Phil, Again, you're talking about Drain-Source inrush, or gate-source inrush current? I understand that the gate resistor creates a higher RC time constant, which slows the turn on time of the mosfet, but I just don't understand how the Drain-Source current affects the driver? On the other hand if you are talking about the gate-source current, well, that's another story. The driver is rated for 4A per output, and I have 2 outputs in parallel(as per the datasheet), for a total inrush capacity of 8A. My previous 4.7 Ohm resistor would have limited the gate-source inrush to 12.6V/4.7 = 2.68A. Clearly well below the maximum output.
Either way, I'm still confused.
Just a quicky from the datasheet. For my mosfet, the silicon is rated for 80A continuous, and 320A for a 500us pulse. I haven't had one fail yet due to all this.
Post Edited (Philldapill) : 12/18/2008 5:03:25 AM GMT
Philldapill said...
... I just don't understand how the Drain-Source current affects the driver ...
Frankly, neither do I. But let's deal with the facts, which are:
1. The only kind of load that kills your driver is one with a high inrush current (drain-source).
2. You've demonstrated that by slowing the turn-on time of the MOSFET (via the series R and gate C) that you can eliminate the problem.
3. A slow turn-on also helps to reduce inrush current, albeit at the expense of switching efficiency.
I say get an inrush current limiter for the load and just fix the problem. If it works, then you can deal with the whys. Like jr, above, I suspect the rapid load transient is causing a supply glitch in the driver. It may even be causing latchup. BTW, you say you have lots of bypass caps, which could help with this problem. Where are they? I don't see any in your layout.
-Phil
Addendum: I agree with Beau's recommendation to split the outputs between the two transistors. There's no reason to parallel them. This, in fact, was one of the first things I noticed, but the datsheet says it's okay (despite having bipolors in its output stage), so I didn't flag it. It's certainly worth a try and easy enough to do with a Dremel tool and some wire and solder. (And now, Beau's post has disappeared. Beau, did you change your mind?))
Beau, I feared the same thing about the outputs being put together. However, I was trying this same thing with a 9A, single output version of the chip, originally. That is where the problem arose. I had about 10 sample DIP chips of the UCC27322, and burned(literally) through all of them trying to fix this. Once I ran out of those, I made a new board, using the UCC27324, which is the 4A, dual output version that is in the PCB schematic I attached earlier. The chip may be shorting the outputs, but two points about this: A. the datasheet explicity states a few times that this method of paralleling outputs for higher peak current is perfectly acceptable, and B. The problem arose from the single output version to start with.
Phil, Thanks for clarifying which inrush current. I guess I have some research to do as to WHY the drain-source inrush current causes the driver (which should be insulated by the gate), to fail. I may have to look into those current limiters you suggested, and like you said, be done with it. I've spent 2 solid days tinkering with this stupid thing and I need to be back on schedule of getting this whole thing working.
Beau, you may know something about this, which I've been turning over in my head recently...
I'm thinking that one possible cause of this whole driver destroyer, as I like to call it, is inadequate bypass capacitors. My reasoning goes something like this. Let's say the driver turns on, turning on the mosfet, which creates a fair voltage drop around my driver IC. This causes a negative dV/dt which in turn screws with the logic inside the driver. This jitter in the logic states somehow causes something to go screwy with deadtime management that the chip datasheet keeps boasting about, which makes the high and lowside to both be on simultaneously. This causes a direct short and a huge current to flow, destroying the chip catastrophically. I've noticed with each burst, it's pretty much instantaneous when I connect the power, and there is virtually NO heat from the destroyed chip. This backs up my idea about a direct short through the transistors, causing a rapid expansion that blows apart the silicon, extinguishing the short before it really has any time to generate any heat. Eh?
EDIT: Nice Beau. Thanks for deleting your post. Now I look like crazy, responding to nothing. Thanks alot. LOL
Philldapill, · I think you nailed the answer to your own question in your last post.· The sudden voltage drop is basically causing the UCC27323 to mismanage the outputs. · This was actually a common design problem I experienced working in prosthetics, where you had a·single battery power supply delivering power to a motor in a prosthetic hand, wrist, arm, or elbow and still maintain enough power to the processor. · A couple of effective solutions were to implement an inductor in series with the load as has been mentioned... another and/or approach is to use a simple diode-capacitor filter on ONLY the power supply·to the micro processor and other required IC's that needed to stay a float during the surge. · PS... I deleted my previous post·only because I wanted to think about my response a little bit more... you were just too quick for me.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ Beau Schwabe
IC Layout Engineer
Parallax, Inc.
(1) The FETs are connected correctly.· So is everything else on the PC board.
(2)· You don't tell us how the headlights are connected.· Do they go directly to the battery?· How about the PC board +12v connection.· Does it go the same place as the +headlights?· If so, does it have conductors separate from the headlight wires?
The problem seems to occur only under unexpectedly high current drain, like cold incandescent headlamps, which may draw hundreds of amps until the filaments warm up.
Let's suppose that the conductor from the GND of the PC board to the battery·is too small.· Let's suppose, also, that the isolator has a much better path to ground.
When you turn the lamp on, it connects PC board GND directly to +12 through an extremely small resistance (the cold lamps and the FETs in series).· But perhaps there's too much resistance from PC board GND to the negative battery.· The PC board GND will go to +12 volts!· The drop between the PC board GND, at +12 volts, and the actual battery negative, will all be across the little narrow trace on the PC board.· 2/3 of that trace is between the driver GND and the egulator GND, so perhaps the driver's GND pin will go to +8 or so, very briefly.· Who knows what's happening to the driver +12?· We don't know, because we don't know how the headlamps are connected.· Does their current go through the same path that the driver's +12 goes through?
This is all conjecture and brainstorming, but I'd look at what is happening to all your +12 points and ground points with respect to battery negative.· Since the PC board is carrying all of the lamp current through that one tiny trace, and we don't know what currents are common in other parts of the circuit (parts, including wires,·not on the PC board), the voltages may be momentarily all over the place.· Have a look, and report back.
And oh, by the way, the Prop is not killing anything.
I wish I had that luxury of a digital scope that I could save images from, but I only have an old 20Mhz analog scope. I can tell you what the output looks like with various loads. With the headlight load I have now, the output is a nice square wave, with about a -2V small spike on the high-low transition, and a 2V small spike on the low-high transition. Nothing extraordinairy that I can see. However, even on the 1000uF capacitor, I can clearly see the dip in voltage when the mosfet transitions. I'd say it drops by about 0.2V, and there is a spike on the low-high transition, which is actually the high-low transition of the mosfet, but the capacitor is the inverse of it. Not a big spike, but a clear one.
Well this is interesting... There is spike about 3 times the magnitude of the voltage(12V) across the mosfet drain-source when it turns off... I'd say that spike is nearly 40V in magnitude! Any thoughts? BTW, I've now changed the gate resistor to 220Ohm and it's holding steady...
Carl, that's some good thinking. I think I may be having a "well duh..." moment - excuse me for a second.
You're right, the propeller isn't killing it, obviously. I guess that title is a little misleading, eh?
On my diagram, that big green circle is actually a 4AWG solid wire I have soldered in as a post to connect my positive to. Again, this is all quick and dirty since I don't want to design an elaborate PCB until I get this out of the way. However, it seems this whole quick and dirty aspect may be the problem all in itself. I'll check this stuff in the morning. For now, I'm going to bed. Thanks again, everyone, for the input and insight.
We all tend to think of wires as circuit nodes sometimes.· But they have resistance and inductance and capacitance, and sometimes they're connected with corroded screws or cold solder joints, and sometimes they're just too skinny.· Unlike girls, who cannot be too thin, too rich, or too young, wires can be too thin.
Your +12 is 4 AWG, which ought to be thick enough unless it's really long.· How big is the ground wire?
Anyway, scope all that stuff (a 20MHz scope ought to be fast enough), making sure that the scope ground is connected directly to the negative terminal of the battery so you can see all wiring drops.· You're looking for what happens at all the supposed GND points, and all the supposed +12 points, everywhere on the PCB and off.
Well, Ale, I like to make non-functional circuits. It's a hobby of mine. No, actually, I realized what I did AFTER I made the board. Not a problem though. I just attached a little jumper wire from the 12V supply, to the regulator leg.
Are you saying that the circuit is working now? What did you do?
I can't see were you would be getting 40V spikes unless you have an inductive load. That spike would be coupled back through the drain-gate capacitance into the driver so it's always a good idea to at least allow for a resistor while testing, there is nothing stopping you then from changing the value to anything you like even zero.
You must have an inductive load in those headlights for some reason, perhaps to slow the inrush current. You know if you have inductance in the circuit if you connect +12V to the headlights and hold the terminals with your bare hands when you disconnect the +12V. This is the "bush" engineer test procedure to check for inductive kick (or the poor substitute for missing a morning coffee!).
I would also try disconnecting one of the driver outputs just to make sure there is nothing peculiar about the way the driver turns the outputs on and off.
If you really want to drive the load hard with minimal resistance then you should place a clamp diode on the gate. The driver is rated for 16V absolute max so I would use a 15V fast zener (tranzorb) before a 2R2 resistor to the gate. The other way to do this is to use a schottky diode from the drain to the +12V.
This is all based on the assumption that there is inductance somewhere even though there shouldn't be but the circuit should be designed to handle inductive loads anyway. Nothing to lose.
I was thinking inductance, too. But I don't think he's using any (yet) to quell the inrush current. My thinking was that a short-duration pulse of 40V may be normal, just with the wire lead and filament-coil inductance into a nearly infinite-impedance load like a scope probe. But I doubt there's much energy in it, so I wasn't too concerned about it.
You're right, though, it may be coupling back through the gate via capacitance. Assuming it's a low-energy pulse, a resistor from the gate to ground might be adequate to kill its effects without compromising the turn-on/off speed.
I'm still putting my money on power supply transients from the inrush current causing a latch-up in the driver.
Ok, still, no luck. My original circuit works just fine, still. My new incarnations still don't work. Here is my latest circuit. I am back to using the UCC27322, 9A, single output version of the driver chip from TI. Datasheet link attached.
I am still getting the chip destroyed when I supply power to the circuit. I am still using the same mosfets, regulator, isolator, but now the UCC27322. Any thoughts? Please, for the love of God, someone come up with the magic solution. This makes absolutely no sense that one circuit works, the other doesn't.
My money is on the headlights. You said that a resistive load works fine. Having inductors integrated into the headlight assembly makes sense to limit the inrush current. Put a diode across the headlight and see if there are any more problems.
I'm with Peter on this. The inrush current is the source of your trouble, so get rid of it. Use inductors or the inrush current limiters I mentioned above. (If you use inductors, the diode is a critical addition.)
I think I can rule inductance out. I've tried this on a DC motor... A big one at that, drawing about 2A, and it runs ok (sometimes). The headlight blows it every time unless I have at least 300 Ohms between gate and driver. I did, however, try it with the diode across the load, with no luck. POP... again. 9 more tries left!
How about this... Can anyone recommend a driver? Or maybe even show a GOOD schematic of how to place the driver and components?
What we're suggesting (and have been suggesting for several posts now) is to ADD inductance in series with the lamps. This will help to limit the inrush current. More inductance is not your problem in this case; it's your friend.
Ummm Phil, I was actually suggesting the reverse , that there is a buried inductor in the headlamp assembly or even the lamp itself as it makes sense for two reasons:
1) Limit cold filament inrush current to preserve the life of the bulb
2) That there is a 40V spike present that can kill the driver and 40V like that can only come from a fair measure of inductance
@Philldapill
Also, your diode needs to be fast so use a schottky. Have you tried using just one output, you have got to try this as to you really need to rule out that it's not a weird reaction by the driver to some spike on it's outputs
Brownouts and spikes are the norm on such a supply. Always assume your supply is somewhere between 0 volts and 55 volts in this particular setup. You need to manage this full range of variation between the battery and the driver chip.
Philldapill, · Something that strikes me a little odd... It seems as though from the PDF that you posted that Ti is aware of a gate current driving issue in dealing with the Miller effect, but in my opinion don't document it very clearly.· From the PDF that you posted earlier... · hhttp://forums.parallax.com/attachment.php?attachmentid=57436· · ... Starting on page #7 "source/sink capabilities during miller plateau"·and also page #8 "operational circuit layout" continuing through page #10 · Have you E-mailed Ti with the problems that you are observing·to see what their recommendations are?· There may be other suggestions that they can provide that aren't necessarily in their datasheet. ·
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ Beau Schwabe
IC Layout Engineer
Parallax, Inc.
Peter Jakacki said...
Ummm Phil, I was actually suggesting the reverse, that there is a buried inductor in the headlamp assembly or even the lamp itself as it makes sense for two reasons:
1) Limit cold filament inrush current to preserve the life of the bulb
2) That there is a 40V spike present that can kill the driver and 40V like that can only come from a fair measure of inductance
I think we're in agreement that there needs to be inductance (or some other inrush limiter) present, along with a protection diode. It's just a matter of how much is already there and whether it's causing the trouble. I guess I'm not convinced yet that the observed 40V spike suggests enough inductance to be causing the problem. If the spike, observed with high impedance scope probes, results from the very low inductance inherent in the wiriing and filament itself, it shouldn't pack enough energy to be an issue. In the transistor he's using, the drain-gate capacitance is only 25pF which, in itself presents a rather high-impedance path for the spike to reach the gate. But I could be wrong about the amount of inductance present, and testing with a reverse diode across the load should help to sort things out.
What do you guys think about adding a 12V regulator(fast transient response type) to supply the power to the UCC27322 chip? I've looked at the voltage across the bypass capacitors, and it DOES dip a little bit. I have a 1000uF electrolytic, 10uF tantalum, and a 0.1uF ceramic. My thinking is, the regulator would maybe cut out some of those transients if they are causing the problem...
Another question that I'm hoping someone can answer - Why does the large inrush current cause driver problems even when the driver has a diode from the +12V to it's input rails and bypass capacitors? With this setup, there won't be much of any voltage drop because the diode blocks any charge from leaving the capacitors.
Ok, I tried a little something today. Phil, you've been been the proponent of adding an inductor or some means of limiting inrush current. The theory is, when the headlight is first cold and I try to turn it on with the mosfet/driver, there is a HUGE inrush current and somehow that is frying my driver. Once the headlight warms up, the current goes to a fairly steady state value of around 6A @ 100% duty cycle. If that's the case, then if I let the headlight warm up, I should be able to remove the resistor between the gate and driver because no large inrush would occur. Apparently not.
I tried doing this. I have a 220 Ohm resistor between mosfet and driver. My setup is able to sense the current through the load, and my program adjusts the duty cycle so that a certain current, which I define in the program, goes through the load. Essentially a constant current source using an automatically adjustable PWM signal. Anyway, I had the headlight pulling exactly 5A, it was definately warmed up, so I shorted the resistor between gate and driver, and POP. That same familiar sound and smell.
Anymore theories? I'm literally going insane and losing my hair at an astonishing rate.
That's a pretty conclusive experiment. So it seems that theory's been put to rest for now. The TI docs tout an "industry-standard pinout". Can you find an equivalent driver from another manufacturer to compare with? That would confirm or rule out any anomalies within the TI component. Also, it wouldn't hurt to add a fast protection diode to kill the observed voltage spike and confirm or rule out any effects from it.
You also mention that the driver's supply dips a little bit. How much? If the MOSFET's gate capacitance is fully-charged when this happens, I wonder if there could be a reverse current draining the gate back through the driver's output stage somehow. This is wild speculation with no clue as to the failure mode, BTW.
Thanks Phil. I've been at a loss for days now. Welcome to the club.
That's an interesting theory about the reverse current flowing from gate BACK THROUGH the internal FETs in the driver. I have no real way of seeing exactly how much the voltage drops, but I'd say it drops MAYBE 0.25V, 0.5V maximum.
Again, I reiterate... The driver works just great with no resistor between gate and driver WITHOUT a load. It works fine without a resistor WITH my large Stepper motor pulling 5A. It works fine with a 1 Ohm resistor as a load without a resistor between gate and driver. It FAILS EVERYTIME without a resistor between gate and driver, and with the headlight attached as a load.
Basically, if I don't use a headlight, all is well. However, I want this to be as versitle as possible and I want to get to the bottom of this already non-propeller issue. [noparse]:)[/noparse]
Comments
Either way, I'm still confused.
Just a quicky from the datasheet. For my mosfet, the silicon is rated for 80A continuous, and 320A for a 500us pulse. I haven't had one fail yet due to all this.
Post Edited (Philldapill) : 12/18/2008 5:03:25 AM GMT
I haven't seen anyone mention spiking from the output FETS causing
the supply to contain some serious peaks, taking out the driver.
I've had designs that have easily seen 60V spikes on 12 Volt supplies.
Certain brand FET's (Moto Tmos) were more susceptable than others (Siliconix Vmos)
Suggest running the project at 30% of the supply and looking for spikes
that can be snubbed with appropriate external circuitry.
It may be enough to isolate the spikes from the driver supply.
Send us the print and data sheet links, we're flying blind.
jr
1. The only kind of load that kills your driver is one with a high inrush current (drain-source).
2. You've demonstrated that by slowing the turn-on time of the MOSFET (via the series R and gate C) that you can eliminate the problem.
3. A slow turn-on also helps to reduce inrush current, albeit at the expense of switching efficiency.
I say get an inrush current limiter for the load and just fix the problem. If it works, then you can deal with the whys. Like jr, above, I suspect the rapid load transient is causing a supply glitch in the driver. It may even be causing latchup. BTW, you say you have lots of bypass caps, which could help with this problem. Where are they? I don't see any in your layout.
-Phil
Addendum: I agree with Beau's recommendation to split the outputs between the two transistors. There's no reason to parallel them. This, in fact, was one of the first things I noticed, but the datsheet says it's okay (despite having bipolors in its output stage), so I didn't flag it. It's certainly worth a try and easy enough to do with a Dremel tool and some wire and solder. (And now, Beau's post has disappeared. Beau, did you change your mind?))
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Just a few PropSTICK Kit bare PCBs left!
Post Edited (Phil Pilgrim (PhiPi)) : 12/18/2008 5:38:57 AM GMT
Phil, Thanks for clarifying which inrush current. I guess I have some research to do as to WHY the drain-source inrush current causes the driver (which should be insulated by the gate), to fail. I may have to look into those current limiters you suggested, and like you said, be done with it. I've spent 2 solid days tinkering with this stupid thing and I need to be back on schedule of getting this whole thing working.
I'm thinking that one possible cause of this whole driver destroyer, as I like to call it, is inadequate bypass capacitors. My reasoning goes something like this. Let's say the driver turns on, turning on the mosfet, which creates a fair voltage drop around my driver IC. This causes a negative dV/dt which in turn screws with the logic inside the driver. This jitter in the logic states somehow causes something to go screwy with deadtime management that the chip datasheet keeps boasting about, which makes the high and lowside to both be on simultaneously. This causes a direct short and a huge current to flow, destroying the chip catastrophically. I've noticed with each burst, it's pretty much instantaneous when I connect the power, and there is virtually NO heat from the destroyed chip. This backs up my idea about a direct short through the transistors, causing a rapid expansion that blows apart the silicon, extinguishing the short before it really has any time to generate any heat. Eh?
EDIT: Nice Beau. Thanks for deleting your post. Now I look like crazy, responding to nothing. Thanks alot. LOL
Even though on page 8 of the document, it says that the outputs can be tied together, I don't think that it is a good design practice to do so.
Can you scope the output of the mosfet driver with various loads and show us what that looks like?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe
IC Layout Engineer
Parallax, Inc.
·
I think you nailed the answer to your own question in your last post.· The sudden voltage drop is basically causing the UCC27323 to mismanage the outputs.
·
This was actually a common design problem I experienced working in prosthetics, where you had a·single battery power supply delivering power to a motor in a prosthetic hand, wrist, arm, or elbow and still maintain enough power to the processor.
·
A couple of effective solutions were to implement an inductor in series with the load as has been mentioned... another and/or approach is to use a simple diode-capacitor filter on ONLY the power supply·to the micro processor and other required IC's that needed to stay a float during the surge.
·
PS... I deleted my previous post·only because I wanted to think about my response a little bit more... you were just too quick for me.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe
IC Layout Engineer
Parallax, Inc.
(1) The FETs are connected correctly.· So is everything else on the PC board.
(2)· You don't tell us how the headlights are connected.· Do they go directly to the battery?· How about the PC board +12v connection.· Does it go the same place as the +headlights?· If so, does it have conductors separate from the headlight wires?
The problem seems to occur only under unexpectedly high current drain, like cold incandescent headlamps, which may draw hundreds of amps until the filaments warm up.
Let's suppose that the conductor from the GND of the PC board to the battery·is too small.· Let's suppose, also, that the isolator has a much better path to ground.
When you turn the lamp on, it connects PC board GND directly to +12 through an extremely small resistance (the cold lamps and the FETs in series).· But perhaps there's too much resistance from PC board GND to the negative battery.· The PC board GND will go to +12 volts!· The drop between the PC board GND, at +12 volts, and the actual battery negative, will all be across the little narrow trace on the PC board.· 2/3 of that trace is between the driver GND and the egulator GND, so perhaps the driver's GND pin will go to +8 or so, very briefly.· Who knows what's happening to the driver +12?· We don't know, because we don't know how the headlamps are connected.· Does their current go through the same path that the driver's +12 goes through?
This is all conjecture and brainstorming, but I'd look at what is happening to all your +12 points and ground points with respect to battery negative.· Since the PC board is carrying all of the lamp current through that one tiny trace, and we don't know what currents are common in other parts of the circuit (parts, including wires,·not on the PC board), the voltages may be momentarily all over the place.· Have a look, and report back.
And oh, by the way, the Prop is not killing anything.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
· -- Carl, nn5i@arrl.net
I wish I had that luxury of a digital scope that I could save images from, but I only have an old 20Mhz analog scope. I can tell you what the output looks like with various loads. With the headlight load I have now, the output is a nice square wave, with about a -2V small spike on the high-low transition, and a 2V small spike on the low-high transition. Nothing extraordinairy that I can see. However, even on the 1000uF capacitor, I can clearly see the dip in voltage when the mosfet transitions. I'd say it drops by about 0.2V, and there is a spike on the low-high transition, which is actually the high-low transition of the mosfet, but the capacitor is the inverse of it. Not a big spike, but a clear one.
Well this is interesting... There is spike about 3 times the magnitude of the voltage(12V) across the mosfet drain-source when it turns off... I'd say that spike is nearly 40V in magnitude! Any thoughts? BTW, I've now changed the gate resistor to 220Ohm and it's holding steady...
You're right, the propeller isn't killing it, obviously. I guess that title is a little misleading, eh?
On my diagram, that big green circle is actually a 4AWG solid wire I have soldered in as a post to connect my positive to. Again, this is all quick and dirty since I don't want to design an elaborate PCB until I get this out of the way. However, it seems this whole quick and dirty aspect may be the problem all in itself. I'll check this stuff in the morning. For now, I'm going to bed. Thanks again, everyone, for the input and insight.
Your +12 is 4 AWG, which ought to be thick enough unless it's really long.· How big is the ground wire?
Anyway, scope all that stuff (a 20MHz scope ought to be fast enough), making sure that the scope ground is connected directly to the negative terminal of the battery so you can see all wiring drops.· You're looking for what happens at all the supposed GND points, and all the supposed +12 points, everywhere on the PCB and off.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
· -- Carl, nn5i@arrl.net
Are you saying that the circuit is working now? What did you do?
I can't see were you would be getting 40V spikes unless you have an inductive load. That spike would be coupled back through the drain-gate capacitance into the driver so it's always a good idea to at least allow for a resistor while testing, there is nothing stopping you then from changing the value to anything you like even zero.
You must have an inductive load in those headlights for some reason, perhaps to slow the inrush current. You know if you have inductance in the circuit if you connect +12V to the headlights and hold the terminals with your bare hands when you disconnect the +12V. This is the "bush" engineer test procedure to check for inductive kick (or the poor substitute for missing a morning coffee!).
I would also try disconnecting one of the driver outputs just to make sure there is nothing peculiar about the way the driver turns the outputs on and off.
If you really want to drive the load hard with minimal resistance then you should place a clamp diode on the gate. The driver is rated for 16V absolute max so I would use a 15V fast zener (tranzorb) before a 2R2 resistor to the gate. The other way to do this is to use a schottky diode from the drain to the +12V.
This is all based on the assumption that there is inductance somewhere even though there shouldn't be but the circuit should be designed to handle inductive loads anyway. Nothing to lose.
*Peter*
I was thinking inductance, too. But I don't think he's using any (yet) to quell the inrush current. My thinking was that a short-duration pulse of 40V may be normal, just with the wire lead and filament-coil inductance into a nearly infinite-impedance load like a scope probe. But I doubt there's much energy in it, so I wasn't too concerned about it.
You're right, though, it may be coupling back through the gate via capacitance. Assuming it's a low-energy pulse, a resistor from the gate to ground might be adequate to kill its effects without compromising the turn-on/off speed.
I'm still putting my money on power supply transients from the inrush current causing a latch-up in the driver.
-Phil
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Just a few PropSTICK Kit bare PCBs left!
I am still getting the chip destroyed when I supply power to the circuit. I am still using the same mosfets, regulator, isolator, but now the UCC27322. Any thoughts? Please, for the love of God, someone come up with the magic solution. This makes absolutely no sense that one circuit works, the other doesn't.
*Peter*
-Phil
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Just a few PropSTICK Kit bare PCBs left!
How about this... Can anyone recommend a driver? Or maybe even show a GOOD schematic of how to place the driver and components?
-Phil
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Just a few PropSTICK Kit bare PCBs left!
Post Edited (Phil Pilgrim (PhiPi)) : 12/19/2008 5:27:38 AM GMT
1) Limit cold filament inrush current to preserve the life of the bulb
2) That there is a 40V spike present that can kill the driver and 40V like that can only come from a fair measure of inductance
@Philldapill
Also, your diode needs to be fast so use a schottky. Have you tried using just one output, you have got to try this as to you really need to rule out that it's not a weird reaction by the driver to some spike on it's outputs
*Peter*
Leon
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Amateur radio callsign: G1HSM
Suzuki SV1000S motorcycle
Beau has good advice. Read it carefully.
·
Something that strikes me a little odd... It seems as though from the PDF that you posted that Ti is aware of a gate current driving issue in dealing with the Miller effect, but in my opinion don't document it very clearly.· From the PDF that you posted earlier...
·
hhttp://forums.parallax.com/attachment.php?attachmentid=57436·
·
... Starting on page #7 "source/sink capabilities during miller plateau"·and also page #8 "operational circuit layout" continuing through page #10
·
Have you E-mailed Ti with the problems that you are observing·to see what their recommendations are?· There may be other suggestions that they can provide that aren't necessarily in their datasheet.
·
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe
IC Layout Engineer
Parallax, Inc.
-Phil
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Just a few PropSTICK Kit bare PCBs left!
I tried doing this. I have a 220 Ohm resistor between mosfet and driver. My setup is able to sense the current through the load, and my program adjusts the duty cycle so that a certain current, which I define in the program, goes through the load. Essentially a constant current source using an automatically adjustable PWM signal. Anyway, I had the headlight pulling exactly 5A, it was definately warmed up, so I shorted the resistor between gate and driver, and POP. That same familiar sound and smell.
Anymore theories? I'm literally going insane and losing my hair at an astonishing rate.
You also mention that the driver's supply dips a little bit. How much? If the MOSFET's gate capacitance is fully-charged when this happens, I wonder if there could be a reverse current draining the gate back through the driver's output stage somehow. This is wild speculation with no clue as to the failure mode, BTW.
Other than those suggestions, I'm at a loss...
-Phil
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Just a few PropSTICK Kit bare PCBs left!
Post Edited (Phil Pilgrim (PhiPi)) : 12/20/2008 8:57:10 PM GMT
That's an interesting theory about the reverse current flowing from gate BACK THROUGH the internal FETs in the driver. I have no real way of seeing exactly how much the voltage drops, but I'd say it drops MAYBE 0.25V, 0.5V maximum.
Again, I reiterate... The driver works just great with no resistor between gate and driver WITHOUT a load. It works fine without a resistor WITH my large Stepper motor pulling 5A. It works fine with a 1 Ohm resistor as a load without a resistor between gate and driver. It FAILS EVERYTIME without a resistor between gate and driver, and with the headlight attached as a load.
Basically, if I don't use a headlight, all is well. However, I want this to be as versitle as possible and I want to get to the bottom of this already non-propeller issue. [noparse]:)[/noparse]