T
Tim Williams
Getting back to my induction heater, well first I removed the two dead
transistors from last time I had it out, and then I got to work
investigating why they died. The circuit seems just peachy at 130VDC,
putting out a clean 230W into a light load (1uH coil, resonant around
35kHz). But hmm, something I hadn't noticed before, trash on the falling
edge. Let's take a closer look at that, hmm and only on the falling edge
you say?
Hey, the high side gate drive is flying with the output, isn't it. If
there's enough capacitive coupling in the pulse transformer -- and with a
40ns edge, there is -- it could turn on again. Or off, as the case may be.
Ya know, I recall the gate drive has a couple hundred ns propagation
delay...isn't that curiously near the bouncing frequency? Gack...
For pics see ABSE.
Now I'm gonna go put in TTL-style outputs (i.e., emitter-follower-buffered
collector output with diode from emitter to collector) and see if that
helps any. Then maybe complementary emitter followers, although I don't
think the 1.2V class-C offset is quite good enough, and I don't want to
bother biasing the things. Hopefully it's just an impedance issue at the
transformer and nothing worse...
Another option is to intentionally slow down the oscillator outputs,
literally bypassing the HF-trash with a so-called bypass cap. The emitter
followers will still give sharp edges, but I don't need sharp turn-on, I
need sharp turn-off, so maybe I would do that with the collector driven
layout instead.
I could also slow down the output's edge itself by, pffbt, easily a
microsecond, drastically reducing HF-trash. (Which is a good thing anyway,
since my digital panel meter is reading bonkers half the time and I see
little tone bursts on everything.) Old fashioned snubber capacitor stuff.
So how much and how to connect? Well, right now I've got 20A peak current,
so for delta 130V / delta 1us, that's 0.15uF, with a peak rating over 20A
obviously... one of those MKPs from my tank cap would do it. I wonder how
happy the bridge would be switching into that though. I remember having a
hard time getting the desat to switch into even a resistive load, though
that was an earlier implementation. Maybe a series resistor could be
used -- say, 10% drop at peak current, or 13 / 20 = 0.65 ohms. Gee, rather
small, and still a short-circuit current of 200A from this modest supply
voltage...
Oh, and something Terry Given might like to hear: my desat protection DOES
work (at least on the low side): I was prodding around the high side, saw a
blue spark and thought OH SHIT IT'S FUCKED, slowly coming to realize that
1. supply voltage is still there and 2. the reset light is on... hmm turn
off and check continuity, the IGBTs are still good... woohoo!
Tim
transistors from last time I had it out, and then I got to work
investigating why they died. The circuit seems just peachy at 130VDC,
putting out a clean 230W into a light load (1uH coil, resonant around
35kHz). But hmm, something I hadn't noticed before, trash on the falling
edge. Let's take a closer look at that, hmm and only on the falling edge
you say?
Hey, the high side gate drive is flying with the output, isn't it. If
there's enough capacitive coupling in the pulse transformer -- and with a
40ns edge, there is -- it could turn on again. Or off, as the case may be.
Ya know, I recall the gate drive has a couple hundred ns propagation
delay...isn't that curiously near the bouncing frequency? Gack...
For pics see ABSE.
Now I'm gonna go put in TTL-style outputs (i.e., emitter-follower-buffered
collector output with diode from emitter to collector) and see if that
helps any. Then maybe complementary emitter followers, although I don't
think the 1.2V class-C offset is quite good enough, and I don't want to
bother biasing the things. Hopefully it's just an impedance issue at the
transformer and nothing worse...
Another option is to intentionally slow down the oscillator outputs,
literally bypassing the HF-trash with a so-called bypass cap. The emitter
followers will still give sharp edges, but I don't need sharp turn-on, I
need sharp turn-off, so maybe I would do that with the collector driven
layout instead.
I could also slow down the output's edge itself by, pffbt, easily a
microsecond, drastically reducing HF-trash. (Which is a good thing anyway,
since my digital panel meter is reading bonkers half the time and I see
little tone bursts on everything.) Old fashioned snubber capacitor stuff.
So how much and how to connect? Well, right now I've got 20A peak current,
so for delta 130V / delta 1us, that's 0.15uF, with a peak rating over 20A
obviously... one of those MKPs from my tank cap would do it. I wonder how
happy the bridge would be switching into that though. I remember having a
hard time getting the desat to switch into even a resistive load, though
that was an earlier implementation. Maybe a series resistor could be
used -- say, 10% drop at peak current, or 13 / 20 = 0.65 ohms. Gee, rather
small, and still a short-circuit current of 200A from this modest supply
voltage...
Oh, and something Terry Given might like to hear: my desat protection DOES
work (at least on the low side): I was prodding around the high side, saw a
blue spark and thought OH SHIT IT'S FUCKED, slowly coming to realize that
1. supply voltage is still there and 2. the reset light is on... hmm turn
off and check continuity, the IGBTs are still good... woohoo!
Tim
