Maker Pro
Maker Pro

Optoisolators

T

Tim Williams

I'm hooking up the desat detectors on my induction heater. I'm using PV702V
optos for coupling the desat signal to the shutdown latch, both high and low
side, for symmetry's sake. I've got half an LM393 doing the desat detection
itself (what a pretty slope it sees from these MOSFETs ;), which means I can
only pull a few miliamperes for the error signal, in this case a 4.7k
resistor from +18V. I have this ANDed, with a diode, to the gate signal, so
that only when drain is above the desat threshold AND the gate is positive,
a positive signal is sent out. Good so far. But pushing this into the
opto's photodiode directly is really crappy, so I've got an emitter follower
to the diode, with 1k current limiting resistor. That's a maximum of
10-15mA or so into the diode and should provide reasonable response, no?

It seems to work, with the phototransistors paralleled pulling on a 2.2k
resistor. I can't really make a go/no-go decision with MOSFETs though, what
with the Rds(on) characteristic.

But that said, what the hell am I supposed to do with the base on these
optos? And why does the high side waveform tickle it (as near as I can
tell)? Seems rather shitty to me that what little capacitive coupling is in
these things is able to tweak the phototransistor by a whole miliamp or so.

Tim
 
T

Terry Given

Tim said:
I'm hooking up the desat detectors on my induction heater. I'm using PV702V
optos for coupling the desat signal to the shutdown latch, both high and low
side, for symmetry's sake. I've got half an LM393 doing the desat detection
itself (what a pretty slope it sees from these MOSFETs ;), which means I can
only pull a few miliamperes for the error signal, in this case a 4.7k
resistor from +18V. I have this ANDed, with a diode, to the gate signal, so
that only when drain is above the desat threshold AND the gate is positive,
a positive signal is sent out. Good so far. But pushing this into the
opto's photodiode directly is really crappy, so I've got an emitter follower
to the diode, with 1k current limiting resistor. That's a maximum of
10-15mA or so into the diode and should provide reasonable response, no?

It seems to work, with the phototransistors paralleled pulling on a 2.2k
resistor. I can't really make a go/no-go decision with MOSFETs though, what
with the Rds(on) characteristic.

you can, its just a bit more arbitrary. Vds = 75V while Vgs > Vth would,
for example, be a good time to desat.

But that said, what the hell am I supposed to do with the base on these
optos? And why does the high side waveform tickle it (as near as I can
tell)? Seems rather shitty to me that what little capacitive coupling is in
these things is able to tweak the phototransistor by a whole miliamp or so.

Tim

Hi Tim,

sometimes lopping off the base pin can help. Or use an opto with a
built-in faraday shield, eg sfh6345. pretty much any "gate drive" opto
has such a shield.

ultimately it'll be CdV/dt. you can measure dV/dt pretty easily, and
measuring the resultant output current will give you an idea of the
magnitude of the problematic capacitance. and yeah, bunging the current
up the base of the bjt really makes things worse.


Cheers
Terry
 
F

Frithiof Andreas Jensen

It seems to work, with the phototransistors paralleled pulling on a 2.2k
resistor. I can't really make a go/no-go decision with MOSFETs though, what
with the Rds(on) characteristic.

Why Not?? "Too High D-S Voltage" is A Failure whatever the reason, 7-10 V
for a few usec when supposed to be On is good IMO.
But that said, what the hell am I supposed to do with the base on these
optos? And why does the high side waveform tickle it (as near as I can
tell)? Seems rather shitty to me that what little capacitive coupling is in
these things is able to tweak the phototransistor by a whole miliamp or
so.

Ground it with a suitable resistor - the lower the value, the lower the CTR
(and the lower the sensitivity to other things). Between 470k and 470R is
about right ;-) i.e the lowest value that still gives some margin on the
lower-bound CTR from the datasheet is what you want.
 

Similar threads

Top