For those planning chassis with large heasinks, a couple of not too theoretical or academic articles on optimising heatsink geometries might be useful (don't take the thermal resistance values of suppliers at face value, because they are quoted on several assumptions which may not apply to your case) :
http://www.thermalsoftware.com/vert_vs_horz_sink.pdf
http://powerelectronics.com/mag/power_optimize_fin_spacing/
http://www.aavidthermalloy.com/technical/correct.shtml
http://www.aavidthermalloy.com/products/euro_extrude/correct.shtml
The second article is actually quite useful, but there are many mistakes in the calculation examples. It is best to get the original from a library and use the figures there. Note that Fig. 2 also has degF & degC the wrong way round.
Patrick
http://www.thermalsoftware.com/vert_vs_horz_sink.pdf
http://powerelectronics.com/mag/power_optimize_fin_spacing/
http://www.aavidthermalloy.com/technical/correct.shtml
http://www.aavidthermalloy.com/products/euro_extrude/correct.shtml
The second article is actually quite useful, but there are many mistakes in the calculation examples. It is best to get the original from a library and use the figures there. Note that Fig. 2 also has degF & degC the wrong way round.
Patrick
Hi Patrick,
Thanks, I found those articles very interesting. I would have never realised that a horizontal heatsink with fins up would be so efficient. I wonder if you were forced to use horizontal fins, if cutting vertical slots (say 5 mm wide) or holes every 30 mm would make a significant improvement.
Have you researched the black body aspect of heatsinks? If I understand correctly, it is as significant (10%) as some of the points you have raised in the attached articles.
regards
Thanks, I found those articles very interesting. I would have never realised that a horizontal heatsink with fins up would be so efficient. I wonder if you were forced to use horizontal fins, if cutting vertical slots (say 5 mm wide) or holes every 30 mm would make a significant improvement.
Have you researched the black body aspect of heatsinks? If I understand correctly, it is as significant (10%) as some of the points you have raised in the attached articles.
regards
> I wonder if you were forced to use horizontal fins
No. I use vertical fins, but I have VERY long fins because I build "Mini" Towers. It is a problem to get suitable heatsinks within fin spacing optimised for heatsink height of 500mm (20").
> Have you researched the black body aspect of heatsinks?
If your temperature rise is less than 30degC, the radiation accounts for much less than 10%, so there is no need to use black anodising.
Also rippled fin surfaces are IMHO only effective for forced cooling. For natural convection, the boundary layer is probably thicker than the depth of ripples.
Patrick
No. I use vertical fins, but I have VERY long fins because I build "Mini" Towers. It is a problem to get suitable heatsinks within fin spacing optimised for heatsink height of 500mm (20").
> Have you researched the black body aspect of heatsinks?
If your temperature rise is less than 30degC, the radiation accounts for much less than 10%, so there is no need to use black anodising.
Also rippled fin surfaces are IMHO only effective for forced cooling. For natural convection, the boundary layer is probably thicker than the depth of ripples.
Patrick
Geoff Moss (Geoff here at DIY) found some information relating to the anodizing/paint/color of paint and posted it here once upon a time. If I recall correctly, there were some really strange things...green being better than black, for instance. Drove me crazy just to look at it. Other than searching Geoff's posts, I can't think of a good way to look for it. It was a long time ago, I can tell you that much.
The Aavid/Thermalloy online calculator is a very useful thing. I was thinking about using comparatively tall heatsinks for a project until I started playing with the numbers. Cured me of that idea. Basically, it comes down to this--the degree/watt figures are usually quoted for 3 inch lengths. It goes downhill rapidly from there with increasing height. Once you get past 6 or 8 inches high, you might as well quit because you're paying money, but you're not dissipating very much heat per dollar spent.
Exceptions made for aesthetic reasons, I suppose, but it's a pretty expensive way to build an enclosure just to have it look neat.
I've seen a number of pictures where members have attached their devices towards the top of the heatsink. This isn't particularly efficient. You're much better off attaching them about one-fourth to one-third of the way up from the bottom.
I'm sure some or all of this is covered in the links given, but I've got two screaming boys and have to run. I'll try to read the links later, but thought I'd toss in a few things for people who just glance at the thread and run.
Like me.
Grey
The Aavid/Thermalloy online calculator is a very useful thing. I was thinking about using comparatively tall heatsinks for a project until I started playing with the numbers. Cured me of that idea. Basically, it comes down to this--the degree/watt figures are usually quoted for 3 inch lengths. It goes downhill rapidly from there with increasing height. Once you get past 6 or 8 inches high, you might as well quit because you're paying money, but you're not dissipating very much heat per dollar spent.
Exceptions made for aesthetic reasons, I suppose, but it's a pretty expensive way to build an enclosure just to have it look neat.
I've seen a number of pictures where members have attached their devices towards the top of the heatsink. This isn't particularly efficient. You're much better off attaching them about one-fourth to one-third of the way up from the bottom.
I'm sure some or all of this is covered in the links given, but I've got two screaming boys and have to run. I'll try to read the links later, but thought I'd toss in a few things for people who just glance at the thread and run.
Like me.
Grey
> It goes downhill rapidly from there with increasing height. Once you get past 6 or 8 inches high, you might as well quit because you're paying money, but you're not dissipating very much heat per dollar spent.
That is not the entire truth.
The Aavid correction tables should be taken only as a guide, and not the replacement for proper 3-D thermal simulations. The effectiveness of a tall heatsink depends a lot on the fin spacing (see 2nd reference). You CAN optimise a heatsink to work at lengths up to 20 inches / 500mm.
Take a look at the XA100. It has long, horizontally oriented fins (against textbook recommendations). But it can still dissipate some 250W per side effectively, thanks to large fin spacings (to allow decent air flow). I estimate the spacing between fins to be about 20-25mm, perpendicular to fin walls.
Take another look at the heat resistance curve of e.g. R-Theta profile 60815. There is a significant difference in thermal resistance between 3"and 15". Again, it has a relative large fin spacing of about 19mm.
Could one of the Moderator please kindly correct the spelling mistake of the title ? It should of course be heaTsink.
Patrick
That is not the entire truth.
The Aavid correction tables should be taken only as a guide, and not the replacement for proper 3-D thermal simulations. The effectiveness of a tall heatsink depends a lot on the fin spacing (see 2nd reference). You CAN optimise a heatsink to work at lengths up to 20 inches / 500mm.
Take a look at the XA100. It has long, horizontally oriented fins (against textbook recommendations). But it can still dissipate some 250W per side effectively, thanks to large fin spacings (to allow decent air flow). I estimate the spacing between fins to be about 20-25mm, perpendicular to fin walls.
Take another look at the heat resistance curve of e.g. R-Theta profile 60815. There is a significant difference in thermal resistance between 3"and 15". Again, it has a relative large fin spacing of about 19mm.
Could one of the Moderator please kindly correct the spelling mistake of the title ? It should of course be heaTsink.
Patrick
Just because company XXX, YYY, or ZZZ does something doesn't mean it's the most efficient. People are far too easily impressed by commercial endeavors, thinking that if it sells for multiple thousands of dollars it's "obviously" best/optimized/wonderful. Ain't so.
Take the heatsinks Nelson uses, cut them in half (better yet, thirds), lengthwise, then turn them on end and see how much more heat you can dissipate. It's a big duh that they dissipate heat the way they are. But could they be used more efficiently? Yes. Nelson has not only electronic and thermal goals for his circuits, but aesthetic ones. To the extent that he's chosen to run his heatsinks "sideways" I daresay that it's for product differentiation in the market...and because he likes doing things differently simply for his own amusement. That's not the same thing as saying that it's the optimum thermal arrangement.
Jeez, Levinson had--maybe still for all I know--an amp that had heatsinks that must have been 18 or 20 inches high. Convection becomes nearly useless after the first few inches because the air rising along the heatsink surface becomes too hot to accept much more in the way of heat. After that it's pretty much limited to what they can radiate, which isn't nearly as efficient as having radiation and convection working together.
Grey
Take the heatsinks Nelson uses, cut them in half (better yet, thirds), lengthwise, then turn them on end and see how much more heat you can dissipate. It's a big duh that they dissipate heat the way they are. But could they be used more efficiently? Yes. Nelson has not only electronic and thermal goals for his circuits, but aesthetic ones. To the extent that he's chosen to run his heatsinks "sideways" I daresay that it's for product differentiation in the market...and because he likes doing things differently simply for his own amusement. That's not the same thing as saying that it's the optimum thermal arrangement.
Jeez, Levinson had--maybe still for all I know--an amp that had heatsinks that must have been 18 or 20 inches high. Convection becomes nearly useless after the first few inches because the air rising along the heatsink surface becomes too hot to accept much more in the way of heat. After that it's pretty much limited to what they can radiate, which isn't nearly as efficient as having radiation and convection working together.
Grey
Grey,
I don't think I said in my previous post that the XA100 is the best example to follow. I meerly pointed out a commercial product which has been successful in using long fins to dissipate heat effectively.
Of course it is true that MOST commercially available heatsink profiles are more efficient with vertical fins up to about 150mm, because they ARE designed for those lengths in mind. There is NO reason that long heatsinks cannot be made effective. Just that they have to be dimensioned accordingly, as in some industrial applications (and not hobbyist amplifiers). The only reason why most heatsinks do not dissipate more heat after 150mm is that the air space between the fins are too small and the air flow resistance becomes too high. I suggest you have a look at the second link in my first post, and maybe you might want to agree with me then.
Of course you can design your chassis around a 150mm high heatsink. I myself happen to have other design constraints.
Radiation cannot be relied upon to dissipate heat for low temperature difference.
Anyway, I hope the links will help people to understand a bit more about heatsinks and their thermal resistance ratings as specified by the manufacturer, and how that relates to their own requirements and applications.
Cheers,
Patrick
I don't think I said in my previous post that the XA100 is the best example to follow. I meerly pointed out a commercial product which has been successful in using long fins to dissipate heat effectively.
Of course it is true that MOST commercially available heatsink profiles are more efficient with vertical fins up to about 150mm, because they ARE designed for those lengths in mind. There is NO reason that long heatsinks cannot be made effective. Just that they have to be dimensioned accordingly, as in some industrial applications (and not hobbyist amplifiers). The only reason why most heatsinks do not dissipate more heat after 150mm is that the air space between the fins are too small and the air flow resistance becomes too high. I suggest you have a look at the second link in my first post, and maybe you might want to agree with me then.
Of course you can design your chassis around a 150mm high heatsink. I myself happen to have other design constraints.
Radiation cannot be relied upon to dissipate heat for low temperature difference.
Anyway, I hope the links will help people to understand a bit more about heatsinks and their thermal resistance ratings as specified by the manufacturer, and how that relates to their own requirements and applications.
Cheers,
Patrick
I think this answers the anodize question:
http://www.aavidthermalloy.com/products/extrusion/anodize.shtml
"As a thumb rule, if anodize is not required for aesthetic or corrosion protection, we suggest it only for small, open finned heat sinks in natural convection."
http://www.aavidthermalloy.com/products/extrusion/anodize.shtml
"As a thumb rule, if anodize is not required for aesthetic or corrosion protection, we suggest it only for small, open finned heat sinks in natural convection."
For those who want to understand the science behind a bit more, here are some further references :
1. Heat transfer coefficient chart for a rectangular finned flat surface
Belentepe, Y.C.
Conference Record of the 1986 IEEE Industry Applications Society Annual Meeting
(Cat. No.86CH2272-3), 1986, pp. 1632-5 vol.2, 2 vol. 1807 pp.
2. Fin thickness for an optimized natural convection array of rectangular fins
Bar-Cohen, A.
Transactions of the ASME. Journal of Heat Transfer, Aug. 1979, vol.101, no.3, pp. 564-6
3. Heat transfer from vertically finned plates by natural convection
Grubskii, E.V.
Journal of Engineering Physics, March 1972, vol.22, no.3, pp. 285-9
4. Optimum arrangement of rectangular fins on horizontal surfaces for free convection heat transfer
Jones, C.D.; Smith, L.F.
Transactions of the ASME. Series C, Journal of Heat Transfer, Feb. 1970, vol.92, no.1, pp. 6-10
Patrick
1. Heat transfer coefficient chart for a rectangular finned flat surface
Belentepe, Y.C.
Conference Record of the 1986 IEEE Industry Applications Society Annual Meeting
(Cat. No.86CH2272-3), 1986, pp. 1632-5 vol.2, 2 vol. 1807 pp.
2. Fin thickness for an optimized natural convection array of rectangular fins
Bar-Cohen, A.
Transactions of the ASME. Journal of Heat Transfer, Aug. 1979, vol.101, no.3, pp. 564-6
3. Heat transfer from vertically finned plates by natural convection
Grubskii, E.V.
Journal of Engineering Physics, March 1972, vol.22, no.3, pp. 285-9
4. Optimum arrangement of rectangular fins on horizontal surfaces for free convection heat transfer
Jones, C.D.; Smith, L.F.
Transactions of the ASME. Series C, Journal of Heat Transfer, Feb. 1970, vol.92, no.1, pp. 6-10
Patrick
The link below points to a very useful (free) site for calculating natural convection heatsink performances based on geometries :
http://www.frigprim.com/online/natconv_heatsink.html
Patrick
http://www.frigprim.com/online/natconv_heatsink.html
Patrick
I ran across this thread and happen to be doing some heat sink tests so I'll add what I have found.
One thing that should be noted about heat sink length correction factors is the assumption that the heat load is a point source. Aavid states;
"The published extrusion data shows natural convection performance for a three inch section with a centrally located point source heat load. Because the heat load is assumed to be at a point rather than uniformly distributed, thermal resistance does not change linearly with length. (The ends of a very long extrusion would be cooler than the center and therefore the transfer of heat to the surrounding air is little, if any)."
I think in most cases we parallel several fets so distribution is very doable.
I generally find the published data is not very reliable because it is all about how they are used. I came across surplus 16 inch Aavid 61570 sinks and tested them to see how they perform. If you go by the Aavid correction for temperature rise and length the estimated thermal resistance would be;
0.81 C/W/3" (published 61570 data)
1.257 (30C rise vs 75C temp correction)
0.49 (16" vs 3" length correction)
for a total Rth = 0.5 C/W
My test places two aluminum resistors clamped near each end. I applied 125W (62.5W ea) and measured a 30C rise, thus Rth = 0.24 C/W, about twice as good as calculated. I determined this by measuring the temperature next to each resistor and a spot in the middle of the sink, then take the average. I plan on using 6-8 devices per sink (better distribution than two resistors) this is why I took the average. The three temperatures from top to bottom were 57.5C, 51.2C and 53.4C at a 24C ambient.
I don't find it very convenient to mount the sinks vertically so I tested them horizontally as well. They measure about 32% worse or Rth = 0.31 C/W.
I don't think people should dismiss long heat sinks and I think this goes for wide sinks as well (you should get better performance as well for the same reasons). I encourage everyone to test their own. It is a simple test. All you need is sink mountable resistors, a variac and a temperature probe.
Post your results!
One thing that should be noted about heat sink length correction factors is the assumption that the heat load is a point source. Aavid states;
"The published extrusion data shows natural convection performance for a three inch section with a centrally located point source heat load. Because the heat load is assumed to be at a point rather than uniformly distributed, thermal resistance does not change linearly with length. (The ends of a very long extrusion would be cooler than the center and therefore the transfer of heat to the surrounding air is little, if any)."
I think in most cases we parallel several fets so distribution is very doable.
I generally find the published data is not very reliable because it is all about how they are used. I came across surplus 16 inch Aavid 61570 sinks and tested them to see how they perform. If you go by the Aavid correction for temperature rise and length the estimated thermal resistance would be;
0.81 C/W/3" (published 61570 data)
1.257 (30C rise vs 75C temp correction)
0.49 (16" vs 3" length correction)
for a total Rth = 0.5 C/W
My test places two aluminum resistors clamped near each end. I applied 125W (62.5W ea) and measured a 30C rise, thus Rth = 0.24 C/W, about twice as good as calculated. I determined this by measuring the temperature next to each resistor and a spot in the middle of the sink, then take the average. I plan on using 6-8 devices per sink (better distribution than two resistors) this is why I took the average. The three temperatures from top to bottom were 57.5C, 51.2C and 53.4C at a 24C ambient.
I don't find it very convenient to mount the sinks vertically so I tested them horizontally as well. They measure about 32% worse or Rth = 0.31 C/W.
I don't think people should dismiss long heat sinks and I think this goes for wide sinks as well (you should get better performance as well for the same reasons). I encourage everyone to test their own. It is a simple test. All you need is sink mountable resistors, a variac and a temperature probe.
Post your results!
Attachments
- Status
- Not open for further replies.