6-19-2007

An additional part may be added to the PNP
feedback transistor on the isolated side to improve it.
Place a switching diode
in series with the upper resistor of the
voltage divider for temperature compensation.
That transistor can have the double duty of also
providing current limiting if a very ;low value resistor
is placed in the positive output current path and sensed
with that PNP transistor.


3-12-2007

I have discovered two additions to the ZVS USMPS circuit
which may help performance in many cases.  The first
change resulted from my trying to improve loop stability.  
The second was to address the weakness of the
charging/discharging of the resonant capacitor on the
primary side during light output load.

I have traced much of the loop stability trouble to the slow
response of the optocoupler.   A couple of years ago I
placed an extra transistor after the optocoupler.  That
made the optocoupler drive a lower impedance at its
collector.  It also increased the gain of the circuit, allowing
tighter regulation.  But it was the harder voltage tug
on its collector which helped the frequency compensation.

Recently, I decided that there still was not enough frequency
compensation.  I then had to concentrate on the diode side
of the optocoupler.  I simulated the addition of a TL431 but
was not happy with the result.  Instead, I have used a PNP
transistor combined with a zener diode.  An NPN transistor
should also have worked with a rearranging of the
ancillary components, but I have a lot of extra PNPs on
hand.

The old circuit configuration on the diode side provided a
certain asymmetrical application of the frequency
compensation (often referred to as phase compensation and
phase lead).  The asymmetrical action tends to
damp oscillation and provide quick over-voltage protection.
The phase compensation capacitor can charge through the
transistor base but it can discharge only through the upper
voltage divider resistor.  That situation is somewhat like the
one before I added the new components.  Then, the
capacitor charged directly through the optocoupler diode, but
could only discharge through the feedback resistor in
series with the zener diode.

I did simulate the feedback improvement on both the ZVS
and non-ZVS versions. I hope to also get the opportunity to
try it on the real bench-test circuit.  In fact, I plan on adding 
both of the improvements to the bench-test circuit to see
how they really work.

The second main recent improvement to the USMPS
pertains only to the ZVS version.  The problem was that
under light load there was not enough energy stored in the
leakage inductance of the transformer to charge/discharge
the resonant capacitor in parallel with the primary all the
way to each DC voltage rail.  Then the MOSFET had to
finish the task when it turned on, wasting energy as heat.

The best solution is not only to place an inductor in parallel
with the secondary of the power transformer, but to place a
capacitor in series with that extra inductor.  Without
the series capacitor, the inductor draws more energy linearly
when we do not want it to.  The extra energy then only
works to increase the slew rate on the primary side of the
transformer under heavy load when we do not want it.

The extra capacitor in series with the inductor limits the
energy the inductor can store.  What's more, it saves it until
the end of the duty cycle when the primary side
resonant capacitor needs its energy boost!

I have also added two anti-resonant diodes to go along with
that extra secondary-side capacitor as I have likewise done
for the coupling capacitor in series with the primary
transformer winding.

Now, I must admit that I have studied these two major new
ideas only on the LTspice simulator.  I hope to bench test
both of them during the summer of 2007 by adding them.
to the circuit I have been trying different configurations
on.  But already as of now, I am very pleased with the
ZVS USMPS circuit, at least in its theory.

The ZVS addition to the original USMPS has the main
purpose of reducing electromagnetic interference(EMI) production.  A remaining source of such interference
from it now is the diodes.  They produce interference mainly
when their junctions are recovering during current reversal.
To reduce that problem, they still may need
snubbing in some cases.  They don't need extra snubbing if
they are sandwiched between two capacitors as in the case
of the primary coupling capacitor clamping
diodes.   Additional high frequency harmonic generation in
the circuit, as with all square wave transformer drive circuits,
occurs when the drive waveform hits each power rail.  The
slower rise times of the ZVS USMPS reduces this
interference source.

The clamping diodes on the primary side prevent
resonance of the primary winding with the coupling
capacitor.  Besides, without these, the capacitors cannot be
used in current limiting since they will be able to
charge/discharge beyond the power rails.  The current would
be able to gradually increase as the voltage
swings progressively extend further and further beyond the
power rails with each half cycle.  Eventually the voltage
rating of the capacitor would be exceeded, too.  


10-27-06

My bench test circuit became more stable in the feedback
loop when I connected the phase compensation capacitor
to bypass only the feedback resistor instead of both the
feedback resistor and the zener diode series combination.
I was surprised at the improvement and am unsure if it holds
for all circuit implementations.

I still see no real need for active current limiting in the ZVS
USMPS.  The output inductor alone can provide enough
current limiting.  Additional current limiting action can occur
if the transformer primary winding coupling capacitor is
small enough in value.  In that case, shunt diodes need to
clamp capacitor/transformer resonance peaks to the power
supply rails.  If implemented well, as current through the
primary winding increases, the slew rate of the primary
voltage can actually begin to decrease,
as the circuit moves toward zero current switching,
theoretically lowering EMI.  However,
the switching noise of  the extra
diodes may negate this reduction to a certain degree.  
