For you heat transfer and temperature measuring experts: I'm designing a PCB into this enclosure: http://media.digikey.com/photos/BUD%20Industries%20Photos/PN-1323-C.jpg It will be mounted on a wall so the clear cover is facing the user. The environment is similar to residential so no harsh environment. Standoffs will be screwed into the bosses, and a PCB will be mounted on those. There will be various indicators visible through the clear cover. The PCB needs a sensor to measure the outside ambient temperature. The PCB itself will generate some heat, maybe on the order of 1W max from a micro, a few relays, etc. My thought is to drill a matrix of vent holes holes on the top and bottom for two reasons: 1. allow the electronics to keep as cool as possible, and 2. hoping the resulting convection pulls in enough ambient air to flow over a temperature sensor mounted on the bottom of the PCB. I was thinking of a TO-92 package temperature sensor, mounted straight off (perpendicular) the PCB and utilizing the full length of the leads to get it as far as possible from the PCB. So, I know this is a complex question and testing is the only way to know for sure. Anyone have any experience or have any opinions on this type of thing? Specifically, what is the likelihood the sensor will accurately measure the ambient temperature within a few tenths degrees C with only convection? I know the heat producing parts need to be kept away from the sensor (i.e. on the top) and copper traces conduct heat very well, and the TO-92 sensor leads conduct heat very well too.
You could try to model your problem with an US NSF-funded free, open-source, heat flow modelling application Energy2D at http://energy.concord.org/energy2d/ and see if that gives you any answers. It's designed with high-school students in mind, so it shouldn't be difficult to learn. It's not claimed to be 100% accurate for convection and radiation problems, but it might give you some ideas. Explore the website to see if it might be suitable for you. Else, build up your circuit and power it up, and then look at it through an IR camera or IR pyrometer to see how much heat it produces and where, and what sort of convective flows are set up. Depending on your convection flow results, you might want to think about physically segregating your temperature sensor away from your circuitry, by mounting the sensor in a plastic or cardboard tube within your enclosure, say, so that ambient temperature air flow does not mix with the convective flow your circuit is producing. You may need to extend your sensor's leads for that. The tube would be mated to openings in your enclosure at top and bottom.
In the last paragraph I mean to say "so that ambient temperature air flow to_the_sensor does not mix with the convective heat air flow your circuit is producing."
You can have a look at a ~17 min on demand video about a professional thermal flow package called FloTherm at the Mentor Graphics website: http://www.mentor.com/products/mechanical/products/flotherm Unfortunately I couldn't see any evaluation/demo copies of this package. Do a search for "computational fluid dynamics" and you might find some packages with evaluation copies, or that are open source. There is an open source CFD package that runs on various Linux flavours at http://www.openfoam.com You may do well to check out the Australian CSIRO online CFD resources at http://www.cmis.csiro.au/cfd/
: Edited by User
jm wrote: > Specifically, what is the likelihood the sensor will > accurately measure the ambient temperature within a few tenths degrees C > with only convection? Hello, depending on the sensor used you have to regard self heating of the sensor. E.g. a LM335 used at 1mA will heat around 0.5-1 degrees (celsius) from self heating in still air. If you put the sensor at the bottom of the case and cut some slots into the PCB to further isolate the sensor and use only low power for the sensor (or pulsed usage) you should get the environment temperature with below 1 deg C error. Regards Anja