Go Down

### Topic: help with mosfet heat dissapation with pwm (Read 3346 times)previous topic - next topic

#### winner10920

##### Dec 01, 2012, 03:30 pm
Does anyone know a simple formula to estimate heat dissapation in a switching mosfet? Looking online it gets kinda confusing for me and im not the greatest at math or all the formula stuff, maybe there's a simpler formula out there just for estimates sake?
the variables I know and I guess are important are rds(on), input capacitance, the drain/source current, the gate resistor/driver output impedance
I know the heat dissapation without switching, but I wanna see the different heat dissapation at different switching frequency's/different driver capabilities  so I can see if I can get away without a mosfet driver at the frequency I want to switch at, and/or the frequency cutoff where I can just use a resistor

#### Krodal

#1
##### Dec 01, 2012, 05:19 pm
I don't know if this is even possible.

If an Arduino output is used, a resistor should be used for the gate capacitance of the mosfet (could be 1nF for some mosfets).
That causes a slow edge. The Mosfet has a non-linear curve which also depends on the temperature.
The formula can perhaps be simple for a single mosfet type, but I haven't seen such a formula.

Suppose 40mA output for the Arduino. Ouput resistor of 150 ohm. Mosfet with 500pF gate capacitance.
The cutoff frequency of 150 ohm and 500pF is 2kHz.
To have a least something that looks like a square wave, I would choose 1/10th of that, so 200Hz as maximum.

The mosfet drivers are much better. Some have a mosfet and transistor combined to deliver large peak currents.

#### dc42

#2
##### Dec 01, 2012, 05:25 pmLast Edit: Dec 01, 2012, 05:27 pm by dc42 Reason: 1
Working out an exact value for the power dissipation is difficult. However, it's not hard to calculate an approximate value.

First, you need to calculate the switching time. This is the total gate charge (which you can read from the datasheet) divided by the gate current. The gate current is the excess gate voltage divided by the gate series resistor (including an allowance for the driver output resistance if necessary). The excess gate voltage when switching on is the driver output voltage less the gate voltage needed for the device to be fully on. Likewise, when switching off, the excess gate voltage is the gate threshold voltage minus the actual driver output voltage (typically 0V). For a 5V Arduino driving a logic-level mosfet, it is reasonable to assume the excess gate voltage is 1V in both cases.

For example: IRLU8726PBF mosfet driven from Arduino pin through 100 ohm resistor (plus 30 ohms pin output resistance). From the datasheet, gate charge is 15nC typ @ Vds=15V, ID=20A. So switching time is 15/(1/130) ns = 1.95us. To get a faster switching speed we need to use a higher gate drive current, which requires a driver because 100 ohms as as low as we can safely go on an Arduino output.

Next, calculate the total energy dissipated in the mosfet during switching. If the load is resistive, then this is about 0.17 * voltage * current * switching_time. If the load is inductive with a flyback diode connected, then it's about 0.5 * voltage * current * switching_time.

Now multiply that energy by twice the switching frequency, and you have an estimate of the dynamic power dissipation.
Formal verification of safety-critical software, software development, and electronic design and prototyping. See http://www.eschertech.com. Please do not ask for unpaid help via PM, use the forum.

#### Krupski

#3
##### Dec 01, 2012, 06:37 pm

Does anyone know a simple formula to estimate heat dissapation in a switching mosfet? Looking online it gets kinda confusing for me and im not the greatest at math or all the formula stuff, maybe there's a simpler formula out there just for estimates sake?
the variables I know and I guess are important are rds(on), input capacitance, the drain/source current, the gate resistor/driver output impedance
I know the heat dissapation without switching, but I wanna see the different heat dissapation at different switching frequency's/different driver capabilities  so I can see if I can get away without a mosfet driver at the frequency I want to switch at, and/or the frequency cutoff where I can just use a resistor

The important thing to consider is the switching time. The longer the rise time of the gate, the more power is dissipated.

A FET has a "problem" called the "Miller effect". The inherent capacitance between the drain and gate couples changes in the drain voltage to the gate. When you pull the gate up, the drain goes down and that capacitance "fights" the gate driver (tries to push the gate down). The actual capacitance is "multiplied" by the gain of the FET so the gate driver sees an effective capacitance a lot higher than it actually is.

That's why you need a low impedance driver for a gate. Using a simple resistor in series with the gate is the worst thing to do... because you are making a nice low pass filter that increases the time the FET is in it's linear region and therefore increase the power loss (dissipation).

If the max Vds of the FET is above 3.3 or 5 volts (depending on your processor), then simply connect the processor output directly to the gate (of course, be sure that's enough to completely turn the FET on!).

Hope this makes sense.
Gentlemen may prefer Blondes, but Real Men prefer Redheads!

#### winner10920

#4
##### Dec 01, 2012, 07:58 pm
That makes sense, unfortunetly in one case I need a nice sized power mosfet to switch up to 10amps worth of led strips, and preferably id like to not use a heatsink and so keep the total heat dissapation<1W, but calculating that had seemed daunting so far, I get different results sometimes with formulas I found that I think are right, but really unsure
I have some stp40nf12 mosfets id like to use, but id rather buy a more expensive mosfet than add a driver, also my pwm freq. is 15khz, which I could lower but id rather not to completelly reduce the flicker effect(which I hate alot)
But I have some other projects at different voltages,currents,frequencys so id like to get this all down and learn it

So for the stp40nf12
Ciss=1880pf
vgs=4v max
rds(on)= .032 max
Qg= 80 nC

And well say load of 6 amps, so I know I alrready hit my 1W heat just from static resistance assuming 100% duty cycle but anyway ill carry on just to figure it out
Now for that formula before

80/(1/130)=10.4us switchin time

So .17*13v*6a*10.4us= 137.904 ? uC?

137.904*15000*2= 4137120?? What does this mean, it seems quite large, even if I cut down the switchin time dramatically this seems way too high unless its really a small unit

I tried figuring out a formula from a pdf online and using 10amp load and I get something less than 1W, but it only used crss,voltage,load and gate current so idk if its even right for what im doing or if my math is even right

#### dc42

#5
##### Dec 01, 2012, 08:08 pm

So .17*13v*6a*10.4us= 137.904 ? uC?

138uJ (microJoules)

137.904*15000*2= 4137120?? What does this mean, it seems quite large, even if I cut down the switchin time dramatically this seems way too high unless its really a small unit

You've multiplied 138 microJoules by 15000 Herz, so the units you end up with are microWatts. So just over 4W.

15khz is way higher than you need to avoid flicker, 70Hz is high enough for that.
Formal verification of safety-critical software, software development, and electronic design and prototyping. See http://www.eschertech.com. Please do not ask for unpaid help via PM, use the forum.

#### dc42

#6
##### Dec 01, 2012, 08:14 pm
BTW:

1. btw total gate charge for a mosfet increases with Vds due to the Miller effect.

2. You might like to consider the IRLU8726PBF mosfet. Gate charge is just 15nC typical (23nC max) and Rds(on) is 5.8mohm max. It's also inexpensive. It comes in an SMD package, but with 0.09" lead spacing, it's usable with stripboard and perfboard if you bend the leads slightly.
Formal verification of safety-critical software, software development, and electronic design and prototyping. See http://www.eschertech.com. Please do not ask for unpaid help via PM, use the forum.

#### winner10920

#7
##### Dec 01, 2012, 08:58 pm
So I already know the stp40nf12 wont do what I want, but Its not so much flicker effect I guess more pov kinda thing, the default pwm freq. On the attiny85 was 200hz and the flickering was really bad, atleast to me, changing it to 15khz fixed that
the problem isn't while looking straight at it, but if you move your eyes quickly it tracks across your vision, which since this is being designed to possibly be of automotive/truck use where while driving your eyes move around quickly as well as other drivers seeing you pass quickly(I hate most stock led taillights because a few manufacturers use pwm to dim the lights and on some cars its really annoying to see that pov effect while driving)

Now would that mosfet u suggested be able to do what I want? Its input capacitance is higher than the other, but I guess with the ultra low gate threshold its better?

#### winner10920

#8
##### Dec 01, 2012, 09:15 pm
I think perhaps this can do what I want? http://www.irf.com/product-info/datasheets/data/irliz44npbf.pdf
More a educated guess than proof yet, and its cheaper
ill do the calculations on the two when I get more time and have a pen and paper in front of me, changing from browser to pdf to calculator on my phone gets really confusing, lol

#### dc42

#9
##### Dec 01, 2012, 09:17 pmLast Edit: Dec 01, 2012, 09:21 pm by dc42 Reason: 1
I'm very surprised that you can detect 200Hz PWM. Were you using one of the hardware PWM pins of the attiny? How had you programmed it, and are you certain that the PWM frequency was 200Hz? The Attiny defaults to a clock speed of 1MHz unless you change the fuses, so perhaps your PWM frequency was lower than you think.

The critical parameter for mosfet switching speed (given the same gate drive resistance) is total gate charge at whatever Vdd you are switching. The mosfet I suggested has lower Rds(on) and lower gate charge than the one you quoted - although as the gate charge is quoted at a higher voltage for the STP40NF12, we can't be certain how they compare at 12V.

If you really can detect 200Hz PWM then try 1KHz instead of 15kHz - this should reduce the dynamic power dissipation to a reasonable value.
Formal verification of safety-critical software, software development, and electronic design and prototyping. See http://www.eschertech.com. Please do not ask for unpaid help via PM, use the forum.

#### dc42

#10
##### Dec 01, 2012, 09:20 pmLast Edit: Dec 01, 2012, 09:22 pm by dc42 Reason: 1

I think perhaps this can do what I want? http://www.irf.com/product-

More expensive than IRLU8726PBF, much higher Rds(on) and higher gate charge too. But as it's in a TO220 package, you can add a heatsink if you need to.
Formal verification of safety-critical software, software development, and electronic design and prototyping. See http://www.eschertech.com. Please do not ask for unpaid help via PM, use the forum.

#### winner10920

#11
##### Dec 01, 2012, 09:25 pm
Hardware pwm, 8mhz internal osc. I used an actual frequency counter to check therefore frequency and it was just over 200hz, I guess since im only 19 my eyes are still sensitive to things like that, anyway I think I changed the prescaler two steps so I can go back one, but idr I programmed that a while ago
That mosfet you suggested isn't that much more expensive actually, I think I may use that one

But so that formula u think will be accurate enough atleast to decide if I can get away without a heatsink

#### dc42

#12
##### Dec 01, 2012, 09:38 pm
OK, so you must have reprogrammed the Attiny fuses to turn off the CLKDIV8 bit. Try a PWM frequency nearer 1KHz, that way you should manage without a heatsink.
Formal verification of safety-critical software, software development, and electronic design and prototyping. See http://www.eschertech.com. Please do not ask for unpaid help via PM, use the forum.

#### Grumpy_Mike

#13
##### Dec 02, 2012, 06:55 pm
Quote
problem isn't while looking straight at it, but if you move your eyes quickly it tracks across your vision,

If this is your pet hate then you pay for it in increased switching dissipation. In fact is 15KHz fast enough? This is not a problem that worries most people. I have never seen any commercial LED strings do this, infact you can get quite good effects by moving your camera during the exposure and letting the PWM split it up into dashes.

#### winner10920

#14
##### Dec 02, 2012, 08:17 pm
I think I got it down pat now for calculating switching losses, so I guess to calculate total loss I can just add the static heat dissapation at 100% duty cycle for worse case scenario heat?  Im definetly probably gonna use a mosfet driver, I've worked it out with and without the driver for the mosfet I have and the one suggested, and for this scenario atleast im gonna go with the suggested mosfet with a driver, @ 20khz its switchin loss is negligible

Go Up

Please enter a valid email to subscribe