On 24.02.2016 22:37, Cursitor Doom wrote:
On Wed, 24 Feb 2016 21:15:08 +0100, Dimitrij Klingbeil wrote:
[...]
Oh, Dimitrij, I meant to say your diode replacement has made a bigger
improvement to the psu than we expected. The hissing noise has almost
gone and the power resistor now takes almost a minute and a half to
reach 50'C instead of less than 15 seconds using the old incorrect
BY134 diode. I now have sufficient time to do some probing around
under mains power! First up I plan to test the rectified outputs from
the long secondary winding to see if they are anywhere near the
6V-60V range they should be. I'll report back with the results
tomorrow. Many thanks for that!
Wow, that's quite something! I wonder how this thing ever worked in its
previous state. Given the change you describe, that square-to-sine wave
circuit must have been as "good" as completely non-operational.
Now that it looks like it's "almost working", the supply might even run
again in the scope (to some extent at least), but the fact that the
resistor still slowly heats up a little, may indicate that it's slightly
out of resonance now. I mean, the frequency is not completely wrong, but
it might be just somewhat off-center.
No exact idea yet, but I think I'm beginning to see a pattern:
Here's my guess, not sure wild or not, so take it with a grain of salt
and the usual precautions of a power supply repairman
It looks like the thing may have drifted a little bit out of resonance
over the years. That can happen, electronic parts age and tolerances
slowly increase. Running out-of-resonance, the power factor of the
resonant circuit was probably no longer close to one, but instead the
resonant circuit began to pull reactive power. If you try to drive an LC
circuit with a frequency that is slightly wrong, the driving source will
still force the LC into its frequency, but there will be a phase shift
between voltage and current. The further off the frequency, the larger
the phase shift will become. A phase shift means that a load is no
longer purely resistive, but also reactive (either capacitive or
inductive depending on direction) and so the power factor gets lower. As
the power factor gets lower, the total current draw increases (imagine a
constant current due to resistive load plus an additional current due to
the reactive part of the load, which increases).
Now my guess, what may have happened:
The thing drifted over the years, and the total RMS current was slowly
increasing because the frequency wandered away from resonance and the
power factor of the resonant circuit was going down.
With the RMS current becoming larger, the load on the diode also became
larger (it depends on the total RMS current of the LC circuit, no matter
whether that's resistive or reactive), and the diode heated up more.
Sooner or later, after many years, the diode finally overheated and
shorted out, immediately blowing the fuse. Somebody saw this and
replaced it. But he did not know, with what speed grade to replace it
properly, so he put in a particularly slow one without thinking.
Now with the slow diode in place, it no longer blew the fuse, but
instead the energy recovery circuit became barely operational.
It still ran for a while, but without the square-to-sine conversion, it
was driving the (now no longer really resonant) main transformer with a
rather square-ish looking waveform.
That waveform caused large current spikes in the resonant capacitors
(you know, capacitors don't like being driven with a square wave, charge
current peaks go through the roof if one tries to do that).
These peaks, plus possibly the low power factor and reactive current
(the frequency may or may not have been re-adjusted after the diode
repair) were now stressing the "new" diode (that was wrong anyway) and
also the resonance capacitors. The capacitors did not like this
additional stress (and they may have drifted over the years already).
When you stress film capacitors with large repetitive current spikes, it
slowly erodes and embrittles the foil electrodes inside. The capacitor
still tests OK on an LCR meter and even on an ESR meter, it may still
look like working, but under full load conditions it can no longer
sustain large currents. It becomes like as if someone has put an inrush
current limiting device on it, it can no longer supply peak loads
(people who repair photoflash units often find this fault in the HV
trigger capacitor, it tests with a correct capacitance, but can no
longer supply a strong current pulse for triggering).
Now there were probably two "processes" going on, accelerating each
other. The resonant capacitors (C1807, C1808) were degrading and letting
the frequency drift ever more out of resonance. The wrong diode degraded
too. When you try to feed a slow diode with large high frequency current
peaks, it can also degrade even more and become even slower and more
like a high-frequency short circuit. Both things probably started
slowly, but were accelerating each other until something really broke.
Now you've replaced the diode, so it should be OK again from the diode
point of view. But the resonant capacitors may have degraded and may now
be in a pitiable state. They are difficult to test because the problem
usually means good LCR meter readings, but much reduced power handling
capability only when running at full power.
My advice would be to replace them anyway. This sort of degradation is
difficult to test, so better safe now than sorry later.
With new capacitors, the resonance frequency will somewhat change (you
know, component tolerances, degradation of old ones, slightly different
values of new ones...). The change won't be drastic, but it may be
significant. Therefore I would advise you to measure the new resonant
frequency and then readjust the power supply's working frequency if it
happens to be different (R1827 is the FREQ trimmer).
To measure, you sweep the primary winding with a signal generator (with
the transformer in circuit and connected, but the transistor V1806
disconnected and no loads attached to the outputs). Look for maximum
amplitude with a scope and measure the frequency with a counter. Compare
the measured value with the one that the circuit runs at (fully
reassembled, with dummy load attached to avoid "light-load" mode).
If they deviate, VERY SLOWLY and VERY CAREFULLY readjust R1827*. The
service manual says how to do it. See chapter 3.4.4.2.1
"This potentiometer is a factory adjustment control. THE SETTING OF THIS
POTENTIOMETER MUST NOT BE DISTURBED UNLESS IT IS ABSOLUTELY IMPOSSIBLE
TO SET THE 12.7 V WITH THE AID OF POTENTIOMETER R1826* (FEEDBACK).
Adjusting procedu
- Set the main input voltage to 220 V.
- Turn R1827* (FREQ) fully anti-clockwise.
- Check that the voltage on the positive pole of C1831* is 12.7 V +/-
100 mV; if necessary; readjust potentiometer R1826* (FEEDBACK).
- Set the main input voltage to 170 V.
- Check that the voltage on the positive pole of C1831* is 12.7 V +/-
100 mV; if necessary; readjust potentiometer R1827* (FREQ)."
If you've had to readjust "FREQ", better re-test with 220 V afterwards,
when you are finished, and double-check the "FEEDBACK" setting again.
Readjust "FEEDBACK" again if there is a voltage mismatch on 12.7 V.
Note that in the service manual that you linked to, the part names are
different, like C1843 instead of C1831 on the schematic. But the
descriptions are reasonable and the meaning seems to be the same. I've
changed the text a little and written the schematic numbers instead.
I've marked the changed item numbers with an "*" and written the
descriptive names (FREQ and FEEDBACK) next to them.
Regards
Dimitrij