+/-45V PWM supply - terrible inefficiency and feedback inaccuracy

[mw]

Ref: my Dec 05 post, "transformers for cheap low current flyback supply,
+/- 45V ? "

After holidays I was able to devote some time to this and I have a pair
of independent supplies built. They basically work but there are two
problems and they are likely related.

1) Low efficiency, GIANT current spike at turn on, immediately followed
by a sag in the sense resistor current, then it rises back up at a more
expected ramp.

I have a theory this is due to the transformer not having good
inter-winding coupling. It is a cheap, small, versatile transformer,
the J.W. Miller PM620-07, 14 uH inductance per winding, with six
windings. My hookup has 3 of the windings in series as primary, and 3
in series for the secondary.

The odd thing is that there is a lot of wasted power but nothing is
getting hot.

2) the calculated feedback resistors do not give the expected output.

I have used the PWM chip, UCC3801, and similar parts a few times before
but have never seen this: the resistors I put into the feedback network
do not cause the supply to run at the correct voltage. I had to resort
to putting in a pot, adjusting it to cause the output voltage to
decrease to the right voltage, then measuring the pot to find out the
right value of resistance. Before doing this I was getting output
voltages of 112V or so, not too safe when using 63V electrolytics.

It could be that the big current spike is somehow confusing the chip's
cuurent sense. Maybe there is a minimum pulse width that always causes
the chip to run with too high of a duty cycle.

 

[Bob Eldred]

Ref: my Dec 05 post, "transformers for cheap low current flyback supply,
+/- 45V ? "

After holidays I was able to devote some time to this and I have a pair
of independent supplies built. They basically work but there are two
problems and they are likely related.

1) Low efficiency, GIANT current spike at turn on, immediately followed
by a sag in the sense resistor current, then it rises back up at a more
expected ramp. Snip.......

A high turn on current transient may be due to too long of a an on time when
the the circuit comes up before the oscillator stabilizes. You need a
storage or digital scope to capture and observe the turn on switching events
to see what's happening. Low efficiency means power is being dissipated
somewhere, most likely in the power switch (FET). Again a careful analysis
of the switching waveforms will be required to discover the problem. I
suspect that the switch comes on too early or goes off too late while high
currents with high voltage are present. Both at the same time causes
excessive dissipation. It's best to switch when both current and voltage are
low or zero. Depending on the design that may not be possible. I can't speak
for the transformer but one thing to look for is winding capacitance. It can
cause resonant waveforms that are quite different than expected causing
excessive dissipation. Also be sure the transformer gets nowhere near
saturation during the on time current run up. What is the frequency, what is
the max on time, what is the max flux density, what is the core material.
What are the core losses? Many questions. Power supply design, while simple
in concept is complex and difficult in practice.
Bob

 

[Tam/WB2TT]

Ref: my Dec 05 post, "transformers for cheap low current flyback supply,
+/- 45V ? "

After holidays I was able to devote some time to this and I have a pair of
independent supplies built. They basically work but there are two
problems and they are likely related.

1) Low efficiency, GIANT current spike at turn on, immediately followed by
a sag in the sense resistor current, then it rises back up at a more
expected ramp.

I have a theory this is due to the transformer not having good
inter-winding coupling. It is a cheap, small, versatile transformer, the
J.W. Miller PM620-07, 14 uH inductance per winding, with six windings. My
hookup has 3 of the windings in series as primary, and 3 in series for the
secondary.

The odd thing is that there is a lot of wasted power but nothing is
getting hot.

2) the calculated feedback resistors do not give the expected output.

I have used the PWM chip, UCC3801, and similar parts a few times before
but have never seen this: the resistors I put into the feedback network do
not cause the supply to run at the correct voltage. I had to resort to
putting in a pot, adjusting it to cause the output voltage to decrease to
the right voltage, then measuring the pot to find out the right value of
resistance. Before doing this I was getting output voltages of 112V or
so, not too safe when using 63V electrolytics.

It could be that the big current spike is somehow confusing the chip's
cuurent sense. Maybe there is a minimum pulse width that always causes
the chip to run with too high of a duty cycle.

Just a guess, but I think the tipoff is the grossly wrong output voltage.
Check, and double check your circuit against the data sheet. Also, you may
have a bad chip. Another thing to look at is that any external transistors
get a turn off signal, and are not left floating.

Tam