A really great Mosfet to use with the Propeller... especially when it's grounde
ElectricAye
Posts: 4,561
Hi all,
I've been using an IRF3708 MOSFET with the Propeller and the PWM Object running at about 2KHz, and I must say it works really well for switching current (Thank you, PhiPi!). But I've recently fried a couple of them and I suspect it's because I extended the wires that lead to/from them to about 6 feet, so I'm guessing that the inductance of the long wires might be zapping them. So I suppose I need a "flyback diode" to keep them happy, but I have no idea what to look for in such a diode. Anybody have any good suggestions?
thanks,
Mark
Post Edited (ElectricAye) : 10/3/2009 5:31:13 AM GMT
I've been using an IRF3708 MOSFET with the Propeller and the PWM Object running at about 2KHz, and I must say it works really well for switching current (Thank you, PhiPi!). But I've recently fried a couple of them and I suspect it's because I extended the wires that lead to/from them to about 6 feet, so I'm guessing that the inductance of the long wires might be zapping them. So I suppose I need a "flyback diode" to keep them happy, but I have no idea what to look for in such a diode. Anybody have any good suggestions?
thanks,
Mark
Post Edited (ElectricAye) : 10/3/2009 5:31:13 AM GMT
Comments
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe
IC Layout Engineer
Parallax, Inc.
Oh.
Um, yes, I am directly driving the gate. I guess that's not a good idea. For in-rush protection, what resistance is best?
thanks,
Mark
The gate resistor is optional but it protects the Prop in the event of a catastrophic failure on the part of the mosfet. If you make the value too high it will limit the frequency you can drive the mosfet at as the combination of the resistor and the gate capacitance form a filter. I usually use around 1K but I would go much lower for higher frequency although you should read on before you start sticking a resistor in the circuit.
We don't have any idea of your circuit or the load you are driving but there are a couple of typical mechanisms by which many mosfets fail.
In your case though thermal failure due to either insufficient gate drive or just too much current for too long. A perfect switch doesn't dissipate any power and that is what we try to do when we use the mosfet as a switch. We want Rdson to be as low as possible and the 3708 does a very nice job with very little drive so that shouldn't be the problem.
The trouble with a slow gate drive formed from any resistance (even I/O) is that the mosfet will go through a linear region when it is turning on and off during which it gets hot like crazy. If we keep repeating that as in high frequency PWM then we are asking for trouble. That's why PWM circuits have special gate drivers that can deliver the peak currents that are needed to charge and discharge the gate capacitance quickly and thus keep the mosfet out of it's linear region as little as possible and acting more like a perfect switch. Yes, we know mosfet don't need any actual current to drive them but there is that capacitance.
So the question is, what are you driving and what is the voltage and current rating of your PSU?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
*Peter*
Peter,
thank you for your insights into this problem. I have attached a sketch of my circuit that appears to be frying these Mosfets after a while. R1 is 10K ohms and R2 is a non-inductive 10 ohm power resistor being used as a heating element controlled by PWM. The mosfet is the aforementioned IRF3708, data sheet attached. The power supply is +15 volts, 10 amps, data sheet attached.
The wire from R1 to the Mosfet is about 12 inches long right now. The wires from the Mosfet to R2 and to the ground are about 6 feet long.
When these Mosfets die, I end up with roughly a 10.5 volt drop across R2, and a 4.5 volt drop across the Mosfet. Mosfet and R2 then heat up. The PWM signal from the Propeller will still affect the Mosfet somewhat even after it fries.
Some weeks ago, when I was given advice on using Mosfets, Philldapill suggested I use a Mosfet driver like a ucc27321, data sheet attached. So if capacitance is the problem, would that solve it do you think??? Also, any chance this could be an ESD problem? - I worry about that because the heat sink tab isn't linked to ground, it's hooked to the Drain.
thanks for your help,
Mark
The other thing to do would be to just try the MOSFET driver. It can only help you.
Thanks, Mike, I think that's what I'll do. I guess it's possible that different frequencies of PWM might allow the Mosfet to live longer, but it could be only a matter of time before I burn them up again.
cheers,
Mark
If all you're doing is creating heat, 10Hz — or even 1 Hz — would be plenty fast for your PWM, obviating any need for a MOSFET driver or any worries about slow switching due to a series gate resistor.
-Phil
You have a current of 15V / 10 Ohm = 1.5 A, this should be no problem for this MOSFET and a 3.3V Gate voltage.
I would try the reverse diode parallel to the Load, because MOSFETs are sensible about overvoltage. You can use another IRF3708 as a diode to test it (connect the gate to source and use the diode between drain and source), or any diode with a max current > 1.5 A.
A gate resistor can protect the Propeller, but makes the state between ON and OFF of the FET longer.
I use a similar circuit with a IRLZ14 and a shottky diode to switch a valve with 10 kHz PWM and 1 A max. and it works well since months.
Andy
Okay. I've still got some Mosfets laying around. I'll try that out while I wait on these new parts. I guess I should have known, when it comes to heating, there's no reason for fancy frequencies.
thanks for helping me narrow down what I might be doing wrong,
Mark
That's a very good question. I haven't been monitoring that, but if I keep zapping these things, I'll hook a DS18B20 to it and watch what happens.
Hey, that's pretty clever! I love ideas like this because my time lag between part shipments can be almost a week.
Thanks Andy!
The extra diode across the load is redundant I would think, I can't see how it would make it work any better or worse and especially since your load is resistive. Maybe Andy might be able to explain his reasoning on this matter, I could be missing something. Also, I don't think you can count the tiny tiny bit of inductance that the wire might have as we are not talking about RF frequencies here. You could run at a higher PWM frequency like Andy has but only with a light load of an amp or two as the mosfet will not be switching effectively unless you use a gate driver as Mike suggested. I find it a little strange that you have wired up the mosfet so far from the Prop which makes me concerned that your logic and power grounds may be the problem. It probably would help if we had a photo of the relevant parts.
But now that we know it's for heating and relatively low current then Phil is correct in suggesting low low PWM frequencies which depend upon the thermal inertia of your heating element.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
*Peter*
Hi Peter,
for now, the wiring is made long because the heaters have to go to various places inside a large (100 cubic foot) freezer that can sometimes go down to -20F. For now, I'm using the Demo board version of the Prop and it's running the VGA, so I can't yet (ever?) put the Propeller inside the deep freeze. I don't have a picture of the set-up on hand, but for now just imagine Medusa's hair mingling with Gordian knots of Gorgon's intestines. Ah, and your point about my groundings being crazy is well taken: I'm ashamed to admit how many cables we have criss-crossing the floor right now on their way to various outlets. Consolidating all of that is on the to-do list, of course, but for now I guess I'm looking for the surest alternative(s) to solving the problem.
thanks for your help,
Mark
Also I doubt that the inductance of the (wire-)resistor and cable is high enough to have any influence. But sometimes it's exactly what you exlude that is the reason for the fault [noparse];)[/noparse].
But if the inductance is high then the diode inside the FET does not always help: When the FET switches off, the Load-Inductivity produces a high voltage spike, if the reverse current is not drained off. And this current can only be drained off when the whole circuit has a low resistance for a reverse current. This includes the power supply and the Load and the insides of this two are not always known. So if you want be shure connect a diode direct at the Load which produces the spike.
It's always hard to find the fault if you don't have direct access to the build-up.
Andy
-Phil
That's a good point about the discharge path through the power supply which *should* be low impedance so that's always something to keep in mind then although there would be insufficient energy in the stray inductance in this particular case. But I know what you mean about not ruling things out when chasing a fault. That was a good idea too about configuring another mosfet as a fast diode which can be handy.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
*Peter*
I wired up a new IRF3708 mosfet this morning and attached a digital thermometer to its heat sink tab and employed my oscilloscope to watch the signals going from the Propeller to the mosfet gate. I've been running this test at 100 Hz and so far all looks well. The grounding looks good and the transitions all seem very crisp. The mosfet has not been heating up. The signals from the Propeller to the mosfet gate appear to be safely above +3 volts.
One thing I feel really stupid about. The Prop Demo board comes with its EEPROM loaded with some sort of Demo software. I never purged this software because it always gave me an indication that my VGA was turned on. But this morning I discovered that this demo software also happens to pulse Pin 1, almost mimicking a human heartbeat, and that's the pin I was using for my mosfet test. So even though the transitions on that demo look crisp, I feel stupid that I never thought about that as a source of snafuing my circuit. Duh on me.
Anyway, I've ordered some mosfet drivers and will keep all of your comments close to my heart as things progress (or not).
I greatly appreciate all of your help.
many thanks,
Mark
I really don't think you will need the MOSFET drivers for 100Hz. Also, when you measured the gate voltage, did you measure from gate directly to source or from gate to Vss. Even though source and Vss are connected together, you may get different readings when the MOSFET is "on".
-Phil
Oh, that's a good point. I was only measuring from Pin 1 to the Demo board's Vss. I'll check to see what the prop's gate signal looks like between the mosfet gate and the mosfet source. I've got to learn that ground here doesn't necessarily mean ground there sometimes.
thanks Phil
Wow! I put my Oscope probes between the mosfet gate and the mosfet source and there's RINGING! The ringing is about +/- 2 volts and lasts for about 0.3 milliseconds. So if the gate signal is diving from about 3 volts to 1 volt, that means the mosfet is spending time in its linear region, which might explain the eventual frying. This ringing is NOT seen when I measure the Prop outputs at the demo board (from Pin 1 to Demo board Vss.). I'm totally astonished! The Oscope, power supply and Demo board are all powered off the same power strip, too, so I'm totally freaked out by this. The ringing shows up whether I have voltage on the mosfet drain or not.
Big question is: why's it happening and what to do about it? Will a mosfet driver solve that problem?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Watching the world pass me by, one photon at a time.
Post Edited (ElectricAye) : 10/2/2009 6:08:23 PM GMT
-Phil
Thanks, Phil. I'll try that tonight.
I have one concern about that. The sketch that I've attached shows my present configuration except there is no connection at the dotted line. The dotted line is where I think I need the connection to be to solve this ringing problem. However, my power supply that runs the Prop Demo board is a wallwart weenie 9 volt, 0.9 amp output (120 volt, 0.3 amp input), so could there be a problem with large currents now running through the mosfet (from the 15 volt, 10 amp power supply) and then through the dotted line and then through the Demo board Vss and thus burning up the Demo board or its power supply???
If that's a worry, would it be best to insert a small resistor of say, 10 to 100 ohms, to limit current while maybe sufficiently dampening the ringing?
I thought I'd ask before I try it, rather than burn down the lab bench.
thanks,
Mark
Post Edited (ElectricAye) : 10/2/2009 10:34:19 PM GMT
Be sure to connect both the 15V return and Vss directly to the MOSFET source and nowhere else. With both Vss and the 15V return connected directly at the transistor, there won't be any weird currents to freak out the logic side of things.
I'm really amazed your circuit worked at all without that connection!
-Phil
The ground resistor should go on the Prop side, not the MOSFET side. Otherwise, you're forming a divider which will lower the available gate drive. The reason for the ground resistor is to keep the MOSFET "off" when the pin is floating during reset. The series resistor is to protect the Prop from driving an excessive capacitive load. 330R should be more than adequate protection.
-Phil
Bless you, O Pilgrim, Bless you.
Now that you have got that sorted out you should be a happy chappy. All you need is a good "grounding" in electronics
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
*Peter*
I'm really sorry about that. I didn't even know enough to know how profoundly non-knowing I was in this department.
much obliged,
Mark
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Watching the world pass me by, one photon at a time.
-Phil
The datasheet lists the thermal resistance, junction to case as 40 C/W for a PCB mount. I am driving an 8 amp load directly from the prop. I see that the Rds on is 0.029 ohms. So the power dissipated in the MOSFET is 0.029ohm*8amp*8amp = 1.856 W. If I mount the MOSFET to a PCB board, then the temperature will be 1.856W*40C/W +25C = 99 C. Is that correct? The data sheet seems to indicate that this is an acceptable operating temperature for this part.
I'm currently using an aluminum plate for a heat sink and I'm wondering if I can get away with board mounting on a PCB design.
Thanks,
David