|Elliott Sound Products||DIY Heatsinks|
Copyright © 2006 - Rod Elliott (ESP) (See note)
Published 11 February 2005
Note: The basis for this article was originally written by John Inlow, and was available on his website up until 2002 (when the entire site disappeared). Drawings have all been substantially revised (and colour added to make them clearer), and are based on John's originals. The photo in Figure 1 is a cropped and cleaned version of John's original picture. Parts of this document may still retain original copyright. The text has been almost completely re-written.
Build your own heatsinks? Absolutely - this page shows you how to build a heavy duty heatsink and chassis, suitable for Class-A or high power Class-AB amplifiers. While there is no claim that the end result will be cheap (nor are store-bought heatsinks), the performance can be as good or better than anything you can buy.
At only 20% efficiency for a Class-A amplifier, the wasted heat is enormous! Out of 100 watts of input power, only about 20 of those Watts are available as useful sound. The rest of the power (80 Watts) is dissipated as heat, and must be removed - all the while keeping the transistor junctions at a safe operating temperature. It takes big heat sinks to remove the excess heat from the power transistors. It can also be unbelievably difficult to locate affordable, large heatsink stock (and this seems to be worse in the US). So, this page describes a solution to the problem.
Figure 1 - Photo of a Completed Chassis
The pictured chassis (John Inlow's photo) will easily dissipate well over 160 Watts of wasted power - all as heat. If this meets your requirements, get ready to head for your workshop in readiness to make a significant amount of mess . The design is totally adaptable, so any size of heatsink can be produced, designed to fit any opening size or application. The chassis, although time consuming, is easily reproduced by anyone with basic construction skills. Use a drop saw (also known as a 'chop' saw in the US) with a carbide tipped blade designed to cut soft metals.
The saws and blades are readily available from most hardware stores, and are relatively inexpensive - actually, one can purchase the power saw for less than the blade (a very odd situation indeed). Be sure to use plenty of cutting fluid designed for machining aluminium. Denatured alcohol (methylated spirit) is excellent, but is highly flammable. On the positive side, there is no oily residue that can spontaneously combust if incorrectly stored. Take extreme care with all cutting fluids, as there are risks with all of them - especially anything that works well with aluminium.
Before going into great detail, it is important to look at the following drawings so you can see what is involved. There is no doubt that you will need some fairly serious tools to be able to tackle something of this nature. Apart from the drop saw and metal-cutting blade mentioned above, you will need a drill press - a power hand drill could be used, but the results will be disappointing or very, very time consuming (or both). You also need taps (for cutting threads, not whatever you might have been thinking ). A decent sized work area is also needed, and be aware that you will create a vast amount of aluminium chips, powder and swarf. The kitchen table is not recommended.
You also need the usual array of hand tools and miscellany - a hacksaw, files, drill bits, G-cramps, and various bits of scrap material that can be used to construct drilling and assembly jigs, etc., etc.
Figure 2 - Rear View Drawing
Looking from the rear, we see the two heatsinks, top and bottom covers, as well as the front panel. If you insist on using imperial measurements, then divide all measurements shown by 25.4 to obtain inches. The dimensions are not critical for the most part - adapt to suit your specific application.
Figure 3 - Top (Plan) View Drawing
The plan view gives you the rest of the picture. Again, adapt the dimensions to suit your application. The heatsink sections can be as large (or as small) as you need, limited by available funds and your patience (as always).
It is suggested that you use 6 x 25mm (or 0.25" x1.0") bars for the spacers and 2 x 75mm (0.08" x 3.0") bar or cut sheet for the heat dissipation fins. The length depends on your intended height - all drawings here are based on a 150mm (6") heatsink height. Starting with a spacer section on the outside, the two sections are repeated (spacer-fin-spacer-fin ... spacer) until you have the length you need. To achieve the desired depth for the chassis, the plates are bolted together onto four (two for each heatsink) 10mm (3/8") threaded rods. Drilling the holes is a tricky and time consuming job - they must be aligned perfectly. The difficulty of the process is reduced if you create a jig to position the material as you drill the holes. Although it adds time to the whole process, partially pre-drilling each section with a centre drill (a special drill bit with a thick shank and small stubby tip) will ensure that the drill bit does not wander, causing holes to be off-centre. It is very important to clean up the holes after the drilling is completed. Using either a larger drill bit or a de-burring tool, chamfer the holes to remove the jagged edges that form when drilling. Don't even think about avoiding this step - the plates will not touch one another evenly so heat transfer will suffer, and your assembled heatsink will never be straight.
If you use 10mm threaded rod, the holes should be drilled to 12mm. This allows for the occasional hole that is slightly off centre, and also lets you align the assembly perfectly before it is finally clamped up tight.
Figure 4 - Fin and Spacer Detail Drawing
Keep repeating the above pattern until you achieve the desired depth for your chassis. The same procedure is used for both sides of the chassis. The threaded rods must be long enough to pass through all of the spacers and heatsink fins, and still have sufficient left over for nuts. Optionally, the rods may be extended further to create a mounting for handles at the front of the chassis.
Make sure that all surfaces are flat after drilling. De-burr all edges before assembly to get the best possible contact between the spacers and fins.
When everything is machined, cleaned and ready to assemble, begin by screwing a half nut onto one end (to become the front) of the four threaded rods. Make sure that the distance from the end of the rod to the face of the half-nut is correct for your handle sections (allow for front panel, tube (or long nut if you prefer), handle section and acorn nut). Then slide on the first spacer, followed by a fin,then a spacer (etc.) in alternating fashion. When completed, screw a full nut onto the remaining end (rear) of the rods - this should only be finger tight at this stage! You need to get the distance at the front dead right if you are planning on adding handles - the final acorn nut does not have much depth. Feel free to add washers at the front as well as the back, but make sure they are allowed for in the length of the threaded rod.
Alignment is critical. Because your heatsink is made from many separate pieces of aluminium, it is possible for it to assume many different and entertaining shapes, within the constraints of the threaded rod. None of the potentially entertaining shapes is useful - you need flat and square. Period.
To achieve this, you'll need a flat surface with a piece of scrap angle the same length as the heatsink assembly screwed to one edge. You also need a small hammer, a square, a scrap piece of aluminium and a piece of timber. First, lay the assembly onto the surface, with fins pointing upwards. Slide it across so that it contacts the angle. The alignment is by nature repetitive - each step will need to be repeated - possibly several times ...
Once completed and the nuts are firm, you can clamp a piece of wood on top of the assembly, thus clamping the heatsink to your flat surface. Be careful that you don't disturb any of your alignment during this process. Now the rear nuts can be tightened fully. They should be tight, but not ridiculously so - a stripped nut or threaded rod is not a bonus at this stage.
Upon removal from the assembly jig, the heatsink should be flat and square, requiring the minimum effort to obtain excellent thermal contact between the reinforcing bar or mounting plate (see below). If the nuts are tight enough, firm effort on your part to bend or twist the completed assembly will result in nothing more than heatsink imprints on your hands - the heatsink itself should remain nice and flat, with no bending or warping. If it does warp, you will have to repeat the final step, after loosening the nuts just enough to allow the assembly to be made flat once again. Although this is a frustrating step if things move, it is highly recommended - you will find out for yourself how rigid the heatsink is, and whether (or not) you need to add a reinforcing bar (or perhaps just tighten the nuts a little more).
If it is possible to distort the assembly (or it comes out pre-warped), then you must use reinforcing bars screwed to the back of the heatsink (all drilling and tapping must be into the spacer strips only. Alternatively, a full-length flat plate will also provide reinforcement, but at somewhat greater expense. These options are discussed below. Note that one or the other needs to be used anyway - at issue is how long it must be to achieve rigidity and good thermal contact.
After assembly and prior to fitting the heat spreader (either a reinforcing bar and/or flat spreader plate), it is advantageous to mill the rear of the heatsink to present the flattest possible surface. Given that few hobbyists have access to a milling machine, an alternative is to carefully file the entire rear surface - this will be an extremely tedious job, but will improve performance. A linisher (belt sander) can also be used, but the surface must be finished with a fine grit so it is as smooth as possible. The photo in Figure 6B shows the rear surface of a small heatsink I made, and the linishing marks are visible. I deliberately did not complete the job so the anomalies could be seen clearly.
Note Carefully - Once the heatsink is fully assembled and the back has been linished (or milled) flat, you must not disassemble it. If you do, you will have to re-surface the back of the heatsink (where heat-spreaders and then transistors are mounted) because it will be impossible to re-assemble the fins and spreaders to obtain the original surface finish. You must make sure that everything is the way you want it to be before the base is machined, and after that it's no longer possible to make changes unless you are prepared to re-surface the back again.
Although the surface shown is very flat (better than 25um / 50mm), the surface finish will almost certainly not be good enough for direct mounting of semiconductors. This makes the heat spreader mandatory. You can check surface flatness by trapping a thin hair (for example) at various places on the surface with the edge of a steel rule. It should be not possible to pull the hair from beneath the edge of the rule at any position under the heat spreader location.
Figure 5 - Transistor Mounting Options
Before assembly of the heat spreader to the heatsink itself, apply heatsink compound between the spreader and the heatsink back. Apply a thin layer, and check that the two surfaces mate well by pressing the bar onto the heatsink. Remove it, and check that the heatsink compound is evenly distributed and shows signs of full contact. This will be immediately obvious upon inspection.
The transistors may be mounted directly to the reinforcing bar as shown. Normal transistor mounting procedures apply. Alternatively, attach a section of flat bar as shown in the right-hand drawing. This method produces a nice flat surface, ideal for mounting boards where the transistors (or MOSFETs) are under the PCB.
It is imperative that all holes for the bar or plate line up with the centres of spacer sections - drilling and tapping such that a hole is part way between a fin and a spacer will cause deformation of the assembly, resulting in potentially dramatic loss of performance.
Figure 6 (A and B) - Photo and Scan of Small Demonstration Heatsink
The above photos are of a small demonstration heatsink I built. The left side (A) shows the general construction, while B shows the surface finish on the underside (the latter was scanned to get the best image of the surface). I deliberately didn't complete the machining process so you could see the aberrations that you will get when the heatsink is assembled. The dark areas are actually the shiny (not sanded) areas of the fins. This was despite my following the instructions listed above, so you will also have the same problem. Attempting to remove every imperfection is futile unless you have access to a milling machine - it will be too boring and frustrating for words. The last 5% of the surface could easily take 90% or more of the total construction time.
This is the reason for using a heat spreader - it distributes heat over a much larger surface area than you will get with any transistor, but the surface still needs some basic attention or heat transfer between spreader and heatsink will be badly affected. This will lead to an excessive temperature on the spreader, and an even higher temperature for the transistors. This machining is the most important part of the exercise!
I used a linisher (essentially an upside-down large belt sander), but careful filing will also work very well. Yes, it will be tedious and hard work, but the results will be worth your efforts.
You will also note that I didn't follow the procedure, in that I have fins (rather than spacers) at each end of the heatsink. Feel free to break the rules too, provided you work out exactly what you need. The whole idea is that this process allows you to make a heatsink that exactly fits your needs. In case you were wondering, the overall size of the heatsink pictured in Figure 6 is 100mm (h) x 60mm (w) x 52mm (d). There are 8 fins, each is 32mm deep, giving a thermal resistance of about 1.16°C / W (if black enamelled) or 2.0°C / W (polished aluminium, as pictured).
Once the heatsink sections are drilled and tapped to accept the top and bottom covers, the heat spreader has been drilled and tapped for the transistors, and no further work needs to be done in that area, you may start the final assembly. The bottom needs to be drilled for external feet and for all internal hardware that will be attached to it. As shown in Figure 3, the heatsinks (as well as top and bottom covers) are drilled to accept mounting screws. The heatsink holes are tapped to accept the size screws you will use, as will the heat spreader.
The front panel simply attaches to the two heatsink sections as shown in Figure 3, and the top and bottom panels attach to the heatsink and front and rear panels. Although no additional mounting for the rear panel to heatsinks is shown, this can be added if desired. You can use 12mm square section, or a piece of angle if you prefer. The extra is not really needed, but the rear panel will flex a little when the top is not in place without it.
If you find that you need the reinforcing bar(s), first assemble the heatsink as described above, then drill and tap the holes for the reinforcing bar (in both heatsink spacers and bar). Assemble the heatsink as described, carefully turn it over and file, sand or otherwise machine the rear surface so it is as flat as possible. Then, attach the reinforcing bar, and screw down lightly. You will quickly see if there is any misalignment, and this must be corrected before you permanently attach the bar to the heatsink. In general, it is expected that if the assembly process is carried out carefully, the heatsinks will be very rigid indeed - almost as if they were one solid piece of aluminium.
The unit as described is very heavy, so file and sand all edges and outside corners of the ribs to reduce the risk of being cut. It will be much easier to assemble if you drill and tap all the heatsink holes for mounting the reinforcing bars and / or heat spreader before final finishing.
Paint the finished heatsink with a spray can of flat (matte) black enamel. It should be a self-etching type designed for aluminium finishing for best results. Don't be tempted to try for a perfect finish between the fins, as you will end up applying too much paint which will reduce performance. You could get all the sections black anodised before assembly, but that would be an expensive option because of the number of pieces. Don't even contemplate pre-finishing the heatsink components - any paint between fins and spacers will degrade performance dramatically, and may cause the heatsink to bend or twist when it is assembled.
The front panel arrangement may look rather suspect, with the 12mm bar attached using screws into blind holes in the panel. With a 6mm panel, this allows 5mm hole depth, and about 4mm of this can be threaded. Having done exactly this on many occasions, I can assure the reader that it works perfectly. You do need to be extremely careful to ensure that the holes don't go all the way through, but other than that, the process is straightforward. Tapping blind holes is a cow, but you can cheat and use self-threading screws. The latter must be square-ended - conventional tapered self-tapping screws will not work! If you are concerned, use a good epoxy resin (24 hour setting type - not 5 minute) as well as the screws for a permanent bond.
As described, each heatsink section will have a thermal resistance of around 0.12°C / W. This is assuming a middle-of-the-road value for emissivity of about 0.8 - typical of a flat black surface finish you will be able to apply at home. This is a very good figure, but it will be degraded if the heat spreader is too small, or makes ordinary (as opposed to excellent) thermal contact with the heatsink, etc. The distribution of transistors will also have a bearing on the thermal performance.
However, even if we were to assume the worst case and the thermal resistance is effectively doubled, it is still very good at 0.24°C / W. The 'typical' figure (0.12) was calculated using the ESP heatsink calculator program (see the downloads section of the ESP website). Heatsink temperature was taken as 60°C, at an ambient temperature of 25°C. Even using the worst case figure, this means that when dissipating 80W (as discussed at the beginning of this article), the heatsink temperature will stabilise at around 20°C above ambient (45°C at 25°C ambient) ... and that's worst case!
Because of the large thermal mass (as well as actual mass), this heatsink will take a significant time to reach full operating temperature. Each heatsink will weigh in at about 5.9kg for aluminium alone, and even the baby one I made (see Figure 6) weighs 480 grams (admittedly, that's with the steel threaded rod, washers and nuts). Both heatsinks will weigh about 12kg (over 26 pounds in the old measurements), and you have yet to add the panels, transformer(s) and other components.
The basis for this article was originally written by John Inlow, and was available on his website up until 2002. Drawings have all been substantially revised (and colour added to make them clearer), and are based on John's originals, as is the photo in Figure 1.
|Copyright Notice. This article, including but not limited to all text and diagrams, is the intellectual property of Rod Elliott, and is Copyright © 2005. Reproduction or re-publication by any means whatsoever, whether electronic, mechanical or electro- mechanical, is strictly prohibited under International Copyright laws. The author (Rod Elliott) grants the reader the right to use this information for personal use only, and further allows that one (1) copy may be made for reference. Commercial use is prohibited without express written authorisation from Rod Elliott.
Parts of this page (in particular Figure 1 and parts of the redrawn diagrams) may also be the intellectual property of John Inlow.