Top waveform is the ripple voltage (not absolute; 10x scale!) on the half-bridge's supply, which is about 1000uF of electrolytics behind some wire, plus 6 x 0.47uF film caps at the bridge. Pic here:
formatting link
Except I have split the electrolytic cap bank in two, with 250uH between. (The big bank of thick caps in the middle is the coupling capacitor.)
Well anyway, the bottom channel is centered at 0V = 1 div from bottom. That means the square wave shown is about 20V above ground at the lowest.
You may recall that my topology involves inductance. This requires that voltage drop below 0V, so the MOSFET body diodes can conduct the flyback current.
So, um, what the f*ck am I seeing exactly?! MOSFETs cannot conduct negative forward current!
Also, dissipation is nowhere near 20V times whatever current is flowing (I'll check current next, I'm winding a current transformer). At maximum power output (maybe 3-4A average draw at 200V = 600-800W), the transistors get luke warm in a minute or two (after which the (uncooled) work coil is dripping solder and the load is glowing yellow hot).
As for the positive half of the waveform, that shows the same signs -- about
20V shy of the rail. I don't have two 10x probes so I can't show them simultaneously.
Tim
-- Deep Fryer: a very philosophical monk. Website:
Welcome to the world of measurements. You're measuring voltages, not currents. And voltages have a reference point, which, when positive, make others appear negative that still are positive.
If you're trying to tell me that voltage is only a difference, you're ten years too late. I'm measuring with respect to the half bridge negative (ground) terminal.
Even if my absolute potential were off, that can't explain how the positive half also appears to be below the rail.
Tim
-- Deep Fryer: a very philosophical monk. Website:
a) That's quite a hunk of aluminum the MOSFETs (??) are mounted on. It won't get hot right away. b) Careful -- if the work coil's that hot, those film capacitors soldered to it will melt !
More facts would help, e.g. 1) was the coil loaded when the pics were taken? (I assume 'yes'.) 2) I assume your oscillogram (1st link) is showing an H-bridge output voltage, but you didn't say. 3) schematic ? You mention MOSFETs, but this diagram shows IGBTs
formatting link
??)
Absent more data, I'd assume the H-bridge transistors are dropping 20 volts when 'on,' just like the 'scope says they are.
Don't understand the above...doesn't the 50v/div trace show the positive half-cycle too? It sure looks like it does.
Suggestion: tack a ~0.1 ohm resistor in there and measure the pull-down transistor current.
Also -- was the 'scope probe as pictured for the measurements? You may have to move it closer to the transistors, remove the extension/clip thingie, and reduce its ground-lead length to get a good picture of what's really happening ... there are a number of inductances, high currents, high magnetic fields, and large-area loops in your gadget that might produce false signals in the 'scope probe.
Naah...you've got a fine 'scope, and the DMMs won't help you here.
The cool thing about electronics is that there *are* objective answers. My motto: Why wonder when you can measure? Cobble together a non-inductive 0.1 ohm resistor, tack it in, and find out what's really happening.
True, but even if the MOSFETs were dissipating say 100W each- I'd know, the dies would be getting warm. They're mounted on insulating pads, which I don't know for sure but I suspect they add quite a bit of thermal resistance.
Nah, they'd fall off first ;)
Actually the caps stay pretty cool from all the open space, but whatever the case, rest assured I don't feel like melting $40 of caps, they'll be fine.
Ya. It was near resonance (maybe, 1/4 power? WAG).
Half bridge. Bottom trace is output voltage, top trace (not sure I made this clear) is the AC component of the supply voltage to it, at 10x scale. Negative side of the bridge is grounded.
Ya, the current circuit is in flux so it's kind of pointless to draw you a schematic; the current setup is essentially as shown though, except:
- Power supply is 100VAC (MOT "iso"lation transformer) to a doubler, yielding 200VDC or so (260V O/C; haven't measured regulation precisely);
- Oscillator has an R-S latch for shutdown;
- Control circuit is open-loop, by potentiometer only;
- Output transistor drivers (high and low side) essentially as shown, minus desat detectors.
But if they were, they'd be running at best what, 70% efficiency? 20V * 20A is a generous instantaneous power dissipation.
Also, the voltage drop is almost completely constant -- as I would expect from the relatively low Rds(on) transistors (0.8 ohms max; 0.25 ohms for three paralleled). If it were actually 20V, it would be in the 25A range
*per transistor*, or 75A total. Since it doesn't change, it would reflect a resistive load for most all frequencies, clearly impossible given the honking coil in series with the output.
about
Ya,...I mean the top supply rail. Ideally I would have three traces showing
+V, output and GND simultaneously.
Last night I looped a ferrite core with 75 turns 28AWG on it, with a 0.56 ohm (probably inductive, sigh) shunt on the secondary. It shows a reasonably straight slope from some negative value of current to some positive value, plus a fuckload of hash when the diode slams on. When the opposing transistors are doing their job, the current waveform is flat, as you would expect. I wasn't able to test it very well because 1. I got tired, being 5am, and 2. it did some strange oscillation sounding thing in certain situations.
I really need another scope or three and about five DMMs...
Tim
-- Deep Fryer: a very philosophical monk. Website:
The U-U inductor concerned me too, both for radiated magnetic flux, and ... call me queasy, but that sure looks like a lot of (turns * amps) on that core ... might it be saturating? Measurements will tell.
If it weren't compensated, it would show a completely different error (mainly rise and fall times). It looks just peachy comparing to my Wavetek
193 generator's squarewave output delivered via 50 ohm BNC cable, minus what a 50ns rise/fall time does to a few inches of non-50 ohm wiring inherent in a probe.
Tim
-- Deep Fryer: a very philosophical monk. Website:
Well, the main circuit is on something of a backplane, hard to tell if it's aluminum or steel or what. The rest would be kind of inconvienient to build on plate, what with all the exposed wires. ;) Even so, I have the large signal and small signal circuits reasonably seperate.
Huh? That thing's doing an HF filter now.
...Oh, you must've missed the part where I said "Not as pictured". Fine...
formatting link
The current transformer is 75T 28AWG on a black ferrite toroid (probably a nice high mu). The resistor is 0.6 ohm stainless steel wire (7") wound on a copper tube core. Signal is going to scope channel 1.
Ya. That's positive reverse current, not negative forward current. It's notoriously hard to get -20A out of a device with +20V apparent terminal voltage.
And it's obvious that said diodes are working, because the circuit works period. The flyback energy of the matching inductor has to go somewhere.
Update: since I added the current transformer (negligible inductance??), I don't know what the f*ck is going on: it's oscillating with period as low as
10Hz or so. The sound is between crunchy, buzzing and hissing depending on duty cycle, frequency and the phase of Uranus. It appears to be UVLO on the high side kicking in, which means the UC3842 isn't doing its job for some damned reason, and also that peaks are getting through the gate drive despite the signal then being clamped below the turn-on threshold..WTF...
Tim
-- Deep Fryer: a very philosophical monk. Website:
No pun intended but the whole rig-up looks like crosstalk city. If you'd build it on "islands" glued to a large chunk of copper clad things might come out a lot better.
The loop formed by the scope probe and its long ground clip alone will pick up stuff from the U-U core inductor next to it and from other high current wires.
Yabbut, the grounds aren't connected to that backplane via short distance or probably not at all. Then the plane won't do much good.
But then it will spew around some of the stuff it's supposed to filter out. And some of that can get picked up.
The scope ground is even longer than I thought. Not a good thing in this scenario. At least try to run it in parallel to the probe tip, tied with a rubber band or something. Avoid running the probe cable clear across the control stuff.
High mu could mean it's saturating, as James was hinting.
Ahm, I don't understand. Current only has two possible directions.
Could be 'motor-boating', often caused by loop instability because of a heavy capacitive load.
"Tim Williams" a écrit dans le message de news:YBFqf.257$ snipped-for-privacy@fe02.lga...
Nah. Think a bit more to it. Your time scale is 2us/div. Probes (low) freq compensation works on a much lower frequency range. At high frequencies the capacitive divider will dominate the attenuation and on undercompensated probes you'll see a mostly correct but attenuated ac signal riding on the right DC component (given by the resistive divider).
Did you check your probe? That's one of the most frequently overlooked error in scope measurements. Do that at _low_ frequency, i.e. 1kHz or so.
It looks just peachy comparing to my Wavetek
what
in
Low frequency probes undercompensation has no effects on the rising edges you can see at high frequencies/high sweep rates, except amplitude errors.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.