As mentioned earlier, I wanted to make a new starter solenoid because the one designed around a 40 amp automotive relay wound up much too big. A new one was designed around a power MOSFET whose heatsink set a lower limit of the solenoid's size. In this simple switching application for a MOSFET, it's the internal heat created by the device's ON resistance that needs to be controlled.
Guesstimates based upon earlier starter tests showed the solenoid may have to handle cranking currents as high as 70 amps. A TO-220 packaged Infineon IPP011N03LF2S was selected for its impressive 200 amp capacity and .001 ohm ON resistance (which by the way is an order of magnitude lower than the relay's contact resistance). At a worst-case 70 amps this would work out to be 5 watts of dissipation inside the Delrin solenoid.
The new solenoid was designed, but only its heatsink was initially machined for some preliminary tests. A pair of screws eventually connect the heat sink to the solenoid's mounting bracket and so the MOSFET must also be electrically insulated from the heatsink.
While waiting for the .001 ohm MOSFET to arrive, I mounted a .012 ohm IRF3205 MOSFET (I had plenty of these on hand) to the heatsink for testing. A 20 amp resistive load was set up to generate 5 watts of heat in the test device. Measurements showed its temperature approaching 100C in under ten seconds. More scary was the 20C temperature differential measured between the device and its heatsink. Ceramic, mica, and Kapton insulators augmented with silver grease all behaved similarly.
Plan B was to machine the solenoid body from aluminum rather than Delrin and to use the entire solenoid as a heatsink. With the test device mounted to a large chunk of aluminum the results were only marginally better with the device reaching 90C after 20 seconds or so.
The aluminum bodied solenoid was machined, but in the process of tidying up the test area on my bench, I broke a lead off my test device. The replacement part performed so much better that I thought the test had somehow changed.
To make a long story short, the first device came from a batch of 3205's purchased from Amazon, while the second device was taken from a batch purchased from Mouser Electronics - an actual legitimate supplier. Three devices from each batch were tested, and the results were essentially the same. Under identical load conditions the Amazon parts came close to thermal runaway while the Mouser parts warmed up to only 43C or so. In addition, the temperature differentials across the insulators dropped to less than 2C with the Mouser parts.
After reading a couple one star reviews among some 400 positive reviews for the Amazon parts, it was obvious that others who had actually used the parts had run into similar problems. After disassembling a couple, one user discovered the die size of Amazon parts was just a fraction of a legitimate part.
After wasting several days in this rabbit hole, the Delrin body was machined and the new solenoid completed as originally intended. The partially finished aluminum solenoid is a paper weight reminder of my experience with counterfeit Chinese parts.
Tests on the completed solenoid with the .001 ohm MOSFET included a repeat of the 20 amp tests which showed essentially no temperature rise. The solenoid was also tested with an unloaded motor, because the earlier torque tests couldn't be repeated since the test fixture had already been repurposed.
OK, but wait ...
So what goes on inside a 40 amp automotive relay while carrying 20 amps through typical .030 ohm contacts and generating some 12 watts of internal heat? I cut one open and connected it to my test bed to find out. Sure enough, the contacts get hot - 65C in fact. What keeps the relay from self-destruction is an efficient internal heatsink that includes the relay's entire frame, coil, and five thick connector blades. The contacts themselves are bonded to wide copper straps that carry much of the heat away from them where it's conducted through the blades and into sockets and circuit board traces. It's a nice design that uses all five blades to conduct heat out of the relay. - Terry