Old Honda generator

[BurningTantalum]

I have owned a 2.5kVA Honda GX160 generator for about 35 years. It was in poor condition when I got it having spent its life on building sites. The other day we had a power outage for 24 hours (not uncommon here...) so I fired it up. I ran the fridge back to cut-off, then boiled the kettle, then ran the toaster, and it stopped generating. It was running into a strip board with 10A breaker (standard Aus)
I did a few tests and had a visual of the windings, and then stripped the stator and the rotor out. Both stator windings measure OK - switched in series for 240V, or in parallel for 110V (UK trade sites.)
The rotor has no slip rings and thus no connections at all, but has a winding on each of 2 laminated pole pieces. One end commoned together, and the other ends connected across 3 diodes in series, each with a 20k resistor across on, a small paxolin pcb. All diodes face the same way and check out OK.
The pole pieces have small magnets set into them, which seem to be reasonably strong.
I assume that the magnets are to provide a remanent magnetism at start-up. I am not sure about what the diodes are doing.
I have only looked before at slip ring type gennies with an AVR. I have read various posts about 'flashing' to remagnetise, and not starting the genny under load etc. but it was running happily when it failed. Any ideas, please? I'm stumped.
BT

[Ian.M]

Explain "... measure OK"  What did you measure and how did your measure it?

As there are two identical stator windings, it would be worth measuring their inductance, which if significantly different, would suggest one has developed a shorted turn, which can be very difficult to detect by DC resistance measurement.

If you can isolate the windings on the individual rotor poles, inductance measurement would also be helpful there, but if they are wound in series and you cant get at the center tap, you are S.O.L. unless you've got access to a Growler + some experience using it.

Check any capacitors out of circuit for value, and for shorts or excessive leakage at as close as you can get to their rated working voltage.

[BrokenYugo]

Pics might be helpful, sounds kinda like a capacitor excited type but not quite. Those have no brushes, no AVR, and a third winding in the stator with a pretty big cap across it.

[floobydust]

GX160 is the general purpose engine model number, need the generator model number to find wiring diagrams.
The rotor diodes might have a MOV across them, I've never seen a resistor used.

I did mention some of the brushless AVR generator theory:
https://www.eevblog.com/forum/chat/noisy-power-from-typical-backup-generators/msg3476406/#msg3476406

[tpowell1830]

There are several things to look for on the stator windings, one being dark patches in the wires, which indicate a hot spot due to shorting. If the stator windings have a short, this could be a problem. As far as the rotor, what you are describing is a n exciter circuit used to magnetize the rotor windings, possibly using the magnets, however it needs DC, thus the diodes.

This youtuber only repairs small gensets and has some good insights, however, most of the gensets are modern with AVRs, etc. Check his videos out, it may be helpful.

https://www.youtube.com/user/jcondon2007/videos

[BurningTantalum]

OK, thanks for all the info. I had reassembled the genny, but will dismantle it again and do some inductance tests (which I should have done in the first place...) and also take some pictures. I forgot to mention that it does have a capacitor wired into the stator windings. It is 13uF and tested OK at 11uF. I clipped on a 16uF but it made no difference.
 - tpowell1830 The windings look perfect with no loose windings or discoloured lacquer.
I do have a 'growler' but as there is no commutator I didn't think of using it. I suppose I could energise one laminated pole piece and measure the voltage in the relevant coil.
Please bear with me and I will report back.
BT

[bdunham7]

Do the rotor winding and the stator winding with the capacitor both have continuity?

[BrokenYugo]

Are you certain the trouble is in the alternator itself and not some failure in the wiring harness, outlet, circuit breaker, etc? If the cap and all the windings are there I would think it should be doing something.

[BurningTantalum]

Hot out of the workshop:
The stator windings measure 1R26 and 1R23.
The rotor windings measure 11R57 and 11R54.
The capacitor winding measures 3R35.
All are insulated from the frame at 250V.
The cap winding is NOT connected to any other winding (and I see from Floobydust's link that it should be thus.)
The stator inter-winding capacitance is 7nF.
My LCR meter will not measure the inductance.
I cannot find my Growler anywhere, but will search again.
The faces of the two embedded magnets are about 20mm x 10mm and will hold the weight of a pair of standard pliers easily.
The rotor laminations are not magnetised, but passing 0.5A through each winding in turn, in the correct magnetic polarity to match each embedded magnet, the laminations became strongly magnetic.
The diodes check OK at 530mV forward. The resistors check out as 20k.
The layout resembles the schematic in Floobydust's EEVblog link, and apart from being unable to check that the coil inductances are identical in both pairs, I cannot see why the generator will not produce at least some AC voltage, even if the windings got hot in the process due to an internal fault.

[BurningTantalum]

BrokenYugo: I have disconnected the stator leads at the 24V/110V switch, and there is nothing else connected.
I tested the output with a meter, or a lamp load clipped directly to the leads.

[james_s]

There really isn't much to go wrong with those things, the only failure I've ever seen was the capacitor but you seem to have eliminated that. With no load you should get some voltage even if just from the magnets used to bootstrap it. You could try the trick of exciting it with a battery connected to the output (engine NOT running) to charge it up with some residual magnetism. Shouldn't be necessary with the permanent magnets but worth a go.

[BurningTantalum]

I forgot to state: At full governed speed there is 3.4VAC across the 240V leads (1.7VAC across each 110V winding.)
This falls steadily to 1.3VAC as the motor drops to idle. I will attempt to get an inductance figure tomorrow, after perusing my old college text books tonight !

[bdunham7]

You might try just replacing the capacitor on the off chance that it has an issue like a low-voltage breakdown that doesn't show up with whatever you used to test it.  Other than that I'm stumped.

[BrokenYugo]

See track side picture of diode board, upper connecting pin, track appears to be gone for a couple mm where it meets that pad?

[floobydust]

The rotor has three diodes in series, each with 20k equalizing resistors. That unterminated pcb trace is to the first resistor, the diode is still soldered to the board's leads but would take all the reverse voltage instead of the string. Oops.
I don't always trust readings on individual components, the whole thing should read 60k ohms in reverse, and 1.5-3V forward and I don't rely on a multimeter to test it, I'd put current even a few hundred mA through it with a car light bulb for example. I think the Growler would hurt the diodes if in circuit?

I would suspect an open in the rotor circuit, it's not building up flux. But the capacitor and winding voltage there would also be work measuring.

[BurningTantalum]

Thanks all, and especially Flooby for persisting with this. I tested the pcb  for the 60k and the 1.6V but don't recall if it was from the pins or across the diode leads, doh! I have reassembled the whole caboodle to measure the generated voltage. I suspect that the 'missing track' is a discolouration of the soldermask, plus there is a 'berm' of opaque epoxy that held the board in the 'slides' of the nylon carrier. I may be able to get at the pin and the resistor leg in situ but will pull it out again, and I can test the inductance of the windings at the same time.
BT

[BurningTantalum]

I managed to prod the lead of the resistor and the connection pin in situ (using a probe with long bare prod off the house '830' meter...) and it was OK.
I dismantled the genny again and removed the board. The trace was OK - it was an optical illusion. I tested the diodes again and all measured 530mV, but...
 THE STAR PRIZE GOES TO... drum roll... FLOOBYDUST !!! (To be confirmed tomorrow)
Flooby quote: "the whole thing should read 60k ohms in reverse, and 1.5-3V forward and I don't rely on a multimeter to test it, I'd put current even a few hundred mA through it with a car light bulb for example."
It did read 60k across it, and 1.5V but...
BT

[BurningTantalum]

... I connected a current limited supply across each diode in turn and one did not pass current. I checked again with the Fluke diode test and it was fine; repeated several times due to disbelief.
Out came the microscope and I could see a slight fracture where the lead joined the diode. If I injected the current right against the glass body with a sharp prod it was fine. As I unsoldered the diode the gap opened and the lead separated. The fracture face on the diode was a slight concave dish and was a matt grey colour; the leg end was a convex match. I have never seen this before, and don't understand how it could test as a PN junction (passing small current) but not pass a 200mA current. I was looking at the analogue meter on the PSU so didn't spot that it may have been passing a few mA.
The components were covered in epoxy so I had to destroy them to remove the from the board. I cleaned it up on a belt sander but decided to make a new one from glass fibre perf board. I had some 20k resistors and 1N5403 diodes. Not sure of the voltage spec of the originals but these are 300V 3A.
I fitted Vero pins instead of using the diode leads as leadouts
Thanks everyone, the generator lives for another stormy day !!
BT

[james_s]

That's a tricky one, I'd guess maybe a thermal crack?

I suspect what happened is the broken lead was touching at a very tiny point, resulting in a resistance that was not enough to upset the diode tester but was enough to prevent the diode from passing enough current to work in the circuit. Either way it's pretty obvious that you've found the problem, with that fixed you'll probably get another 30 years out of the thing provided you keep oil in it and change it once in a while.

[bdunham7]

... I connected a current limited supply across each diode in turn and one did not pass current. I checked again with the Fluke diode test and it was fine; repeated several times due to disbelief.
Out came the microscope and I could see a slight fracture where the lead joined the diode. If I injected the current right against the glass body with a sharp prod it was fine. As I unsoldered the diode the gap opened and the lead separated. The fracture face on the diode was a slight concave dish and was a matt grey colour; the leg end was a convex match. I have never seen this before, and don't understand how it could test as a PN junction (passing small current) but not pass a 200mA current. I was looking at the analogue meter on the PSU so didn't spot that it may have been passing a few mA.
The components were covered in epoxy so I had to destroy them to remove the from the board. I cleaned it up on a belt sander but decided to make a new one from glass fibre perf board. I had some 20k resistors and 1N5403 diodes. Not sure of the voltage spec of the originals but these are 300V 3A.
I fitted Vero pins instead of using the diode leads as leadouts
Thanks everyone, the generator lives for another stormy day !!
BT

I've seen those type of diodes fail completely at that spot, but never, ever had one test good and be bad!  Good work.

[floobydust]

Sinterglass diodes are really tough. I would guess it was fractured possibly from damage during lead-bending, assembly and wave solder. Any temperature shock is not good for them.
Neat electronic component failure analysis with SEM: https://www.semlab.com/glass-rectifier-diode/
You might have just had a fractured die and thus a micro-diode.
I find multimeters have such low test currents that are inadequate in a world of moisture, corrosion, oxidation, soot or damaged semiconductors etc. so sometimes it's advisable to push a component beyond a few mA.

[BrokenYugo]

My guess would be vibration, there isn't anything smooth about a 2 pole genset. 

[Zero999]

I've had a similar thing happen before, in an instrumentaion amplifier circuit, which had a much higher gain, than it should, causing a huge offset and the output to saturate, with the tiniest signal. The fault was was a 64R65 precision resistor, which measured the correct value, when removed from the board, but dropped to 5R, when soldered in.  After wasting a lot of time, I noticed the value could be change from 64R65 to 5R manipulating one of the leads. I still don't understand how this could have happened. My guess is the connecting was short-circuiting when the lead was moved, but it seems odd, considering it was in an axial, plastic, through hole package.

[PaulAm]

Yeah, I had a precision ceramic resistor network go intermittent in a Tek scope once with a fracture in the lead right at the body.  Took a week to track it down and it drove me crazy.

Always instructive to hear new stories of how things fail; it might help in the future.