Chassis as a heatsink. How good is it?

Discussion in 'General Electronics Chat' started by daviddeakin, Jul 19, 2010.

  1. daviddeakin

    Thread Starter Active Member

    Aug 6, 2009
    207
    27
    I have been looking around for an answer for this, but so far nothing conclusive.

    Does anyone have any idea roughly how good a typical chassis is as a heatsink?

    I'm thinking of, say, a 10x10cm section of 0.8mm thick metal, exposed to the air on the outside of the chassis, but with natural convection.

    I'd be interested to know how much power you would expect a typical TO-220 transistor to handle when bolted to that bit of chassis made of, say, stainless steel versus aluminium. Let's say with a junction temp rise of 75 degrees, and a junction-to-case resistance of 2 degree per watt,

    Since a small clip-on heat sink can usually muster about 3.5W dissipation, intuitively I want to say the steel chassis could manage perhaps 5W, and maybe 10W for aluminium, but really I'm just guessing.

    Any advances?
     
  2. Bernard

    AAC Fanatic!

    Aug 7, 2008
    4,170
    395
    Just my gut: SS .8mm- about as good as free air- no heat sink, Al .8mm 5W, 2mm 15W??
     
  3. beenthere

    Retired Moderator

    Apr 20, 2004
    15,815
    282
  4. daviddeakin

    Thread Starter Active Member

    Aug 6, 2009
    207
    27
    Does anyone know *roughly* how much heat transfer is improved by adding a typical 12V fan to blow air across a sink? I saw on a forum someone said that a fan can improve the net thermal resistance by 4 times, but to be honest, I'm sceptical.
     
  5. kingdano

    Member

    Apr 14, 2010
    377
    19
    when in doubt, test.

    hard to predict, thats why FEA programs are so expensive.

    i dont think rule of thumb stuff applies.
     
  6. Markd77

    Senior Member

    Sep 7, 2009
    2,803
    594
    I can believe it, my Semperon processor uses 60W but is running at 15 centigrade above ambient with a pretty small heatsink and fan. I'm not going to try stopping the fan but I think the temperature rise would be drastic.
     
  7. Ghar

    Active Member

    Mar 8, 2010
    655
    72
    Here's an example curve of thermal resistance vs air flow
    http://www.htrdltd.com/Performance_charts.htm

    That 'dimpled fin' one is almost 4 times better with massive air flow.

    A standard rpm 80mm computer case fan is about 30 to 40 CFM
     
  8. Papabravo

    Expert

    Feb 24, 2006
    10,135
    1,786
    Things like mass, surface area, color, material, and air flow all have their effects on heat dissipation. Experiments are just about the only inexpensive way to be sure. It is also hard to criticize overdesign in this area.
     
  9. someonesdad

    Senior Member

    Jul 7, 2009
    1,585
    141
    I took an engineering heat transfer course about 30 years ago and learned the ways MEs handle heat transfer situations like this. It wouldn't be terribly hard to come up with some reasonable estimates. The trickiest part would be estimating the convection transfer coefficients. Of course, all the details of doing that stuff have long since been lost out of my poor brain, so I couldn't give an off the cuff answer. But I vaguely remember things like Nusselt numbers and Grashoff numbers and other similar things...

    To solve the steady state problem, you'd have two problems: the two-dimensional conduction problem in the metal (solve the heat equation) and the convective losses to the atmosphere, which could probably be modeled with a number of small areas considering the convective loss in the small areas constant. I'd make the problem circularly symmetric to turn it into a one dimensional problem -- which probably is a good approximation.

    Since you'd want to check your answers anyway, you'd be building a prototype and measuring the temperatures vs. powers -- so the easiest thing would probably be to go straight to the prototype and forget the paper models.
     
  10. timrobbins

    Active Member

    Aug 29, 2009
    318
    16
  11. daviddeakin

    Thread Starter Active Member

    Aug 6, 2009
    207
    27
    Thanks for the insight guys! So it looks to me as though a four-fold improvement is possible, but only with very rapid air flow and the right sink. With a standard computer fan and a bog-standard sink, a 30% improvement looks reasonable.

    I don't suppose you know of a simple way to measure the thermal resistance of a hunk of metal? Say if I attached a power resistor to a chassis, and let it dissipate a known power level and measured the temperature rise of the resistor with a thermocouple, can I put a figure on the thermal resistance of the chassis?

    (Yeah, approximate I know, but not orders of magnitude off, surely?)
     
  12. loosewire

    AAC Fanatic!

    Apr 25, 2008
    1,584
    435
    Don't some design go Into a heat sink.
     
    Last edited: Aug 6, 2010
  13. marshallf3

    Well-Known Member

    Jul 26, 2010
    2,358
    201
    Plenty of standards of metals and formulas for that but sometimes we tend to get way too far into the engineering side of design when we should be focusing on attacking the problem more as a technician.

    Get a chassis box in the size you esitmate you'll be using then see how hot it gets. You can always inboard or outboard additional heat sink metal or increase the size of the box or where the component mounts. If I recall a 6" x 6" x 2" common aluminum chassis box is good for about 10W in practicality and the thickness is less important than the surface area exposed to the air. Thicker does allow the heat to migrate around the entire surface area of the box faster and I'm a big fan of the cast aluminum boxes, however they're often cost prohibitive. Also remember that even though there's no air flow inside an enclosure an internal heat sink on a component does indeed help to a greater degree than you might think.

    ASIDE: During my first round in college I worked part time at the local TV/Stereo/etc sales and service shop as their head TV tech. My roommate (no college but an FCC license) took care of the town's business band police, fire & commercial two-way radios. The store also carried the full line of GC/Calectro parts, ECG & Motorola HEP transistors for experimenters. If you're old nough you'll remember when Radio Shack actually had a few parts, the GC/Calectro line we carried put them to shame with walls and walls of experiementer parts so we got a lot of student business when project times came around.

    Together it seemed there was little we couldn't fix, and little we couldn't design if we put our heads together. Even though I was in engineering school we couldn't help but silently laugh when an engineering student would come in with a handful of paperwork telling us he nedd a 7.264K resistor, a 9.4 uF capacitor etc and when explained how components were commonly supplied to the industry some would even get upset, waving their slide rules around and insisting they had to have that value. One was so insitent I went through the explanation of how to take a carbon comp resistor of lesser value and carefully filing it with a triangular file until it reached his desired value.

    There exists a fine line between some "by the book" fresh engineers and those which have had enough real world experience to gain the honorable title of also being an electronics technician. It was once explained to me that engineers specify the theory and prototype technicians make the projects work - engineers says 7.264K while the technician uses that as a starting point, puts in a 10K pot and adjusts it to find the closest standard value that can be used.
     
    Last edited: Aug 6, 2010
  14. kingdano

    Member

    Apr 14, 2010
    377
    19
    Never ever underestimate the value of a good electronics technician, or technician in general.

    I did the technician job for 2 years out of school, and have been an EE by title for 1 year.

    I find that it is better to test these things, instead of trying to calculate them or elegantly design an experiment.

    Get a K/J probe attachment for a DVM, turn the thing on and take some data.

    Best way to do it, in my humble opinion
     
  15. Markd77

    Senior Member

    Sep 7, 2009
    2,803
    594
    If you want to get more heat out of an area you can use a heat pipe (I like the simple way they work) or a peltier effect device, although they use extra power to operate.
    Both improve heat flow but also need conventional heatsinks.
     
  16. daviddeakin

    Thread Starter Active Member

    Aug 6, 2009
    207
    27
    That's fair enough, but having measured the temperature, is there a way to estimate how hot the actual junction temperature is? I mean, if the surface temperature reaches 80 degrees, how can estimate whether the junction temp is less than 125 degrees (safe) or more than 125 degrees (lets the magic smoke out)?

    EDIT: Apart from simply burning out a few components with a cut-and-try method ;)
     
  17. kingdano

    Member

    Apr 14, 2010
    377
    19
    you can mount a T/C to the surface of the chip and see how toasty it gets based on the ambient temperature and presence/absence of a fan.
     
  18. marshallf3

    Well-Known Member

    Jul 26, 2010
    2,358
    201
    Component data sheets specify thermal characteristics from junction to case, you just have to do some backwards math to figure out where the junction temp is.

    In my opinion if it feels way too hot in operation then throw more heat sink at it, it's as simple as that.
     
  19. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    A relatively thin, flat piece of metal makes a pretty poor heatsink because of an effect called spreading resistance. It's better than nothing, but not great.

    A naked TO-220 in still air has a thermal resistamce of about 60C/W, so it can handle about 1 - 1.5W in most cases. Screwed down to a chassis I would be surprised if a flat chassis in still air did better than maybe 15C/W ballpark.

    Airflow helps but you have to run it and measure. many variables involved.
     
  20. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    Use any TO-220 transistor wired like a diode. Trickle about 10mA through it at room temp and measure the VBE. Force high current through it and set a known power dissipation (current times VBE) then abruptly flip the current back to 10mA and measure the VBE before it cools much. Use 2mV/C to calculate the temp rise and you will get a good ballpark measurement.
     
Loading...