| Electronics > Projects, Designs, and Technical Stuff |
| Unexpected amount of heat |
| << < (4/5) > >> |
| Mechatrommer:
--- Quote from: OM222O on May 12, 2019, 03:55:43 pm ---thanks for the detailed answer! I can place a hard lock of 5s cool down after each measurement (which again, only lasts for 2 seconds and dissipates about 8 watts). --- End quote --- but you are testing it up to 10 sec in OP and surprised that it got too hot. which engineering stress test you want to satisfy? 2 sec? or 10 sec? and from the impression, you want minimal heatsinking. there are more involved, such as how air is ventilated in your product, more ventilation, less heatsink (metal) is needed, no ventilation? you can calculate maybe from thermal conductivity of various materials in your product... http://hyperphysics.phy-astr.gsu.edu/hbase/Tables/thrcn.html https://www.engineeringtoolbox.com/air-properties-viscosity-conductivity-heat-capacity-d_1509.html i'll just go with try and error method until i'm happy with the system. if you want PTC route, fine, thats another try and error method. ymmv cheers. |
| Siwastaja:
You can look up the basic physical equation for specific heat capacity, weigh the transistor, assume it's close to the specific heat capacity of copper (since the large copper tab would dominate its mass), and simply calculate the temperature rise for your energy of 9W times 2 seconds. The result would be pessimistic, assuming no heat flow out during the pulse, and ignoring the mass of solder added (which will be more significant than the mass of copper in PCB pad). This way you get a rough idea whether you can ignore the pulse length and just calculate using average power, or if there is significant per-pulse heating. For average dissipation, you need: 1) surface area, 2) thermally conductive paths to said surface For peak dissipation, you need: 1) mass, 2) thermally conductive paths to said mass It's well possible that adding extra solder by adding solder stencil openings to the collector copper fill does the trick for practically free. Oh, you have software, you can easily control the duty cycle. If you fear about user rebooting to skip the wait-time, just insert an extra wait right after the boot. |
| OM222O:
--- Quote from: Siwastaja on May 12, 2019, 09:33:40 pm ---You can look up the basic physical equation for specific heat capacity, weigh the transistor, assume it's close to the specific heat capacity of copper (since the large copper tab would dominate its mass), and simply calculate the temperature rise for your energy of 9W times 2 seconds. The result would be pessimistic, assuming no heat flow out during the pulse, and ignoring the mass of solder added (which will be more significant than the mass of copper in PCB pad). This way you get a rough idea whether you can ignore the pulse length and just calculate using average power, or if there is significant per-pulse heating. For average dissipation, you need: 1) surface area, 2) thermally conductive paths to said surface For peak dissipation, you need: 1) mass, 2) thermally conductive paths to said mass It's well possible that adding extra solder by adding solder stencil openings to the collector copper fill does the trick for practically free. Oh, you have software, you can easily control the duty cycle. If you fear about user rebooting to skip the wait-time, just insert an extra wait right after the boot. --- End quote --- yes that seems like a very reasonable approach! unfortunately I have no way of accurately measuring the mass of copper so I just assumed a 1g estimate (not sure if it's too small or too large). specific heat of copper is 0.385 J/g.k which means with a 2 second on time, it's only about 8.5 degrees temperature rise. more than acceptable! no heat sinking required. I think this is actually a very good estimate as well, as it's inline with initial testing (same estimate results in 46 degrees after 5 seconds which feels hot and 68 degrees after 10 seconds which feels uncomfortably hot). I think 73 degrees is the max you can touch without getting burnt ... not sure if it's accurate, just read it somewhere. So knowing that with pulsed load and low duty cycle I shouldn't require any heat sinking at all since the 8/9 degrees is more than reasonable here! thanks a lot for the help. you saved my day! |
| IanB:
--- Quote from: OM222O on May 12, 2019, 10:16:46 pm ---I have no way of accurately measuring the mass of copper --- End quote --- If you know the plating thickness of the copper and you measure the area you can estimate the mass that way. --- Quote ---So knowing that with pulsed load and low duty cycle I shouldn't require any heat sinking at all since the 8/9 degrees is more than reasonable here! thanks a lot for the help. you saved my day! --- End quote --- The duty cycle matters. If it doesn't fully cool down after each pulse the next pulse will reach a higher maximum temperature. |
| OM222O:
I meant the copper used in the D2PAK since that is similar to junction temperature (1C/W resistance). to measure that I need to cut off the 2 legs, measure the mass of the package with a sensitive scale (don't have anything that's more accurate than 1gram at the moment) and dissolve the package in some sort of acid which then leaves the plastic package and silicon behind. finally measure the mass again and find how much was lost which equals to the mass of the tab. IMHO the 8.5 degrees increase is more than fine which again, is a very bad estimate since it ignores the thermal mass of everything else + the amount pulled away by the PCB copper :-+ |
| Navigation |
| Message Index |
| Next page |
| Previous page |