the output of the second winding is: -Two ER3B parallel
Diode -> Pi filter, the left are two 1000uf/10V capacitors,
the right are two 100uf/10V capacitors
The failure happens on these two 1000uf/10V low-ESR capacitor
I edited the drawing, is this right?
| ___Vcc_Feedback to UC3843
| ER3B |
|--o--|>|---o---o-------o-o-- inductor --o--------o-----VCC
| | | + | + | + | + |
| '--|>|---' _|_ _|_ _|_ _|_
| --- --- --- ---
| | | | |
| 1000uf/10V 1000uf/10V 100uf/10f 100uf/10V
What's your Vcc voltage and load current? What's your
measured output-voltage ripple on the 1000uF caps?
What's the uc3843 output-switching frequency, 50kHz?
Can you tell us the cap's manufacturer and part number?
Wait, I see, HITANO EXR Low-ESR 1000uf, 10V (0.08ohm,
1040 mA ripple current 105 Centi degree, 100KHZ).
The ripple current in each 1000uF capacitor is equal to
half your load current; compare this to the datasheet's
ratings. (If the rating is for 120Hz, you can multiply
this value by about 1.3, for use at 50 to 100kHz.)
A ripple-current rating for a good 1000uF 10V cap,
like nichicon's PW series, is 1.0A. This would limit
your supply to well under 2 amps, say 0.75 to 1.0A max.
Note, a 50kHz smps running at 0.5A per 1000uF cap,
would have 10mV of 50kHz dV/dt = I/C ripple voltage,
plus more from the cap's esr, 32mV across 65 milliohms
for the PW part, or 42mV total, which is a lot. Your
L-C output filter makes me wonder if your capacitor's
voltage ripple isn't rather high? Hmm, the HITANO EXR
cap has 76m esr compared to 65m for the nichicon part.
If your ripple voltage is too high, your best choice is
a larger capacitor: higher value or voltage rating, or
both, to lower the esr and increase the ripple-current
rating. If this isn't possible, you can try increasing
the 3843's operating frequency to reduce the capacitor's
dV/dt ripple voltage, which may help reduce its stress.
Phil's suggestion of a bad electrolyte is sadly a real
possibility. But this scene has been most common with
inexpensive substitutes for premium-grade conductive
polymer or organic-semiconductor high-conductivity-
electrolyte capacitors, like Sanyo's OS-CON types.
It's more likely you're just over-stressing your caps.
To test this idea, remove some "good" capacitors from
power supplies that have not yet failed, but that have
experienced a known stressful operating life so far.
Measure the capacitor's esr at 100kHz, etc., looking
for a significant increase compared to new capacitors.
If you have a supply with one failed capacitor and can
remove and test the other, that'd be an opportunity to
find one going bad before it self destructs. I measure
capacitors with an hp 4192A impedance meter, purchased
on eBay for $3k. That's what capacitor manufacturer's
use. But you may be able to get by with a Dick Smith
Electronics ESR meter. I bought mine from Canada,
let us know if you need help finding one.