On 24.02.2016 00:23, Cursitor Doom wrote:
On Wed, 24 Feb 2016 00:04:13 +0100, Dimitrij Klingbeil wrote:

On 23.02.2016 19:43, Cursitor Doom wrote:
On Mon, 22 Feb 2016 08:49:11 -0500, legg wrote:

If it's slow, it looks like a short when the power transistor
is trying to turn on, stressing the current snubber around
L1804.

I don't think this diode is the culprit, TBH. Just out of
curiosity I hooked it up and tested it this afternoon. The faster
diodes turned up so I thought it might be instructive to compare
them. The main flaw in my test is that I'm unable to replicate
actual working conditions. I just hooked up each diode in series
with a 1k resistor and fed the arrangement from my 600ohm sig gen
using 10VAC p-p. Slow recovery was certainly visible on the scope
with the BY134, but it wasn't *that* bad. In fact it was still
able to function as a viable rectifier right up to nearly 600kHz.
There were no signs of slow recovery with the UF4007 of course,
but the difference at 20kHz, whilst still noticeable, is unlikely
to be causing the issues I've experienced. But as I say, it was
in no way a scientific test and only when the new diode is in
circuit will we know for sure. I won't be holding my breath!

Well, I don't think that it's the main culprit either. But it may
impair the working of the energy recovery circuit far enough to
make it inefficient, forcing it to dump too much power into the
resistor. If everything else was well, that might still have worked
to some extent.

But you're trying to troubleshoot it, and something is obviously
wrong that causes the resonant circuit to appear as too low
impedance. Either the transformer is broken or the output circuits
(rectifiers) or the whole thing is operated on wrong frequency too
far out of resonance.

If the energy recovery circuit was working well, it should be able
to protect the resonant circuit, even at some overload, by
diverting the energy back into the main capacitor. That would allow
you more time to "probe around", checking what is the cause of the
overload.

Slow diodes usually become worse with rising currents, so one that
is able to drive an 1k resistor from a signal generator may just as
well behave like an RF short circuit if one tries to push
significant amps through it. So it's really difficult to compare.

Anyway, while I don't thing that it's enough, I was hoping that
making that part work efficiently again would at least lower the
load on the resistor to some extent, and give you more time for
such more complex things like resonance frequency measurements or
even adjustments.

Also, as for testing the transformer (out-of-circuit, with a poor
man's IWT equivalent), see my other post.

Regards Dimitrij

Many thanks as ever for your thoughts, Dimitrij. One question on
your other post before I forget: your schematic shows pulsing the
transformer input at 100Hz, so we're just testing the primary winding
in this instance, right? We're not concerned in this test about
what's coming out of the secondaries? I assume so because 100Hz is so
far off its intended frequency range but would be grateful if you'd
confirm if I have this right.

I fully agree with your observations on my diode test's
shortcomings.

The only other thing I'm waiting for is some replacement caps for
the original tropical fish types that don't look very healthy. They
test okay at low voltage but may be misbehaving badly at closer to
their working conditions. They're in really poor shape visually and I
could certainly believe THEY might be responsible for the issues I've
had. They should be here tomorrow or Thursday so by the end of this
week, I should have some firm results one way or the other.

As for the transformer test: This is actually quite a generic test that
is often used to test inter-turn winding isolation under high voltage
conditions by the manufacturers of motors, inductors and transformers.
Normally they use a specialized piece of equipment called an IWT, and
the test is routinely performed in production. Unfortunately an IWT is
expensive (like $2500 and up), and rather specialized, so the typical
repairman won't have access to one unless he works at a place where they
are commonly used.

The test frequency actually doesn't matter, and most common IWTs won't
go all that high. In the past, when IWTs still had a tube screen, they
needed a steady repetition rate in order to display the trace. So 50
Hz (or whatever your country's line frequency is) was not unusual.
Nowadays they all have flat screens and lots of sample memory, so the
repeat rates are usually from "single shot" to maybe a dozen a second.

Since you've told us in the past that you have a CRT oscilloscope (you
posted a picture of a noisy signal on a switching transistor that was
shown on such an instrument), a test circuit should have some impulse
rate that is reasonably fast that you'll be able to see a steady trace
on the screen. That's where my 100 Hz came from. You can obviously go
lower, the circuit has no lowest limit, but because of the resistor in
the charging circuit it won't likely be able to go much faster. 10k is
already a low resistance for 300-something volts and 10W is also quite
considerable, so making it faster would mean making it beefy and power
hungry too, and these side effects would outweigh the benefits.

For the actual test, one single impulse would theoretically be enough.

In practice however you'll want a repeating pulse train for 2 reasons.
First, in order to see it (unless you also have a digital scope to
capture a single pulse), and second, in order to see the state of the
isolation properly (usually broken isolation will arc in some sort of
semi-irregular fashion and that may not be visible with only one try).

The test is done in such a way that the coil (under test) and the
resonance capacitor (inside the IWT) are connected together while at the
same time a very fast charging circuit "charges" this LC resonant
circuit to a preset voltage and then immediately disconnects itself. The
LC tank is then allowed to "ring down" naturally without outside
interference and the ringing waveform is observed.

The frequency of the ring wave is determined by L and C, the duration by
the coil's resistive losses. A defective coil (shorted or with arcing
isolation) will have a very low "Q", so there will be basically no ring
wave, just a fast decay from the charging peak down to zero.

In your case, you can use 15 nF for the cap, so that it will match the
same conditions like in the actual power supply. This means that the
ring wave will have not only the full starting voltage but also the
actual "correct" resonance frequency. With these conditions you can of
course take wave shape measurements with an oscilloscope on any output
of the transformer, not just on the primary. They all should show the
same shape and the voltages should all be realistic (but of course
brief, since each pulse doesn't last very long). The cap and charging
circuit should of course be connected only to the primary, to make for
realistic test conditions.

With enough pulses per second you should get a bright steady trace.

The 10 W resistor should allow for continuous duty operation, so you can
take your time looking at the scope. Lower wattages will also do, but
will need limited test duration with cool-down periods for the resistor.

Regards
Dimitrij