Automating adhering TO-220 MOSFETS on heatsinks?

• Automating adhering TO-220 MOSFETS on heatsinks?
  Posted by Rat_Patrol on 26 Feb, 2021 17:20
• Is there a relatively low cost (less than $10k) solution to either take over doing this or dramatically assist an operator?
  
  We put hundreds and hundreds of TO-220 THT MOSFETs onto aluminum heat sinks (8 per heat sink): they are held to the aluminum by either thermal epoxy or thermal tape. Epoxy is preferred for durability reasons, but we tend to use tape due to speed/reliability reasons.
  
  I am thinking a cheap PNP machine can be rigged up to hold the tubes and place the FETS, 3D print a jig to hold the heat sinks (holding several), but is there a relatively low cost way to automate the epoxy with a "cheap" epoxy?
  
  When we've explored the epoxy setting by hand, we 3D printed some jigs to hold the MOSFETS, fill the jig by hand with FETs, using a manual "extruder" to mix/squeeze the epoxy onto the back of the MOSFETS, then set the heat sink on top. We then let it sit like that until the epoxy cured.
  
  Issues doing this:
  Time. It takes too many labor hours.
  Reliability. Sometimes the MOSFETS weren't sitting properly, and they shorted out to the heat sinks (they can't be electrically shorted to the heat sinks. Yeah, I know, not my design, can't be changed). This means we either re-process the entire heat sink for the one that shorted, or by hand remove the shorted FET and manually put on a new one. Big PITA.
  Mess. The jigs end up being disposable after a few uses. This isn't a huge deal, they only cost about 50 cents each to print, but just more time messing with making more.
  
  The tape method works, is reliable, but again takes a decent amount of labor.
  
  We built a jig for spreading the epoxy on the heat sinks like a paste screen machine. It worked OK, but wasted a lot of epoxy. It could be refined and optimized I think. Again, is there cheap thermally conductive (electrically isolating) epoxy or something else to use here?
  
  Basically, either we hire on a low paid worker to do this type of operation or we buy a machine. The company has to pay high-value techs way too much per hour to do this type of low-skill work. Maybe a combo of both?
• #1 Reply
  Posted by Siwastaja on 26 Feb, 2021 17:50
• Did I understand it right, you need electrical isolation and rely on the correct amount of epoxy glue and the exactly correct placement of the part to provide that insulation?
  
  What is wrong with the normal industry-tested solutions to mounting TO-220 packages on heatsinks? I mean, this is a very common thing to do, widely seen in low-cost devices.
• #2 Reply
  Posted by Rat_Patrol on 27 Feb, 2021 15:35
• Quote from: Siwastaja on 26 Feb, 2021 17:50

  Did I understand it right, you need electrical isolation and rely on the correct amount of epoxy glue and the exactly correct placement of the part to provide that insulation?
  
  What is wrong with the normal industry-tested solutions to mounting TO-220 packages on heatsinks? I mean, this is a very common thing to do, widely seen in low-cost devices.
  

  
  Correct, there must be electrical isolation between the FET and the heat sink and so there must be enough epoxy to provide that. Don't ask, not my design, cannot be changed, we just have to deal with it.
• #3 Reply
  Posted by Siwastaja on 27 Feb, 2021 16:07
• You can always change the design if it's completely unpractical and beyond repair. It's your duty to say it out loud: this design is unsuitable and must be changed, the quicker we get to it, the quicker the product is out.
  
  Try asserting that poo must be used to hold capacitors in place and see if that would be accepted as the only unchanged way of doing things. Then see if this can be changed.
• #4 Reply
  Posted by jmelson on 05 Mar, 2021 17:46
• Not sure I want to wade into the insulation/epoxy issue, but it sounds like a P&P machine ought to be able to do this.
  Can you apply the epoxy to the heatsink with a stencil, just like solder paste to a PC board?  Then, a vibratory feeder could bring the TO-220's down to where the P&P could pick them up.  Seems very straightforward.
  
  Now, as for insulation, heavy anodize ought to work, as the epoxy, if not squished too hard by the P&P would maintain a film under the transistor. 
  
  Jon
• #5 Reply
  Posted by fcb on 05 Mar, 2021 18:44
• Could you add one of those old-school TO220 screw insulators (look like a little tophat) on the wrong side of the TO220 then assemble as before, perhaps with a bit more thermal epoxy - sort of like the easy-tile systems plumbers use?  http://www.mecint.co.uk/Item/118
  
  As for applying epoxy - I'd start with a 3D printer as a base for doing the epoxy, and perhaps a 2nd one for doing the TO220.
  
  We have an old CAMALOT XYFLEX PRO we picked up incredibly cheaply for this type of job. I'd recommend one in-a-heartbeat if this is regular work.
  
• #6 Reply
  Posted by brabus on 09 Mar, 2021 14:01
• The heatsinks can be anodized with a thicker layer of oxide, providing electrical insulation. Thus, the insulation will be independent from the glue quantity.
  
  This technique has been used by Linn amplifiers for a long time; I even tested myself, it works fine.
  
  
• #7 Reply
  Posted by Rat_Patrol on 09 Mar, 2021 20:36
• You know, just thinking, what about a nice thick layer of conformal coating, such as urethane? Would be easier for use to do as we don't have equipment to anodize. Maximum voltage is 100DC (nominal) that these FETs deal with.
  
  Couple layers, bake it off, and ready to go. I wonder how well the epoxy would adhere to the urethane? Hmm, may have to experiment with that one.
• #8 Reply
  Posted by Siwastaja on 17 Mar, 2021 14:53
• Conformal coats are soft. TO-220 metal edges are sharp. Thermal pad materials are quite thick (like 0.5mm and up), and often contain fiberglass reinforcements) for a reason. With such thickness, comes the requirement of good thermal conductivity of the material. TLDR; buy existing thermal interface materials, don't invent your own or try to misuse products not designed for that.
  
  Such corner cutting measures are extremely costly to develop, you need to test them well and adjust/control the process. If you save $1 per each product but develop it for a year costing you $100000, you need to sell 100000 units to break even.
  
  If you can order thicker-than-normal anodization and verify it works for you reliably, that's great, but it will take time and resources to verify it.
  
  For small batches, use normal methods.
• #9 Reply
  Posted by Nauris on 17 Mar, 2021 17:24
• There are those epoxies with glass spheres in them to ensure miniumum thickness, maybe try adding some spheres to your cheap epoxy. Mix well.
• #10 Reply
  Posted by T3sl4co1l on 17 Mar, 2021 17:39
• Yes, this. ^^
  
  I have seen BTW, devices glued to heatsinks in plasma TVs -- no idea what jig and placement and all they used, but it's at least possible.  Near-slave labor, likely.
  
  And as mentioned, hard anodize is good to a few hundred volts, enough for functional in a lot of applications but don't count on it where some kV or basic or reinforced is called for.
  
  Tim
• #11 Reply
  Posted by voltsandjolts on 17 Mar, 2021 17:47
• My pref is
  TO220 + silpad + spring clip
  
  But you could go fancy-pantsy with directfets and heatsink on top
  
  
  
  Edit: I have never used directfets and had always assumed you could put silpad and heatsink on top but I can't see any pictures of that, so maybe I'm talking crap.

  ---

  
• #12 Reply
  Posted by T3sl4co1l on 17 Mar, 2021 19:13
• Those, and CSPs especially like EPC eGaN, oh and also a lot of newer GaN parts with top side thermal pads too, are good with thermal contact on top, yup.  I'm fond of using DFNs that way, the 3x3mm and PDSO-8 sizes for example are good for a few watts, especially on heavier copper or multilayer board (the plastic package isn't very conductive by itself, most of your heat dissipation comes out through the pads, spread by PCB copper).  Probably not enough though, if you need a TO-220 in the first place.
  
  But hey, that can be an argument in favor of exactly this kind of technology.  Work smarter, not hotter.  The most expensive watt is the watt you have to dissipate, don't dissipate it in the first place, that sort of thing.  A lot of these parts are more expensive, and harder to work with, but maybe that's a payoff worth considering.
  
  Tim