This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LMT70A: temperature accuracy outside spec

Part Number: LMT70A
Other Parts Discussed in Thread: LMT70

One temperature probe in a batch of 40 is consistently measuring lower temperatures than all the others.

The probes consist a flex PCB that is 4mm wide where the LMT70A is mounted, but the width increases to 10mm at the other end for a connector. Total length is 150mm. The boards have a 10um paralin overcoating.

I used one probe as a "reference" and dipped each probe tip in turn into an oil bath together with the "reference" probe to get an idea of the spread of temperature differences. Typically, measurements were within 0.2 degC of each other at 100 degC which sit nicely within the LMT70A spec. However, one probe consistently measures lower than the others:

At 100 degC, error = -0.64 degC

Around room temperatures the error is within +/-0.1 degC, so there is at least a gain error.

What could cause the accuracy variation to beyond the spec?

The contract manufacturer did not like mounting the small BGA package to the skinny flex board. Could something in the reflow temperature profile (eg. a local hot spot) cause a parametric change?

Could ESD cause a change?

Any help would be appreciated since we anticipate many more batches in the future.


  • Hi Tony,

    Let me think about this one.
  • Tony,

    It is important to note that the accuracy specified in the LMT70 datasheet assumes minimal error contribution from other system components such as voltage reference error, ADC gain and offset error, adc quantization error etc  

    You will have to consider all of the component variation over process, temperature, and voltage.

    On top of that, the pcb manufacturing process could also be a factor.  For example, flux could be an issue.

    There are many moving parts in this solution and that is a challenge.



  • Hi Kelvin,

    Thanks for your suggestions.

    I am comparing multiple LMT70 based temperature probes using common electronics. The low temperature reading consistently follows a single probe, so I'm fairly confident the measurement circuit is not the issue.

    When I consider PCB leakage currents, the output resistance of the LMT70 is quite low (< 80 ohms) and requires a "strong" pull-up current to upset the accuracy:

    If the probe is reading too low at elevated temperatures, then the measured voltage is too high. Any leakage current would then be a path from Vdd (+2.7V) to the output. At low temperatures the readings are more accurate so the leakage currents must be smaller. I calculate perhaps 30k at 100 degC (5.6mV offset), and 1.5M at 5 degC (0.1mV offset).  I don't know if this is a plausible explanation.

    This is proving to be a bit tricky. I will try some unpowered pin-to-pin resistance measurements of the probes and see if there is anything noteworthy.


  • Tony,

    You mentioned that the low temperature reading follows a single probe. Does this mean that you have checked this "bad" LMT70 device on another good probe and get the same behavior? You could also test a new LMT70 device on this "bad" probe and see if the reading is still off.

  • Hi Kelvin,

    I have only been swapping complete probe PCB assemblies to verify operation and to isolate the one "rogue" assembly out of twenty probes.

    I have not attempted to swap devices between different probe PCBs because the LMT70 BGA package is so small and delicate our chances of successful removal and re-soldering are minimal. Even our previous experiences on replacing devices with new ones on other probes that we damaged were not totally successful. However, as a last resort we could try replacing the device with a new LMT70, but we first need to complete our life test with the existing probes, and then perhaps send out to our PCB assembler for surgery in another week or two.