Hi again,

I know I said I was going to mothball this scope for the time being, but
I keep thinking of new things to try and just can't leave the damn thing
alone. It struck me I ought to next check out the main transformer
because if that's toast, the whole psu might as well be binned. I think I
*may* be on to something.

Here's the circuit again:


https://www.flickr.com/photos/128859...in/dateposted-
public/

The transformer in question is T1801. I've pumped in a test signal into
its primary winding (2V p-p sine wave @20kHz Zs=600 ohms) and scoped all
15 remaining pins. Every single one of them is producing clean sine waves
at various amplitudes just as you would expect, except one, which is
giving a really thick, noisy trace. On the diagram, this is the 5th one
down on the right hand side (immediately below the 1kV out pin.) It's not
marked on this diagram, but is supposed to be 0V from another schematic I
have. Someone has replaced C1826 & C1827 at some time and the
replacements don't look good for more than 50V max to me, and that's not
enough since the pd across them will be much more (the parts list
specifies a 500V rating!) They're not showing short-circuit, so I'm
wondering if they've failed open maybe that could account for the
symptoms I've been getting: terrible noise in the primary circuit and the
rapid heating of R1814 (the 20 ohm power resistor intended to dissipate
the switching energy).
Any thoughts?

"Cursitor Doom" wrote in message
...
Hi again,

I know I said I was going to mothball this scope for the time being, but
I keep thinking of new things to try and just can't leave the damn thing
alone. It struck me I ought to next check out the main transformer
because if that's toast, the whole psu might as well be binned. I think I
*may* be on to something.

Here's the circuit again:


https://www.flickr.com/photos/128859...in/dateposted-
public/

The transformer in question is T1801. I've pumped in a test signal into
its primary winding (2V p-p sine wave @20kHz Zs=600 ohms) and scoped all
15 remaining pins. Every single one of them is producing clean sine waves
at various amplitudes just as you would expect, except one, which is
giving a really thick, noisy trace. On the diagram, this is the 5th one
down on the right hand side (immediately below the 1kV out pin.) It's not
marked on this diagram, but is supposed to be 0V from another schematic I
have. Someone has replaced C1826 & C1827 at some time and the
replacements don't look good for more than 50V max to me, and that's not
enough since the pd across them will be much more (the parts list
specifies a 500V rating!) They're not showing short-circuit, so I'm
wondering if they've failed open maybe that could account for the
symptoms I've been getting: terrible noise in the primary circuit and the
rapid heating of R1814 (the 20 ohm power resistor intended to dissipate
the switching energy).
Any thoughts?



Leaning towards the HV multiplier...

Mark Z.

On Tue, 16 Feb 2016 19:13:09 -0600, "Mark Zacharias"
wrote:

"Cursitor Doom" wrote in message
...
Hi again,

I know I said I was going to mothball this scope for the time being, but
I keep thinking of new things to try and just can't leave the damn thing
alone. It struck me I ought to next check out the main transformer
because if that's toast, the whole psu might as well be binned. I think I
*may* be on to something.

Here's the circuit again:


https://www.flickr.com/photos/128859...in/dateposted-
public/

The transformer in question is T1801. I've pumped in a test signal into
its primary winding (2V p-p sine wave @20kHz Zs=600 ohms) and scoped all
15 remaining pins. Every single one of them is producing clean sine waves
at various amplitudes just as you would expect, except one, which is
giving a really thick, noisy trace.

As this terminal has no ground reference, with the loads removed, you
can only measure its winding voltage by measuring between other pins
that are connected to the winding, or by introducing the ground
connection on one of them.

The amplitudes measured will not be various - they will be scaled as
you expect them to be, as reflected by schematic output voltages or
labels. Measuring with a high impedance scope, it is possible to see
a waveform that is simply capacitive transfer, if you're not making
rational measurement observations.

On the diagram, this is the 5th one
down on the right hand side (immediately below the 1kV out pin.) It's not
marked on this diagram, but is supposed to be 0V from another schematic I
have. Someone has replaced C1826 & C1827 at some time and the
replacements don't look good for more than 50V max to me, and that's not
enough since the pd across them will be much more (the parts list
specifies a 500V rating!) They're not showing short-circuit, so I'm
wondering if they've failed open maybe that could account for the
symptoms I've been getting: terrible noise in the primary circuit and the
rapid heating of R1814 (the 20 ohm power resistor intended to dissipate
the switching energy).
Any thoughts?

C1827 is 'Select On Test' (SOT) to trim some operating parameter in
the functioning unit - possibly the HT tolerance under load or even
just display noise. You'd need a manual to figure this out. The more
often you mention the actual model number (PM3264 ?), the more likely
it is that you'll get someone to cough one up. It's $20 from jetecnet.

Ceramic disc caps of this size were typically 500VDC rated at that
time. Only smaller square plate types were less (and some offshore
stuff). Lower voltages will generally be marked as such, or colour
coded for identification in the mfrs spec sheet.

You still haven't identified the diode substitutions made in the
primary, or reported on secondary winding amplitudes loaded/unloaded.

You still need to report IC pin3 voltage when loaded. (it is reported
as normal, when unloaded, indicating a regulated condition exists.)

RL

On Tue, 16 Feb 2016 19:13:09 -0600, Mark Zacharias wrote:

Leaning towards the HV multiplier...

Nope! Disconnected in this instance.

On Wed, 17 Feb 2016 01:09:27 -0500, legg wrote:

C1827 is 'Select On Test' (SOT) to trim some operating parameter in the
functioning unit - possibly the HT tolerance under load or even just
display noise. You'd need a manual to figure this out. The more often
you mention the actual model number (PM3264 ?), the more likely it is
that you'll get someone to cough one up. It's $20 from jetecnet.

Ceramic disc caps of this size were typically 500VDC rated at that time.
Only smaller square plate types were less (and some offshore stuff).
Lower voltages will generally be marked as such, or colour coded for
identification in the mfrs spec sheet.

You still haven't identified the diode substitutions made in the
primary, or reported on secondary winding amplitudes loaded/unloaded.

You still need to report IC pin3 voltage when loaded. (it is reported as
normal, when unloaded, indicating a regulated condition exists.)

RL

Thank you for your observations.
I can only reiterate that testing under mains power is not practically
possible, loaded or unloaded, given that I only have about 15s before
that power resistor starts to burn up in either case. I really have to
try to crack this by other means.
There is no way on earth the subbed caps are anywhere remotely close to
500V rating; they'd be lucky to withstand a tenth of that. I'm going to
pull them out later today and replace them with a properly rated part
then I'll scope for noise again. I suspect it's this noise, fed back,
that's causing the primary circuit to fire falsely regardless of the
signal from the pwm chip; the excess switching energy therefrom showing
up as excessive heat in the power resistor which is only rated to
dissipate the *expected* switching energy from a properly functioning
primary circuit. Anyway, we'll see....

On Wed, 17 Feb 2016 10:05:38 -0000 (UTC), Cursitor Doom
wrote:

On Wed, 17 Feb 2016 01:09:27 -0500, legg wrote:

C1827 is 'Select On Test' (SOT) to trim some operating parameter in the
functioning unit - possibly the HT tolerance under load or even just
display noise. You'd need a manual to figure this out. The more often
you mention the actual model number (PM3264 ?), the more likely it is
that you'll get someone to cough one up. It's $20 from jetecnet.

Ceramic disc caps of this size were typically 500VDC rated at that time.
Only smaller square plate types were less (and some offshore stuff).
Lower voltages will generally be marked as such, or colour coded for
identification in the mfrs spec sheet.

You still haven't identified the diode substitutions made in the
primary, or reported on secondary winding amplitudes loaded/unloaded.

You still need to report IC pin3 voltage when loaded. (it is reported as
normal, when unloaded, indicating a regulated condition exists.)

RL

Thank you for your observations.
I can only reiterate that testing under mains power is not practically
possible, loaded or unloaded, given that I only have about 15s before
that power resistor starts to burn up in either case. I really have to
try to crack this by other means.
There is no way on earth the subbed caps are anywhere remotely close to
500V rating; they'd be lucky to withstand a tenth of that. I'm going to
pull them out later today and replace them with a properly rated part
then I'll scope for noise again. I suspect it's this noise, fed back,
that's causing the primary circuit to fire falsely regardless of the
signal from the pwm chip; the excess switching energy therefrom showing
up as excessive heat in the power resistor which is only rated to
dissipate the *expected* switching energy from a properly functioning
primary circuit. Anyway, we'll see....

Your concern for this resistor's health is touching, but it really is
misplaced, as the part is easily replaced, relocated, temporarilly
beefed up or whatever.

If you know what you want to do, in 15seconds, you can do a lot,
particularly if you're prepared to repeat the process at intervals,
until all the necessary info is accurately recorded.

RL

On Tue, 16 Feb 2016 18:27:50 -0000 (UTC), Cursitor Doom
wrote:

Hi again,

I know I said I was going to mothball this scope for the time being, but
I keep thinking of new things to try and just can't leave the damn thing
alone. It struck me I ought to next check out the main transformer
because if that's toast, the whole psu might as well be binned. I think I
*may* be on to something.

Here's the circuit again:


https://www.flickr.com/photos/128859...in/dateposted-
public/


By the way......I very strongly doubt that this assembly was issued
from the factory with transistor sockets. I would be suspicious of any
substitutions made here, as they are ripe for error, even if only for
part lead identification.

The low voltage square-plate ceramics (63V to 150V) referred to
elsewhere are visibly in use on this unit, in the photos supplied. The
location of the two suspect parts is not visible/identifiable.

If you have a bill of materials, listing a 500VDC requirement, perhaps
you might distribute this as well? Wondering where it came from, if
not from a service manual? It's possible that tempcos may be
specified, if only indirectly, in the part numbers listed.

Philips didn't pick part numbers out of a hat, or junk box.

RL

On Wed, 17 Feb 2016 08:48:08 -0500, legg wrote:

Your concern for this resistor's health is touching, but it really is
misplaced, as the part is easily replaced, relocated, temporarilly
beefed up or whatever.

Nice idea, but it's in series with the main transformer primary winding,
and I'd rather that resistor go up in smoke than the transformer. If I
replace it with a higher power one of the same value, it puts the primary
winding and other components at risk of fatal damage.

If you know what you want to do, in 15seconds, you can do a lot,
particularly if you're prepared to repeat the process at intervals,
until all the necessary info is accurately recorded.

It takes about 5 minutes to fully cool down and then give me another 15s
of probing. I could of course let it cool for just 90s, but then I'd
hardly have time to get a probe in place before it got too hot again. I
simply haven't the patience to work that way - and the fumes off that
resistor are not pleasant.

On Wed, 17 Feb 2016 09:16:07 -0500, legg wrote:

By the way......I very strongly doubt that this assembly was issued from
the factory with transistor sockets. I would be suspicious of any
substitutions made here, as they are ripe for error, even if only for
part lead identification.

Maybe. But the main signal boards are all socketed, so the fact that the
psu board is too is not suspicious.

The low voltage square-plate ceramics (63V to 150V) referred to
elsewhere are visibly in use on this unit, in the photos supplied. The
location of the two suspect parts is not visible/identifiable.

Yes, they're tucked away tightly next to the side of the transformer with
the top prongs sticking out of it. You can only see one, but someone has
used two in parallel to make up the required 560pF I would guess, but I
can't see without pulling them - which I will of course. BTW, there's
another variant of this schematic which uses just one 560pF 500V ceramic
instead of the fixed and variable combo shown on the schematic I've
previously posted, so perhaps Philips themselves had problems with this
part of the design before!

If you have a bill of materials, listing a 500VDC requirement, perhaps
you might distribute this as well? Wondering where it came from, if not
from a service manual? It's possible that tempcos may be specified, if
only indirectly, in the part numbers listed.

I could find no further info on the actual model I have, but I discovered
the PM3262 is identical except it has only 2 channels, so I downloaded
the plans for that instead. It contains a rough description of how the
switcher section works. Here's the link (ignore the 4112 bit, it's an
error, you get the 3262 trust me!)

http://www.download-service-manuals.com/en/manual.php?
file=Philips-4112.pdf

On Wed, 17 Feb 2016 14:56:20 +0000, Cursitor Doom wrote:

BTW, there's another variant of this schematic which uses just one 560pF
500V ceramic instead of the fixed and variable combo shown on the
schematic I've previously posted, so perhaps Philips themselves had
problems with this part of the design before!

The manual I linked to in my earlier post contains *both* of the above
variants of the SMPS section, btw.

On Wed, 17 Feb 2016 14:39:49 -0000 (UTC), Cursitor Doom
wrote:

On Wed, 17 Feb 2016 08:48:08 -0500, legg wrote:

Your concern for this resistor's health is touching, but it really is
misplaced, as the part is easily replaced, relocated, temporarilly
beefed up or whatever.

Nice idea, but it's in series with the main transformer primary winding,
and I'd rather that resistor go up in smoke than the transformer. If I
replace it with a higher power one of the same value, it puts the primary
winding and other components at risk of fatal damage.

If you know what you want to do, in 15seconds, you can do a lot,
particularly if you're prepared to repeat the process at intervals,
until all the necessary info is accurately recorded.

It takes about 5 minutes to fully cool down and then give me another 15s
of probing. I could of course let it cool for just 90s, but then I'd
hardly have time to get a probe in place before it got too hot again. I
simply haven't the patience to work that way - and the fumes off that
resistor are not pleasant.



One thought has come to mind. Common advice on powering up a scope
after long terms out of service involves the use of a variac to
'reform electrolytics'. This advice may be counterproductive with some
switch-mode power supplies, which may never operate normally (for more
than ~20mS) at their low-line dropout voltage. It's just not something
that they are required to do, nor do they see this condition in normal
service. It's usually avoided by introducing considerable hysterisis
in the UVLO and start-up thresholds.

If you're confident that the high voltage bulk electrolytics are
functional, you should probably just apply normal mains input to the
fully loaded psu. This may allow it to function normally, if no stupid
substitutions or errors have been created by earlier meddlers.

Based on your response (or lack there of) to earlier suggestions, I
can only suggest now that you review previous posts on this subject.
I personally couldn't go farther in troubleshooting this unit without
running through those steps or gathering that information.

Best of luck anyways

RL

On Wed, 17 Feb 2016 10:53:18 -0500, legg wrote:

If you're confident that the high voltage bulk electrolytics are
functional, you should probably just apply normal mains input to the
fully loaded psu. This may allow it to function normally, if no stupid
substitutions or errors have been created by earlier meddlers.

Already double-checked that yesterday, just as you suggested above.
Exactly the same thing happens as when the smps board is out and on the
bench and tested separately.

Based on your response (or lack there of) to earlier suggestions, I can
only suggest now that you review previous posts on this subject. I
personally couldn't go farther in troubleshooting this unit without
running through those steps or gathering that information.

The original thread was packed with suggestions I'd already tried from
people who clearly hadn't read all the posts. I don't blame them for
that, of course, it became rather long and convoluted and I had trouble
keeping track of it myself.
I shall continue to work on it as time permits, but it's a challenge for
me having zero experience with switchers before and not knowing what's
been done to it by previous technicians. The fact that it's been 'got at'
by others adds an extra layer of unknowns to an already tricky set of
problems.

Best of luck anyways

Thanks, I'll need it.

did you disconnect everything frmo the secondary side and fire up the primary side and see if the resistor still get hot.

if not, you have a fault on one of the secondaries, reconnect them one by one.

if it still gets hot, you have a fault on the pri side or a bad xformer perhaps in the form of a shorted winding. Use yopur 2v drive on the pri to try to isolate a shorted winding,,,how??? I'm not sure. See if there is any winding that when you intentionsally short it, nothing much happens.

On Wed, 17 Feb 2016 14:15:08 -0800, makolber wrote:

did you disconnect everything frmo the secondary side and fire up the
primary side and see if the resistor still get hot.

if not, you have a fault on one of the secondaries, reconnect them one
by one.

if it still gets hot, you have a fault on the pri side or a bad xformer
perhaps in the form of a shorted winding. Use yopur 2v drive on the pri
to try to isolate a shorted winding,,,how??? I'm not sure. See if
there is any winding that when you intentionsally short it, nothing much
happens.

Yes, I'm pretty much on the same page as you with those suggestions. I've
sorted through about 500 caps today and not one of them was a 560pF!!
Could not believe it but still have a couple of thousand or so left so
maybe there's still hope.
I'm using an old valve HP sig gen for this purpose as it puts out a
healthy 75V p-p maximum thereby getting closer to the operational voltage
than a modern solid state one would be capable of.
Anyway, I'll report back in a day or so....

On 16.02.2016 19:27, Cursitor Doom wrote:
Hi again,

I know I said I was going to mothball this scope for the time being,
but I keep thinking of new things to try and just can't leave the
damn thing alone. It struck me I ought to next check out the main
transformer because if that's toast, the whole psu might as well be
binned. I think I *may* be on to something.

Here's the circuit again:


https://www.flickr.com/photos/128859...in/dateposted-


Hi

It looks like the circuit is tricky and you aren't getting anywhere yet.
The caps you mentioned won't do anything with the board outside the
scope. Even if they were completely shorted, as long as the board is on
your table and with no loads connected, they won't be noticeable. So you
can skip them for now.

I think that you should first try to pinpoint, in which general
direction to look for the fault, that is:

- in the power circuitry
- in the feedback, control and regulation circuitry
- in the output circuitry (rectifiers and capacitors)
- in the auxiliary supply (diodes V1816 to V1819 and caps)

If the fault appears in the power circuitry, it is likely

- due to dielectric breakdown (capacitors and transformer)
- due to diode breakdown
- due to operation outside of resonance (also control related)

To this end, you'll need a way to slowly power up this thing at
controlled, limited power conditions and try to see what works
(to a limited extent, given the conditions) and what doesn't.

I'll try to include some step by step instructions, but please be aware
that this is mostly instructions for the desperate, so please take care
to exercise more caution than you might otherwise.

You'll need a 12V load, an isolated variac, and a signal generator
capable of outputting a square wave with variable duty cycle. The 12V
load should be easy to observe visually, therefore preferably a lamp. It
should be substantial (not a christmas tree light), but still sort of
within the capabilities of this supply under heavily reduced voltage
conditions (so not a car headlight either). A car taillight lamp is, I
think, Okay, but a brake light lamp is maybe already too much.

You will also (actually first and foremost) need to pay attention to
good safety practices, and also don't increase any voltage too fast and
don't "take shortcuts". Patience is a virtue and "I'll just quickly do
...." has no place when working with a defective and live power supply!

---------------------------

First, disconnect L1803 and R1807 and R1804. This will make the
controller inoperative and allow you to drive V1806 externally.

Connect the signal generator to V1806 B-E (in parallel with R1812) A
signal generator with 50 Ohm output, set to +/- 5 V peak square wave
should easily drive the transistor into full turn-off and into full
saturation.

Connect the board (mains input) to the isolated variac (or to an
isolation transformer that is itself being fed from a variac), but keep
the voltage at zero first.

Set the generator to the nominal resonance frequency that this board is
expected to run at (value was mentioned somewhere in the previous
thread), square wave, duty cycle somewhere around 30 - 35 %, 5V peak.

Connect the lamp to the 12.7 V output and some voltmeters to other
outputs for reference.

Turn on the signal generator, make sure the waveform is right.

VERY SLOWLY start raising the input voltage with the variac. Don't go
higher than about 70 V at any time.

Watch out for unexpected problems (heat, significant increases in
primary current consumption, other signs of overload!

Try to power it up so far that you can see the lamp faintly beginning to
light, but not much further than that. Watch for overload conditions. If
you notice shorted rectifier diodes in the output (they would heat up)
or capacitors (also), replace them and try again. Don't let the
dissipation resistor (R1814) overheat, reduce voltage if needed. If you
can't get the voltage up enough for the lamp to glow (because something
else overheats first), try a small 3 volt lamp from a pocket light
instead. It should allow the test to continue at much lower voltages.

Now, try to adjust it into exact resonance (it won't automatically be
there with your "first try" coarse signal generator setting). Slowly and
carefully adjust the signal generator frequency (still with 35 % duty
cycle constant) a little up and a little down, try to get a feeling for
the direction and sensitivity, and then set it for maximum lamp
brightness (maximum brightness = resonance frequency peak). Keep the
frequency there. If you (at any point) notice the lamp becoming too
bright, wind the variac down. Don't ever let it burn out!

If you cannot get good resonance, even at low voltage, check C1806 and
C1807. They may be open-circuit. Capacitors do not always fail shorted,
sometimes they fail open, and sometimes they even fail with a weak (too
high resistance) connection inside (this won't be noticed on a LCR
meter). Especially foil capacitors at high impulse loads sometimes will
do that. If in doubt, replace always both together.

Make sure that you really have resonance - check the waveform at the
transformer. It should be a sinewave. Possibly with some distortion,
that's okay, but still mostly a recognizable sinewave. This test is
better performed with the small 3 volt lamp at a low variac voltage. If
you cannot get a reasonable waveform, it means that some component in
the power circuit is most likely shorted out completely. It may (or may
not) be a diode. Also don't forget V1816 and its mates, C1819 and C1821.
If you found a problem in this area, just disconnect all those 4 diodes
for now, you can replace them later.

Always remember to switch it off (wind the variac full down) when
changing lamps, in this mode the board is not supposed to run with no
load at all, at any time and for any time duration. Not even for a
fraction of a second.

Now put the 12 V lamp back in and dial the duty cycle on the signal
generator very low (preferably around 3 %, at least smaller than 5 %).
Make sure that you have the correct polarity (so that the switching
transistor duty cycle is really 3 % and not 97 %)!

Now slowly raise the variac again. Keep an eye on the lamp, another on
heat dissipation. Make sure the variac is isolated (or connected through
an isolation transformer) because you will want to try reaching the
nominal mains voltage now.

Keep an ammeter in the power input. Watch out for sudden increases in
current draw. Keep an eye on the lamp and on the voltmeter(s) on the
output(s).

Watch out for power "inversion" behavior. This is important! Watch out
for behavior, where the lamp suddenly becomes darker instead of brighter
as you increase the input voltage, although the current draw has
increased. If this happens, then it's likely that you have a problem in
the power circuit due to dielectric breakdown (or due to a diode
breaking down). Set the voltage where this "just happens" and identify
the part that is in process of breaking down (it may start hissing,
smoking, getting hot and showing similar signs of distress).

If you found one, replace and repeat, looking for others. Whenever you
have replaced something, go back to low voltage and 33% duty cycle and
readjust for best resonance (because it may have changed when you
replaced a part).

If you cannot get the input voltage back to full mains, even with
minimal duty cycle (3 % or slightly less) on the signal generator,
because something overheats too fast, and you cannot get the lamp to
light (even a little), then something is wrong in the power section. It
is most likely some isolation in some part breaking down or a part in
the power section being shorted. Disconnect all rectifier diodes from
all transformer outputs except the ones where the lamp is (V1831 +
V1834) double-check or replace those and see if the problem goes away.

If you can now get it back to full mains without anything smoking, count
yourself lucky. This would mean that the power circuitry is basically
okay, and that therefore the problem must be elsewhere (in the control
circuitry most likely).

At full mains, but still at 3 % duty cycle, the lamp will probably light
very dimly (maybe hardly visible). That's ok.

Now try very slowly and carefully (and considering isolation and good
safety practices of course, after all there are high voltages now) to
increase the duty cycle in order to get the lamp to light up with
"normal" brightness. Remember that there is no regulation and no
feedback, so don't switch anything in "hard" steps and don't burn the
lamp out. With the load suddenly disappearing (burned out lamp) the
board would fail catastrophically and immediately, faster than you can
turn the power off!

Carefully set the duty cycle to normal lamp brightness and let it run at
nominal input voltage (variac at 100 %) and normal lamp voltage (around
12 V) for a while and watch out for signs of electrical and thermal
overload. Don't touch parts now, they are at 310 V!

If you can let it to run in this way for a while, then the power
circuitry is definitely good and the original problem is not in the
power part. You will have to troubleshoot the control and regulation
circuits later.

Drop the variac voltage somewhere low, measure the resonance frequency
in this (semi-working) state and write the value down. You may need it
later.

Carefully wind everything down, first variac to zero, then signal
generator off, and reassemble the board back.

-------------------------

Post here, how far you came and if you could really find a problem in
the power circuitry or if it was good and the control part is the next
in line.

@other regulars he If you notice something amiss with the steps
above, or something that is likely to be wrong, please indicate that.

Regards
Dimitrij

In article , says...

I've
sorted through about 500 caps today and not one of them was a 560pF!!
Could not believe it but still have a couple of thousand or so left so
maybe there's still hope.

Wow! When riffling through mixed resistors, long before that many I get
a powerful urge to pick one at random and paint it in the right
colours...

Mike.

On Thu, 18 Feb 2016 06:10:08 +0000, MJC wrote:

Wow! When riffling through mixed resistors, long before that many I get
a powerful urge to pick one at random and paint it in the right
colours...

LOL! I like your thinking. Actually when I returned to the task today and
fished out another tub of caps, the very first one came out at 535pF -
close enough to 560pF IMV - and didn't go up in smoke when I put 270VAC
across it. Neither did its capacitance change afterwards, so I've got
myself a replacement part at last. Funny how these things eventually work
themselves out.

On Thu, 18 Feb 2016 00:47:43 +0100, Dimitrij Klingbeil wrote:

[...]

Dimitrij, you've gone to an awful lot of trouble here for which I'm
extremely grateful. The main impediment at this time is that my fancy,
high-end Marconi sig gen is temporarily out of order. Perhaps I could rig
up something to do the same job from a 555 timer, but that's another
story.
I followed everything you said - except for one important thing: you say
this board has to be under load at *all* times when mains powered or it
will be catastrophically destroyed (or words to that effect).
I'm a complete switcher novice, so there have been *many* times when I've
probed this board under power with no load connected; I didn't know any
better. Before I go any further I need to know if I've therefore toasted
the board. Which components exactly catastrophically fail under no-load
conditions?
Thanks again for all your help.

Well this afternoon I removed the dodgy looking caps someone had soldered
in. They crumbled into dust in the process; certainly not 500V rating no
way. Soldered in a replacement. Anyway, as predicted by two of our gurus,
no improvement.
So I've just pumped in 10VAC p-p sinewave @ 20kHz to the primary winding
and scoped all the secondaries. From these recorded voltages I shall now
work out the implied winding ratios at the various tapping points and see
if anything shows up as requiring further investigation (shorted turns
and whatnot). Never give up, eh?

'Laterz' as they say nowadays.

On 18.02.2016 12:35, Cursitor Doom wrote:
On Thu, 18 Feb 2016 00:47:43 +0100, Dimitrij Klingbeil wrote:

[...]

Dimitrij, you've gone to an awful lot of trouble here for which I'm
extremely grateful. The main impediment at this time is that my
fancy, high-end Marconi sig gen is temporarily out of order. Perhaps
I could rig up something to do the same job from a 555 timer, but
that's another story. I followed everything you said - except for one
important thing: you say this board has to be under load at *all*
times when mains powered or it will be catastrophically destroyed (or
words to that effect). I'm a complete switcher novice, so there have
been *many* times when I've probed this board under power with no
load connected; I didn't know any better. Before I go any further I
need to know if I've therefore toasted the board. Which components
exactly catastrophically fail under no-load conditions? Thanks again
for all your help.

Hi

As for the board being loaded at all times...

In that rather long description I posted a method for slowly bringing up
the board "from zero up to where a lamp lights up". Please note that the
first step was to make the board's internal switching controller
inoperative (by disconnecting its start-up power supply and the power
transistor driver transformer) and using a signal generator for timing.
In this particular state of affairs the board (obviously) has no
automatic control, feedback or regulation whatsoever. The only feedback
"control" is through your eyesight (lamp brightness) and the reaction of
your hands ("Turn the duty cycle and the voltage down quick NOW!"). This
is not a particularly fast nor a particularly reliable method of SMPS
control, so I put in some extra words about exercising caution. When
running the SMPS in this "mode", you need to understand that the power
flow in it is governed by two aspects.

The switcher at a fixed duty cycle (that was set on a signal generator
manually) and at a fixed voltage (also manual from a variac) will
basically be pumping out a fixed amount of power (a fixed amount of
energy per unit time). Normally the power would be controlled (and
automatically reduced very fast if needed) by the SMPS controller chip,
but in full manual mode that's obviously not possible.

Now this "flow" of power is balanced by the dissipation in the lamp -
and only there, because there is nothing else where to put the power
into and no immediate way to reduce it either. As long as the lamp is
still OK, whenever the SMPS power "flow" gets higher, the output
capacitors will charge up, the output voltages will therefore rise, the
lamp will light brighter at higher voltage, pulling more power. By its
increased dissipation, the lamp power will balance the SMPS power, and
restore a net zero balance, preventing the voltage from rising further.

But if the lamp breaks, the SMPS will suddenly find itself without a
load, and with the power having "nowhere to go". Without automatic
control the power won't get reduced automatically, so it will continue
flowing into the output capacitors and charging them up. But they don't
have a very big capacitance and in a fraction of a second, their
voltages would rise way outside their ratings, causing some spectacular
bang somewhere. This happens so fast that you won't have a realistic
chance to turn down the operating parameters (voltage and duty cycle)
manually in time before something breaks. Power supply controller ICs
can handle power control and shutdown on a sub-millisecond basis, but
human reaction times and hand movements - unlikely.

As you see, the "keep it loaded with no interruption" is a particular
precaution that applied only to a particular mode of operation - full
manual power setting with no automatic feedback control. Yes, it's
tricky, and when you have an "uncontrolled" (unstable) SMPS waiting to
do something stupid as soon as the power balance gets upset, that's
where caution is definitely needed.

In "normal" circumstances, with the controller IC working and performing
its natural protective functions a power supply is way more robust and
can work with no load just fine.

Regards
Dimitrij

On Fri, 19 Feb 2016 21:39:35 +0100, Dimitrij Klingbeil wrote:

Brilliantly clear explanation again, Dimitrij. Many thanks indeed.

Just to let you know I just found this 3212 to have a bad transformer. It is a little different as it runs on 24 volts, but it is bad.

I finally hooked it up to all its loads and it let out a little smoke when excited by a power amp at 10 KHz. Couldn't see where the smoke came from so I had a rally good look and looked at the transformer. The outside insulation appears burnt, not too bad but bad enough I guess.

Thing is, this thing has no shorts in it. Being a self oscillator of course the bad transformer stops it, not enough feedback.

I know it is a different unit, but this proves the transformers can go bad. I'd've never thunkit. Like, flybacks in TVs went bad but they had 30 KV floating around in there. The highest supply on this one is 1,500 volts.

Anyway, just lettin' y'know that it is probably more of a possibility than I thought before. If you find out unequivocably that your trasnsformer is bad let us know and I will take that into consideration if buying another one that does not have basic operation.

I don't like dead scopes anyway, 'I like to see a trace before putting the money down, but this one got thrown in on a package deal type thing so it won't break the bank.