Chinese commodity power supplies have tended to use recognizable
configurations from times gone by. In doing so, it's easy to
miss some of the 'small stuff' that actually produced a reliable
product, in the day.

Even more so, when pricing reaches the 'replace vs repair' threshold
- why even bother with burn-in, in that case? If no burn-in or field
return failure analysis is ever consudered, the small errors persist, particularly if vendors play wack-a-mole with the same hardware
offered under different brand names and paperwork.

Case in point is a 5V 40A unit advertised 'for use in LED sign',
commonly used in Onbon product. In the application where a repair
or replace decision was made, actual consumption was in the 35W
range, though a test sequence could draw much higher power.
replacement with an identically rated unit was Cdn$22.00.

The replacement was physically and schematically identical, but
relaid as a mirror image for component placement. Different
brand name.

Anyways - a basic self-oscillating bipolar transistor half bridge
with forced beta, synchronized/steered and pwm'd by opening and
shorting the resistor-limited, center-tapped 'drive' winding.
Open collector drive out of a KA7500.

What's a KA7500 ? Turns out to be pin compatible to TL494, but
mfrd by Samsung/Fairchild/ONS.

http://ve3ute.ca/query/TL494_vs_KA7500.pdf

Oodles of data and apps for the 494, not so much for the 7500.
If anyone's got app info published for the KA7900, in any
language, I'd be interested to see it.

The TL494 was an interesting choice for a chip to clone,
considering the perceived importance of pulse-by-pulse
current limiting in subsequent control chip designs.
It modulates the turn-on time in the drive period.

It's possible to turn the switch off, after turn-on, before
the end of a conduction period. The error amps are extremely
fast, but you have to latch this decision somehow, in order
not to produce multiple pulses on the phase, before the period
ends.

The commodity app simply adds a slow control loop for average
current limit from a crude output sensor. I was unable (and
unwilling) to provoke a current limit below 70A of test load,
prefering to adjust the circuit to get some kind of limiting
response before component and fuse ratings were exceeded.

The actual failure in the pulled unit was an electrolytic
capacitor in the bootstrap housekeeping supply. If ESR rises
above a certain level in this part, the unit cannot start.
In 24/7 service (or static burn-in), you'd not notice till
the last power cycle, power failure or cold snap.

I used to be quite sniffy when it came to specifying parts
for this kind of position - ratings seldom reflecting the
standard use; it was hard to ensure ESR below 10 ohms (the
practical upper limit for guranteed start-up) in small
electrolytics over their intended environmental range and
lifetime.

RL

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)