Hi Swordfishy,
Do not worry about the anodising, it will have negligable effect on thermal performance (infact, depending on the colour and size it can offer an improvement.....but again, its very small)
DO NOT start to sand the anodising away, you will introduce a rougher surface finish, requiring a thicker thermal pad or more thermal paste that before......and thermal pads / pastes are a thermal resistance to the link between case and sink. You will probably find the heatsink (im assuming an extrusion here) will have a surface finish in the region of RA 0.3 - 0.8 if its from a good company, such as Bluecore Heatsinks
This will allow you to use as thin a thermal pad as possible (or less grease)
Do not worry about the anodising, it will have negligable effect on thermal performance (infact, depending on the colour and size it can offer an improvement.....but again, its very small)
DO NOT start to sand the anodising away, you will introduce a rougher surface finish, requiring a thicker thermal pad or more thermal paste that before......and thermal pads / pastes are a thermal resistance to the link between case and sink. You will probably find the heatsink (im assuming an extrusion here) will have a surface finish in the region of RA 0.3 - 0.8 if its from a good company, such as Bluecore Heatsinks
This will allow you to use as thin a thermal pad as possible (or less grease)
Pass DIY Addict
Joined 2000
Paid Member
Hmmmm... I don't think this advice can be universally followed. On the heat sinks I purchased, the anodize left rather noticeable ridges on the otherwise flat surface of the heat sink. Sanding a smooth spot with 400 grit paper and some WD-40 worked very well. I then used a sil-pad between the transistor and the sanded spot on the back of the sink.
The difference between sanding and not sanding was very clear by scratching the surface with your finger nail.
Eric
Yeah, this is what I am dealing with. However the ridges seem to be the result of a poor extrusion not the anodising which is a very thin (almost transparent) layer.
Hi Guys,
The anodising will be microns in thickness, so I very much doubt it is leaving ridges.
The ridges are more than likely coming from very cheap extrusion meathods. I am the thermal engineer at a heatsink company, and standard extrusion practises should achive RA of around 0.3 - 0.8 for a high quality extrusion, just over 1 for a standard extrusion. Both will carry a max flatness tolerance of 0.4. Our standards are used acoss mass consumer electorincs.
I have attached a image of one of our heatsinks with black adnodise for your reference
The anodising will be microns in thickness, so I very much doubt it is leaving ridges.
The ridges are more than likely coming from very cheap extrusion meathods. I am the thermal engineer at a heatsink company, and standard extrusion practises should achive RA of around 0.3 - 0.8 for a high quality extrusion, just over 1 for a standard extrusion. Both will carry a max flatness tolerance of 0.4. Our standards are used acoss mass consumer electorincs.
I have attached a image of one of our heatsinks with black adnodise for your reference
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Pass DIY Addict
Joined 2000
Paid Member
Perhaps I mis-spoke. I agree that the ridges are most likely a result of the extrusion process, not the anodize process. Another set of higher quality sinks that I have are very, very smooth. Either way, the back of my sinks have ridges that are larger than they should be for good thermal contact. Sanding for just a few minutes works wonders in this situation.
I wonder what reaction you will get from the receptionist when you ask for a flatness specification/guarantee?
I wonder if the machinist can even measure his resulting flatness?
It depends greatly on the system to the advantage of a spreader plate. To simplify a spreader plate (assuming is the size of the heatsink base area), its merely an extension to the heatsink base thickness.....although you are introducing a second thermal resisance via the interface.
Increasing the thickness of a heatsink base will improve the spreading resistance, BUT...that will not nessacerially improve the performance of the heatsink. As an example, I have ran a data driven CFD model of the black heatsink in the images above, the below graph shows the resistance of the heatsink when compared against the base thickness. So the addition of a spreader plater could be an added cost with no benefit.
As Itsmee said, skimming it will certainly help if its that bad, although fly cutting will probably be cheaper.
An externally hosted image should be here but it was not working when we last tested it.
If its a skimming shop they should be meeting flatness tolerences and so will have the equipment to measure it.
Having said that, the heatsink base surface should also be manufactured against flatness and RA tollerences, their is not a drawing that passes our manufacture that does not have this on!! lol
The heatsink in that model is just one I had to hand (from taking the pictures yesterday, its still on my desk!!), its 200x250x65mm, 16fins @ 3.6mm (max) tapered. I simplyed strapped a single centralised heat source kicking out 10w as a demonstration.
That 200 by 250gives a maximum radius from Device to farthest corner of 153mm. (sqrt[95^2+120^2] for a 10*10 thermal interface on the device)
Using the 10:1 ratio of radius:thickness that I regularly recommend, results in 15.3mm for that heatsink.
I see that 13mm almost matches that optimum performance. This is a ratio of ~11.8
It was for that comparison that I was interested in the heatsink size.
Hi Andrew,
The heatsource size was 18*18
I have seen your method used before, and its a nice simple way to roughly approximate the base thickness providing the heatsource area is relatively small in comparison to the heatsink base area, and the heat source is centralised. As soon as the heatsource and heatsink base become closer in size, the percentage of error in this method can increase, up to a point where it then becomes unsteady in its accuracy. I had a more concise report on this somewhere but cant put my hand to it at the moment, I will post it when i dig it out, but for now I have quickly ran error analysis as below.
Once the heat source moves out of a centralised position, or you look at optimising the size of the heatsink (therefore reducing the size difference between heatsink and heatsource) you have to start adding correction factors to this method to approximate the results, alternatively you can start to get more indepth and build up a model using a mass of formulas.....mainly deriving from a chap whos name very annoyingly escapes me, Benzell or Byenzl. You also need to consider the whole heatsink, and the fin structure as a small part in the base's performance as well.
The heatsource size was 18*18
I have seen your method used before, and its a nice simple way to roughly approximate the base thickness providing the heatsource area is relatively small in comparison to the heatsink base area, and the heat source is centralised. As soon as the heatsource and heatsink base become closer in size, the percentage of error in this method can increase, up to a point where it then becomes unsteady in its accuracy. I had a more concise report on this somewhere but cant put my hand to it at the moment, I will post it when i dig it out, but for now I have quickly ran error analysis as below.
Once the heat source moves out of a centralised position, or you look at optimising the size of the heatsink (therefore reducing the size difference between heatsink and heatsource) you have to start adding correction factors to this method to approximate the results, alternatively you can start to get more indepth and build up a model using a mass of formulas.....mainly deriving from a chap whos name very annoyingly escapes me, Benzell or Byenzl. You also need to consider the whole heatsink, and the fin structure as a small part in the base's performance as well.
An externally hosted image should be here but it was not working when we last tested it.
I will add that it all comes down to the confidence and preference of your analysis, as the graph I have post is based on the CFD code I used to the calculations you use.......who is actually right and who is wrong can only be proven with real life test data. Both our models are recognised and I have seen and used myself your model in the past (although when i have used in many years ago we took a 11:1 ratio), the CFD code I use has been proven greatly also in complex industry applications. Its really all down to the accuracy you wont and your beliefs in the many formulas that surround physics
I'm a believer in the value of modeling.
If the model can be shown to closely represent the actual and that examining the model saves time/expense then it it valuable.
My model gives a very quick check on heatsink suitability. It is not any more powerful than that. It cannot, for instance, predict interface temperature and surely no one would try to use it to prove that another formula is more or less accurate in that respect.
I simply wanted to see if your results came close to what I had observed.
If the model can be shown to closely represent the actual and that examining the model saves time/expense then it it valuable.
My model gives a very quick check on heatsink suitability. It is not any more powerful than that. It cannot, for instance, predict interface temperature and surely no one would try to use it to prove that another formula is more or less accurate in that respect.
I simply wanted to see if your results came close to what I had observed.
Hi Andy,
No problem, I will try to dig out the correction factors for you to apply to your model for other instances as well.
The only thing I will say, is becareful using your model when the heatsource footprint becomes closer to the heatsink footprint, if I can assist in any calculations in the future dont hesitate to contact me
Cheers
Russ
No problem, I will try to dig out the correction factors for you to apply to your model for other instances as well.
The only thing I will say, is becareful using your model when the heatsource footprint becomes closer to the heatsink footprint, if I can assist in any calculations in the future dont hesitate to contact me
Cheers
Russ
Well, you guys have both extremes covered.
To get down to practical solutions, for dressing an area as small as the footprint of a TO247, a large very fine mill file will get you reasonably flat, then you can polish to your hearts content. Sanding without flattening gives you a depression. Whether sanding without flattening will be sufficient depends on the amount of heat you need to transfer. I always hit them with a file.
That's my 2 cents. YMMV
To get down to practical solutions, for dressing an area as small as the footprint of a TO247, a large very fine mill file will get you reasonably flat, then you can polish to your hearts content. Sanding without flattening gives you a depression. Whether sanding without flattening will be sufficient depends on the amount of heat you need to transfer. I always hit them with a file.
That's my 2 cents. YMMV
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.