MOSFET motor controller heat problem

Discussion in 'The Projects Forum' started by russpatterson, May 4, 2010.

  1. russpatterson

    Thread Starter Member

    Feb 1, 2010
    351
    16
    Hi,

    I'm using an irf530 to control a pump motor. I PWM it at 5Khz to limit the current to the pump and control water flow. At 1 amp everything works fine but with a 3-5 amp draw the mosfet overheats to the point of failure.

    When I look at the gate on the scope the on/off transitions on the PWM wave look pretty clean. Do I need to discharge the gate more cleanly than simply driving the PIC IO pin low (e.g. adding a ~10K pull down or even a transistor that opens and grounds the gate when in the off cycle)? The RDS(on) is 0.10 - 0.16 ohms. I can heat sink it by putting a bolt through the tab and my PCB but there is too much heat for that right now.

    I need to be able to pull about 9 amps through it eventually (drive three pumps). So I probably need to move to another part or put some mosfets in parallel or something. Before I get there I need to solve this heat issue.

    Here's the schematic. The irf530 is in the upper right hand corner.

    Any input is appreciated.

    Thanks!
     
  2. rjenkins

    AAC Fanatic!

    Nov 6, 2005
    1,015
    69
    The IRF530 has a gate threshold in the 2 - 4V range, it's not a 'logic level' device.
    It needs at least 10V gate drive to switch properly.

    Try a logic level power FET and it should work better.
    Are you using a fast recovery diode across the motor? (I'm pretty sure it's not a 4148 as on the schematic..)

    I'd also put the FET gate series resistor value down to something like 100 Ohms or less.
     
  3. russpatterson

    Thread Starter Member

    Feb 1, 2010
    351
    16
    Hi Robert,

    Thanks for the information. I thought if I was 2-4V above the source voltage (negative in this case) I would be able to switch the FET on fully. Not so huh? I'll look for a logic level FET that is rated for 20 amps or so. Any suggestions?

    What rating do I look for to find a fast recovery diode? Also, does that diode need to be rated at the amperage I plan to pull through the FET?

    I'll lower the value of the series resistor to 100 ohms or less. Thanks.

    -Russ
     
  4. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    If your load is inductive, then the diode needs to be rated for the current that will be flowing at 50% duty cycle.

    Think about using a gate driver IC. Have a look at the FOD3180. It will really "snap" that gate on and off, minimizing power dissipation in the MOSFET.

    Look for a MOSFET that has maybe 1.3 to 2 times the voltage you're operating the motor at, with a low Rds(on) and low total gate charge, often specified as Qg. The IRF5xx series are really a bit "long in the tooth", as they've been on the market many years now.
     
  5. russpatterson

    Thread Starter Member

    Feb 1, 2010
    351
    16
    Thanks for the reply. I ordered a couple of these:

    STP80NF03L-04

    http://search.digikey.com/scripts/D...rch_go&lang=en&site=us&keywords=497-2676-5-ND

    They're rated at 30V and 13V is the most I'll be sending through to the pumps.
    The Rds(on) is 4 mOhm @ 40A, 10V
    The Qg is 110nC @ 4.5V

    It looks pretty good but not cheap at $3.00.

    The gate driver looks interesting. The test schematic in the datasheet shows it connected with 2 resistors, and a couple of caps. If I can avoid the extra parts I'd like to. It just provides a really clean on/off for the FET?
     
  6. mik3

    Senior Member

    Feb 4, 2008
    4,846
    63
    Remove the 1.9k resistor (R1) from the circuit to allow the MOSFET to switch faster and minimize switching losses. Also, put a 10k resistor across the gate-source to turn off the MOSFET in case the PIC fails. If it overheats again use a heatsink or a gate driver to minimize switching losses even more.
     
  7. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Mik3,
    The PIC won't like it very much with a capacitive load being switched that quickly with no resistance between the PIC I/O pin and the gate. It also won't help resolve our OP's current Vgs problem.

    Using a gate driver IC will help to avoid a lot of problems.
     
  8. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    The Rds(on) is darn low, but the cost of that is the high Qg @110nC.
    Take a look at these:
    http://search.digikey.com/scripts/DkSearch/dksus.dll?

    You could run three of those in parallel to spread the heat around, and still wind up with a lower Rds(on) and a much lower Qg (39nC for 3 vs 110nC for 1) - for roughly the same price. The gate charge has an affect on how quickly you can switch them on and off with the same driver.

    Your PIC can supply up to 20mA. The gate driver I mentioned can supply 100x that current. It also completely isolates your PIC from the MOSFET gate. Sometimes when MOSFETs fail, the gate shorts to the drain. This would put 13v at high current on your PIC pin, which would fry it.

    Better to fry a gate driver than your PIC.
     
  9. russpatterson

    Thread Starter Member

    Feb 1, 2010
    351
    16
    Thanks again for all the info. I like the idea of spreading the load across the 3 mosfets. That way I can only populate with one if I'm running a small 500 GPH pump (~ 1 amp @ 12V). I ordered some of those IRLU8721PBF's and a couple gate drivers to test on the breadboard.

    I'll try with and w/out the driver and see how things run. I'm only switching at 5Khz which should be pretty slow for these things. Can you point me to some reading so I can understand what the charge units, nC's, on the FET's mean? Also my PIC costs $1.19 at qty. of 1 so burning up that or the driver means I've killed a board either way.

    How many mA's do I need to put on the gates of the FET's to get them to switch? Isn't 20mA enough? Just trying to understand the trade-off's.

    I still need to find suitable diodes. What numbers, swithing speed, etc. do I need to look for? So if I plan to pull 10 amps I need a diode rated at 5 amps or should I just double that and get a 10 amp diode? That's gonna be a big part. Can I spread the load around to 3 diodes as well?

    Thanks!

    -Russ
     
  10. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    I suggest that you do NOT breadboard it. The breadboard will add lots of parasitic L and C. You will get much better results if you design a PCB with very short traces.

    5kHz may not seem very fast, but you have to keep in mind that an ideal square wave is composed of the fundamental frequency (5kHz), plus ALL of the odd harmonics of the fundamental frequency. The situation becomes more complex when you are operating at a PWM rate that is not exactly 50%.

    Can you point me to some reading so I can understand what the charge units, nC's, on the FET's mean?[/QUOTE]

    Here's some reading material:
    http://www.analogzone.com/pwrt1208.pdf
    http://www.microchip.com/stellent/idcplg?IdcService=SS_GET_PAGE&nodeId=1824&appnote=en020629
    http://focus.ti.com/lit/ml/slup169/slup169.pdf

    Better to not kill anything. ;) "Thou shalt not fry boards.".

    You're going to need a resistor between the driver and the MOSFET gate, as close to the MOSFET as you can get it. This is to suppress the tendency to "ring" when the gate charge state changes.

    It's the gate charge over a period of time. The more current that you can source to/sink from the gate, the shorter the turn-on/turn-off times will be, and the less heat you will dissipate in the MOSFETs.

    Here's a dual Schottky diode in a TO220 package rated for 10A:
    http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=MBR2030CTLGOS-ND

    Here's one rated for 30A, and it's less expensive:
    http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=MBR2535CTGOS-ND

    Be sure to use a heat sink on them, as they will dissipate heat.

    If you want to get more sophisticated with things, you could look at using a MOSFET as an "ideal diode"; you'd have to time your switching carefully though. Look up "Synchronous Rectification" to get an idea of what I'm talking about.
     
  11. russpatterson

    Thread Starter Member

    Feb 1, 2010
    351
    16
    I'm still waiting for the FOD3180S gate driver and the smaller IRLU8721PBF MOSFETs to show up from Digikey. In the mean time I dropped in one of those big STP80NF03L-04 's to my current board. Which drives the gate from the pic thru a ~2K resistor. I ran the thing for the day and had some interesting results. I'm not sure I understand the results though.

    The MOSFET ran one pump ~2.7A at any duty cycle in the lab but would generate some heat. So I put a 1 1/4" machine screw in the top of the TO-220 heat sink tab and tightened it up against the board. I ran it outside all day testing with between 1 and 3 ~3A pumps at various duty cycles.

    It ran great at first but the heat sink screw and part would generate some serious heat. Like I could only hold my finger tip on the top of that 1 1/4" screw for about 2 seconds when I ran at ~50% duty cycle. It seemed to run cooler at full ~99% duty cycle. Anyway it ran, hot, for a number of hours and is running right now but it doesn't seem to allow as much current through as when it was new this morning. It doesn't get hot like that either. So I think I've burned through another part but my question is...

    Should I expect to have to dissipate that kind of heat for my 12V-10A pull through the transistor(s)? I noticed that the IRLU8721PBF's don't have a big heat tab with a screw hole. Should I expect to have to heat sink them or will they run cool once I get the gate driver working correctly and they are switching on and off cleanly? I need to figure in the heat sink space before I send another board out to fab.

    Thanks,

    -Russ
     
  12. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    2k is far too much resistance.

    See the attached for a comparison of a vaguely similar MOSFET (Qg=125) having it's gate charged via a 2k resistor, and then being charged via a 240 Ohm resistor.

    The datasheet for your MOSFET recommends a 4.7 Ohm gate resistor, but your PIC is not capable of driving that much current; you'd fry the I/O pin. This is why I've been trying to tell you that you need a gate driver IC.

    With 10A current, they will dissipate about 1W power, as they have a Rds(on) of 10m Ohms. P=I^2R, so 10*10*0.01 = 1W. If you're running multiple of the IRLU7821's in parallel, they will divide the current. You can use a copper pour area on the board and solder the tab to it to use as a heat sink.

    One square inch of exposed copper pour will dissipate 1W of power with a 20°C rise.
    Four square inches of exposed copper pour will dissipate 1W of power with a 10°C rise.

    Positioning the PCB vertically will assist in heat transfer via convection. Fan cooling increases the efficiency.
     
  13. russpatterson

    Thread Starter Member

    Feb 1, 2010
    351
    16
    Hi,

    Thanks for all the great info. Sorry for the delayed response. Life sometimes gets in the way of electronics.

    I'm not sure I understand what the graphs are showing exactly. Is the green the PWM input, the red the gate charge? and the purple the time the mosfet is conducting? The smaller resistor is clearly better.

    In fact the 100 ohm gate series resistor I soldered in works much better. No overheating at all now. I chose that 1.9K value because I was trying to get rid of a ringing problem I could see on the scope and just tried a bunch of values and 1.9K seemed to work the best.

    Here's where I have some more questions. My PIC can supply .020 Amps on the I/O pin. So the math for figuring out the smallest series resistor says the smallest resistor I can run is 250 ohms, (if I've done the math right). Here's how I got there:

    R = V/I
    R = 5/.020
    R = 250 ohms

    So the 20 microamps is probably a best conditions thing so 10 or 15 seems like a better number to use. However I'm already using a 100 ohm and it's been working great for quite some time. What am I missing?

    Thanks for the info on thermal dissipation with the copper on the board. My device is for backyard solar power (fountains, irrigation, lighting, etc.) So it runs in a small, waterproof, case. Which means I can't put in a fan and that heat issues make that waterproof case a little oven quickly. Space on my board is an issue as well so maybe I should solder the MOSFETs in vertically and put a heat sink on the tab?

    Concerning the current limiting behavior I was posting about earlier; it was a software issue with sensing the battery below 10V, turn the pumps off, then the battery voltage would bounce right back, I'd turn the pumps back on,
    which draws current and pulls the battery down below 10V again, many times a second.

    It would drive one pump down to a trickle before finally deciding to stay off for the night. Which could be used an energy harvesting feature if done properly by decreasing the duty cycle on the PWM.

    I now have a question about the switching speed and thermal issues with my panel MOSFET. I'll start a new post for that.

    As far as the gate driver goes. I have the parts and just need to put together a design and build a board. Would using perf-board, spread things out and point to point wire be suitable? I understand that breadboards are not desireable due to all the unwanted capacitance and inductance created by the proximity of the strips. I'd like to avoid making a custom PC board just for the test of the gate driver. What I'd like to do is run a test and measure power loss. Test two boards, one board with just the series resistor driven off the PIC I/O pin and one board with the gate driver. Then calculate the yearly power loss, use the cost of the system and see if the gate driver and discreets would pay for themselves in a year or not.

    Thanks again for all the great info. It's kept me busy. BTW I have this project showing at our county fair right now. It's part of a garden display that a local nursery built. It's just a small 500 GPH pump that pulls < 2 amps.
     
  14. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    V(v1) (green) is the signal input to the gate (and the source for the load resistor) for M1.
    V(gate1) is the voltage on M1's gate. Notice the slow rise and fall times.
    V(P002*Id(M1)..etc. is the power dissipation in the MOSFET. Notice it peaks at 120W.

    The lower plot pane is similar for M2. Note that peak power disspation for M2 is less than 13W, and it's only that high for a brief period of time.

    Well, you sure got rid of the ringing, and vaporized the MOSFET at the same time. ;)

    You might get away with it for awhile, but you're really exceeding the ratings of the device. If you're trying to switch it really fast, you will probably burn out the port fairly quickly.

    You need SOME way to get the heat out of the case. If you don't, you will burn it up. A copper thru-bolt sealed with RTV that connects to some copper flashing painted black on the outside would likely work pretty well.

    You need more hysteresis.

    If you are going to be switching a good bit of power, then perfboard isn't the way to go. Can't you make your own boards using toner transfer or the like?

    Note that you will either need to use really wide traces for the power bus, or scab on some copper stock or copper wire to prevent the board from overheating. Download PCBtemp (google is your friend) and use it to determine the trace widths you'll need for the current vs thermal rise vs copper thickness.

    Cool :)
     
Loading...