Hello all,

I have a DAQ system which will only allow me to put a scaling factor in the form of xxxV = 1C
Using a 10k NTC thermistor, I am trying to linearise its charastics from 0C - 110C so I can get at good as possible accuracy.

Currently I am just trying to work this out on paper & spreadsheets. I have tried modelling a parellel + series resistor but the numbers I have worked out start getting wildly inaccurate as I move away from the centrepoint of 60C.

Can anyone offer a suggestion please?

p.s I am using a 5V supply, but this can be altered.

Thank you

 

For starters, please provide the brand and part number of the thermistor and any links to the spec sheet and characteristics. Manufacturers usually provide a table showing resistance vs temperature.

 

Sadly the part is no longer manufactured, its a EPCOS B57860S103H, but a list of specs:

Min -40C
Max 150C
Res at room temp = 10K Ohms
B-Value 3988
Temp Therm R

-40.0 336500.0
-35.0 242589.0
-30.0 177000.0
-25.0 130370.0
-20.0 97070.0
-15.0 72929.0
-10.0 55330.0
-5.0 42315.0
0.0 32650.0
5.0 25388.0
10.0 19900.0
15.0 15708.0
20.0 12490.0
25.0 10000.0
30.0 8057.0
35.0 6531.0
40.0 5327.0
45.0 4369.0
50.0 3603.0
55.0 2986.0
60.0 2488.0
65.0 2083.0
70.0 1752.0
75.0 1481.0
80.0 1258.0
85.0 1072.0
90.0 917.7
95.0 788.5
100.0 680.0
105.0 588.6
110.0 511.2
115.0 445.4
120.0 389.3
125.0 341.7
130.0 300.90
135.0 265.40
140.0 234.80
145.0 208.30
150.0 185.30

 

For NTCs, I always use a simple voltage divider. Fixed R on top, and the NTC to ground.

I generate a spreadsheet to plot the curve of Vout vs. T. You can use either the NTC formula for the part, or the mfgrs. R vs. T table.

Then, I adjust the fixed R value until the temperature range of interest is within the most linear portion of the curve.

Finally, using the spreadsheet, I generate a table of digital counts vs. T that the processor will see for the desired range. Usually, one data point for 1C is acceptable for accuracy and linearity.

I program this table into my code, and use an interpolation algorithm to compute the final temperature.

Make sure you take into account NTC self heating in your final accuracy analysis.

 

For the data you gave, I'd use about a 3K ohm resistor:

R    3000  
      
T    Rntc    Vout
-40    336500    99.12%
-35    242589    98.78%
-30    177000    98.33%
-25    130370    97.75%
-20    97070    97.00%
-15    72929    96.05%
-10    55330    94.86%
-5    42315    93.38%
0    32650    91.58%
5    25388    89.43%
10    19900    86.90%
15    15708    83.96%
20    12490    80.63%
25    10000    76.92%
30    8057    72.87%
35    6531    68.52%
40    5327    63.97%
45    4369    59.29%
50    3603    54.57%
55    2986    49.88%
60    2488    45.34%
65    2083    40.98%
70    1752    36.87%
75    1481    33.05%
80    1258    29.54%
85    1072    26.33%
90    917.7    23.42%
95    788.5    20.81%
100    680    18.48%
105    588.6    16.40%
110    511.2    14.56%
115    445.4    12.93%
120    389.3    11.49%
125    341.7    10.23%
130    300.9    9.12%
135    265.4    8.13%
140    234.8    7.26%
145    208.3    6.49%
150    185.3    5.82%

 

That would also be my first way for measuring. Sadly I only have a DAQ system with only allows me to add a scale factor in, in the manner "xxx mV = 1C" So I'm trying to recreate a linear temp to voltage using only analogue electronics.

 

Look here
http://mathscinotes.wordpress.com/2011/07/22/thermistor-mathematics/