When to use a heat sink?

Hi,

I am new to electronics, so this is probably a very basic question.

I know the short answer is 'when things are getting hot', so I've got that down. What I would like to understand are the parameters or scenarios that you can identify during planning where a heat sink will be required, and how to decide on the type and usage of a heat sink.

Thanks!

You work out the wattage a component generates and then decide how hot you think you can live with then find a suitable shaped and fitting heatsink with an equivilent W/degC rating to achive the temperature you want. Alternatively you do a trial and stick you finger on the component when its running, if it burns your finger you stick a heatsink on it…

Heat sinks are fairly inexpensive so get some when ordering components. My rule of thumb is that if it sizzles spit, you can smell it, or it makes a small blister on your finger, a heat sink is in order.

Most linear regulator chips have an internal self-protection circuit if the device is going over temperature or over it's maximum output current rating. If it gets too hot is just starts cutting off, folding back output voltage or whatever to keep it in it's safe operating area.

So as long as the output voltage is being maintained at the rated regulation voltage then the self-protection is not being activated. It is still good to not run the regulator hotter then needed for long term reliability, but that is best done by lowering the input voltage to the regulator's input such that the regulator has less heat to dissipate to begin with. The best input voltage is that which is just higher then the minimum drop out specification while still maintaining the nominal rated +5vdc output voltage.

Lefty

If you want to go beyond "feel it and see", you get into some pretty major physics calculations that include the thermal conductivity of metals, mass, surface area, air flow, junction compounds (thermal grease, etc) and a few other areas I really don't understand. I've seen very expensive software packages that do nothing but design heat management systems for electronics applications.

Ironically, it's all totally unnecessary for most DIY applications. Overkill is usually acceptable and may only affect the physical size of the project. As long as you are keeping it cool enough, you should be fine.

There's probably some sort of middle ground here, but I haven't seen it.

I think the short answer is work out the power P that you need to dissipate, and then the maximum temperature rise above ambient (dT) you can tolerate (20 C is a reasonable default), then you need a heat sink rated at

dT / P degs C / Watt or less

degrees C per Watt is a measure of thermal resistance, so smaller values are a better heat-sink.

I have answered this a couple of times on different sites
so here is my “off-the-shelf” answer for thermal calculations.
IIRC the question was for temperature rise of a MOSFET
or BJT but the concept applies to any device.

To determine the temperature rise you need to calculate the junction
to ambient thermal resistance (Tja) —

Tja = Tjc + Tcs + Tsa all the dimensions are DegC/W

Tjc thermal resistance between the IC junction and IC case. This is
from the IC datasheet.

Tcs thermal resistance between the IC case and your heatsink. This
should also be in the IC datasheet. Usually you place a thermal
compound between the IC case and the heat sink to decrease the
thermal resistance. The decrease occurs because the thermal
compound fills the air gaps between the two surfaces and conducts
heat better than the air.

Tsa thermal resistance between the heatsink and ambient air. This is
in the heatsink datasheet. Most heatsink datasheets specify Tsa
with and without airflow. You can determine if you need a fan by
reviewing this number.

For a single heatsource the junction temperature would be

Tj = Tja * P + Tamb (DegC)

where P is the power dissipated in the IC and Ta is the ambient
temperature. For a MOSFET P = rds(on) * Id (NB: use the rds(on)
value at the actually operating temperature). For BJT
P = Vce(sat) * Ic (NB: use the Vce(sat) value at the operating
temperature).

The maximum juction temperature is listed in the datasheet (usually
in the “Absolute Maximum” section). I would not run the device at
a temperature greater than 80% of the the absolute maximum rating.

Unless you have very short on times PWM is not going to help
much with power dissipation. To see how much look at the transient
thermal response graphs in the datasheet.

To determine the temperature rise you need to calculate the junction
to ambient thermal resistance (Tja) —

Tja = Tjc + Tcs + Tsa all the dimensions are DegC/W

Tjc thermal resistance between the IC junction and IC case. This is
from the IC datasheet. You need to adjust this number using
the data in the transient thermal response curves. For short
on times and low duty cycle there is a reduction in thermal
resistance.

Tcs thermal resistance between the IC case and your heatsink. This
should also be in the IC datasheet. Usually you place a thermal
compound between the IC case and the heat sink to decrease the
thermal resistance. The decrease occurs because the thermal
compound fills the air gaps between the two surfaces and conducts
heat better than the air. Very thin layer just enough to fill
the gaps.

Tsa thermal resistance between the heatsink and ambient air. This is
in the heatsink datasheet. Most heatsink datasheets specify Tsa
with and without airflow. You can determine if you need a fan by
reviewing this number.

For a single heatsource the junction temperature would be

Tj = Tja * P + Tamb (DegC)

where P is the power dissipated in the IC and Ta is the ambient
temperature. For a MOSFET P = rds(on) * Id (NB: use the rds(on)
value at the actually operating temperature). For BJT
P = Vce(sat) * Ic (NB: use the Vce(sat) value at the operating
temperature).

The maximum juction temperature is listed in the datasheet (usually
in the “Absolute Maximum” section). I would not run the device at
a temperature greater than 80% of the the absolute maximum rating.

(* jcl *)


www: http://www.wiblocks.com
twitter: http://twitter.com/wiblocks
blog: http://luciani.org