wrobinson0413 said:
Camlight, I think that you are not being very clear about the FIT. It is only a relative number that is dependent on the sample size that they tested. You should also mention that it is defined as the number of failures in 1 billion hours ie 114,155 years. A typical FIT for the 100V mosfets in a TO-220 case at fairchild was ~1.5 at 55C, so even if you apply your 10C doubling of the FIT number, you are still going to have a low failure in time number if operating the devices within their published specs. I think that it was somewhere around 528 years per failure at a 60% confidence. So it doesn't really mean anything other then to say I don't expect very many failure if the device is operated within spec. What ryanvm777 is trying to figure out is how hot can his controller safely get. No one can say that the inside devices are way too hot because you don't know if where he was measuring it happen to coincide with the end of the internal heat spreader or not. Since his controller looks almost flat black or at least an egg shell black, the IR temperature measurement is probably not that far out, so getting a temperature sensor on the heat spreader or a fet tab would at least give an indication of what is going on. In terms of reliabilty, you are correct in saying that there are other mechanism that could affect the life expectancy of the controller with elevated temperatures, but many of those are related to operating components above their specified ratings. If his controller temperature, voltage, and current stays within the component specifications, then he should expect it to last many years for a typical ebike application. If you have the white paper that derives the 10C relationship to the rise of FIT, I would appreciate it if you could post the document.
I agree with a lot of your post, and I agree that I could have posted a clearer message.
It's a ridiculously tough subject to summarize so briefly.
I agree, there are very, VERY few failures for MOSFET operated within their specs.
I agree, if it's operating within spec it should result in years and years of perfect operation.
However, in my opinion (and that's all we can give with the limited information available), if the external case is at 59C then the MOSFETs are probably way too hot. There's the case-to-spreader thermal resistance, spreader-to-insulator, insulator-to-FET-case, and case-to-junction thermal resistances that are 'in the way" of those MOSFETs cooling themselves. I feel that a controller that typically operates close to, or even above, a MOSFET's specs in 25C ambient conditions will be operating quite a bit above it at 59C. I just don't think that any of these controllers have that much headroom to spare. They're typically designed to operate right at the edge to hold down costs.
You could be right about the IR temperature reading being pretty close to the actual temp. But, we don't know. Add on that these meters typically read low when encountering an emissivity they're not calibrated for and I felt that the temperature had to be
at least 59C and possibly quite higher. Reading the temp of the FET tab can notoriously inaccurate because of that emissivity difference. Also, many IR meters have too wide of a receiving angle and will read the cooler epoxy case or heat spreader temperature along with the tab temp. With some math and specs for the MOSFET insulator used (Sil-Pad, mica, whatever), reading a black heat sink'ss temp can lead to a pretty good estimate of the junction temperatures of the MOSFETs though.
All that added up to my opinion that it was too hot in there and I still strongly believe that's true.
But, this is just my opinion as it's the only thing I can offer.
A quick check didn't lead me to the ones I've downloaded but the reliability papers and testing reports from IR, TI, Toshiba, etc., are a great source for temp. vs. failure info.
