Heat sink or no heat sink?

My nano will run 24/7. It has a 3A FET and voltage regulator to power a 1.5A LED strip at night. Im using PWM, but may output at 100% PWM.

Without advice, i'd heat sink the FET, regulator, and nano.

Are heat sinks, for this situation, overkill, appropriate, better safe than sorry...?

Test the temperature with your finger.
If it gets very hot, heatsink.

If a heatsink is needed, properly attach the component.

The finger-test is a good check - If you can hold your finger on it, it's not too hot. If it's too hot to touch, you might need a heatsink. Of course the problem is, sometimes the part burns-out by and by the time you realize you need a heatsink it's too late...

The Nano shouldn't be a problem.

The other two are hard to predict... Sometimes the datasheets will give you the maximum power dissipation without a heatsink, and sometimes they they'll give you information about the maximum internal die temperature, power dissipation, and thermal resistance. Calculating the thermal resistance and temperature drops can get "messy".

You get a temperature rise related to power (Watts) so yo might need a heatsink on a hot day even if you don't need on a cold day. Power is calculated as Voltage x Current.

A well-designed switching regulator probably won't need a heatsink. A linear regulator will probably overheat even with a heatsink.

Most regulator chips are thermally protected and they'll shut-down safely without burning-up. Then, they'll work again after they cool down. (Transistors & MOSFETs will just permanently die.)

The power dissipated by a linear regulator is the voltage drop across the regulator x the current. If you feed 24V into a 12V regulator it will get hotter than if you feed-in 14V. The amperage rating can be misleading. A 1A voltage regulator can overheat at less than 1A depending on how much voltage is dropped across it.

The same goes for the FET. The amperage rating doesn't tell the whole story. The heat is determined by power dissipation. When the FET is saturated the voltage dropped across it is determined by Rds(on) and the current (Ohm's Law). With Ohm's Law you can derive another way of calculating power: Power = Current2 x R.

The bottom like is, a MOSFET with lower Rds(on) will run cooler.

If the MOSFET (or transistor) is partially-on (not saturated) it will run hotter.

Zip_Ferndale:
My nano will run 24/7. It has a 3A FET and voltage regulator to power a 1.5A LED strip at night. Im using PWM, but may output at 100% PWM.

Without advice, i'd heat sink the FET, regulator, and nano.

Are heat sinks, for this situation, overkill, appropriate, better safe than sorry...?

Since you've neglected to provide the FET part number, datasheet, on-resistance or package-type, there's no
way to answer. The current rating is about the least important FET rating there is, BTW.

Lots of great info, thanks!

Here are the parts:
LightEver LED strip:
4100058 (12VDC, 18 W)

KA378R12C voltage regulator fed by a 12v AGM battery

MOSFET: IRLB8721PBF or 30N06 (i have no idea how to choose)

The IRLB8721 has a voltage of 30V and an on-resistance of 13 milliohms, so at 1.5A will dissipate 1.5 x 1.5 x 0.013 = 29mW, no problem at all.
The RFP30N06LE has a voltage of 60V and an on-resistance of 47 milliohms, so at 1.5A will dissipate 1.5 x 1.5 x 0.047 = 106mW, no problem at all.

Both have adequate voltage rating for 12V, the on-resistance at 5V gate drive is low enough.

Above about 0.5W dissipation a MOSFET will get really hot and need heatsinking, roughly speaking, at 0.1W
it'll only get warm and be fine without help.

The other parameter you need to consider is the total gate capacitance as that affects how fast PWM can be used (depending on the gate drive current).

The other parameter you need to consider is the total gate capacitance as that affects how fast PWM can be used (depending on the gate drive current).

How would i do this?

Very often, Mosfet gates are driven directly by a logic signal pin... which means you are not limiting the current to the Gate pin. The reason being is that the gate pin is not driven by current... ( its driven by voltage, unlike a BASE pin on other types of transistors.

But, the very thing that isolates the gate from the rest of the transistor (a capacitive GAP) can become a concern when the gate is constantly driven by high frequency. Its that capacitance that may be enough to to hang on to a gate charge.. and then you have it not doing what you asked. Normally, its not an issue... but you should consult the datasheet to see the relevant values.

Zip_Ferndale:
The other parameter you need to consider is the total gate capacitance as that affects how fast PWM can be used (depending on the gate drive current).

How would i do this?

By calculating the RC time constant and thus the rise/fall time of gate voltage - during rise and fall the
device is changing from on to off and back again and can dissipate a lot of power (upto about 1/4 of the
load power). You want this time to be a small proportion of the total time, and thus at higher PWM
frequencies you need to ensure shorter RC time constant in the gate-drive circuit. This usually means
using a MOSFET gate driver chip at higher PWM frequencies as gate drivers have low resistance (a
few ohms upto a few tens of ohms, typically).

Say you have 1k ohm gate resistor and a MOSFET with 1nF gate capacitance, it's time constant is 1
microsecond, which would be fine for 1kHz PWM, but pushing it for 50kHz PWM where there's only 10us
on average between transitions.

Capacitance isn't quite the right thing as gate charge is non-linear, and more thorough calculations
make use of the gate charge/voltage graph in the datasheet.

Factoring the Nano will drive the LED with PWM at 99.9% its capacity, what are good resistor and capacitor values for testing?

Full disclosure: I'm terrible with analog, and I don't know how to use the data or graphs in the datasheets.
There is no value for the "gate capacitance" in the datasheet. There's input, output and reverse transfer. Is it one of these?

irlb8721pbf.pdf (271 KB)

I guess the question I would ask is: do you have an idea what speed you want for the PWM?

I did some research for an LED project and found above 120 Hz was generally acceptable. I ran at ~253Hz to be sure the PWM could not be seen and to stay away from multiples of 60Hz (or 50Hz in the UK)

The odd number was just how the Arduino timer / counter setup worked out.

If you can run and these relatively low frequencies then I think it is safe to not worry about gate capacitance. If you need a much higher frequency then you'll have to MarkT's lead.

John

I prefer links to a datasheet over having it attached. With it attached, I must download it and open it, then delete. Much simpler with a link, it opens in my browser and the browser will push it out of the cache.

Anyway... Gate Charge is what you are looking for. Top of page 1, QG. Also on page 2, about halfway down. 7.6nC (nano coulombs).

A coulomb passing a point in one second is a current of 1 ampere.

So if you provide 20mA of charging current from an Arduino digital pin, 20mA is 20mC/s so 7.6nC/20mC/s = 3.8e-7 or 380ns. That is an extreme simplification. The maximum allowed output source or sink from an Arduino pin is 40mA. If you limit that with a resistor, it is not a constant current but will taper off and so the time to charge/discharge will be longer. As a rough guess, I'd say call it 5x that as we're not trying to calculate exact numbers, but instead to prevent excess heat.

If this was just a square wave, you'd be done. Maximum frequency set by making sure the switching time is no more than 1/10 of the period as an absolute. However, with PWM, your switch may be on for 1% of the time, which means timing equal to 100 times the PWM frequency. So a 490Hz PWM at 1% or 99% has the switching time requirements of a 49kHz square wave.

I'm spitballing here. Anyone want to jump in with more exacting math?

JohnRob:
I ran at ~253Hz to be sure the PWM could not be seen and to stay away from multiples of 60Hz (or 50Hz in the UK)

Interesting.

What consequence do you anticipate from the LEDs blinking at a multiple of the line frequency?