

if you carefully design a clc filter, it shouldn't oscillate.

Thanks Ian,

Yes, the filtering is probably overkill, which is fine as a learning exercise. What amount of PS ripple would you consider excellent, but not obsessively low, for a preamp?

Any tips on carefully designing a CLC filter?

[GeorgeF]

Originally Posted by NickG

I bet (IMHO) that a simple CLC or LC followed by a regulator will sound better.

Thanks Nick!

If it's not too much trouble, could you point me to a thread/webpage/whatever that describes a regulator that you think would be ideal for this application-- 50mA at 250V?

[NickG]

Well MJ has some good examples using three pin SS regs with mosfets to hide them from excessive voltage. For valve regulators, I would start with Steve Bench.

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

[colin.hepburn]

Nothing to do with MJ But I found this one maybe of interest to you works with voltages up to 350 volts
http://www.tubeaudiokits.com/LIBRARY/Marantz-7.htm
More Details @

http://www.tubeaudiokits.com/Modules/modules-init.htm

[NickG]

Shame about the 800R resistor in the output though.

[GeorgeF]

Thanks all for the comments--I found them interesting and educational.

Any suggestions on what you'd consider a reasonable ripple figure to aim for? Something silent, but not obsessive?

Thanks!

[NickG]

Remember though under about 1mv, noise starts to become as important as ripple, thats why I say that ripple isn't everything, and also you have to consider how the supply behaves under dynamic load, thats my I thought the 800R resistor was a shame.