Simple HT supply build

[Mike Watterson]

The DC was from 2x PSUS in series.

Next I'll wire up the transformer (which has a core based current limit and centre tapped 1:1 secondary), rectifiers and capacitors for one diode regulated negative supply with a pot for a -Vgs and several +V rails.

I think I'll use a stock 12V PSU via pot to gate test FET bias with a single 26W lamp in series to the nominal 300V rail and see what sorts of voltages happen. May be possible to overrun lamp. Maybe one on 150V HT and two in series on 300V HT?

[Mike Watterson]

So at 121.5V, the new 26W lamp is a 9.2W load approximately.
If it was a linear resistance then it would be only about 6.5W at 120V, because power is V squared.
Thus if HT was 243V, the FET would also have 9.2W to dissipate when there is no load.
An EL84 is about 50mA anode max and 6mA screen max.
At 50 mA load (FET and external load added) the voltage drop on the lamp would be about 45V.
The 6 ma screen would be via a a separate lamp with a separate parallel FET. Almost full HT would be available when FET off

Also the FETs sink slightly less as they heat up. No thermal runaway.

The unreadable current on the new 26W lamp was 56.8mA at 60V.

The FETs might be about €2 each.
.

[Mike Watterson]

I found alternate oven lamps made by Solas (Irish for light. Irish had Latin loan words before the Christians came). They are 15W medium ES. Current product. Likely there 15W alternates to the 26W Phillips also in UK and elsewhere.

I also ordered 2 panel cheap LED meters that do volts and current. The current is 10A FSD, but may have 1A option. Or the shunt may be accessible to change to 100mA FSD.
Voltage measurement range: DC 0V-100V
Current test range: 0-10A
Working voltage: 4.5-30 V DC
(Note: if the voltage exceeds 30V, an external independent power supply is required)
Voltage error: ±0.1%
Current error: ±1%
Working current: <20mA
Refresh rate: ≥300mS / times
Display: 0.28" Two color blue and red
(the current is measured in the negative rail) Not sure what that means!

Likely I'll power them from 12V DC. Increasing the voltage range is trivial. I'll add a x10 switch.

[Refugee]

(the current is measured in the negative rail) Not sure what that means!

Likely I'll power them from 12V DC. Increasing the voltage range is trivial. I'll add a x10 switch.

That looks like you can power both meters from one common power rail.
It would require a second floating power supply if it was done in the positive rail.

[Mike Watterson]

Arriving Monday. Under £4 each inc shipping and VAT.

One option seems to be to power it from the voltage being measured. That's OK for a car or truck, but not generally sensible. However it does seem to have a "separate" power option. If the current -ve is common to voltage meter negative, then the current + terminal is the load negative. Awkward for a dual HT supply feeding some device or circuit under test with a common negative. Like a Tetrode, Pentode, Hexode etc.

I have some LM3900 so called Norton mode "op amps" that can sense drop in a small series resistor at HT. They then can drive a 0V referenced current meter. Obviously they can directly measure current in the FET shunts.

[Mike Watterson]

Checked winding resistances on the bare transformer (from a Shaving adaptor).
About 55 Ohms Black to White. That's the 115V outlet
About 117 Ohms White to Red, so likely the outer.
Black to Red is the 230V. So Black is common. Actually I think in the US the white is neutral and black is live, but the transformer likely has 4KV isolation.

I have a drawer full of KPB 210G bridge rectifiers. 2A 1000V (though bag erroneously has 4000V. I always check several sources).

Warning, danger Will Robinson!

One hand in pocket and NO TOUCHING with fingers or iron till working meter shows safe!
No sort of trip or fuse is any help for safety with an isolated transformer. Ironically the initial test setup is safer with nothing earthed (except soldering iron).

Put 330 K resistors across two ex-SMPSU 470 μF 200V caps and wired in series. Added to bridge and get about 365V no load. Added white wire to centre tap as it's nominally 1/2 way. Still same voltage.

Considered how to add a pair of voltage doublers. Same circuit as a peak detector. The white wire even unconnected won't work as relatively it has no AC compared to ANY point on the rectified side.

So for layout convenience I put 1 μF 400V met-polys in series with red and black wires (input to bridge) and completed the two voltage doublers with 2 x 1N4007 each. A 1 μF 400V met-poly on each output.

Remeasured everything:
Bridge gives 370 V DC
Negative Doubler with its first diode cathode and output cap (1 μF 400V met-poly) connected to OV gives -230V.
Plus Doubler with its first diode anode and output cap (1 μF 400V met-poly) connected to the +370V gave OL. Got out AVO on 1KV DC. Read 720V. Realised there was another range on DVM 0000 instead of 000.0 and it read 709V.

Don't use the cheap DMMs with a 750V range at over 300V! I've seen those die with a blue flash at 460V on the 750V switch setting!

Obviously that's all with nearly no load. Works as expected. Nothing went bang. No-one died etc.

Next is the lamps, FETs and 12V supply for adjustable gate voltage and powering panel meters etc. I see a drawer full of giant FETs that make TO200 look like wimps. I've a box of old heatsinks from scrapped CPUs. Worst case should be less than 15W.

No PCB or tag strips. A tag strip would be best. Stripboard isn't smart at these voltages. Currently bridge, resistors, 4x 1N4007 and 4x 1 μF 400V met-poly hanging off the pair of tied together 470 μF 200V caps with centre tap to transformer centre tap.

[Mike Watterson]

It will be nice to have a leakage tester that does more than 350V. Should I add more doubler stages?

[Mike Watterson]

We need to sense the load current at the supply side because we might have a device that takes two HT voltages and only has one common return connection.

Looking at all the cheap combo voltage & current meters on Amazon etc it seems the 10A shunt is obvious, so I will cut it and power module from 5V or 12V (maybe 5V to reduce waste). Then I will measure what current gives 10.0 on the scale. I will call that 10mA.

So I will have a switchable 100 Ohms or 10 Ohms in series with HT to load. That will be 1V at 10 mA or 100mA (which the supply won't do, but it might manage more than 10mA). A 1/4 of an LM3900 will measure the voltage and with a PNP transistor on the output it will be a current source such that with shunt on meter cut, 10mA through the 100 Ohm resistor gives 10.0 on the display.

The voltage part is apparently 100V max. However I will scale the HT, by measuring volt meter input resistance, by a series resistor so that 300V reads 30.0 V. Maybe I can blank the dot, or just write x10 beside it.

Loads of people sell these and the sellers may be optimistic about the voltage the circuit can use (I'll assume 5V to 12V rather than claimed 4V to 30V. The voltage measurement, module supply and common 0V are on a 3 pin connector with a supplied plug having black, yellow and red wires. I'll use 5V to verify black + yellow is power. The current uses a fatter red and black wire and the thick 10A shunt is obvious beside the connector.

The only kind of affordable DC current at a floating HT is a battery powered DMM. So the series sense resistor and amp to current convert is the simplest. A voltmeter would only save a PNP transistor and two resistors.

[Mike Watterson]

Obviously measuring the negative supply is also less simple. One approach is an isolated power supply for the digital meter and reference the PSU 0V to the negative output, then measure the system 0V rail. A situation where a mechanical meter is simpler.

The LM3900 can connect to negative voltages. So can a regular op-amp with suitable biasing. The LM3900 has four "Norton" amps in it. I might use the fourth with a 99 M Ohm (3 x 33 M Ohms) input as buffer to measure the "Breakdown" voltage out, which may be 700V via 4x 1M resistors. Regular common leaded resistors are only 200V rated. Could be the voltage rating of the trimming spiral. Resistors are not just about power rating, tolerance, noise and temperature stability, because they have a breakdown voltage. I bought a bunch of 33M resistors to make an EHT probe. I forget how many kV each they are. The theory was that three is 99M so nothing for a 1M DMM and a bit more than 1M final for a 10M, 11M or 20M input DMM.


I've some basic LED voltmeters coming on Monday (tomorrow) too. I'll see what input resistance they are.

[Mike Watterson]

700V via 4 megaohms is 175 μA into a short. Probably about 140 uA to dry skin. A capacitor charged via 4 M Ohms is not safe. Perhaps the leakage / breakdown sockets should have a spring loaded switch/button that disconnects the supply and puts a 1 K ohm resistor load.

The 175 μA max is 150 μA on a diode or transistor with 100V breakdown (PIV or VCE open circuit base, or VCB). Generally it's non-destructive. Things like a 1N4148 can be selected on test as 120V reference. With capacitors it's simpler to have a series neon with an almost zero leakage capacitor across it. The slow flash speed roughly measures gigaohms. Rapid or apparently solid means the cap is useless in valve set for "tone correction", isolation, g2 decoupling or anode to grid coupling. There is no one "that capacitor". So there will be at least a 3 way switch between the 1/2 HT (about 185V no load), full HT (about 350V no load) and doubler volts (about 700V no load). I might add another stage to have 1000+ volts as some capacitors need that. I'll see if I have a four way switch and stare at my stock on 1μF 400V met-polys! The switch would have 1 M ohm on output. The 350V would have a 1 M Ohm feeding switch, the 700V having 3x 1 M ohm feeding switch. A 1000V range would need 5x 1M or maybe I might have some real HT resistors. I can test really high value resistors for breakdown by DMM in series with 0V and check the microamps are correct.