Using wiper motors. To-220 power dissipation. Building a motor driver.

[David97]

I'm wanting to design some motor drivers for controlling windsheild wiper motors for robots. One of my mates wants to build some for a robot and I thought I would give him a hand because I can see myself using some wiper motors for some projects myself.

So I have some  IFRz44 and IRF4905 to-220 fets I'm planning on using. There's plenty of old computer heat syncs that can be used to cool them. I'm trying to work out how many fets I need. I'm not sure weather Ill be doing a full H bridge yet, or using a relay to switch the direction. Really depend on how much power each one can dissipate. 

I haven't measured current of a wiper motor. Hopefully should be able find someone else who's done a project and can tell us, or could load some up and test it. . Should be around 10kg and have about 24cm diameter wheels (purley guessing) wiper motor at 60rpm. So that should give about 0.8m/s). I would be thinking about 7A or current. But at this point can't be sure.
IRFZ44 = 17.5milliohm
IRF4905 = 20milliohm

P = I^2* R
P = 7^2 *0.02
P = 0.98W

This is just a guess and ill come back once I have some better data. I know the info Ive given if pretty vague. . But wondering how much power can each fet dissipate? They will all be mounted to a old computer cpu heat sync. ?

[Ice-Tea]

Just look at the datasheet. 107W in the case of the IFRz44. This would obviously require something like an infinitely low thermal resistance heatsink. So, a hefty satefy margin (and good thermal interface material) is required.

That said: keep in mind that thermal losses in FETs are NOT only static power dissipation. Depending on switching frequency, driving method, voltage etc etc... dynamic losses can euqal to or surpass static losses..

[ajb]

How much power each FET can dissipate depends on how good your heatsinking is.  Sometimes you'll see a max dissipation figure in the datasheet, but as a rule these numbers are completely divorced from real-life limitations and are best ignored.  Instead you need to look at the thermal resistance to the case, usually denoted as something like ?_jc and measured in degrees (celsius or kelvin) per watt.  You will also need to know the thermal resistance of the interface between the package and the heatsink (consult the datasheet for whatever thermal interface material you use) and the thermal resistance of your heatsink, which is a little trickier as it depends on airflow.  If you need an accurate idea of your heatsink's performance, you can get TO-247 resistors with monstrous power ratings--bolting one of those to your heat sink and applying a known power to it and them measuring the temperature of the heatsink relative to ambient will give you the numbers you need.  This also makes it easy to determine the impact of fan speed and so on has on the heatsink's performance.

So you should wind up with two or more terms that include the thermal resistance from the die of the FET to the case (from the datasheet), from the case to the heatsink, and from the heatsink to ambient.  Since these are resistances in series, they can all be summed up to give the total thermal resistance from the die to ambient (still in °C/W).  Now you need to determine the maximum operating temperature of the die (from the datasheet, and might be something like 120°C) and your maximum operating ambient temperature.  If the driver will be mounted in a warm enclosure with a few other drivers, be sure to take this into account!  Subtract the ambient temperature from the maximum die temperature, and divide the resulting number by your aggregate thermal resistance determined earlier.  This will give you the maximum amount of power that the device can dissipate while keeping the die below its maximum temperature....in theory.  In practice, you will need to include some headroom for transients and overloads, incorporate any necessary derating factors, and then add some margin for error.  But you need to start with the thermal resistance calculations.

[rstofer]

The majority of the heat doesn't come from Ron, is comes during switching, assuming you are doing PWM.  Consider the current waveform to look like a trapezoid.  Both the leading and following edge have slow ramps where the device isn't fully ON.  It is in this region that most of the heat is built up.  Given sloppy rise and fall times, don't switch very fast.  The less time the device spends in this switching region, the better.  You want to get fast turn-on and turn-off whenever possible.

For the small motors you are talking about, you can still probably get by with a logic level MOSFET
https://www.sparkfun.com/products/10213

If you get heating, add a heatsink.  If you still get heating, you aren't controlling the rise/fall time well enough.

When you get to the H-bridge, things get more difficult!  You have a choice of using N-channel devices or P-channel devices for the top MOSFETs.  If you pick N-Channel, the gate voltage needs to get several volts above the motor + rail.  That's often hard to do; you clearly can't drive that voltage directly from a uC.  If you pick a P-Channel device, controlling the gate is easy but finding a suitable device is a little more difficult.

There is a reason that semiconductor manufacturers make MOSFET drivers.  They want to be able to make an H-bridge using all N-Channel Mosfets and they want to use a capacitor dump to force the MOSFET through the transition region.  Google for 'H-bridge driver'

[max_torque]

How cheap does the finished product need to be??

These days, i just use off-the-shelf drivers from companies like Infineon. They come in various flavours (current/ voltage capability, layout RDSon  etc) and include over temp, over current, under voltage protection and often built in current sensing!

They are expensive in low quanties, but compared to the effort and time spent in rolling your own they make a lot of sense imo:


For example:  http://www.infineon.com/dgdl/bts7960b-pb-final.pdf?folderId=db3a3043156fd5730116144c5d101c30&fileId=db3a30431ed1d7b2011efe782ebd6b60

[Siwastaja]

If you need a lot of heatsinking when driving a wiper motor, you have something horribly wrong.

Get proper FET drivers (something that suppliest amp or two peak, specified to rise/fall times below 50 ns) and use low enough frequency (like around 5 kHz, or just over 20 kHz if the audible noise matters) so that you get switching losses down.

If Rds(on) is too high, parallel two of those FETs to get it down. Try to keep dissipation under a watt so that you are fine with a small heatsink (or none at all).

Absolutely use synchronous rectification, don't even consider using a freewheeling diode (7A times 0.7V = 5 W of dissipation, and driver efficiency approaching 0% at low speeds); which brings us back to the FET driver. It's cheapest and simple to use a bootstrap half-bridge driver (which means you can't run at 100% duty; you are limited to about 95%). If you run at low rpms (low voltage) with heavy load (high current), you may need to watch the Rds(on) of the synchronous FET (the one in parallel to the motor) and parallel even more of those. This is a DC/DC converter and the current through the synch FET will be more than the input current!

Is your 7A some kind of typical figure or worst case? How much current does it take if you completely stall the rotor and apply full voltage briefly?

[T3sl4co1l]

TO-220 max 50W, TO-247 max 100W. And that's with a damn good heatsink.

If you need more, parallel up more devices. Transistors are cheaper than heatsinks, and both are way cheaper than labor!

Better still, design it so you don't need more of either.  Design a proper switching circuit.  Motor control is pretty easy, even including bidirectional motion and regenerative braking.

Tim

[SeanB]

Wiper motor expect running current to be around 15A, and a peak when starting or when stalled approaching 30A. Design the bridge to survive this 30A for a few minutes, and it will run with any load from zero ( which will be around 5A for many motors) to just short of stalled. Drive them as hard as you can, just to get as close to that idealised on resistance. Remember the PCB traces might add the same again, and the connectors the same. Just always have a 30A automotive fuse inline with the battery lead per motor driver, it saves on the smoke, plus gives a convenient current measurement point.

[Siwastaja]

Design the bridge to survive this 30A for a few minutes

Agreed. This would mean two IRFZ44's in parallel. They can easily handle it, and will also have enough voltage rating at 12V supply against stray inductance.

When designed to sustain the stall current, defined by motor winding resistance, you can go without current sensing. This is a sensible design choice sometimes.

Things to google and learn about:
- Half-bridge
- Synchronous rectification
- FET driving, esp. integrated half-bridge drivers.
- Minimizing DC link parasitic inductance and using bypass capacitors
- Permanent magnet brushed DC motor control (the easiest type).
- Synchronous buck converter (what a motor controller is, in essence, inductor being the motor!)
- Current sensing in motor control (if you want to go advanced, really recommended, but maybe not for the first proof-of-concept prototype)

[rstofer]

Or just Google around for 50A H-Bridge.  There are thousands!  Everybody make 'em...

http://www.ebay.com/itm/like/291333937938?lpid=82&chn=ps&ul_noapp=true

$20 is a lot cheaper than I can design one and I can have it in a few days.

The first link I clicked is for an outfit in China so delivery is a long time out.  There are many other places with similar products.

[rstofer]

I used an IRF540N to drive a wiper motor attached to a plastic chair.  The chair held a skeleton covered in a blanket and there was a PIC used to generate a snoring sound.  It was a Halloween decoration sitting on our front porch.  Pretty cool effect!

I didn't need reversing and the IRF540N stayed cool even when driven directly from a microcontroller pin.  But the load was fairly small so I'm not making any predictions on this project.

[Siwastaja]

Or just Google around for 50A H-Bridge.  There are thousands!  Everybody make 'em...

http://www.ebay.com/itm/like/291333937938?lpid=82&chn=ps&ul_noapp=true

$20 is a lot cheaper than I can design one and I can have it in a few days.

I'm wanting to design some motor drivers

[emphasis mine]

---

If you just use the FET to turn the motor on/off, then a logic-level fet directly driven from an MCU pin will generally work because of negligible switching loss. Of course, if the switching is catastrophically slow, FET might stay in the linear region too long, causing die hotspotting. But an MCU pin can supply like 40-50 mA peak easily, which is about 10 times too small for good switching, but good enough for very low frequencies. Heck, you can even get the PWM working from an MCU pin, if you are happy with running the PWM at about 500Hz max.

For absolute simplicity, you'd use a "logic-level" NFET directly on an MCU pin (with about 10R in series with gate), keep the frequency below 500Hz, and use a freewheeling diode. Diode would require more heatsinking than the FET.

But with IRFZ44, you need to use a fet driver to get it work at all because it requires at very least 6-7V on the gate.

Using a relay for reversing is not a totally bad idea. Many small relays have rather good conduction rating (as opposed to switching rating), and this is one of the rare cases you can utilize that spec.