EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: EEEnthusiast on December 11, 2019, 12:45:38 pm
-
Folks,
What is the best way to mount a power transistor (TO-220 non isolated tab) to the metal case of the power supply in order to act as a good heat sink. It should be electrically isolate from the case as well.
It should be assembly friendly as the TO-220 package is PCB mount. The transistor cannot be connected to the board using wires. I was thinking of using two heatsinks, one on the transistor and the other on the case. After the case is closed, the two heat sinks would sit close to each other acting like a small heat exchanger. Like below. Any experience with this, or a better solution?
-
Air gap is one of the best heat ... insulator.
Whats wrong with thermal pad ?
-
The thermal pad on the PCB would mean more heat dissipation on the PCB, causing temperature fluctuations. I need to keep the PCB at the most stable temperature and hence need to take the heat off the board.
Air gap could be a good insulator, but with enough fins interacting with each other, should the thermal gap place a big restriction in the heat exchange?
-
Thermal pad to the metal body, then secure the TO220 on it.
-
Linn Amplifiers use the case as heatsink, by screwing the transistors to an anodized block of alluminium. The thick anodization acts as electrical insulator and good heat conductor at the same time.
In your case I would definitely use a traditional sil-pad, it's an industry standard. I don't see how you could induce temperature fluctuations on the PCB with a sil pad: if it's correctly placed you will notice no difference in comparison with a direct alluminium mount.
-
How exactly does a thermal pad increase the power (watts) dissipated by the circuit?
And how would an air-insulated heatsink not do the same thing..?
Typical solution is thermal pad with bolts and shoulder washers, or even better: spring clips, which can be screw mounted, press-in, or lever actuated.
Tim
-
Well, the thermal pad on the PCB would increase the effective area of the radiator allowing the component to cool faster. My assumption was that the gap between the two heat sinks could conduct the heat well as the effective area of heat exchange is larger.
-
https://www.picclickimg.com/d/l400/pict/183650763104_/20-x-Thermal-Insulator-Kit-Silicone-Pad.jpg (https://www.picclickimg.com/d/l400/pict/183650763104_/20-x-Thermal-Insulator-Kit-Silicone-Pad.jpg)
-
Mica (0.1mm thick):
http://duckduckgo.com/?q=mica+to220+site:ebay.com (http://duckduckgo.com/?q=mica+to220+site:ebay.com)
Alumina has the best thermal conductivity (20-25W/m.k), but it's thicker (0.635mm):
http://duckduckgo.com/?q=Alumina+Ceramic+t0220+site:ebay.com (http://duckduckgo.com/?q=Alumina+Ceramic+t0220+site:ebay.com)
-
http://duckduckgo.com/?q=mica+to220+site:ebay.com (http://duckduckgo.com/?q=mica+to220+site:ebay.com)
I strongly suggest to not use mica for mains-case insulation as it sometimes cracks. Thick-ish silicone pad is much safer solution. Also don't use screw+plastic washer. It's a very weak point. Use square silicone pad without any holes + clip. Or even better, use silicone cap over whole transistor instead of pad.
(https://uk.farnell.com/productimages/standard/en_GB/1892353-40.jpg)
(https://image.made-in-china.com/202f0j00ewmQLsidkfku/Heat-Resistant-Flexible-Silicone-Rubber-Insulation-Cap-with-to-220.jpg)
-
Aside from agreeing with the others that the correct way to do it is to use a thermal pad between the TO220 package and the case...
Assuming that you did manage to achieve sufficient thermal coupling between the elements of your large area air-gap heat exchanger, how would you manage to ensure that the mechanical clearances were maintained well enough to achieve proper electrical isolation?
-
The electrical isolation is guaranteed by the precision in the assembly process only. The TO-220 package height above the PCB is fixed and so is the heat sink height. The other heat sink on the case is also fixed at a constant location by marking it.
-
Just make a suitable metal insert between the TO220 and the case, use good thermal pad to insulate it electrically to the metal case, but thermally coupled thru the pad.
This has been done for many many decades, why you sound like you are trying to reinvent the wheel, but yet using non circle shaped wheel ? :-//
-
My issue is with the case style. The case I am using is a aluminium extrusion with open ends on two sides. The PCB slides into the case from the sides. So there is no easy way for me to access the package once the PCB is inside.
-
My issue is with the case style. The case I am using is a aluminium extrusion with open ends on two sides. The PCB slides into the case from the sides. So there is no easy way for me to access the package once the PCB is inside.
Well, you've boxed yourself in pretty good, then <rimshot>!
You might be able to transfer enough heat from a board-mounted, but physically separate, heatsink to the aluminum enclosure by using a stirring fan - basically, a fan inside the box that blows the air around inside. This can be surprisingly effective, but you do need to have a relatively large enclosure surface area and it needs to made of aluminum, not steel and definitely not plastic.
-
Thermal pads can be used in those enclosures, with moderate success, but yes, you've found one good reason why they are to be avoided where much power dissipation is required.
Tim
-
SIL pads and similar are great. Their conductivity seems to be a bit lower than mica + grease, but mica + grease is delicate and messy. I’m guessing you’re not after maximum performance given you’re using an enclosure heatsink.
Could you put the TO220 right at the end of the board so that you can slip a clip over it once the PCB has slid all the way into the enclosure?
How much voltage does the insulator need to stand off? If it short circuits, would you expose a user to dangerous voltages?
-
The electrical isolation is guaranteed by the precision in the assembly process only. The TO-220 package height above the PCB is fixed and so is the heat sink height. The other heat sink on the case is also fixed at a constant location by marking it.
My issue is with the case style. The case I am using is a aluminium extrusion with open ends on two sides. The PCB slides into the case from the sides. So there is no easy way for me to access the package once the PCB is inside.
As you said you can do precision assembly process, although not used at extrusion style case, below are examples used a lot at common switching psus, this is from Meanwell.
I'm guessing this technique can be implemented in your case, all you need is just to precisely drill a hole at the case for the screw, and provide a metal insert sandwiched between the case and the TO-220 body as below.
The gray box is a TO-220 encased inside a box style thermal pad (cap) as above Wraper's example post, as this is a HV transistor, its secured using a strong "L" shaped steel with a hole (threaded) for the screw, when fastened will make the L steel pressing the TO-220 body, and its secured/fastened from "outside" of the case.
(https://www.eevblog.com/forum/projects/adding-heatsink-with-electrical-isolation/?action=dlattach;attach=888368;image)
Same as above, this is the schottky diode using the same securing style. You see clearly a big piece of metal insert behind.
(https://www.eevblog.com/forum/projects/adding-heatsink-with-electrical-isolation/?action=dlattach;attach=888364;image)
-
Rather than complicating the process by precision engineering the “heat exchange” type h/sinks, why not mount the transistor (with proper electrical insulation-I would use mica/plastic washer/thermal compound) in a convenient position that does not interfere with the PCB and have the leads soldered to 3 wires that end up in a plug? This is a common practice in a lot of things. Examples can be seen in most Tek scopes. As long as you have a tight fitting plug, I don’t see any issue.
See attached image where you can see the transistor on the left of the EHT transformer on a 465M scope.
-
SIL pads and similar are great. Their conductivity seems to be a bit lower than mica + grease, but mica + grease is delicate and messy. I’m guessing you’re not after maximum performance given you’re using an enclosure heatsink.
Could you put the TO220 right at the end of the board so that you can slip a clip over it once the PCB has slid all the way into the enclosure?
How much voltage does the insulator need to stand off? If it short circuits, would you expose a user to dangerous voltages?
The voltage is well below 24V. I think the idea to move the TO-220 to the edge of the board would work. I would still be able to access the TO-220 after the PCB is slid into the slot.
-
Thank you all for the great inputs.. I would try out these and let you know which one worked the best in this "case".
-
Rather than complicating the process by precision engineering the “heat exchange” type h/sinks, why not mount the transistor (with proper electrical insulation-I would use mica/plastic washer/thermal compound) in a convenient position that does not interfere with the PCB and have the leads soldered to 3 wires that end up in a plug? This is a common practice in a lot of things. Examples can be seen in most Tek scopes. As long as you have a tight fitting plug, I don’t see any issue.
See attached image where you can see the transistor on the left of the EHT transformer on a 465M scope.
Not possible as the case style is where the PCB slide in type thru front / rear openings.
-
I would rather use a case which has a top side open access. Let me not compromise the thermal integrity over aesthetics. This would enable me to mount the TO-220 on the TOP lid and connect that to the PCB using wires.
-
Instead of this kind of style ...
(https://indianextrusions.com/image_article/1566408787-aluminium-box-extrusion.jpg)
Why not just use this below 2 pieces splitted extrusion style enclosure, your problem will be solved nicely.
(http://www.electrostock.com/images/Bop/build_a_box.gif)
-
Thanks, I was going to use that same one you recommend...
-
Rather than complicating the process by precision engineering the “heat exchange” type h/sinks, why not mount the transistor (with proper electrical insulation-I would use mica/plastic washer/thermal compound) in a convenient position that does not interfere with the PCB and have the leads soldered to 3 wires that end up in a plug? This is a common practice in a lot of things. Examples can be seen in most Tek scopes. As long as you have a tight fitting plug, I don’t see any issue.
See attached image where you can see the transistor on the left of the EHT transformer on a 465M scope.
Not possible as the case style is where the PCB slide in type thru front / rear openings.
Why not? Look at his original drawing.
-
Thanks, I was going to use that same one you recommend...
You still have to decide how you are going to mount the transistor. Changing the type of enclosure does not resolve your original issue.
-
I strongly suggest to not use mica for mains-case insulation as it sometimes cracks. Thick-ish silicone pad is much safer solution. Also don't use screw+plastic washer. It's a very weak point. Use square silicone pad without any holes + clip. Or even better, use silicone cap over whole transistor instead of pad.
The only time I have seen mica fail is when it was damaged by being folded. It is pretty delicate but there are other insulating pad materials if something more rugged and easy to use is required. I have seen alumina crack even in normal service.
A clip to apply pressure at the center of the package is one of the best mounting attachments because it is least likely to bend the die frame but it can be difficult to get enough force.
The screw attachment requires insulation and tends not to have enough compliance but works very well if a Belleville washer is used. Split washers do not apply enough force. Because of poor availability of Belleville washers, I tend to use several wave washers stacked together to get enough force.
Motorola published an excellent application note discussing power semiconductor mounting considerations:
https://www.nxp.com/files-static/rf_if/doc/app_note/AN1040.pdf (https://www.nxp.com/files-static/rf_if/doc/app_note/AN1040.pdf)
-
Another option is to drill a (countersunk) hole in the flat bottom of the case and have a slot in the PCB so the TO220 can be mounted with the legs up onto the solder side. That way you can slide the PCB in and use a screw to fix the TO220 to the case (with whatever isolator you want to use).
-
Another option is to drill a (countersunk) hole in the flat bottom of the case and have a slot in the PCB so the TO220 can be mounted with the legs up onto the solder side. That way you can slide the PCB in and use a screw to fix the TO220 to the case (with whatever isolator you want to use).
That works but long term reliability will be compromised if too much tension and compression force is applied along the length of the leads. Solder joints are not mechanically strong and will eventually crack if this happens. Changing force on the leads should always be in bending so lots of mechanical compliance is available and forces are low. This is a real problem with TO-3 parts mounted rigidly between a printed circuit board and separate chassis.
There are some heavy leaf spring type sockets for TO-220 style packages but they are of questionable long term reliability. Better is to solder a three wire stranded pigtail directly between the leads and printed circuit board although this brings up lead length issues. Collet socket pins are available which will fit a TO-220 lead but I have never tried them.
-
Another option is to drill a (countersunk) hole in the flat bottom of the case and have a slot in the PCB so the TO220 can be mounted with the legs up onto the solder side. That way you can slide the PCB in and use a screw to fix the TO220 to the case (with whatever isolator you want to use).
That works but long term reliability will be compromised if too much tension and compression force is applied along the length of the leads.
For clarity: I meant bending the leads and soldering them into through-holes but I see this wasn't clear from my text. I'm aware bends are required to have some form of strain relief from thermal cycles.