How hot are your key components getting? There’s a good chance you’ve built a project and thought: “Well I guess I better slap a heat sink in there to be safe”. But when working on a more refined build you really need to calculate heat dissipation to ensure reliability. This is actually not tough at all. The numbers are right there in the datasheet. Yes, that datasheet packed with number, figures, tables, graphs, slogans, marketing statements, order numbers… you know right where to look, don’t you?
Hackaday has you covered on this one. In under 10 minutes [Bil Herd] will not only show how easy these calculations are, he’ll tell you where to look in the datasheets to get the info you need quickly.
Above, [Bil] used his bench as a whiteboard to illustrate the thermal resistance equation. In this case each resistor symbol represents part of the heat dissipation. You must consider all places where heat can be transferred: (from left to right) the component die (junction) to the component case, the component case to a heat sink, and the heat sink to ambient air. He illustrates each of these dissipation points in the video.
An example of the junction-to-case is shown to the right. This is a TO-3 case which has had the lid cut off. It’s a much simpler way to look at a chip die than trying to decap a component with a plastic case.
Make with the Math Already!
Okay, okay, we’re getting there. The math is not hard… just multiplication and addition, so hang on a minute more.
Gather the following values: maximum power you plan to use with this component, maximum heat rating of the part, maximum ambient air temperature in which this component will be used, and the theta values from the datasheets. Theta, which is a measure of degrees per watt, is often listed as a symbol: Θ Multiply theta by the max wattage and you will know how much temperature to add to your equation
Datasheets: Finding Θ and Temperature
Because [Bil] does such a great job in the video we’re giving you the quick version here. Temperature generating components will include a maximum operating temperature like the one shown below (click through for full datasheet) which is for a linear regulator:
The theta for “Juntion-to-Case” is found a bit further down the same datasheet in the Electrical Characteristics table. Datasheets will also provide a “Junction-to-Ambient” value (also shown below but not used in our calculations) used to calculate how much power you can use without any type of active or passive cooling. This answers the question of: “do I need a heat sink?”.
Finally, you want to look at values from the heat sink being used. [Bil] looks at the datasheet of a heat sink which lists a thermal resistance of 25.8Θ with the chart below on the left showing how that number may be altered with moving air (a fan). The chart to the right covers the use of interface agents like thermal grease, and a mica pad (for electrical insulation) with thermal grease. Both of those values are circled but only one will be used in the calculation.
Putting It All Together
If we assume an ambient air temperature of 38 C (100 F) and a maximum power of 2 W all of the numbers we need have been collected.
Max Temp = Junction + Mica/Grease + Heat Sink + Ambient
Max Temp = (4Θ * 2W)º + (0.4Θ * 2W)º + (25Θ * 2W)º + 38º
Max Temp = 8º + 0.8º + 50º + 38º
Max Temp = 96.8º
The maximum temperature rating for this part is 125 C, which means that this part is being properly cooled. [Bil] goes one step further in the video, showing how to calculate how much more reliable the properly cooled part will be.
- Texas Instruments LM317-N Datasheet (PDF)
- Texas Instruments App Bulletin for mounting TO-3 packages (PDF)
- AAVID THERMALLOY 7173DG heat sink example
44 thoughts on “Hot Or Not? Find Out How To Calculate Component Heat And Why You Should”
If in doubt, double the heatsink area.
Sounds like a wise policy to me.
Wasting money on extra heatsink capacity is a wise policy? Making your project bigger than it needs to be is a wise policy? i think doing the math and getting the heat sink right is the wise policy. Heat is not voodoo magic.
It’s also the reason most old things lasted almost forever, while the new stuff almost doesn’t survive warranty, if a screw needs to be at least 1 mm thick to hold a part, that’s what will be used, if the screw rusts a bit or was 0.1 mm too thin from a manufacture defect it will break, but hey you saved the company $0.01 per screw, that will look good on the report, the warranty returns can be blamed on the crappy chinese assambler, that did you asked them to do, so at least you are safe, the company can take the bad rep if the bean counters are happy, no loss here
And in the case of laptop computers, it’s EXACTLY this reason: They only last for a few years because they save on cooling capacity by using less heatsink metal, which is of course bulky, expensive and heavy.
Computers used to be able to encode movies all day long, now it’s normal for your (very) expensive ‘gaming’ laptop to overheat without external fans, even while doing mundane tasks, blame a fruit company for starting a heavy fad of form over function and the general public for going along with it
I was totally with you until you brought up Apple. What do they have to do with this? My previous laptop was a Lenovo Ideapad that literally melted from the heat. I am now using a MacBook Pro and it is extremely well-designed. I would argue that the thermal system on Apple products is consistently better than the alternatives.
In fact, I would say the same is true for audio hardware, displays, and most other hardware components.
There’s a lot of reasons stuff “doesn’t last as long.” Some of it nostalgia – tubes burn out super fast, you know. Some of it is that old transistors were physically larger. More silicone, more insulating layer, more metals… Overvoltage and transient conditions are nowhere as damaging. They have more mass to dissipate instantaneous heat (from short bursts of activity) and keep their temperature down.
And of course, some of it is the reduction in weight of supporting material. The heatsinks, the fasteners, the frame, it’s all pared down. THAT is because we like being able to carry our electronics with our hands, and not a bloody forklift. The idea that laptop companies use super tiny fasteners ‘to save money’ is laughable to me. They do it because it lets them build the bloody things smaller and lighter.
Sorry did retail tech support for a while some years ago, still have a sore image of the company over the reasons for denial of warranty and general assembly quality. screens going gray or orange and they blame the manufacturer while doing nothing for over a year, hdds with only 1 year warranty even when the law here is 2 years, aging white plastic macbooks that melted, deformed or discolored for lack of proper cooling, where they replaced the fan but left the plastics has they where only cosmetic even if the part of the keyboard is deformed, supplied a batch of defective i7 macbook pros and requested that our company do a firmware ‘update’ that lowered the i7 to i5 specs to avoid overheating. it was pretty pathetic. But its true it’s not only them, saw a lot of companies that i respected lowering their quality and having shady practices, some brands (toshiba and asus if memory serves me well) would take a returned laptop and for a small fee would repackage it to look new.
You mentioned audio and displays, and i have to agree with you and tekkieneet (bellow), looks like nobody makes a decent psu anymore, crappy chinese capacitors, sometimes even touching heatsinks, way underrated diodes, lack of distance between primary and secondary sides of main transformer, passive cooling in completely closed spaces with ICs running > 90C, it’s nuts how these get approved.
tl;dr Never buy a laptop/anything from retail from a store where you don’t have a friend to check the system for unusual return levels (bad batches) and run the serial number in the support system to make sure you are actually receiving a new machine
I think another thing that needs to be touched on is the fact that electronics seem to be produced as being disposable these days. So I’m sure it would be reasonable to assume that companies these days don’t manufacture their products to last because they believe people will already by lining up to buy the newest greatest thing. It seems like a vicious circle to me. They decrease the retail of the product so people can “afford” something new. But by decreasing the retail they are essentially are decreasing the quality because there is no way they would ever want to shrink their profit margin.
Not so sure if crappy components are the only reason why things fail a lot sooner these days.
A lot of things fail because of they are considered to be “mundane” stuff for a typical designer’s or parts purchaser’s these days. e.g. power supply related issues or bad caps or bad mechanical part. The rest of the product that don’t have crappy alternatives tend not to be the cause of failures (although they could be victims).
Most designers don’t even know about (nor seem to care about) Temperature rating, ESR, Ripple current rating, Lifetime etc nor they are aware to not put the capacitors right next to or down stream of hot objects like heat sinks inside a power supply. etc Experience is irreplaceable in that area, but companies opt to have new grads as they are “cheaper”.
In the long run, making crappy products damages the “Brand” that (only) the marketing people cares about. That costs a lot more than the $0.01 you are trying to save on crappy capacitors. They should at least try to make they last beyond the attention span of the consumers.
Waste money? Who buys heatsinks? I got a whole box of them I’ve salvaged. I’ve got acres of space too so volume isn’t really an issue for me either. Quit projecting your inadequacies onto others and life might improve.
even just building 10 units that need reliability we buy heatsinks instead of using the junk parts bin, unless we can find 10 same model units that can do the job, it’s simply not practical to make a pcb that can fit all the different heatsinks and then test all of them seperatly
@nelsontb I do not usually make multiples of projects but when I do I make custom heatsinks out of larger ones. All it took was sawing one
What can I say? I have a really nice band saw.
The units in OP are inconsistent. Some uses “38 C” and some uses “38º”. 38º is for angle measurement or is it referring to C or even F?
My personal preference to use ºC/W (and not Θ) in my calculation (for dimensional analysis).
your example: (4Θ * 2W)º – Not even sure what that means at all???
on the other hand: (4ºC/W * 2W) gives 8ºC
The W units cancels out in the calculation, the temperature shows up as ºC which makes a lot more sense.
Dimensional analysis really is the best thing.
Excellent point, the units are always always always more explicit this way. Too often I’ve hosed something up employing shorthand / abstractions. The voice of my 3rd grade math teacher comes to mind “Don’t skip steps and always show your work”. :)
BTW (4Θ * 2W)º is numerically correct, but it make a mess of units.
> degrees per watt, is often listed as a symbol: Θ
So that expression have a unit of ºCº (an extra º) which make absolute no sense for the equation. Either you leave out all the units inside the bracket (which I don’t recommend especially when you are explaining things to people) or you leave everything in and *not* add units to the whole thing.
There are only so many Greek alphabets and some of them are overloaded with different meanings. Looking hip can make a mess of things.
I sort of learn some of that in grade school doing factions. I went into it again when the professor in a Physic class was very much into it. Kind of disappointed to see them dropping the units as soon as they plug in an equation in my engineering classes.
Just being able to pick out nonsense from a equation *without* even knowing enough about it is very useful in real life. I found a few of those in datasheet!
Can anyone verify that the calculations in this video are correct? I question it, because if accurate it would mean that even if you had a perfect heat sink on the chip, the chip temperature would still rise to ambient + Jc.
Also, what if the ambient temperature is below Jc? The temperature wouldn’t actually decrease until the ambient temperature was negative. But what’s special about a negative temperature versus a temperature that’s simply below that of the chip?
I didn’t bother to watch the video.
From your context, You are talking about a delta T(JC) for first paragraph.
A “perfect” (or infinite sized) heatsink would have a Theta Heatsink to Air = 0ºC/W. So basically temperature at heatsink is the same as that of the air surround it.
The junction temperature by definition be the ambient temperature + Junction to ambient difference. The temperature difference is caused by thermal resistance and the amount of power between transfer between the two. There are still the thermal resistance between the die (junction), the case, heat sink compound that is not accounted for.
For the second paragraph, you’ll need to clarify what you are talking about. JC by itself only say Junction to Case and nothing about whether it is resistance or temperature difference. I can only guess here.
If your chip is colder than the ambient (e.g. the cold side of a peltier cooler). then there will be heat flow into the chip. The power would therefore have a negative sign and the equation should work out.
Not too sure about Theta heatsink to ambient whether it would be the same as air current (convection) is going a different direction.
Indeed, even with a perfect heatsink, the junction would still be hotter than ambient, due to the junction-to-case thermal resistance.
You’re right. I guess I didn’t really have a clear picture in my head before about what R0jc really meant.
Read the Flir thermal imaging camera hacks on this website, buy a camera for a grand, and just point and shoot.
We bought one for our lab at work and it more than paid for itself already. There is theory and paperwork and then there is actual world. I’ve been doing this for almost 30 years and I hate to tell you this but sometimes (GASP!) Datasheets lie, or if the components come from China, they just cut and paste and the data does NOT line up.
I would recommend doing both math and real temperature measurement. If the two are not even close you know something is wrong with the system.
Nicely done Bil, explanation was accurate. Glad you threw in the reliability discussion though in my experience most temperature accelerated mechanisms are 0.7eV activation energy rather than 0.99eV. 0.99eV will give you worst case degradation in reliability for operation at elevated temperatures versus a lower temperature while 0.7eV provides most conservative improvement of reliability for operation at lower temperatures versus a higher temperature.
Also like the 2112 t-shirt!
Thanks phase. Basically what you said about ,7 was going through my head when I mentioned not going into Boltzman’s. In this case I scarfed the example directly from page 3 of http://www.ti.com/lit/an/snva509a/snva509a.pdf as a clean example which is probably why they chose .9 also. This way I am quoting TI rather than proving I need new reading glasses when extracting from the graph. >:)
Or you apply the heatsink compound the wrong way (*) and the numbers would be very different.
Still having the basic knowledge and telling people to read more section of the datasheet instead of blindly using parts or getting technical info from 3rd party like youtube video or a blog is bad.
One of the places, I have access to a FLIR camera and I get very similar results on my board as my paper calculations. (There were people who thought my calculations was pessimistic.)
I have access to my fingertip. I can usually tell with it if something is really hot, or not.
Can I borrow your fingers next time when I am trying to find out the temperature of a transistors running on the primary side with 170V DC *while* the power supply is on or when one that is under fault conditions?
There is also the documentation side of things that I can put the picture of whether your finger is burnt in a design and verification report. BTW I put an outline of board components overlay on top in the FLIR picture. It took about same amount of time for a full set of data under various load vs just one datum doing the old way of measuring individual critical components with taping thermal couples over various place without shorting them out.
Shut it off then touch it if you’re that big of a pussy.
That doesn’t sound like a professional approach way of writing a verification report. Measurements with uncalibrated methods and non-reproducible results are useless and waste of time.
There is nothing gain for chest pounding and you are being a liability for people you work for. You “recommendation” is of no use.
BTW I got an even better trick for the HaD crowd.
Stick a piece of thermal printer paper onto each of the parts on a board to tell if the chips are getting hot. That doesn’t tell you how hot the actual temperatures are.
True on the finger tip probe but one thing the math shows us it’s possible to burn your finger on something that is working correctly. (been there, done that. Had an imprint on one finger tip for a couple of years)
There are some cases where you intentionally run a component hot.
e.g. Schottky diodes in DC application like OR’ing or reverse polarity protection. The forward voltage can drop a few tens of millivolts if you keep the junction temperature hot. If the power feed drops, the diode junction would cool off so the leakage current would drop as well.
60C case temperature isn’t “hot” for some embedded applications that use parts with automotive/mil specs. I would expect that your finger probes would start top feel pain above that as you can get a 3rd degree burn at 60C with 5 second contact. I don’t know how well the burn chart translate to finger probes.
I don’t do jackass shit like that.
What if you have a heatsink that doesn’t have a datasheet? Can you calculate some usable specification from its dimensions and material?
Yes you can, I couldn’t find the TI Appnote that I used to use that was fairly straight forward approximation by taking the size and finish of the metal.
With that said, the reason I showed a “standard TO-220 heatsink” was to help get an intuitive idea of what the effectiveness of a common heatsink is and when you might need one. Next time you see that something may run hot it may be useful to remember that you can probably deal with 2cWatts and that you need a little extra room for your heatsink.
Also if you go to Thermalloy’s website (or other heat sink mfgs) you may be able to pick a heatsink that matches your mystery heatsink and get an idea. And then multiply time .75 in case your off a bit.
Theoretically yes, but all the easy ways to calculate it are quite inaccurate. The best way is to bolt a power resistor on it and measure the temperature rise.
In principle, you take the mean path lenght through the bulk material of the heatsink to all points along the surface to calculate the thermal resistance of the heatsink, or some approximation of it. Then you add to it the metal-air interface thermal resistance according to the total surface area exposed to moving air as follows:
I should have also included that by all means you can get a temperature probe out and measure it as a system/environment to take into account other variables such as airflow, etc. If you know the power being dissipated you can then just solve from case to junction.
I once had a boss/president who decided that we would put a fan inside an airtight Hoffman enclosure. We tried to explain to him that all he was doing was stirring the same air and heating it further in the process. In that example I would have had to take a measurement as tangible.
Although a fan inside a closed box can mean the difference between overheating and working, because stationary air is an insulator and moving air carries heat between the components and the inner surface of the box.
If it is a DIY one off project, measure it. I bought a thermometer with type K thermal couple probe for $3 something recently. The cheap probe is decent and can easily do 175C or so for quick measurement. I tested it against melting ice/boiling water and it seems accurate enough for DIY project.
For anything else, it is worth while to spend the $$$ and buy something with a datasheet.
instead of “maximum power you plan to use with this component” should be “maximum power _dissipated_ with this component”. It’s a HUGE difference.
E. g. Sharp S116S02 optorelay. The datasheet says “Thermal resistance between junction and ambient 40ºC/W”. Load it with 100W (e.g. halogen lamp). Room temp. 20ºC. Calculations would be: 40ºC/W * 100W + 20ºC = 4020ºC (!)
um..(please pardon a potentially dumb question and tell me why? why? why?). When I look at this datasheet: http://www.nxp.com/documents/data_sheet/74HC_HCT125.pdf
it does not have the thermal resistance value stated. It gives max ambient temp (125C) as well as total power dissipation (500mA for the chip I’m looking at). So does that mean as long as the current is < 500 mA, no heat sink is needed? (and thus I don't have anything to calculate…) or is there additional info I'm not considering?
I think they don’t have it in there because there is no situation where that IC (or AFAIK any of the 7400 or 4000 series) would need a heatsink. If you are doing something to it that makes it get that hot, you’re using it wrong.
Please be kind and respectful to help make the comments section excellent. (Comment Policy)