# Weather Station remote sensor temperature correction

#### CrustyS

Joined May 20, 2019
27
I have a radio controlled weather station which has always shown 2°C less than the environment in which it is placed for the remote sensor, verified by good quality mercury thermometer and a couple of other systems I also have.

I particularly like this station as it shows sun rise/set and moon rise/set for my city and the overall data presentation is quite good. So I decided to try and improve the accuracy.

The remote module uses an NTC for temperature so my first step was to remove this and do a calibration curve. I recorded 10 points from 5°C to 35°C and did an X/Y plot with Excel. Data shows linear relationship with R² of 0.98 which is a pretty go fit. The NTC value for 25°C is 44K which is little odd but could be a custom sensor.

Looking to my data set a 2°C increase would need a reduction in resistance of 3K (I wonder if the software is expecting a 47K NTC as mine 44+3 would be 47K coincidence ?).

I think the measurement is using a voltage divider.

Calculating a resistance in parallel I need to get the 3K reduction I get 1 Meg ohm, or so I think.

I would be most thankful if some one with more knowledge could cast an eye I over what I have done and point out any mistakes or erroneous assumptions I have made before I start soldering.

I haven't posted the data but if any one is interested I will.

TC

#### MrChips

Joined Oct 2, 2009
30,988
I have a radio controlled weather station which has always shown 2°C less than the environment in which it is placed for the remote sensor, verified by good quality mercury thermometer and a couple of other systems I also have.

I particularly like this station as it shows sun rise/set and moon rise/set for my city and the overall data presentation is quite good. So I decided to try and improve the accuracy.

The remote module uses an NTC for temperature so my first step was to remove this and do a calibration curve. I recorded 10 points from 5°C to 35°C and did an X/Y plot with Excel. Data shows linear relationship with R² of 0.98 which is a pretty go fit. The NTC value for 25°C is 44K which is little odd but could be a custom sensor.

Looking to my data set a 2°C increase would need a reduction in resistance of 3K (I wonder if the software is expecting a 47K NTC as mine 44+3 would be 47K coincidence ?).

I think the measurement is using a voltage divider.

Calculating a resistance in parallel I need to get the 3K reduction I get 1 Meg ohm, or so I think.

I would be most thankful if some one with more knowledge could cast an eye I over what I have done and point out any mistakes or erroneous assumptions I have made before I start soldering.

I haven't posted the data but if any one is interested I will.

TC
I don’t see much harm in trying your 1MΩ hack. Try it and see the effects.

#### CrustyS

Joined May 20, 2019
27
Data and graph

#### Attachments

• 21.6 KB Views: 4
• 23.3 KB Views: 4

#### CrustyS

Joined May 20, 2019
27
Update.
I reattached the NTC and added some extension wires so I could try different resistances. First one 1 M ohm only gave 1°C increase, half of what I was expecting. I then changed this for a 1.5 M ohm potentiometer and adjusted it down from 1.5 M slowly allowing for the value to settle. I was rather surprised that the highest value was actually a little lower than the 1 M fixed resistance I had tried. I continued to reduce resistance and the "measured temperature" increased until I reached the +2°C I was looking for. The resistance was way less than I had anticipated and measured 495K when detached from the circuit. I used a fixed 500K next and this confirmed the +2°C. Hmmm.
Last experiment was to put the pot back in the circuit and wind it down further. The temperature duly increased but in a non linear fashion and hit "HH" at minimum resistance.

I guess my assumptions about he voltage divider were wrong. Any insights are most welcome.

TC

#### Alec_t

Joined Sep 17, 2013
14,385
Putting a fixed resistor in parallel with the NTC can correct the reading for one particular temperature, but not for the whole range.

#### MrChips

Joined Oct 2, 2009
30,988
Try finding the parallel resistance required at 25°C.
Then find an NTC thermistor at that value and use that in parallel instead of the fixed resistor.

#### CrustyS

Joined May 20, 2019
27
Try finding the parallel resistance required at 25°C.
Then find an NTC thermistor at that value and use that in parallel instead of the fixed resistor.
I will give it a go, thanks for the advice

#### CrustyS

Joined May 20, 2019
27
Putting a fixed resistor in parallel with the NTC can correct the reading for one particular temperature, but not for the whole range.
Understood as the as the relationship is as shown in the graph equation.

#### CrustyS

Joined May 20, 2019
27
Final update.
I ordered some mixed value NTC's and did a little experimenting and found that a 300K ohm NTC in parallel gave me the offset I needed and was 1.7 - 2.0°C from 5-30°C, I was looking for 2°C.

Thank you to all those who helped and especially to MrChips

#### MrChips

Joined Oct 2, 2009
30,988
Final update.
I ordered some mixed value NTC's and did a little experimenting and found that a 300K ohm NTC in parallel gave me the offset I needed and was 1.7 - 2.0°C from 5-30°C, I was looking for 2°C.

Thank you to all those who helped and especially to MrChips
Thanks for the update and acknowledgement. It is always appreciated when a member responds with a final update.