I have been following the discussion here and I don't know much about MOSFETs and such to add to that part of the conversation. I was wondering if perhaps some voltage was shorting or going astray in your setup when you use the headlight that you don't get when you use the other equipment. Maybe setting it on the bench in such a way that it is grounded? Maybe something is wrong with the headlight itself? Graping at straws here.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Timothy D. Swieter, E.I. www.brilldea.com - Prop Blade, LED Painter, RGB LEDs, uOLED-IOC, eProto for SunSPOT, BitScope www.tdswieter.com
Well, slightly good news. All my other loads work just great with my circuit. My big fat stepper motor is pulling exactly the amount of current I specified in my code; the 0.1 Ohm resistor was pulling the proper current until I blew it up; and last, but not least, my other low resistance resistor bank(0.5 Ohm, 50W) was pulling a big current of 10A... All through one of these tiny mosfets switching at 50kHz. The catch diode got a little smokey, but everything worked just fine - except the headlight which is now in the trash, never to be heard from again.
Yes, for some reason, it is. I've put it in series with other loads, and it works fine. By itself(no resistor between gate and driver), and the driver blows without fail.
Leon, as Phil said, that was a the leading theory for a while, but after doing a little expirement with a near deadshort(0.1 Ohm resistor), I'm pretty confident that the huge inrush current that we thought was causing the failure, isn't that bad. I tested the resistance of the cold lamp, and it's about 0.4 Ohms. It still COULD be the problem, but I'm moving on and hoping that the people I sell this device to, never connect it to a headlamp [noparse]:)[/noparse]
Every other load works great. Even large inductive loads, with a diode of course. As of now, the headlamp problem is still a mystery, but I'm not to concerned about it anymore. If anyone has a new theory, you're more than welcome to re-open this thread to discussion. I'd love to get to the bottom of it for curiousity's sake.
You never did say what type of headlamp you were using did you? Have you tried other types?
I did some research on headlamp assemblies as normally you would expect simple bulbs but there is much more to it. There are HID Xenon headlamps which employ a high-voltage ignitor which may very well be integrated into the bulb.
Without going into the possibilities in any great detail the point is that a headlamp is not always what you expect. But a 40V spike can never be dismissed in a power circuit as just some aberration, even if it's not the culprit directly, it indicates that something is different.
I'm surprised the fast diode across the load didn't kill the spike. Just because you may have seen around 40V doesn't mean it wasn't a much larger spike because the MOSFET's diode will clamp that to around 55V anyway but some scopes might not show the narrow top of the spike anyway. Usually it's good to isolate sections of the circuit until you find the culprit. In this case simply reducing it down to the offending load and pulsing the headlamp directly via a switch without any MOSFET or clamping in the circuit while scoping it would reveal the full extent of the load abnormalities.
Another thing I THINK we can rule out - voltage transients on the supply of the driver. I put a large inductive load on an old version board. This board would again, blow every time I put the headlight on it, but work fine with others. Well, I only have a 2.2uF capacitor(no supply diode) as the bypass capacitor for the driver. The load was set to pull 10A @ 40kHz, and WHOA. The capacitor got smoking, literally, hot in a matter of seconds. I looked at the voltage of it on the scope and it swings from about 12V, all the way down to 5V, and back again. Maybe the driver can internally compensate for a fairly slow transietn like this and not higher freq transients, but I found it odd.
@Peter, sorry about not elaborating on the bulb. This was the cheapest headlight they had at Autozone(7.99) and It is a simple resistive element, nothing fancy. It takes a full second for the bulb to reach it's full brightness, and another second to extinguish. Other than that, I don't know what to tell you about it.
You have got me intrigued. Are you absolutely sure the bulb is a simple resistive type? If I had been blowing all those drivers I would have been inclined to break the base of the bulb open just to check anyway. Either way I'd feel better [noparse]:)[/noparse]
Now I did have an idea earlier that could throw further light on the problem. If you place a diode across the drive resistor so that the anode goes to the driver we should be able to rule out whether it is some transient upsetting the driver chip. But of course I would be very much inclined to do this with only one driver as the cross-conduction between pairs has not been ruled out. If it works then reverse the diode, if that works couple the drivers etc until it fails. TI do recommend a drive resistor to dissipate some of the energy (thermal considerations).
*Peter*
P.S. that 2.2uf should be a tantalum but always use a 0.1uf across the driver as well.
Peter, I've read a couple of things about people using Tantalum capacitors, but why? Do they have a lower ESR than typical aluminum electrolytics? And yes, I think my final design WILL have a 4.7 Ohm resistor between gate and driver.
As for the diode, I'm not sure I totally understand. Is this diode there so that when the driver sources current, it goes through the resistor like normal, but when it sinks, it goes through the diode with minimal resistance?
The tantalum haver lower ESR and inductance and so they are especially required for medium to fast response as decouplers. Electrolytics are fine for the bulk supply filtering or for use in coupling analog signals. The nominal 0.1uf ceramic is mandatory though in all these circumstances for high frequency performance and of course this needs to be coupled close to the driver pins avoiding long thin traces.
My idea with the diode is that it will block any positive going transients coming back or else it doesn't which may shed light on what is happening. If it fails with the anode on the driver then you may have to check for thermal drive problems or alternatively a negative spike (unlikely). If it doesn't fail then reversing the diode will let those transients through and should blow the driver. Not that we want to blow any more drivers but at least we getting some answers. I'm starting to think that the problem is related to the duty cycle/frequency and the thermal limitations of the driver.
Don't compromise the test by leaving the drivers paralleled though, there is no guarantee that a spike back on the outputs or supply won't cause pair cross conduction..
BTW, the regulator is not able to pull down the supply if it should rise so it may be an idea to allow for a 15V tranzorb across the supply.
Yup, zenor/tranzorb across the driver's power rails and another one across source-gate pins of the mosfet. I'm not sure how fast the transorbs go but you do want to keep all diodes as fast ones. Make use of the mosfet's built in transorb for suppression.
As mentioned earlier, at these voltages you can also use Schottky diodes for rectification. They are very fast. And very cool running of course.
Make sure the negative track for all these suppression diodes and the mosfet and supply capacitors is fat and short. Keep that ground bounce to a minimum. And keep the high current paths away from the driver.
I was out the weekend and cannot pinch the thread untill now that it is almost dead. But I'm courious enough to adopt the same setup and continue from where Pilldapill give up. Unfortunatelly, my Propeller's skills are not enough yet. But here's my theory: The relative high freq. PWM (40khz) aplied on a component designed for pure DC can induce the circuit to oscilate fast and strong·enough to the drivers lost its logic control and produce an internal short circuit with the known results. I think if you risk another driver and install a .1 cap in paralell with your headlamp, you'll validate my theory or ...
SkyKit, I hate to put a dent in your theory, but I've tried this setup at many frequencies, all the way down to 20Hz. The result is the same - instant driver destruction. All is well if I add a 0.1 Ohm resistor in series with the headlight.........
Leon said...
What is the cold resistance of the bulb?
He's already said it's 0.4 ohms. He also claims to have driven a 0.1 ohm resistive load without a problem. Moreover, when he shorts the gate resistor after the lamp has incandesced, the driver blows. Apparently inductive loads work fine.
Leon, Philldapill eliminates concerns about inrush current with an experiment shorting a driver's resistor with the headligt on. The mistery's flag is already on. In behalf of my almost derailed theory, I'd try with 100% PWM (pure DC).
100% DC works fine... 99.9% Fails. This is withOUT a gate resistor.
I swear, this is no longer an engineering issue, but an issue of magical demons out to get me and make me go insane and kill people. For the sake of my sanity, and the lives of others, I'm going to brake open this lamp, take some pictures, and throw it away. If it solves the mystery, great, if it doesn't, I'll still have my sanity. [noparse]:)[/noparse]
Don't break the lamp. You have a very good test unit with it.
Good to see the 100% PWM setting work fine. This mean you are dealing with flyback issues.
Which could either be supply surge/spike, or capacitive coupling punching back through the mosfet into the driver chip and causing latch-up.
I gather you are using a catching diode across the lamp already? Hope it's a fast one or it will cook also. At any rate this diode should keep the supply spikes to something reasonable.
Try throwing a 12volt zenor across the source-gate pins of the mosfet to suppress capacitive spikes from reaching the driver chip.
Comments
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Timothy D. Swieter, E.I.
www.brilldea.com - Prop Blade, LED Painter, RGB LEDs, uOLED-IOC, eProto for SunSPOT, BitScope
www.tdswieter.com
This should be the end of this thread... finally.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Timothy D. Swieter, E.I.
www.brilldea.com - Prop Blade, LED Painter, RGB LEDs, uOLED-IOC, eProto for SunSPOT, BitScope
www.tdswieter.com
Leon
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Amateur radio callsign: G1HSM
Suzuki SV1000S motorcycle
That's been my pet theory throughout this thread, but it's been pretty convincingly debunked.
-Phil
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Just a few PropSTICK Kit bare PCBs left!
Every other load works great. Even large inductive loads, with a diode of course. As of now, the headlamp problem is still a mystery, but I'm not to concerned about it anymore. If anyone has a new theory, you're more than welcome to re-open this thread to discussion. I'd love to get to the bottom of it for curiousity's sake.
You never did say what type of headlamp you were using did you? Have you tried other types?
I did some research on headlamp assemblies as normally you would expect simple bulbs but there is much more to it. There are HID Xenon headlamps which employ a high-voltage ignitor which may very well be integrated into the bulb.
Without going into the possibilities in any great detail the point is that a headlamp is not always what you expect. But a 40V spike can never be dismissed in a power circuit as just some aberration, even if it's not the culprit directly, it indicates that something is different.
I'm surprised the fast diode across the load didn't kill the spike. Just because you may have seen around 40V doesn't mean it wasn't a much larger spike because the MOSFET's diode will clamp that to around 55V anyway but some scopes might not show the narrow top of the spike anyway. Usually it's good to isolate sections of the circuit until you find the culprit. In this case simply reducing it down to the offending load and pulsing the headlamp directly via a switch without any MOSFET or clamping in the circuit while scoping it would reveal the full extent of the load abnormalities.
*Peter*
@Peter, sorry about not elaborating on the bulb. This was the cheapest headlight they had at Autozone(7.99) and It is a simple resistive element, nothing fancy. It takes a full second for the bulb to reach it's full brightness, and another second to extinguish. Other than that, I don't know what to tell you about it.
Now I did have an idea earlier that could throw further light on the problem. If you place a diode across the drive resistor so that the anode goes to the driver we should be able to rule out whether it is some transient upsetting the driver chip. But of course I would be very much inclined to do this with only one driver as the cross-conduction between pairs has not been ruled out. If it works then reverse the diode, if that works couple the drivers etc until it fails. TI do recommend a drive resistor to dissipate some of the energy (thermal considerations).
*Peter*
P.S. that 2.2uf should be a tantalum but always use a 0.1uf across the driver as well.
As for the diode, I'm not sure I totally understand. Is this diode there so that when the driver sources current, it goes through the resistor like normal, but when it sinks, it goes through the diode with minimal resistance?
My idea with the diode is that it will block any positive going transients coming back or else it doesn't which may shed light on what is happening. If it fails with the anode on the driver then you may have to check for thermal drive problems or alternatively a negative spike (unlikely). If it doesn't fail then reversing the diode will let those transients through and should blow the driver. Not that we want to blow any more drivers but at least we getting some answers. I'm starting to think that the problem is related to the duty cycle/frequency and the thermal limitations of the driver.
Don't compromise the test by leaving the drivers paralleled though, there is no guarantee that a spike back on the outputs or supply won't cause pair cross conduction..
BTW, the regulator is not able to pull down the supply if it should rise so it may be an idea to allow for a 15V tranzorb across the supply.
*Peter*
As mentioned earlier, at these voltages you can also use Schottky diodes for rectification. They are very fast. And very cool running of course.
Post Edited (Ale) : 12/23/2008 6:27:34 AM GMT
I just found this:
www.automotivedesignline.com/201807702
Of course, the 55A in a car is when all the lights are switched on.
Zetex has this interesting doc:
www.zetex.com/5.0/pdf/Improving_reliability_in_auto_applications.pdf
Leon
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Amateur radio callsign: G1HSM
Suzuki SV1000S motorcycle
Post Edited (Leon) : 12/23/2008 2:49:12 PM GMT
-Phil
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Just a few PropSTICK Kit bare PCBs left!
Post Edited (Phil Pilgrim (PhiPi)) : 12/23/2008 5:14:49 PM GMT
I swear, this is no longer an engineering issue, but an issue of magical demons out to get me and make me go insane and kill people. For the sake of my sanity, and the lives of others, I'm going to brake open this lamp, take some pictures, and throw it away. If it solves the mystery, great, if it doesn't, I'll still have my sanity. [noparse]:)[/noparse]
Good to see the 100% PWM setting work fine. This mean you are dealing with flyback issues.
Which could either be supply surge/spike, or capacitive coupling punching back through the mosfet into the driver chip and causing latch-up.
I gather you are using a catching diode across the lamp already? Hope it's a fast one or it will cook also. At any rate this diode should keep the supply spikes to something reasonable.
Try throwing a 12volt zenor across the source-gate pins of the mosfet to suppress capacitive spikes from reaching the driver chip.
Evan