Integrated Temperature Sensor Accuracy Question

Hi Max,

    I believe, you can find the answer to your question at the MSP430F2132 datasheet, under temperature sensor.

-kel

Instead of reading datasheet as suggested in your previous post you just start new thread with literally the same question? You are hoping that someone else will read datasheet for you and report here? I think it's pathetic.

Hi llmars,

Aside from the unprofessionally rude comment, I did read the datasheet and the family user guide. Go troll elsewhere.

Thanks Markel,

I did review the spec sheet and the family user guide quite a few times. They were not clear as to what the typical accuracy is.

Max Tibbetts said:

I did read the datasheet and the family user guide

On page 46 there's table 10-Bit ADC, Temperature Sensor and Built-In VMID. Check "Sensor output voltage" data, it contains information needed to estimate sensor (initial) accuracy. For instance at 0 degree Celsius sensor min/typ/max output is 895/995/1095 mV. For 25 degree Celsius it's 985/1085/1185 - huge overlap with 0 degree numbers. So as you can see without calibration this temperature sensor is useable only to estimate - msp430 core is really hot or really cold :D I would not recommend to use this sensor for any other functions than temperature correction of msp430 peripherals (like ADC).

[edit] I did read datasheet for you. And in result did not change my mind ;)

Hi llmars,

Thank you for that bit of information. However that did not fully answer my question. Specifically speaking, what is the typical accuracy I can get from this IC with calibration? 

Also, please be professional and courteous in your future posts. Rude comments are generally not appreciated by other people, especially so among other engineers,

Max Tibbetts said:

Specifically speaking, what is the typical accuracy I can get from this IC with calibration?

Unfortunately I don't know. Absence of detailed performance data for tempsensor clearly shows: not recommended for most of the applications, use at your own risk - do calibration and parametrization yourself :) Also consider that msp430 itself have some heating when loaded. Personally I am happy with LM92 - it ispredictable, works out of the box :)

Max Tibbetts said:

Also, please be professional and courteous in your future posts.

LOL thank you :) Peace. In return hints for your future posts:

1) don't post twice. if you accidentally do - there's delete option

2) before you ask - check that documentation/datasheet indeed does not contain answer. if unsure, then briefly tell what you supposedly know - so we see that you are NOT ignorant "give me working code NOW" type, that you did your homework before asking.

Robert Cowsill said:

As you've pointed out, the datasheet gives no information on the tolerance when using the ADC temperature sensor calibration values. This is in stark contrast to the DCO calibration constants, which have their tolerances specified over temperature and/or supply voltage.

Basically, the temperature coefficient is constant, so voltage changes linearly with temperature.
However, neither offset nor gain are closely specified. But with the information of a constant temperature coeeficient, both values an be determined by a simple two-point calibration. Apply two known temperatures, read the resulting voltages and you can calculate the offset adn the gain very precisely (only limited by ADC resolution, as its offset and gain errors are part of the calibration results, sufficient settling time after exposing the MSP to these temperatures, and precision of the applied temperature.

Of course, the measured temperature is chip die temperature, not ambient temperature, so depending on the packages thermal resistance and the momentary power dissipation, the measured temperature might be higher than ambient.

Some MSPs have factory-determined values for 30°C and 80°C. So the difference between these two values is the gain for 50°C etc. However, due to limited testing time (production cost) these values aren't that precise. But they have a tolerance specified :)
(own calibration gives better results with tolerance towards 0)

Jens-Michael Gross said:

Some MSPs have factory-determined values for 30°C and 80°C. So the difference between these two values is the gain for 50°C etc. However, due to limited testing time (production cost) these values aren't that precise. But they have a tolerance specified :)

Where are those tolerances specified? These are the temperature calibration constants I was referring to in my last post, and there's no sign of tolerance specifications for them in the datasheet.

From the 5x family users guide:

1.13.5.3 Temperature Sensor Calibration
The temperature sensor is calibrated using the internal voltage references. Each reference voltage (1.5/2.0/2.5 V) contains a measured value for two temperatures, 30°C±3°C and 85°C ±3°C and are stored in the TLV structure.

Well, with own calibraiton, I'm sure you can do much better :)

Aha! You're right, as usual, Jens-Michael :)

The 2x family user's guide also specifies the calibration tolerance in the TLV section. Interestingly the 2x chips have better specs than the 5x, with a tolerance of ±2°C!

I wonder why these specifications aren't in the datasheets though. Perhaps TI think the temperature sensor tolerance is so bad that they don't want to shout about it, but they do list the VLO drift with temperature and that's woeful!

Robert Cowsill said:

I wonder why these specifications aren't in the datasheets though. Perhaps TI think the temperature sensor tolerance is so bad that they don't want to shout about it,

This is why I said that doing your own calibration will surely result in a much higher precision.

Robert Cowsill said:

but they do list the VLO drift with temperature and that's woeful!

I'm not complaining about the precision (or lack thereof), just wondering why that information is in the user guide rather than the datasheet.

The tolerances for the DCO calibration values are in the datasheet, so it seems logical that the temperature calibration tolerances would be listed there as well.

Robert Cowsill said:

just wondering why that information is in the user guide rather than the datasheet.