• State TI Thinks Resolved
  • Locked Locked
  • Replies 1 reply
  • Answers 1 answer
  • Subscribers 30 subscribers
  • Views 506 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

LM35: LM50 vs LM35 Accuracy performance

Jacob Mader
Prodigy 10 points
Part Number: LM35
Other Parts Discussed in Thread: LM50,

Tool/software:

My company currently uses the LM50 component to evaluate and report a system temperature data stream. The absolute accuracy is not of incredible concern for me, however we use this system temperature data point on some systems to coordinate heaters and fans in order to realize a fairly consistent temperature, despite whatever external environmental conditions a system might encounter. The deadband between the heater turn on/off threshold and fan turn on/off threshold seems to be much larger than needed in order to get reliable performance. On one system I've been testing this week for example, I've seen the fans thresholding behavior occur between 33.5C - 36.5C. This seems to be in line with the LM50's accuracy specification, however colleagues had suggested that the variance in reported temperatures might be based more in noise behavior than the temperature sensor accuracy. 

As I interpret it, the temperature sensor accuracy essentially provides a band of values that might be reported, and could essentially be interpreted as a noise spec in the temperatures reported, and thus I've identified the LM35, as a very comparable component to upgrade to requiring minimal design adjustments, with an accuracy specification that would allow a much tighter band to be seen in temperature reporting and thus better control of the dead bands between the heaters and the fans.

Am I interpreting this accuracy specification correctly, or does it seem likely that this accuracy spec is only applicable to the absolute accuracy, and the variance I see in reported temperatures will likely stay rather similar to what I see with the LM50?

  • 0
    TI__Mastermind 43785 points

    You're not wrong that LM35 is more accurate than LM50. I agree that LM35 is a fine upgrade from LM50.

    Most of the temperature accuracy specification comes from our ability to trim it to zero error in production. When you mentioned noise, you were talking about noise in your system when you use the sensor. That noise is not necessarily reflected in the accuracy specification. It's actually the noise in our production test that drives the accuracy specification we can claim. We want to make sure we never ship a device that violates the spec, so we use statistical analysis of previous measurements to ensure we choose a spec that will enable that to happen. This means the devices should always be a bit better than what we claim they are, and this is due to uncertainty in our production test measurement. However, there is another major consideration.

    The IC behavior is expected to vary over its operating range. This includes both temperature and voltage supply. It is not practical to test every sensor in production at every temperature and/or every possible voltage. Again, we use statistical analysis here to make a decision on worst-case behavior across variation, and write the spec to accommodate that. This error source has to be summed with the error source discussed above: they both represent uncertainty that must be accounted for to ensure we never ship out-of-spec devices. There are other error sources that we must consider, but these are the two main factors.

    Temperature change - thermodynamics - is often slow. Practically speaking, even if you know the current temperature very accurately, you probably can't change it quickly. You don't really want your feedback to oscillate on and off. This is what hysteresis is for. A good system design will intentionally ignore temperature change until it is deemed significant enough to act on. Your system likely has some form of hysteresis, and I'd expect that is part of what you're observing. 

    Finally, a comment on temperature accuracy. Atmosphere is not a great thermal conductor, and heat sources are everywhere. It is not unreasonable to measure a different temperature in different locations in the same enclosure. There can also be subtle differences in our environment that impact a sensor. For example, if one of the systems being tested is closer to a window or an air vent. I expect this is some of the variance your colleagues alluded to.

    thanks,

    ren