Simple HT supply build

[Mike Watterson]

Don't hold a charging gadget or anything else using a "floating" SMPSU with wet hands or bare feet. The isolation on all tested ones with no earth pin is rubbish. All the laptop PSUs I tested with 3 pins (an earth) connect 0V on output to the earth pin.

Can't beat a split bobbin (plastic former) regular mains transformer for isolation. The transformers in most (all?) USB chargers and other small gadget SMPSUs just rely on one or two layers of tape.

So any isolated floating PSU needs to be using a suitable mains transformer or be really certified 2 kV or 4 kV isolation. I've blown up two PSUs now, a 2.5A 12V (off a Modem/router) and 500 mA 5V USB charger. Both had only two pin live and neutral mains connections. No problem at all with the split bobbin mains transformers, even ones from bargain basement Argos and Tesco mains LED clock radios (a very cheap source of parts).

Testing the shunt regulator FETs soon. They are 2SK560 FETs
VDS = 500V
ID = 15A (continuous, 60 A peak)
Pmax = 100 W
VGSt = About 3.0V.
About 6V insures fully on at about 0.3 Ohms.

Worse case gate capacitance about 3nF, which is low for a 500V 15A part. I'm redesigning a PWM that the designer used an IRF1404Z (about 8nF!) even though max current is about 3A. I'm changing it to a an IRF740 (400V @ 10A) with 0.55 Ohms static on resistance and about 1.6 nF gate capacitance. Changing drive from 5V to 12V too. The 5V drive is too low for PWM.

[ppppenguin]

Are Class 2 wall warts and "lump in cord" PSUs really that bad? I know the CE mark is widely abused but double insulated equipment should really be able to pass the high voltage flash test.

I've just looked at the 5V switchmode PSU supplied with a Samsung phone. It's a genuine Samsung part. It has the double square mark for Class 2. I don't plan to test it to destruction with a high voltage.

[Mike Watterson]

The ones I tested so far are that bad.

You don't need to destroy. A leakage tester may work, or 300V via 2 M Ohm. May need to be running on mains to fail. I'll check some more with unpowered leakage test and then powered.

See also RFI via offical tests on LED built in PSUs and electronic Ballasts on CFLs. The test setup is unrealistic compared to separate L & N wires daisy-chained and switch wiring.

Then there is networking over mains. They test either one unit (thus no communication), or without ethernet. They don't have SMPSU interference as that is well filtered so that they work. They can knock-out DSL and faster ones have RF to over Band II (into Aircraft band).

Then also there is no in market testing. So I've seen PC PSUs with wire links for inductors and capacitors missing, so as to save costs.


[You mean Samsung bought it with Samsung branding or did Samsung really make it?]

[Mike Watterson]

I've assembled the miniature 11.5V transformer (double bobbin with obvious plastic, and thermal fuse in primary) to a simple smoothing circuit. Bridge, 220 μF 35V cap, 22 Ohms series and finally bigger 20V cap (1000 μF). No load is just over 16V. I put an 470 Ohms and LED as a load and it's below 15V (maybe 13.9V or something, I forget).

I don't think a regulator is needed.

[ppppenguin]

[You mean Samsung bought it with Samsung branding or did Samsung really make it?]

No means of knowing but it was sold with a new Samsung phone so it's not a fake. Samsung, Apple and other major brands are staking their reputations on their kit being safe. They don't want headlines like "Apple phone kills child". One hopes that they've done the necessary work though you can never be entirely sure.

Ruth has a modern Macbook. The PSU is Class 2. You can feel a slight tingle if you brush your fingers lightly over exposed metal parts of the computer. I reckon to get this with just about all floating Class 2 kit.

[Mike Watterson]

You mean class II, not class 2.

[Mike Watterson]

The two PSUs were in different applications and the issue was failure, not leakage to humans.

1) The 5V 500mA USB charger had 0V connected to slider of a 250K pot across -96V and the 5V powering an LED voltmeter and Vsense connected to 0V. It worked and then the 5V charger (out of case) started smoking. It was rated 110V to 240V so I'd powered it off +150V DC. Input is a bridge rectifier. The tiny meter is still OK. The main +150V and -96 from a transformer with 4KV (minimum) split bobbin. No earth anywhere, no mains test gear connected.

2) The +12V 2.5A PSU (originally two core mains flex and no earth) was powering the transistor side of an opto-isolator driving a buffer. Later I was going to add a FET switch. The main PSU is 19.1V laptop PSU that has the 0V connected to mains Earth. A 3 core clover plug. Mains Earth (as with most of Ireland) is bonded to the neutral and a 1.2m (approx) earth spike near the meter box. You never ever get a "tingle" off earthed equipment here. Some 2 core gear does, but only if it has RFI cap(s) from mains to the output or case. Quite rare, but seen on one Sky box.. The +12V may have been connected to laptop PSU +19V, as that is the cathode and heater + on the tester. The heater - normally connects to an IRF1404Z connected to 0V (which is also earth on all the higher power laptop PSUs). Anyway, something in the +12V SMPSU failed (smoke!) and the main PSU was cutting out. The +12V SMPSU was replaced by a tiny 8V linear PSU (split bobbing transformer, just a bridge and 1000 μF cap) and everything was fine.

I took case off the +12V 2.5A SMPSU and wondered why the PCB terminals were labelled L and N rather than both ~. Also the track space between the mains wires doesn't meet UL spec for 110V to 120V mains, though still enough in reality for 240V as long as dry!

I don't know why either PSU died, but the transformer in the USB charger is the size of a sugar lump. The 12V 2.5A SMPSU mysteriously has two small transformers and seems very complicated.

I will investigate more and do leakage tests first. It may be a subtle problem. The 5V USB charger had never been used because it was only 500mA and was "europlug" styled. It came with some gadget. Maybe it was going to die anyway!

The +12V 2.5A SMPSU had a two core mains flex, but not double PVC. That's legal in USA, but not here or UK now (though it used to be common). Perhaps I had it connected wrong? Putting 19V across the output would not make it happy, though I often design PSUs so they don't die if +30V is across them. Only shunt regulated PSUs can sink external current.

The commonest cause of fires in the home now apart from cooking and smoking is SMPSUs used to charge phones, ecigs, scooters, bikes and power stuff like decorative LED lights or gadgets.

It's a mystery. However with the linear PSUs you can see the transformer construction and insulation. A SMPSU can be a box of mysteries. I think I first designed one over 50 years ago and unless a cheap easy to replace part had obviously exploded, I had no joy repairing them. Now they are scrapped, though I might remove any easily removable and testable part!

[Mike Watterson]

Found a case. Using a scrap PC 450W SMPSU case. The two FET heatsinks will sit nicely at the fan with a plain matrix board vertically on the FETs with the high voltage caps etc on it.
A plastic panel in front with the controls and meters.

Not using the PII 333 MHz heatsinks as they are about x2 the size needed. Found a nice pair. I've no sockets for the 230V filament lamps so I'll cut hole in a piece of plain matrix board.

Also found three boxes of old PSUs (from small linear to laptop style "brick" SMPSUs in the attic.

[Mike Watterson]

Seeing if bits will safely fit?

Bottom and top of live heatsinks will have thick plastic insulation on case.

Controls, displays and sockets on a panel in front of the perforations, spaced well out.
Lamps in a row just behind front perfs.

A 230V AC oven lamp at 122V DC
   
Really quite orange, but using Viber on phone via WiF in Shack to have the images direct on desktop, then drag & drop. Very lazy. Hence mad file names.

I'm hoping the 12V fan goes enough on 8V, the raw volts on the little PSU. It will power 0 to 8V gate drives and the meters.
A real stripboard will have the quad Norton op-amp to measure the two supply currents at "hot" side of HT and convert the variable 0 to -95V to positive volts for a separate meter. I might use +12V via regulator on doubler for IC supply, or maybe no regulator.

[Mike Watterson]

And a simple, cheap, quick to make socket for the ersatz barretters, aka filament lamps.

A cheap fibreglass pcb (2 cm x 8 cm) with just PTH pads on each side (bought to mount rod pentodes) and shiny steel jumbo paper-clips that mysteriously take solder.

It will take 3 lamps.

Mount with standoffs to perforated front of PSU case.