PCB temperature sensing.

[kb1gtt]

There are a variety of reasons whey we would want to measure the PCB temperature. Ranging from diagnostics, to warnings, to adjusting for temperature dependent signal delays. What chips are good for this? On one hand I'm tempted to suggest a thermocouple chips, as you can move the sensor easily. However that's kind of expensive. I know the digital semiconductor chips are known for low cost, limited temperature range and high quality. Does anyone have a suggested chip? Perhaps a MAX6642? That max chip is kind of expensive. Can we find at temperature range that goes above 150C and is low cost?

[abecedarian]

Why not the in-built temp sensor on the STM32 processor?

I'd think the PCB temp sensor would only be useful for de-rating things on the board.

[puff]

i bet the built in sensor is not very accurate. i'd suggest using famous dallas 1-wire sensors? i don't see any sense in sensing high temperatures if these are not temperafures of the power switches.

[kb1gtt]

The goal is to measure the heat generated by the injector and ignition perhaps the power supply. Basically if you don't have a good heat sink, or you mount it next to something hot, we could create a warning before failure, then shutdown before it goes out of spec. As well most MOSFET's have variable Ton and Toff times based on temperature. If you measure the temperature you can compensate for this variation in signal delay. I seem to recall the 1-wire stuff was kind of expensive, and often limited to low temperatures. This one is around $4 to $7 in low qty http://octopart.com/partsearch#!?q=DS18B20 It only goes up to 125C, and we have some components that are 150C rated. I guess we can limit to 125C leaving 25C of safety margin. Basically warn at like 110C, then shutdown if we hit the 125C limit. It would be nice to find a 150C rated device. What's the range of the STM's internal sensor?

I was thinking of something more like this http://www.digikey.com/product-detail/en/LM95071CIMF%2FNOPB/LM95071CIMF%2FNOPBCT-ND/723478 in low qty it's $2.45, and you have to hang it off an SPI bus, which we already have. I tend to feel like there is probably still a better sensor chip. For example, this one is under $1 in low qty http://octopart.com/tc77-5.0mcttr-microchip-516247 but is limited to 125C. I'd really like one that does closer to 150C. I also see this one it's under $2 in low qty, but it produces a PWM signal, instead of a serial stream. http://www.digikey.com/product-detail/en/TMP06AKSZ-500RL7/TMP06AKSZ-500RL7CT-ND/810385

I'm assuming we do not want an analog sensor, as we would have like 3 of these sensors. I'm also assuming we'll have thermocouples on the STI bus, so I think we prefer the SPI interface. Do we care if it's PWM?

[puff]

air functions as a perfect thermal insulator which means that you'll need a separate sensor for each critical component. do you really believe it's needed?

[kb1gtt]

I believe the temperature sensors are a really good idea and I believe that a heat sink will be needed. Per my predictions on this page https://code.google.com/p/daecu/wiki/Injector_driver_theory At around 9kRPM, I expect each injector channel to produce about 5 watts. 5 watts times 12 channels is 60 watts of heat that has to be removed from the board. You will not have enough heat rejected from these boards with air alone. Even with a heat sink, if you don't have proper air flow, or you get a bunch of dirt on the heat sink or something like that you can have issues. If you told me I could prevent those issues for around $6, I'd be all for it.

I see two sensors at the injector heat sink, one as close to the middle as possible, as that will be the hottest point, then another at the edge, between the ignition and the Injectors, then a third should be planned for a stepper or H bridge or something like that I'm sure we'll want a 3rd as time goes on. So we should plan for it to have an enable pin.