Transients connecting and disconnecting SMPS from battery

I am finalizing my design for a fairly simple boost converter to convert 12
VDC nominal SLA battery voltage to 20 to 50 VDC at 750 mA to power a series
string of high power white LEDs. I am using a PIC16F684 with PWM at 100
kHz, a 10 uH inductor, and an HUF75645P3 MOSFET as discussed in the 2N3055
thread. This is to be used in an underwater searchlight.

As an extra feature, I want to be able to toggle the power switch to the
battery so as to provide two or more brightness ranges. The PIC can monitor
the raw battery supply, while its own power is kept up by isolating it with
a diode and large capacitor. The raw 12 VDC is stabilized by a 0.47 uF
capacitor.

I was concerned that there might be potentially damaging transients if the
battery switch is opened while the inductor is still transferring energy to
the output circuit. I confirmed this in LTSpice simulation by using a 12
VDC pulse source for the battery. I got very high currents of several
hundred amps. However, the simulator circuit did not have any series
inductance from the battery, so I added 100 nF. This reduced the current
from a sharp spike to a ringing waveform, but I saw a negative voltage
transient of about -10 volts peak at the capacitor. I added a Schottky
diode across it, and the transient was reduced to about -1.5 volts with a
diode current of almost 20 amps for about 1 uSec.

I have additional filtering from the input voltage divider as well as from
the output voltage divider and the output current sampling resistor, with
time constants of 1 mSec or so, and I think that will eliminate any chance
for latch-up of the PIC. Previously, I had left off a capacitor for one of
the A/D inputs, and the PIC got quite hot, probably because of latch-up.
(But it survived!)

This circuit must be as reliable as possible, as it will be encapsulated
and not repairable. Cost is only a minor factor, as the LED assembly is
something like $50. The PCB is about 1" x 2". In simulation, I was able to
get efficiency of 90% to 95%. In bench testing a rough prototype I measured
about 85%, although none of the components got very warm.

It made a big difference when I used a true-RMS AC+DC meter for output
voltage measurement, and it may still not be accurate enough at 100 kHz.
I'm also using E^2/R for output power, so small errors are significant.

I will probably add the series inductor to the raw battery supply from the
switch, and also the Schottky diode. It seems to work best with the 0.47 uF
capacitor directly across the input, then the small inductor, and finally
the diode to ground. The simulator showed capacitor currents of 150 amps
with input pulse rise and fall times of 10 nS, but was a more reasonable 50
amps at 100 nS. There is probably enough external inductance in the battery
and wires, and also enough variation of resistance as the switch contacts
(actually a reed relay) operate, that there should be no problem.

However, I'd like any suggestions or comments before finally committing
this to copper on a first run PCB.

Thanks,

Paul


.

Loading