Stabilised power supply

Why bother?

What on earth is the point? The tube rectifier with capacitor-input filter and a choke for the screens was good enough for your granddaddy, it should be good enough for you! Well, I love tinkering with electronics. I was also interested to find out how the tone of the amplifier was affected by ripple and sag in the HT supply. Eliminating them seemed to be the obvious experiment.

Practical advantages are: current limiting protects the power tubes from damage, all hum in the speaker is eliminated (even with no negative feedback), the amp is less sensitive to line voltage variations, and the tone becomes more dynamic, because you get no compression caused by sagging screen voltage.

The regulator costs approximately 30 UK pounds to build, including a heatsink (about the price of a filter choke) and fits on a double row tag board (it's a bit crowded mind you) The plate voltage has mains ripple of 0.1V p-p at full load, and load regulation of 2V (no load to 250 mA)

How does it work?

RIght-click to open schematic in separate window

Conventional power supply regulator circuits are not really suitable, because they can't stand the high voltages. I designed this circuit specially for the job.

It's all based around U1, which is a shunt regulator IC. This is placed in cascode with M1, to protect it from the high voltages (it's rated at 30V max) It senses the output voltage via R15-R18, and adjusts the voltage on M2 gate. C4 provides a dominant pole for stability, and improves ripple rejection by smoothing the voltage on M2 gate. Q1 and associated components provide foldback current limiting.

The screen voltage is derived from the divider R19-21 (which also supplies bias for ZD1) bypassed by C8 and buffered by M3. The 250V preamp supply comes from a standard dropper resistor arrangement. It isn't regulated at all, but I think this contributes to the tone.


The circuit headroom is limited by the current drawn through R8,9. Therefore I've done a few things to minimise this. C4 is plastic film for low leakage, and R24, R25 supply a minimum bias current for U1. M2 gate of course draws no current, and neither does Q1 collector, unless the current limiting is active.

C7 could be made smaller (4.7uF?)

You can implement a standby facility by using a switch to break the connection between R9 and the unregulated rail. This works nicely, it causes the HT to ramp up over about a second, so no thuds. Use a SPDT switch so you can ground R9 (via a 1M resistor for slow ramp-down) in the standby position. I couldn't show this because I hit the 50-part limit in my schematic program.

Tips on using the circuit

Mind the high voltages! They are high current too and can easily kill you stone dead! Remember to Switch off - Isolate - Discharge - Earth! Keep one hand in pocket at all times!

The BFC62 is a 600V, 1.5A N-channel Euro mosfet. Any 600V device will do, as long as it can take plenty more current than you will be drawing.

Make sure the resistors are up to the voltage rating! The small metal film ones can only take 350V, if you exceed this, the insulation will break down. I destroyed a few shunt regulators before I sussed what was happening.

Remember to insulate M2,3 tabs from each other, and the heatsink. Ground the heatsink! The dissipation, thus heatsink size, will depend on the output tubes, power transformer, and filter you use. In the Toaster prototype they dissipate about 15W and are fitted on a 1 degC/W heatsink.

When testing the regulator, three 240V 60-watt bulbs (or 6 120V 30-watt bulbs) in series make a great dummy load. They draw almost the same power as two EL34s on full blast.

Have a nice day and don't kill yourself!


e-mail me: "steve at scopeboy dot com"
Home - Toaster top page - Electronics

(C) 1999 Stephen J. Conner (2/11/99)