Transformer shot! (was scope SMPS/ capacitor venting)

[Cursitor Doom]

On Sun, 28 Feb 2016 22:50:19 +0100, Dimitrij Klingbeil wrote:

[...]

OK, Dimitrij, a *huge* quantity of info to digest here and in your other
new postings on the subject and more detail than I can assimilate in a
single sitting! Many thanks as ever; that will certainly be all I need to
know for the time being. I'll get to work on the checks when I return to
base after midweek.
Laterz...

 

[Cursitor Doom]

On Sun, 28 Feb 2016 14:02:33 -0800, jurb6006 wrote:

[...]
> I consider a dim bulb tester a must for this type of work.

Ah, that makes much more sense now; many thanks for that clarification.

 

[Dimitrij Klingbeil]

On 28.02.2016 20:53, Cursitor Doom wrote:

On Sun, 28 Feb 2016 18:55:47 +0100, Dimitrij Klingbeil wrote:

[...]
I think that the most realistic test would be to sweep the resonant
circuit with a signal generator and watch the waveform. If the
resonance frequency looks right (in the 20 kHz ballpark) and a
signal generator is able to drive it from a high 600 Ohm source
impedance to a significant amplitude without much "sagging" (that
is, the resonant circuit presents little load to the generator),
it's probably OK.

Thanks again, Dimitrij. You're obviously an expert on the little
understood world of resonant converters so when you say try this or
that, I make a point of paying extra attention. I liked your theory
on the resistor heating due to this supply running out of resonance
as a result of component values changing over time; in fact I'm
currently pinning my hopes on it. It's a pity I'm stuck here for a
few more days with my revolting in-laws but it'll be the first thing
I do on my return!

Hi

Please don't rely in my advice too much. While I do design electronics,
I'm very far from being an expert in this particular field. I've never
actually designed a resonant power supply, unless you count one little
3W prototype based on a modified Royer / Baxandall structure.

It may be relatively easy to look at a ready-made schematic and try to
guess various upper and lower limits based on parts and topology (like
"signal X cannot be higher than Y volts, otherwise part Z breaks down"
or "ratio of transformer X cannot be above or below A:B, otherwise the
ratings of part Y would be exceeded"), but that's not expertise by any
stretch of the definition. A lot may be intuition, but that's no
expertise either.

Somewhere I have a big old valve/tube capacitor tester capable of
simulating realistic high voltage working conditions. It'd be
interesting to know what kind of checks it's capable of performing
if it's still in working order and if I can find it among the
towering piles of obsolete test equipment I have here (a couple of
million pounds worth of gear at new prices adjusted for inflation) I
may possibly hook it up and give it a shot.

How about those 'Octopus' component testers? They subject the part
under examination to sweeping test voltages over the expected
working range and you look for any signs of breakdown on an
oscilloscope in X=Y mode. I guess this method is about as good as it
gets?

I've had to look up, what an "Octopus component tester" is. Apparently a
transformer with some provisions for routing the voltage and current
signals of the load to an oscilloscope, making a simple AC curve tracer.

I don't think that you'll need one here. It can test for breakdown, but
in your case that's unlikely (the capacitor would be buzzing and arcing
and the supply sure wouldn't work "almost normally"). It won't see the
problems that are likely to be important in an LC circuit.

1. The cap must have the correct capacitance. Any LCR meter or any
common pocket multimeter with a capacitance function can measure this.
This is a basic prerequisite that should always be tested first and if
the capacitance is wrong, no further tests will be necessary anyway.

2. The foils inside the cap must have a reliable connection (deviation
manifests itself as ESR, ESL, and the general inability to supply high
impulse currents). This particular curse will sometimes plague the
trigger capacitors from photoflash units (the flash won't trigger or
will only trigger erratically while the capacitance value is still ok).

This is difficult to measure directly, but can be checked with another
capacitor as a reference. You'll need a known good capacitor with the
same value (in your case: 15 nF), but not necessarily with the same
voltage (you can use a known good, but lower voltage one for testing).
The test is only with a signal generator, so the cap won't be subject to
a lot of stress.

Connect the known good capacitor to the original inductor (transformer
primary) with no other loads attached. Sweep with a signal generator
(use as much voltage as the signal generator can provide without much
distortion, that usually won't be a very high voltage anyway) and look
for resonance on a scope. Note the resonance frequency. Disconnect the
known good cap and connect the original one instead. Check where the
resonance is. If it's in the same place and the amplitude has not become
lower, the cap is very likely good. If it disappears and you can only
measure the inductor's SRF instead, (if the inductor has more or less
the same resonance with or without a capacitor connected), then the
capacitor is basically open-circuit or very high ESR. If the resonance
has wandered away somewhere, especially upwards in frequency, then the
cap is most likely degraded and not a good candidate for full power
resonant use either. Same thing if the amplitude has dropped much.

If your resonant caps turn out to be good, that most likely leaves only
the snubber diodes and a possible frequency misadjustment as the likely
causes.

If you check the resonance with a signal generator and scope against a
known good 15 nF, and it suddenly wanders way, or the amplitude drops,
then you'll need to find replacement capacitors. Fortunately, if you put
"WIMA FKP1 33nF" into ebay search, there seem to be many available.

Regards
Dimitrij

 

[Dimitrij Klingbeil]

On 28.02.2016 22:31, Cursitor Doom wrote:

On Sun, 28 Feb 2016 19:06:38 +0100, Dimitrij Klingbeil wrote:

Also, even with a dummy load connected, the stray capacitance of
an oscilloscope, when hanging off the loose end of a power circuit
with some 800 to 900 V worth of HF on it, would probably cause so
much undue capacitive loading that the power supply circuitry would
hardly handle it.

Isn't this just another example of the unsatisfactory nature of this
resonant converter design? If the thing is *that* fussy that a little
bit of stray capacitance can catastrophically destabilise it, then
AFAICS it's a fundamentally unreliable topology and it would be
better to have used one of the non-resonant forms of converter.
Unless there's some compelling reason I may be unaware of not to for
oscilloscope power supplies, of course.

That's definitely not "a little bit". By very far, not!

Muscling around a scope chassis (not the probe tip, but the probe ground
and the big scope chassis connected to it on the other end of the cable)
from zero to some 800 V in several dozen microseconds is no small feat,
much less doing that 20000 times a second repetitively.

Not many power supplies will do that on an internal node without running
into major stability issues (unless you have a very small
battery-operated "pocket" scope, sitting on a wooden table far away from
any earthed metal, thereby being a "light" load).

The probe tip is not the issue, but the scope itself, hanging from the
probe ground, that is.

 

[Dimitrij Klingbeil]

On 28.02.2016 22:33, Cursitor Doom wrote:

On Sun, 28 Feb 2016 19:23:20 +0100, Dimitrij Klingbeil wrote:

P.S. That voltage estimate has probably surprised you. Unless one
looks at the circuit schematic and adds all the voltages from all
the storage elements (inductors / capacitors), considering timing
and phase, it may not be obvious that the thing was intended to
run at such high voltage levels. But there's a reason why they used
a 1500 V transistor in it.

And yet C1804 is rated at 'only' 630V. Weird!

That's not a problem. It only ever sees 320 V from the mains, plus any
little remains of the mains surges that may come its way past C1802+3.

Even with surges and such, 450 V is likely the highest thing it will
ever see, so a 630 V rating is a good and conservative one.

It won't ever see the 800 V. But the transistor V1806 (collector) will.

Basically, the input caps will "see" only normal rectified mains (320
V). The resonant caps will also see some 300 to 320 V, but because the
sinewave resonance signal is bipolar, and one end is tied to the
positive end of the input caps, there will be times (each half cycle)
where the voltages will add and the result (referenced to the emitter)
will reach some 600 - 620 V. At these same times during the cycle, L1806
will also be reset via V1811, and the reset voltage (some 200 V, also
being in series) will also add to this, plus any little remains (50 V or
less) from the L1804 circuit. So the collector of V1806 will "see" quite
a lot of voltage when V1806 is in the "off" phase. But this high voltage
only applies to the V1806 collector, not to the other parts / signals.

Dimitrij

 

[Dimitrij Klingbeil]

On 28.02.2016 22:59, Dimitrij Klingbeil wrote:

On 28.02.2016 22:31, Cursitor Doom wrote:
On Sun, 28 Feb 2016 19:06:38 +0100, Dimitrij Klingbeil wrote:

Also, even with a dummy load connected, the stray capacitance of
an oscilloscope, when hanging off the loose end of a power
circuit with some 800 to 900 V worth of HF on it, would probably
cause so much undue capacitive loading that the power supply
circuitry would hardly handle it.

Isn't this just another example of the unsatisfactory nature of
this resonant converter design? If the thing is *that* fussy that a
little bit of stray capacitance can catastrophically destabilise
it, then AFAICS it's a fundamentally unreliable topology and it
would be better to have used one of the non-resonant forms of
converter. Unless there's some compelling reason I may be unaware
of not to for oscilloscope power supplies, of course.

That's definitely not "a little bit". By very far, not!

Muscling around a scope chassis (not the probe tip, but the probe
ground and the big scope chassis connected to it on the other end of
the cable) from zero to some 800 V in several dozen microseconds is
no small feat, much less doing that 20000 times a second
repetitively.

Not many power supplies will do that on an internal node without
running into major stability issues (unless you have a very small
battery-operated "pocket" scope, sitting on a wooden table far away
from any earthed metal, thereby being a "light" load).

The probe tip is not the issue, but the scope itself, hanging from
the probe ground, that is.

P.S. Since you indicated that you have some background with radio...

Consider the collector of V1806 as a signal source. As a signal source
that can basically drive a 800 V peak-to-peak square wave.

Consider the whole power supply board (including any cables and the
isolation transformer or variac that you are using to feed it) as one
half of a dipole antenna.

Consider the scope (the whole metal chassis) and the probe cable as the
other half of the same dipole antenna.

Consider the two halves connected in the middle by the probe ground
clip, at that overheating power resistor in the supply.

What you get, is a center-fed dipole, sitting on your table, and being
driven with a 800 V peak to peak fast square wave. Not a light load.

An isolated high voltage differential probe would "separate the halves",
so that the big (parasitic) dipole would no longer exist.

Regards
Dimitrij

 

[Cursitor Doom]

On Mon, 29 Feb 2016 21:47:38 +0100, Dimitrij Klingbeil wrote:

The probe tip is not the issue, but the scope itself, hanging from the
probe ground, that is.

P.S.

There is a simple though unwritten rule about power supply testing:

"Never connect the ground (common, chassis etc.) of any test equipment
to the switching node (power transistor collector, drain or power IC
output pin and its associated signals) of a switching power supply!"

It is valid for all types, no matter if flyback, forward or resonant.

The reason for this rule is that a "switching node" usually drives a
square wave with high voltages (some 500 to 600 V in a flyback, may
happen to be as much as 800 or 1000 V in a resonant one), and that a
significant amperage is readily "available" at that node too, due to the
output transistor's low impedance. Neither is the supply designed to
safely drive that into "RF ground" nor is the test equipment made for
being "muscled around" at that sort of voltages and dV/dt rise times.

Grounding the test equipment would mean that the whole power supply
(plus any safety isolation transformer) is being swung around and
letting the test equipment "float" would mean to also swing around the
test equipment. Apart from the obvious safety hazard, this can also
damage the test equipment and even compromise the test equipment's
electrical safety by frying the "Y" capacitors between mains and
secondary or stressing the isolation barrier in the test equipment's
power supply and / or mains transformer, possibly beyond the level of
stress that it was rated for.

So, whenever you troubleshoot some switcher, take heed of this rule.

It's simple to remember, and it can save lives, test equipment,
and some power supplies under test too

I'm grateful for that expansion, to be honest. I was kind of struggling
to get my head around what you were getting at in your earlier postings;
didn't make much sense to me on the first read through and although after
a second read I was beginning to sense your meaning, it still wasn't 100%
clear.
At least now I think I can finally see where you're coming from.
Naturally I read up on safety precautions when dealing with switchers
from books I have and all sorts of diverse sources on the net, but I can
honestly say that what you have outlined above has NOT been covered by
anything I've seen or read up until now. This would seem to be a glaring
omission on the part of those who we rely on to prime us up on the hidden
dangers and pitfalls of troubleshooting such equipment.
Another good reason for me to avoid dealing with switchers in future if
at all possible!!

 

[Cursitor Doom]

On Mon, 29 Feb 2016 21:47:38 +0100, Dimitrij Klingbeil wrote:

P.S.

There is a simple though unwritten rule about power supply testing:

Oh, I just noticed you did in fact actually state it's an "unwritten
rule" - well my personal experience of searching on the subject can
certainly confirm that!

 

"
There is a simple though unwritten rule about power supply testing:

"Never connect the ground (common, chassis etc.) of any test >equipment
to the switching node (power transistor collector, drain or power IC
output pin and its associated signals) of a switching power supply!" >"

Well now that you wrote it, it is no longer unwritten.

But I know what you mean. It is pretty much RF and if the caps don't short it out it can burn you some. If the voltage is high enough you don't even have to touch it. I got burned by the cathode of a damper tube in a color TV set once. That is half of about a 70 KHz sine wave clocked at 15.734 KHz. Arced to my finger, burnt and cauterized all the way to the bone. On blood, but it sure did smart. And then it felt funny for like six months after.

Can you imagine that on the chassis and probes of your scope ?

 

[Dimitrij Klingbeil]

On 28.02.2016 22:59, Dimitrij Klingbeil wrote:

On 28.02.2016 22:31, Cursitor Doom wrote:
On Sun, 28 Feb 2016 19:06:38 +0100, Dimitrij Klingbeil wrote:

Also, even with a dummy load connected, the stray capacitance of
an oscilloscope, when hanging off the loose end of a power
circuit with some 800 to 900 V worth of HF on it, would probably
cause so much undue capacitive loading that the power supply
circuitry would hardly handle it.

Isn't this just another example of the unsatisfactory nature of
this resonant converter design? If the thing is *that* fussy that a
little bit of stray capacitance can catastrophically destabilise
it, then AFAICS it's a fundamentally unreliable topology and it
would be better to have used one of the non-resonant forms of
converter. Unless there's some compelling reason I may be unaware
of not to for oscilloscope power supplies, of course.

That's definitely not "a little bit". By very far, not!

Muscling around a scope chassis (not the probe tip, but the probe
ground and the big scope chassis connected to it on the other end of
the cable) from zero to some 800 V in several dozen microseconds is
no small feat, much less doing that 20000 times a second
repetitively.

Not many power supplies will do that on an internal node without
running into major stability issues (unless you have a very small
battery-operated "pocket" scope, sitting on a wooden table far away
from any earthed metal, thereby being a "light" load).

The probe tip is not the issue, but the scope itself, hanging from
the probe ground, that is.

P.S.

There is a simple though unwritten rule about power supply testing:

"Never connect the ground (common, chassis etc.) of any test equipment
to the switching node (power transistor collector, drain or power IC
output pin and its associated signals) of a switching power supply!"

It is valid for all types, no matter if flyback, forward or resonant.

The reason for this rule is that a "switching node" usually drives a
square wave with high voltages (some 500 to 600 V in a flyback, may
happen to be as much as 800 or 1000 V in a resonant one), and that a
significant amperage is readily "available" at that node too, due to the
output transistor's low impedance. Neither is the supply designed to
safely drive that into "RF ground" nor is the test equipment made for
being "muscled around" at that sort of voltages and dV/dt rise times.

Grounding the test equipment would mean that the whole power supply
(plus any safety isolation transformer) is being swung around and
letting the test equipment "float" would mean to also swing around the
test equipment. Apart from the obvious safety hazard, this can also
damage the test equipment and even compromise the test equipment's
electrical safety by frying the "Y" capacitors between mains and
secondary or stressing the isolation barrier in the test equipment's
power supply and / or mains transformer, possibly beyond the level of
stress that it was rated for.

So, whenever you troubleshoot some switcher, take heed of this rule.

It's simple to remember, and it can save lives, test equipment,
and some power supplies under test too

Regards
Dimitrij

 

[Dimitrij Klingbeil]

On 01.03.2016 00:21, Cursitor Doom wrote:

On Mon, 29 Feb 2016 21:47:38 +0100, Dimitrij Klingbeil wrote:

P.S.

There is a simple though unwritten rule about power supply
testing:

Oh, I just noticed you did in fact actually state it's an "unwritten
rule" - well my personal experience of searching on the subject can
certainly confirm that!

At least myself, I have not seen it being being explicitly explained or
written anywhere yet, but it's sort of "common knowledge" in a way...

Among engineers who design power supplies, this seems to be taken for
granted - too self-evident to warrant explanation apparently. Others,
among them the many who design low voltage circuits and prefer to buy
their power supplies off the shelf, rarely get to see switching nodes
driven with significant fractions of a kV with fast rise times. That
leaves their awareness of the "tricks of the trade" rather limited.

Power supply design is both a science and an art, and the power supply
"artists"' rites of initiation can sometimes involve strange things

Anyway, never fear, but always exercise logical thinking, conservative
judgement and be aware of side effects - that would be my advice here.

Dimitrij

 

[Cursitor Doom]

On Mon, 29 Feb 2016 16:29:47 -0800, jurb6006 wrote:

> Can you imagine that on the chassis and probes of your scope ?

Not really. It's too weird. Maybe I could simulate it in spice to get a
better idea of what's going on here. Anyone got a model suggestion for a
'hanging scope'?

 

[Chris]

On Mon, 29 Feb 2016 16:29:47 -0800, jurb6006 wrote:

> Can you imagine that on the chassis and probes of your scope ?

It's very simple really. Just remember to keep the scope grounded to the
DUT ground AND the probe clipped to the power node (X1 setting on the
probe) at ALL times and bob's your uncle, you can't go wrong. That way
you are only placing about 15pf || 1M loading on the DUT.
HTH.

 

[Chris]

On Wed, 02 Mar 2016 22:28:47 +0100, Dimitrij Klingbeil wrote:

You must be meaning X10 setting. Common scope inputs as well as probes
are not designed to handle the typical peaks from power supplies at X1.

Fair point! But I use an externally selectable decade attenuator for
anything over 400V so X1 is good for me. But yes, in the absence of that,
X10 would be the way to go.

 

[Dimitrij Klingbeil]

On 01.03.2016 22:29, Chris wrote:

On Mon, 29 Feb 2016 16:29:47 -0800, jurb6006 wrote:

Can you imagine that on the chassis and probes of your scope ?

It's very simple really. Just remember to keep the scope grounded to
the DUT ground AND the probe clipped to the power node (X1 setting on
the probe) at ALL times and bob's your uncle, you can't go wrong.
That way you are only placing about 15pf || 1M loading on the DUT.
HTH.

You must be meaning X10 setting. Common scope inputs as well as probes
are not designed to handle the typical peaks from power supplies at X1.

 

[Cursitor Doom]

On Mon, 29 Feb 2016 21:47:38 +0100, Dimitrij Klingbeil wrote:

There is a simple though unwritten rule about power supply testing:

"Never connect the ground (common, chassis etc.) of any test equipment
to the switching node (power transistor collector, drain or power IC
output pin and its associated signals) of a switching power supply!"

As an aside, I'm just a bit mystified as to why anyone would want to do
this anyway?

Now, apologies for the delay, but I had the usual accumulation of
pressing things to deal with on my return so have only now got around to
carrying out the checks last suggested here.

OK, I measured the resonant frequency of the primary circuit (with the
chopper NOT disconnected, see notes below) by sweeping a frequency range
across the main tranformer's primary input terminals. It's not
particularly peaky, so there's a Khz or so on either side of Fo before we
get to the -3db shoulders. Fo, with no load connected came out as
17.35kHz.

Under power, with frequency counter connected between T1 and T2 with V1812
removed from circuit shows the PWM chip pulsing at 22.55kHz.

Unfortunately I have no idea what the factory figures should be and
whilst it seems like there's a big difference between the PWM chip's
output and the primary circuit's resonance, AIUI, they're not supposed to
be in sync at any time anyway. But are they supposed to be this far apart?

Notes:

1. I know somewhere it was stated that the chopper transistor should be
removed for the resonance test, but I couldn't see the harm in leaving it
in. If it invalidates the test, of course, then I'll whip it out and re-
do it. If you think it's relevant let me know.

2. I pulled V1812 as someone suggested because the noise coming back
down its collector from L1803 might have interfered with the frequency
counter's ability to read the clock pulses.

 

[Cursitor Doom]

On Sun, 06 Mar 2016 13:26:33 +0000, Cursitor Doom wrote:

Under power, with frequency counter connected between T1 and T2 with
V1812 removed from circuit shows the PWM chip pulsing at 22.55kHz.

Sorry, ignore that; copied the wrong piece of paper. It should be
20.64kHz. (This is with the load connected.) I then tried again with V1812
re-inserted and got 20.62kHz. Apologies for the earlier error...

 

On Wednesday, March 2, 2016 at 4:28:56 PM UTC-5, Dimitrij Klingbeil wrote:

On 01.03.2016 22:29, Chris wrote:
On Mon, 29 Feb 2016 16:29:47 -0800, jurb6006 wrote:

Can you imagine that on the chassis and probes of your scope ?

It's very simple really. Just remember to keep the scope grounded to
the DUT ground AND the probe clipped to the power node (X1 setting on
the probe) at ALL times and bob's your uncle, you can't go wrong.
That way you are only placing about 15pf || 1M loading on the DUT.
HTH.

You must be meaning X10 setting. Common scope inputs as well as probes
are not designed to handle the typical peaks from power supplies at X1.

The quote function musta got screwed up. I always use 10X unless I really need the gain, which is rare. I also recommend others use the 10X at all times as well. Not only does it reduce circuit loading, it also protects the scope to some extent.

Not the first time Usenet quoting got screwed up. I expect to see >> on a quote of a quote and > on a direct quote but it seems not to work that way all the time.

 

[Cursitor Doom]

On Sun, 06 Mar 2016 13:28:52 -0500, legg wrote:

When you've got this thing plugged in and running, what is visible in
the display? Can you get a locator dot? Traces in free-run?

I haven't tried this yet as I can guess sod's law making it the case that
I'd have to pull the plug just at the point where the CRT has warmed up
sufficiently. The other slight problem is to test this requires the board
to be completely inserted with every connection made plus a temp probe to
the power resistor which is all rather fiddlesome and not to be done
repeatedly if it can be avoided. I can see a situation arising (sod's law
again) where someone here will post saying - "oh, whilst you still have
the board out, just check this..."
Nevertheless, if nothing is said in the next 18 hours, I will test it all
reconnected and post the outcome here.

 

[Cursitor Doom]

On Sun, 06 Mar 2016 23:34:55 +0100, Dimitrij Klingbeil wrote:

If I understood the service manual correctly, they seem to suggest to
start from the lowest frequency when performing an adjustment and going
up until output regulation is reached. (They write from "fully
counter-clockwise" actually, referring to the "FREQ" trimmer, and
looking at the schematic that would likely be from the "lowest" position
of the wiper, meaning starting at the highest resistance and going
towards lower resistance.)

OK, well I can easily establish that safely by popping V1812 out of
circuit temporarily so sweeping the frequency adjustment pot won't have
any effect.

It's only my guess, but I think that they intended this supply to run
rather somewhere below resonance than somewhere above. This would mean
that adjusting the pulse frequency down to 17.35 kHz should do no harm
as that value would be lower than the setting right now.

If there was any danger of something blowing up by setting the frequency
lower, they would not be recommending to set it to the absolute minimum
before slowly adjusting it back "up" again.

Indeed. That seems to be the key point I have to observe.

Can you adjust the pulse rate to 17.35 kHz and then test the supply with
a dummy load?

Well I could.... But that's spot on resonance. I was under the impression
that they're not supposed to run actually directly at resonance?

Can you test it with a variac and see if it still maintains output in
regulation down to 175 V "mains" (adjusted to 17.35 kHz, that is)?

Yes, no problem. It seems the key regulated output is the 12.7V one and
if that's correct, the rest should follow. There's a trimmer for 12.7V on
the underside of the board.

Thanks again, Dimitrij. I'll report back tomorrow...