6

If we have a heat sink with fins, for example like this enter image description here how does the efficiency vary between horizontal and with the fins vertical in non forced air?

Dirk Bruere
  • 13,425
  • 9
  • 53
  • 111
  • 1
    This depends on the fin depth. See e.g. http://www.thermalsoftware.com/vert_vs_horz_sink.pdf – Janka Jul 15 '18 at 14:59
  • If air is unrestricted in flow , vertical is best – Tony Stewart EE75 Jul 15 '18 at 15:01
  • 3
    The old SWR series of bass instrument amplifiers had a horizontally-mounted heatsink inside, with no active cooling. Needless to say, the output transistors fail extremely often. – rdtsc Jul 15 '18 at 15:34
  • 3
    Per a comment I added to Ali's answer below, blackbody radiation can often account for perhaps 1/3rd of the heat removal of a heat sink in a well-designed, but non-forced-air situation. Anodizing the aluminum is important in these cases. – jonk Jul 15 '18 at 18:59

1 Answers1

10

The difference should be pretty dramatic, due to the specifics of free convection airflow.

When the fins are placed vertically, the air flows along all fin surfaces, hot air rises and provides the best possible heat exchange. The active surface will be on both sides of the fins, and along the surface of the baseplate. So the effective heat exchange surface is the the entire fin surface, which is 5-6 times bigger than the surface of the heatsink's baseplate.

When the fins are positioned horizontally, the rising air has to flow across the fins, with pockets of stalled air between fins. So, effectively, the active surface of the heat sink will be of the size of the heatsink's baseplate. Of course, there will be some escape of hot air from these pockets, but the rate of movement is much slower than when the fins are vertical, with no obstruction for air to escape. So the thermal impedance of a heatsink with vertical fins will be 2-3 times better (smaller) than when the fins are horizontal.

Also please note the distance between fins - the fins are wide spaced, as compared to forced-air sinks. This is done so that the boundary layers around the fins' surfaces don't overlap along the space between fins, and the heat exchange is optimal. A heat sink with tightly-spaced fins would behave as a solid brick under unforced air conditions and won't be very efficient.

ADDITION: The link supplied by Janka in the comments contains simulation data in support of my hand-waving explanation:

Graph showing much better efficiency for heatsinks with vertical fins

ADDITION2: Note that the above difference was modeled assuming the surface being bare metal with emissivity of 0.1. The effect of orientation of this heat sink relative to gravity field will be offset by the fact that about 1/3 of heat flux will be emitted in the form of radiation if the sink is anodized or painted black, which will make the surface emissivity to above 0.9. And the radiation is omnidirectional. The other factor that would diminish the effect of orientation is the material of the sink and construction of fins. Thin fins don't conduct/transport the heat well from baseplate to edges, and the edges will have less temperature difference to ambient, be less effective. Copper will work better and make more difference. There are several factors at play in different directions, so the exact effect of orientation is difficult to calculate, and only a sophisticated software as FlowTherm or similar can give a trusted result. Or an experiment. Still a factor of 2 will be a good ballpark estimate.

Ale..chenski
  • 38,845
  • 3
  • 38
  • 103
  • Thermal conductance will improve upto 5x up to 2m/s over fins regardless of how restricted aperture and CFM rate with forced air so even low velocity convection helps a bit. But if vertical flow is blocked then the gain is small – Tony Stewart EE75 Jul 15 '18 at 16:53
  • 2
    I worked on a project ca. 1988 involving a video card with a forest of vertically mounted (zig-zag inline package) memory chips installed in a portable computer. The card was intended for use in a desktop type PC (horizontal backplane, vertical card installation, horizontal chip packages, poor convection properties), but in the portable, the backplane was vertical, the card horizontal, the memory chips pointed *downward*, and the convection cooling was *abysmal*. Overheating invariably corrupted the video memory contents within an hour of uptime. – Russell Borogove Jul 15 '18 at 17:38
  • 1
    @RussellBorogove, ATX PC tower cases with its placement of PCI cards (components appear on PCB bottom) represent the worst case for cooling by natural convection. It was a really poor thermal engineering. Only a good case ventilation (flow-through) can help. – Ale..chenski Jul 15 '18 at 17:47
  • Ali, it would add further value to your answer (I'm not interested in adding one) to include some information regarding emissivity (anodizing the aluminum, for example) as a significant factor here. In general, if natural convection (not forced) is involved, radiation can amount to 1/3rd of the performance (convection being the remaining 2/3rds of it.) Forced convection makes the radiation component far less relevant, of course. Y. Shabany has published work in this area in 2008, for example: [Radiation Heat Transfer from Plate-Fin Heat Sinks](https://ieeexplore.ieee.org/document/4509379/). – jonk Jul 15 '18 at 18:51
  • @jonk, you are right, the simulations assume bare aluminum with 0.1 emissivity. But I am not sure if this site is a good place for comprehensive thermal engineering course. We can also consider fin geometry and sink material conductivity as another factor (ends of fins have less temperature and thus are less efficient), baseplate heat spreaders (or base thickness, because of point-mounted transistors), etc. But I agree, including this information would increase "awareness and inclusion" of this thermal engineering matter. – Ale..chenski Jul 15 '18 at 19:18
  • My reasoning here is *mostly* that for natural convection, the emissivity ***is*** a significant part of the action. If it were a mere 1% or something, I wouldn't bring it up. But it's a third. Being *comprehensive* is a matter for academicians. Granted. But this is actually ***practical*** and therefore very much ***engineering*** and in this case it's large enough that it shouldn't be missed. – jonk Jul 15 '18 at 20:56
  • @jonk emissivity is typically not brought into a discussion of heatsink orientation. While radiation may contribute up to 25%, emissivity, or more accurately apparent emissivity, is only one factor in total radiation. The more effective the heatsink, based on dimensional characteristics, the less radiation will influence heat transfer. Radiation is most important when dealing with a flat plate or flat surface. As you add fins and optimize fin size and spacing, radiation become less important. Radiation is from one object to another. When the other object is another fin radiation is moot. – Misunderstood Jul 22 '18 at 23:41
  • @Misunderstood I can provide some references which illustrate, though experiment, the differences it can make when natural convection is used and not forced. It's not the dominant factor. But given that it can account for 1/3rd, it isn't to be entirely ignored either. Of course, the circumstances also matter. If it is enclosed within a box, it amounts to an integrating sphere regardless. So it's not important, then. So circumstances do matter a lot with respect to my comments. – jonk Jul 22 '18 at 23:49
  • Your answer is oriented toward horizontal fins and neglects plate orientation. It is covered well in your link. Also the discussion of fin spacing, fin height, and fin length is very dependent upon whether the plate is horizontal or vertical. – Misunderstood Jul 22 '18 at 23:52
  • @jonk I would greatly appreciate any test results on anodized vs. unanodized aluminum heatsinks. I just purchased some inexpensive unanodized heatsinks from heatsinkusa.com (1" wide vs 12" wide cut to 1") and am debating whether to get them anodized and retest to compare the experimental difference. I am trying to model the cost for thermal management of horticulture LED strips. Do I use a 25 mm wide strip vs 9 mm wide with heatsink? When the strip is over the canopy vs. side lighting how important is plate orientation vs. the cost to change orientation and/or top or bottom side mounting. – Misunderstood Jul 23 '18 at 00:27
  • @Misunderstood, did you try Google? https://www.heatsinkcalculator.com/blog/the-importance-of-radiation-in-heat-sink-design/ Table 3 two bottom lines, delta T is improved by 30 C if the surface is anodized. – Ale..chenski Jul 23 '18 at 01:23
  • @Misunderstood, re *"Your answer is oriented toward horizontal fins and neglects plate orientation."* No, pay attention to the submitted chart. It says "vertical[plate]/vertical[fins]" and "vertical[plate]/horizontal[fins]", although they could be more clear on that. Horizontal plate would be the worst case, and "horizontal" fin orientation would have no sense. – Ale..chenski Jul 23 '18 at 01:29
  • Did I try Google???? I have a plethora of research papers and text books on this topic. I am not looking for calculated values, I'm looking for empirical data. Now if I were to make a decision based on your link to heatsinkcalc I'd get myself in trouble. It's a horrible article. No fin height and no orientation. I had an account with them and I did run AB tests on anodized vs. not, the difference was typically less than 10%. If a heatsink is well designed for convection, radiation will have less impact. I have learned calculated data is no substitute for empirical. – Misunderstood Jul 23 '18 at 20:42
  • What I said was, in the words you wrote, you did not address vertical or horizontal plate only vertical fins. I said your link addressed it well and the chart is from that link. Yes I did see the vertical[plate]/vertical[fins] in the tiny print only because I was searching your answer for something on plate orientation before I brought it to your attention. A heatsink designed for vert/vert is not going to work as well as when used horiz/vert. An efficient horiz/vert can be designed. The design criteria for vert/vert is very different that for horiz/vert. – Misunderstood Jul 23 '18 at 21:03
  • @Misunderstood, horizontally-oriented plate makes no sense whatsoever for free convection mode, there could be no "design criteria" for nonsense. Please stop. If you are overly concerned, use omnidirectional heat sinks, https://cdn.shopify.com/s/files/1/1538/8585/products/DSC_0465.jpg?v=1527267517 – Ale..chenski Jul 23 '18 at 21:13
  • A horizontal vs vertical plate is taught in kindergarten physics. Look at Fig. 1 in your link. The middle drawing is a horizontal plate. Fig. 3 and 4 show there is very little difference between a horizontal plate and vertical plate as long as the fins are vertical. Obviously you did not read (or understand) the paper you linked to. BTW, "heat rises" does not explain buoyancy. – Misunderstood Jul 23 '18 at 21:45
  • @Misunderstood, okay, you got me. I managed to read the article. This paper is fairly bad. First, it is not based on 3-D solid-fluid solver, it uses approximation formulas obtained 50 years ago, with unknown materials and measuring methods. So all conclusions are fishy. Here is another largely ignorant article (Master Thesis), where 4mm fin gaps were used, with absurd results as well. file:///C:/Users/PC/Downloads/ubc_2015_may_Rana_Md.%20Ruhul.pdf All research of this kind was done 50-60 years ago, before Internet era, and new generation is re-discovering the stuff in a funny wrong way. – Ale..chenski Jul 24 '18 at 00:49
  • Nice link, C:? And no, the physics of heat transfer have not changed in the past 50 years. The definitive 1948 paper DISSIPATION OF HEAT BY FREE CONVECTION by W. ELENBAAS still holds true today. The paper you linked in is accurate, it is a rudimentary summation paper. It gives some necessary knowledge without going into the why. The why is too complex, and that is also why approximations are used. Knowledge is knowing tomato is a fruit. Wisdom is knowing not to use tomato in a fruit salad. See the 1930 paper on the temperature of a wire: https://electronics.stackexchange.com/a/384261/124009 – Misunderstood Jul 24 '18 at 02:07
  • @Misunderstood , try this, https://open.library.ubc.ca/media/download/pdf/24/1.0074423/1 – Ale..chenski Jul 24 '18 at 05:00
  • Let us [continue this discussion in chat](https://chat.stackexchange.com/rooms/80584/discussion-between-ali-chen-and-misunderstood). – Ale..chenski Jul 24 '18 at 05:15