Real accuracy of TC1047A?

[Coper]

I need to measure the temperature with a resolution better than 0.1C, an absolute accuracy of 0.5C, and at a rate better than 5Hz. Sensor in a small max three-pin case, will be connected with a cable of approx. 30 cm. Measured temperature range from room temperature to 95C, it is mainly a range of 65 to 85C

I was intrigued by the TC1047A, but

• High Temperature Converter Accuracy: ± 2 oC, Max, at 25°C, but according to Figure 2-1, a supply voltage higher than 2.5V causes the main error. Msy be use supply voltage 3,3V over diode 3.3V – 0,7 = 2,6V
• according to Figure 2-1, a supply voltage higher than 2.5V causes the main error and output voltage 0-100C 500mV to 1000mV

When using a 12bit ADC from some STM32Fxxx, I can't quite get to 0.1C resolution. If I use an amplifier with a gain of 2x something like the MCP6021 the output voltage will be 1V to 3V and slope 20mV/C.

If I were to calibrate the entire system at one point, for example 50C, is there a chance that it will also meet the long-term required parameters?

Thanks

 

[FvM]

If you look at the accuracy curve in datasheet, it's rather unlikely that achieve 0.1 K accuracy over temperature with a one point calibration.

Long term stability isn't specified.

 

[KlausST]

Hi,

Expected accuracy is 0.5°C.

use supply voltage 3,3V over diode 3.3V – 0,7 = 2,6V

And you want to achieve this ... while you use VCC as some kind of reference??? Good luck!


Klaus

 

[dick_freebird]

The old school bipolar sensors seem to be around 1 degC
accuracy ("available" - so binned out for higher price if that's
your pony). Curvature is variable and the accuracy blows out,
outside the 0-70C range. That's inherent to a bandgap
reference (although mine with thin film metal resistors have
a fraction of the curvature, the bit left that pertains to the
transistors' Rb, but the Pbase resistors commonly used
contribute most of the nonlinearity in common versions).

Always begin by questioning the difficult attribute specs,
including "why?". In my experience I have not once cared
about decimal points of temperature (in my world we just
sandbag a forcing temp to ensure it meets a MIL method
tolerance window that's more based on ordinary equipment
capability).

Now maybe some clever marketing person thinks that
enhanced temperature resolution & accuracy will win sales.
Or maybe you received the spec from a customer who's
always right. There may well be reasons.

But they are not necessarily good or achievable at a cost
that's going to not-tank your plan, and "self-care" means
not signing up for impossible or stupid stuff.

 

[FvM]

The graphs and tables provided following this note are a statistical summary based on a limited number of samples and are provided for informational purposes only. The performance characteristics listed herein are not tested or guaranteed.