Simple power supply overheating components, shorting zener

It may be better to try an inductive line conditioner with a parallel filter cap if the problem is caused by line spikes. I do have some 18 V Transzorbs that might be able to handle the high current spikes. I have a universal switcher with 12 VDC output in the unit, but I need all 8 pins on the octal device for two SPDT relays and the input voltage sense lines. This unit was designed to replace a simpler circuit that was unreliable and inaccurate, in the same can.

When we redesign this test set, we will probably use a separate board, where the power is drawn from the switcher. Actually, a switching supply would also draw huge input current spikes because it has a FWB connected to large electrolytic caps, although it usually has a line filter ahead of that.

You have all given me some good ideas. I will make more measurements and contact the customer later today.

Thanks,

Paul

[snip]

Actually, the capacitors are 2.2 uF, 250 VDC, self-healing metallized film. They are tested to 150% of that voltage, but that is just a safety margin. With a reasonable sine wave 280 VAC input, the capacitors should never see more than about 200 Volts peak. It may be better to use a 1 uF 250 VAC rated polypropylene, but there is limited room in the enclosure.

I checked the waveforms with a scope, at about 280 VAC input. The resistor R1 has a jagged symmetrical waveform with peaks at about 3 volts, and no signs of spikes. The waveform on zener Z1 (and the filter capacitor, which is actually 330 uF), is just about 12 VDC, with a variable ripple about 200 mV above the 12 VDC, and sometimes as much as

800 mV below 12 VDC.

These readings were on a spare identical unit. The one that had failed is inside the test set, and is difficult to get to. We ran the test set at full rated current (1600 amperes) for about a half hour, and nothing seemed to get hot.

We will open up the test set and look at the voltage relay again. Any suggestions on how to test the capacitors to see if they may have been damaged?

Thanks,

Paul

It is not really that simple. For one thing, it is AC voltage, which I am sampling with the PIC and averaging over 200 mSec periods without rectification and filtering to get a faster response. The PIC flashes an external LED through a solid state opto-relay to indicate if the voltage is outside of either range, and another opto-relay enables a contactor which configures a transformer for series or parallel input depending on supply voltage of nominal 120 or 240 VAC. All of this on a board about 1"x2".

Thanks for the alternate suggestions, however.

Paul

There's your problem, no where near good enough.

It may be

It sure would you should include a 250v varistor accross the line as well.

They may have lost some capacity.

He has got 2 in series ! An X-type might be better though.

Graham

I know, thats only 500v, not enough! and he has no charge balancing resistors nor any transient suppressors. A supply destined to fail time and time again.

Right, it is bound to fail, because of 2 violated max. ratings. He didn't consider the inrush current. When you switch on in the wrong moment there is over 20A running through the zener and 100u in parallel. The zener is probably rated for 3 or 4A non repetitive surge. Also this violated the dI/dt of the caps and it has already shorted one cap, because even with a dead short acroo the O/P the current can not be higher than 70mA, but the OP measured 100mA. It is also a safty hazard to leave a 1u/400V charged capacitor open circuit waiting to zap the technician even after days.

--
ciao Ban
Apricale, Italy

Try this simple mod.

R2 +-------------/\\/\\/---------------------> To A/D circuit | 100K 1W | C1 C2 D2 R1, up to 100 ohms? ACHI-+-)|---)|-----+--->|---+--/\\/\\--+----78L05--->

2uF 2uF | 1N4004 | | | Film | | \\_|_ | D1--- | Z1 /_\\ 12V | 1N4004 ^ 100uF=== | | | | | | | | | | ACLO---------------+--------+--------+-------+----- GND

Zener diodes and big C's never sit well in parallel, and a Zener always safer when it has some sort of source-R.

The circuit is good for about a 10mAdc load and will run cooler than the original.

--
Tony Williams.

That's an improvement, but still:

(1) The capacitors are under-rated voltage-wise. They should be at least 600VDC rated each, at the very least, and preferably 120VAC rated.

(2) Removing all resistance before the first 1N4004 is a bad idea. Worst-case, you could be discharging 1uF at 700 volts across the first

1N4004 on power-up. Or with the opposite phase, discharging 700+ volts across the second diode on power-up. Putting a 1 meg resistor across the capacitors will help somewhat, but you'lll still get the jolt if the guy switches the power off and on fast.

WOEFULLY under-rated. Every negative half-cycle

This is utter nonsense !

There's no resistance in the way at all now !

What the heck are talking about ?

Graham

Yes, one step at a time though.

C1 and C2 should be one capacitor with a series flameproof resistor for current limiting.

That voltage doubler seems to look like a 14mAdc cc source.

--
Tony Williams.

Not to beat a dead horse, but to summarize:

(1) The circuit has multiple problems. You have to fix all of them.

400V film caps can't take the voltage or current. The diodes can't take the 1uF 700V switching on spikes. Putting a capacitor directly across a zener is mostly pointless (the diode will keep the capacitor from storing the charge you need for use during the diode off time-- you need a resistor before the zener). The resistor can't be guaranteed to take the 700V voltage or current spikes.

(2) Just because you can't see spikes doesnt mean they're not there, or won't be at the customer's site. Especially in an industrial environment, there can be 1KV spikes every time some inductive load gets switched on or off.

I would patch the design by using a tiny 240V transformer, available at digi key for $12. About a cubic inch.

Or I'd start over with a quad comparator, about 7 resistors, five

1N4148 diodes, one zener and three capacitors should do the trick. Rough description:

A microwatt power supply-- say a 100K 2Watt resistor, a 1N4007 diode,

10uf 50V capacitor, a 100K 1/2 Watt resistor, then a 25uf 15V capacitor with a 12V zener across it.

A mickey-mouse logic voltage selector:

A five resistor voltage divider across the 50V capacitor, totaling about 1 meg. The four taps go to four of the inputs of a 74C14 directly to the odd inputs, thru 1meg resistors for the even inputs. Resistor values chosen so we got 4 VDC at each tap at the magic AC voltages.

The outputs of the even numbered gates go to the cathode end of a diode, anode ends go to the odd numbered inputs.

The odd numbered outputs go to cathode ends of two more diodes, the anodes go together, up to +12 volts thru a 1 Meg resistor and signal "voltage BAD". That point goes to the input of a fifth gate, the output of that gate is a nice strong "voltage OK".

How it works: the voltage divider trips the gates in sequence at 90,

130, 200, and 270 VAC, at the lower schmitt trip voltage (about 4 volts).

The even numbered gates (that trip at the overvoltage for each range) turn off the respective "voltage ok" gate for that range.

The output diodes negative logic OR together the results of low v ok and high v ok.

You can probably lower the power dissipation even more by upping the value of the first 100K resistor until the 12V barely regulates at

80VAC.

Three other comments. For many approval certifications, the circuit as given, would automatically fail. Being 'online', implies the need for extra money to be spent on insulation on a lot of the parts. Going offline, with a small transformer is generally both simpler and safer. It might be worth looking at minature switch mode topologies as well (look at the 'Power integrations' products). PI, also do the 'LinkSwitch-TN', which for a non isolated supply, does this in an IC, with few extra parts.

Best Wishes

This sort of cicruit is pretty common in cheap low-power mains-fed gadgets. The way I've seen it is actually like this:

carbon C1 220 1W D2 ACHI-----||----/\\/\\/-------+--->|---+------78L05--->

470n X2 | 1N4004 | | | | | Z1--- | | 12V/_\\ 100uF=== | | | | | | | ACLO-----------------------+--------+---------+----- GND

The 220 Ohm carbon resistor should hold up pretty well to abuse. The zener doubles as a rectifier and shunt pre-regulator. If the load is absent, it gets to dissipate all the power. You can extend the idea to a full bridge, with two zeners and two ordinary rectifier diodes. That would halve the ripple, but you'd lose the ACLO to GND connection.

A fuse in one of the AC connections is mandatory for fire safety!

Jeroen Belleman

Ban, For some reason they are not buying changing the 1/2 wave to a full wave rectifier. Just takes two more diodes and drops the coupling cap's value in half. This drops the surge currents in half and allows one cap of 0.47uF. The zener should be a transzorb. cheers, Harry

Sometimes you need one side of the line (hopefully the neutral side) running straight through. With a bridge your DC output is flopping up and down with the AC line. Not cool sometimes.

with differential amp he could sense the mains voltage via a divider directly. His Pic can anyway only see half the wave, this way he could also process the negative part. How many times has this kind of supply been discussed here, with good advise regarding galvanically not isolated mains supplies, and here all important safety measures have been left out, delivered to customers?

--
ciao Ban
Apricale, Italy

Thanks to all for the many good observations and comments. I will be getting some 250 to 305 VAC X2 rated capacitors (1 uF). The larger 220R carbon resistor might be a good idea, but it would need to be 3 watts, or else 82R 1W. I need at least 1 uF (which provides about 110 mA at

280 V, so that I can get enough current at the low end (about 75 VAC) to drive two solid state relays at 10 mA each.

The customer called back and said they use this device on a portable generator, and sometimes they have it connected and turned on before they start it up. There are probably lots of noise spikes as a generator comes up to speed.

The circuit is protected by two 5 amp fuses. Possibly an in-line noise filter might reduce the high frequency spikes which are probably causing the failures. Unfortunately we cannot accurately simulate this condition, and I don't think the customer has the necessary power quality analysis equipment.

I agree that this is not an ideal design. We started with a commercially available octal voltage relay, but it had only one setpoint, and did not provide the versatility of the new PIC based design. We needed a plug-in replacement, so we were constrained by the octal relay can. Our new design will use a separate power supply, which should eliminate the weaknesses in this design.

We will need to evaluate the best way to make the existing units safe and reliable. The ideal solution would be a replacement octal can that the customer could replace in the field. This is inside a test set that weighs over 120 lb, so it is difficult and expensive to ship. We sometimes go to the customers' sites for calibration, so we could add external line filtering, larger resistors, or even rewire the unit with a PC board powered from the existing universal 12 VDC supply.

Thanks,

Paul

yeah how ever, if the zener was a voltage ref type with a temperature compensation diode in it, that wouldn't work! :)

--
Real Programmers Do things like this.
http://webpages.charter.net/jamie_5