Lab quality high voltage power supply
In general, regenerative receivers are notoriously sensitive to power supply noise and my regen is no different. I had experimented with several switchmode power supplies while the regen was under development and they were all noisy across the whole spectrum. I tried installing suppressors and dampers but even these didn't suppress all the noise. At one time, I even considered connecting enough alkaline batteries in series, to get the required 150 plus volts. After all, batteries provide the cleanest power there is. Well, I may have been thinking about, it but not seriously ... connecting twenty 9V batteries is certainly not practical. I knew I had to go with the ordinary 60 Hz step-up transformer and rectifier approach.
So how did the designers do it back in the day? A transformer stepped up the mains voltage, a rectifier (which was a tube, too) turned it into pulsating DC which was then more or less thoroughly filtered. Contemporary filters were either just a large capacitor after the rectifier or a more "deluxe" PI filter with two capacitors and a choke. I wanted to suppress the pulsation electronically although employing a transformer-like choke would have given my regen project a 100% retro look.
Please be careful, be sure you know what you are doing. If necessary, find a qualified person to assist you. High voltage can kill!
After thinking quite a bit and simulating in Spice, I chose the circuit below. Here, I employ deep negative feedback to smooth out the output. The left part of the circuit, consisting of a transformer T1, rectifier bridge D2 and a large electrolytic capacitor C2, forms a classical power supply with a fair amount of ripple on it. The actual "brain" behind the circuit is the ubiquitous TL431 shunt regulator. It "looks" at the output (180V in my case) through the divider R6, R8 and R9. Resistor R1 ensures the minimum of 1 mA current through the U1, as well as serving as the load for the Q3. High voltage MOSFET Q1 works as a source follower, being protected by Q2. Resistor R4 in turn protects the Q2 so that it will never "see" more than 5-6 volts. R5 sets the maximum short circuit current, about 120 mA in my design. Another high voltage transistor Q3 acts as a go-between for the cathode of the TL431 and the gate of the Q1. This way the U1 never "sees" more than 33 volts. Resistor R2 and diode D3 protect the U1 in case of a short circuit. I used the same component, 2SK2750 for both Q1 and Q3 but it doesn't have to be that way. Q3 can be any high voltage MOSFET. It needs no cooling since it passes only a few mA of current.
The circuit works as follows. When the voltage at the output decreases, U1 senses it through R6, R8 and R9. This results in an increase of voltage at its cathode, which is then further amplified by Q3. For the error signal, Q3 acts as a common gate stage. This amplified error signal is then applied to the gate of Q1 with the opposite sign, always canceling out the change at the output. Resistor R3 and Q1 gate capacitance (which can be pretty significant, on the order of 500-1000 pF) create a pole in the loop transfer function, without corrective measures this circuit will most certainly oscillate. My remedy was to place a 0.1 uF capacitor C1 in parallel with R3. This creates a zero much earlier than the pole, which keeps the circuit stable. Diode D1 protects Q1 against reverse polarity.
Work in progress, everything is mounted on a slab of masonite. The pass transistor Q1 needs sufficient cooling because in a worst case scenario it dissipates close to 27 watts. I used an old Pentium II CPU heatsink.
The object in the foreground is a speaker terminal block, "liberated" from a junked boom box. This way, I can connect anything I want with zero soldering effort. Its wire clips look like they can only pass a little current, but this is unimportant here, as the currents drawn do not exceed few hundred mA.
A close-up of my circuit: All components attach to a piece of PCB material. The copper foil has been cut into islands which form the circuit nodes. Component legs are then soldered onto these islands. This is a quickie method for creating simple circuit boards such as my project here.
The other TO-220 device on the left is a LM317 3-terminal adjustable regulator which stabilizes the tube heater voltage. I didn't want to leave anything to chance and I wanted the power supply to perform as quietly as possible. Therefore the DC heating was employed. The LM317 circuit wasn't drawn up here because it's used in it's standard application.
Top view of the components layout.
A close-up of my rewound transformer. If I remember correctly, it came from a discarded boom box. After carefully taking it apart, I rewound its secondaries. An excellent tutorial on how to reuse transformers to suit one's own needs is found on Harry's Home Pages. Since the transformer was the heaviest item on board, it required special accommodations. I employed two aluminum spacers and a home-made sheet metal clamp to secure it to the board. The object in front of the transformer is the 3 Amp fuse.
Unfortunately, I never found time to fully enclose this project in a box. As depicted, this high voltage power supply is pretty dangerous, due to the fact that all the components are exposed. I make you this solemn promise: before using this power supply again to power one of my future tube projects, it will have a proper housing, so nothing can be accidentally touched or dropped in.
This little circuit works very well. While powering a load drawing 80 mA, its output ripple was measured to be less than 2 mV. At the same time the pulsation before the stabilizer was 6 volts peak value. This works out to a stabilization coefficient of about 3000. Not bad at all!
Hope you had some high voltage fun and if you zap yourself, don't blame me ;-), de Brian.