Late 1970s SMPS Design Query

On a sunny day (Tue, 26 Jan 2016 18:43:20 -0000 (UTC)) it happened Cursitor Doom wrote in :

What connection? The transforer wire? Then you have no resonance and a capacitive short of 30 nF / 2 = 15 nF as load.

15 nF at 20 kHz = 1/jwC = 1/(20000*pi*15E-9) = 1061.03 Ohm, say 1 k At chopping 230 * sqrt(2) V DC = 230*sqrt(2) = 325.269 V into 1k power is (325.269 *325.269) / 1000 = 105.8 W.

Of course it is not on all the time, so derate by duty cycle, but that pour L1804 is now the flyback coil, and via that diode and R1814 that flyback pulse is dissipated away.

Or am I dreaming? (Its late here...)

Yea, I dunno at this point. But it clearly is swicthing, else you did not have any current.

On a sunny day (Tue, 26 Jan 2016 19:10:44 GMT) it happened Jan Panteltje wrote in :

I was dreaming, no L no current.

Because it is trying to supply that driver, as the transformer does not do it via R1838.

On a sunny day (Tue, 26 Jan 2016 19:09:36 -0000 (UTC)) it happened Cursitor Doom wrote in :

OK, curious. And what is on the other side (100nF) of that driver transformer? Maybe the driver transistor had its best days?

You may well be right. It was rather rash of me; I just got impatient. I'll look into it.

The driver transistor V1812 seems ok at the simplest level. DC checks reveal its junctions acting as expected. Although socketed, its leads are a little too short to permit full testing for hfe and whatnot. Maybe the problem is with that 100nF cap. It's very difficult for you and others who can't see what's going on for themselves, but like I say, I'll post some pics of the waveforms tomorrow.

** Capacitor ESR meters are designed to never forward bias a junction - diode, B-E or C-E.

If you see a low reading across a diode in a SMPS, that does not have an electro in parallel, it is bad.

.... Phil

On a sunny day (Tue, 26 Jan 2016 20:41:11 -0000 (UTC)) it happened Cursitor Doom wrote in :

OK, here is my (final) take. The likely fault 'train' in this Philips design is:

1) The tuning caps go high. 2) That causes the output pulse to rise, 3) The HV multiplier breaks down because of that. 4) The overload protection comes in, the hissing noise.

The solution:

1) solder out the tuning caps. 2) measure both separately. 3) if defective replace by suitable high voltage high current types. 4) remove HV multiplier. 5) test if supply comes on, measure secondary 12V pulse from transformer for the right amplitude. or measure the 12V or 5 V DC.

If the fault condition is gone it the HV multiplier too.

You can easily measure the value of the tuning caps even if you do not have a capacitance meter, like this: make a 30 nF capacitor (say 3 x10 nF in parallel, or some other combination. Put it in series with one tuning cap (on the bench). Apply 20 kHz at a few volt from your signal generator (sine wave). Scope and test if the voltage on the removed cap is HALF of that on the signal generator. (Capacitor divider).

1/2 V | generator ---||----------||----- GND 20 kHz good 30 nF tuning cap

If not, then the tuning cap is defective. Check both caps.

If it is not that then it is something else. :-)

I say, from what you did so far, both HVcaps and HV multiplier are suspect. The fact that you get a narrow pulse without HV multiplier, points to low load and an OK transformer. Measure the 5 V and 12 V output without HV multiplier to confirm that, goto 'The Solution'

OK?

What is this "HV multiplier" to which you refer, Jan? If you mean the Villard cascade, it's been disconnected from the very beginning so cannot be responsible AFAICS?

Thanks for that clarification.

On a sunny day (Wed, 27 Jan 2016 09:56:36 -0000 (UTC)) it happened Cursitor Doom wrote in :

yes, and do you have the 5 V? Did you measure those caps? if not goto

I think you meant current, not DC.

But I can seriously see this happening. The driver(s) might run a bit hotter, and maybe even the output (chopper). If it is not in Ecuador in July it might be alright.

I'll measure them. And the Vouts. I just want to get some scope screen shots first before I remove them for measurement. I'll report back later....

Right, serious difficulty here getting valid readings, not aided by the fact that the meter hardly settles before the scorching smell arises and I have to let it cool off again. The voltage readings are nonsense. I have 13.8V stable on one set of outputs and 26V on another. The end one is well over 100V and there's no

6V at all. The caps I had to measure in-circuit as it seems impossible to remove them without damaging or destroying them and they came out at 62nF each which seems odd given that they're both precisely doubled compared to their stated value. Don't want to spend much more time on this as I think it's above my competence level at this stage so I'll probably mothball it until I have more experience with switchers. I'd be interested to see what you think about the measurements.

On a sunny day (Wed, 27 Jan 2016 13:00:44 -0000 (UTC)) it happened Cursitor Doom wrote in :

Is that not a hint? The + 6 and -6 come from the same transformer windings, with diodes in opposite directions. I'd think time to check for a short on one of the 6 V lines (or both), that go to the other boards, disconnect those?

No they will be parallel in a way when you leave them in, was it no Thevenin?

Not a very reliable way to measure those, too much otehr stiff i ncircuit, I would take them out anyways, but OK.

What measurements? Did you publish the scope (oops I wrote soap) pictures?

The board is not connected to anything, Jan. I'm working on it by itself just as you see it in the pictures. I've already checked for shorts/low- res looking back into these outputs and found nothing suspicious. The diodes in question both look fine and test fine.

Ideally of course, yes. But unless those clear covers pry off easily (they don't) I can't get 'em out without damaging or destroying them. Anyway, from what you say, they appear to be fine.

Next job! Watch this space....

Unloaded and unbled, the outputs can exhibit double their labelled voltages. In this case they may show the formed voltages of the output elecrolytics, which leak individually at differing voltages.

The unit has 5 low voltage outputs with a common return. You must therefore measure 5 voltages associated with 5 terminals, identified by their schematic-labelled voltges. The common meter lead does not shift between measurements.

They both mount with simple soldered leads. Desolder the leads to remove. They do not have removable plastic covers - this is simply the clear polycarbonate epoxy body of the part.

In-circuit, a low frequency tester will see them as being in parallel, due to DC connection through the transformer winding. To measure individually, you only have to remove one of them. This breaks the connection between the two

If one has discoloration of schoopage, this is the prefered part for removal.

So far, you've reported nothing that indicates that the unit is malfunctioning. All 'abnormal' behaviour can be attributed to lack of load.

RL

OK, here we have three photos I managed to take before R1814 got too hot each time. They're self-explanatory. You can see the way the signal gets progressively worse as it gets closer to the high voltage section. The slight 'turbulence' just above zero volts on the driver b/e picture is probably just the downstream mess coupling its way back towards the pwm chip.

So there's the signal at the driver's base-emitter junction, The signal between the driver's collector & emitter (skinny!) The signal at the chopper's base-emitter junction.

All probed X10 of course. Enjoy. :)

public/

public/

public/

OK - 6 outputs. It's a fine detailed schema on a small screen here.

RL

Thanks. That discolouration was just a shadow, at any rate it's certainly not present on the actual component. The soldered joints of those caps are not accessible from above or below. I'm pretty certain that 20 ohm resistor still rapidly overheats even when the load's applied; I'll double check and confirm.