[NHS3100][tlogger] How temperature sensor resolution mode works?

[aresmarser]

Dear Sirs,

I am now working on NHS3100 kit and studying the whole system.
I read the data sheet and user manual and have some questions about "temperature sensor resolution mode".

1. As my understanding, once some resolution bit mode set, the TSEN will measurement the temperature with the corresponding Celsius.
For example, if 9 bit resolution mode(TSEN_9BITS) using, the temperature Celsius output will be the multiple of ±0.2℃, that is, 0.2, 0.4, 0.6, and so on.
If 8 bit resolution mode(TSEN_8BITS) using, the temperature Celsius output will be the multiple of ±0.4℃, that is, 0.4, 0.8, 1.2, and so on.

Is it correct?

2. How TSEN works for resolution Celsius according to the resolution bit mode?
Is it a software algorithm or a hardware aspect?

3. I tried to update the resolution bit mode such as 8-bit, 9-bit and I can see that SP0 register TRESMODE has been updated, but the temperature output is always the multiple of ±0.1℃.
Why? Anything I omitted?

I have got confused for several days.
Anyone can help to explain?
Appreciated.

Postscript:

--- user manual ---

--- data sheet ---

Thanks,
Arna

[driesmoors]

Hi Arna,

1. As my understanding, once some resolution bit mode set, the TSEN will measurement the temperature with the corresponding Celsius.
For example, if 9 bit resolution mode(TSEN_9BITS) using, the temperature Celsius output will be the multiple of ±0.2℃, that is, 0.2, 0.4, 0.6, and so on.
If 8 bit resolution mode(TSEN_8BITS) using, the temperature Celsius output will be the multiple of ±0.4℃, that is, 0.4, 0.8, 1.2, and so on.

Is it correct?

No. The TSEN HW block uses a zoom ADC as operating principle. First a coarse search is performed; after that a delta-sigma approach is taken to increase the bit resolution. The latter part takes time and can be controlled by the mentioned register field. That value then still has to be converted to a meaningful value. Conversion to a signed 10.6 2-complement fixed point Kelvin value is done in HW; further conversion to a Celsius or Fahrenheit value is done in SW. The resolution will then be accurate to +- 0.2 for a 9-bit resolution, but that does not mean this will result in exact temperature values 0.4, 0.6, 1.2, and so on. 

I assume this answers 2 and 3 as well?