# Modify a thermistor's resistance curve?

1. Sep 2, 2009

### mjl5007

I'm hoping someone with a bit more EE background than myself can answer a question for me.

I've got an outdoor temperature sensor which is a simple thermistor with two lead wires. I'm wondering if it's possible to modify the resistance curve of the thermistor with a simple circuit (i.e. by adding an extra resistor or other such component)? I've attached a simple chart of the actual resistance curve of the thermistor and the desired curve.

http://img200.imageshack.us/img200/3185/rxcurves.th.png [Broken]

Last edited by a moderator: May 4, 2017
2. Sep 2, 2009

### Staff: Mentor

Welcome to the PF. It doesn't look like a simple parallel combination of a resistor or another thermistor will get you the desired curve. Can you just do a calibration table at the device that is reading the resistance? What device are you using to read it?

Last edited by a moderator: May 4, 2017
3. Sep 2, 2009

### skeptic2

I think a table of values would be a lot more useful. Temp vs Res. for both actual and desired. (or even a couple of formulas)

4. Sep 2, 2009

### mjl5007

It's actually an outdoor temperature sensor for an HVAC system thermostat. When the system was installed, it came with the base model thermostat from the manufacturer and the accompanying outdoor temp sensor for that thermostat. I've since installed a more advanced thermostat from a different manufacturer, and the "compatible" outdoor sensor from that manufacturer has a different curve than the one I currently have, so I can't just plug in the old sensor because it will read incorrect temperatures. My thought was that if I could modify the resistance curve of the old outdoor sensor (from the old tstat) to match the curve of the outdoor sensor that's meant to go with the new thermostat on the cheap (i.e. a couple bucks worth of parts from Radio shack and some solder), I could avoid shelling out another $30+ to buy the "compatible" sensor and having the old one just laying around being useless. I do have tables of resistance vs. temp for both the current sensor and the "compatible" sensor, which is what I used to generate the graph in my original post, but I don't have the formulas. If the data tables would be useful I can post them, but it sounds like it might not be possible anyway. Thanks for your input! 5. Sep 2, 2009 ### skeptic2 A perfect match may not be possible but I think we can come close. Sure post the tables. 6. Sep 2, 2009 ### mjl5007 Actual: Code (Text): T (F) Rx (Ohms) -15 138900 -10 117700 -5 99900 0 85100 5 72700 10 62300 15 53500 20 46100 25 39800 30 34500 35 30000 40 26100 45 22700 50 19900 55 17400 60 15300 65 13500 70 11900 75 10500 80 9300 85 8300 90 7300 95 6500 100 5800 105 5200 110 4700 115 4200 120 3800 Desired data: Code (Text): T (F) Rx (Ohms) -20 106926 -18 100923 -16 95310 -14 90058 -12 85124 -10 80486 -8 76137 -6 72060 -4 68237 -2 64631 0 61246 2 58066 4 55077 6 53358 8 49598 10 47092 12 44732 14 42506 16 40394 18 38400 20 36519 22 34743 24 33063 26 31475 28 29975 30 28558 32 27219 34 25949 36 24749 38 23613 40 22537 42 21516 44 20546 46 19626 48 18754 50 17926 52 17136 54 16387 56 15675 58 14999 60 14356 62 13743 64 13161 66 12607 68 12081 70 11578 72 11100 74 10644 76 10210 78 9765 80 9398 82 9020 84 8659 86 8315 88 7986 90 7672 92 7372 94 7086 96 6813 98 6551 100 6301 102 6062 104 5834 106 5614 108 5404 110 5206 112 5010 114 4826 116 4649 118 4476 120 4317 7. Sep 2, 2009 ### skeptic2 Offhand, just judging from the plot, it looks like putting a 240K resistor in parallel with the thermistor would get you pretty close to the desired curve. Edit: I posted this before I saw your tables. I'll still look at your tables. 8. Sep 2, 2009 ### Bob S mjl1507- Based on your first post, you would like to reduce 140k ohms to 100k ohms. This would require a 350k resistor in parallel.  Based on the zero degree numbers in the table, 220k ohms looks better. Bob S [Edit #2] After looking at your tables, I suggest 180k ohms in parallel and 600 ohms in series, or 600 ohms in series and 180k ohms in parallel. Last edited: Sep 2, 2009 9. Sep 2, 2009 ### Topher925 Make sure to use a high tolerance, low TCR resistor. Otherwise the resistance of the resistor you put in series with your thermistor will change as well. 10. Sep 2, 2009 ### dlgoff I'm curious. Does the HVAC look at the full range of temps shown in the curve? Or does it just do some switching at a particular temp? Most applications of thermistors used for temperature measurements I've seen only use a small part of the curve and assume linearity. 11. Sep 2, 2009 ### vk6kro It looks like the new thermistor agrees with the desired one within 10 % over the range of 55 deg F to 105 deg F. They are within 5 % between 65 deg F and 90 deg F. They actually cross over at 80 deg F, so any parallel resistors are going to make it worse at that temperature. You might get that much variation even in the expensive new ones. This seems to be the range you would be interested in. Why not just try it and see how it goes? 12. Sep 3, 2009 ### mjl5007 Actually, the range I'm most interested in is on the "lower end", between say 20 and 50 deg F; I say lower end even though this is actually closer to the middle of the data range I've posted because the temps in my area don't frequently go below 10 deg F for very long, so accuracy below that isn't terribly important. This is because the main benefit of having an outdoor temp sensor attached to the new thermostat is that it can be programmed to lock out the heat pump below certain temperatures where it becomes more energy-efficient to run only the auxiliary electric heating elements instead (e.g. if outdoor temp is 35 or below, do not run heat pump, use auxiliary electric only). The same goes for locking out the auxiliary heating elements if the outdoor temperature is above a certain threshold (e.g. 50 deg F). As for the amount of error, the installer's manual for the old thermistor indicates that a variance of more than 5% in the measured resistance from the expected/calibrated value at any given temperature indicates that the sensor must be replaced, so I'm trying to keep the error within 5% as best I can. Again, though I understand that it probably won't be possible everywhere without a more sophisticated setup, so I'm mainly focusing on accuracy where it matters, and if it's not so accurate when the temp is below 10 deg F or above 80 deg F, i'm not terribly concerned. 13. Sep 3, 2009 ### mjl5007 Looks like a 200k resistor in parallel will get me within 5% of the desired values from about -5 deg F all the way up to 80 deg F, which should be perfectly satisfactory. Now the question becomes: how do I select the correct resistor to purchase? I'm not familiar with things like TCR, tolerance, metal film vs. carbon film, 1/4W vs. 1/2 W, etc, so some input on that would be much appreciated. Should I order it online, or will my local Radio Shack carry something that will work? Any suggestions on the best way to mount and wire up this circuit? Should I get a small board or just hook it all up inline with some solder? Many thanks to all who have lent their knowledge and insight! 14. Sep 3, 2009 ### skeptic2 The best overall simple circuit I could find was a 620 ohm resistor in series with the thermistor and a 200k resistor in parallel with both of those. That will give an average error of 2.5% and an error over the range of 70F to 80F of about 1.8%. If your thermostat will be used mainly between 70F and 80F, the values can be optimized for that range (or any other). The best I could find was to use a 680 ohm in series with the thermistor and a 160k in parallel with the combination. This will give an overall error of 3.9% and an error for 70F and 80F of 0.39% 15. Sep 5, 2009 ### vk6kro Now the question becomes: how do I select the correct resistor to purchase? I'm not familiar with things like TCR, tolerance, metal film vs. carbon film, 1/4W vs. 1/2 W, etc, so some input on that would be much appreciated. Should I order it online, or will my local Radio Shack carry something that will work? Any suggestions on the best way to mount and wire up this circuit? Should I get a small board or just hook it all up inline with some solder? You could get a pair of 100K resistors and put them in series to get 200K or just about any 220 K resistor would be OK. If your thermistor was outside and the HVAC system was indoors, you could put the resistor in the plug where it joined the HVAC system. This would be indoors and eliminate any serious temperature effects outside. Resistors are pretty stable with temperature anyway. 200K gives a good result for low temperatures. It is better than 220K and gives an agreement within 5% for the temperature range -5 deg F to 77 deg F and for most of that range it is a lot better than 5%. In the chart below, the red trace is the desired one and the yellow one is the result of adding 200 K in parallel with the existing thermistor. As you can see they are practically on top of each other. #### Attached Files: • ###### thermistor.JPG File size: 25.9 KB Views: 188 16. Sep 30, 2009 ### kinglerch I have a thermistor with the exact same resistance curve as the above. Do you know the part number of your thermistor? I don't know what mine is, only the resistance curve. I am looking for a digital readout (temperature display) that would be compatible with this resistance curve. Do you know where I could find one? 17. Sep 30, 2009 ### vk6kro You probably need a temperature/ humidity readout with "indoor/outdoor" type display. These come with a nice display and either a cable to take a sensor outside or a radio link to a different box outside. The cable type is better if you have very cold conditions because the radio ones have a battery in them that may not work if it is too cold. These come already calibrated with their own sensors, so you don't need your thermistor. Places like Walgreens and WalMart carry them in the US. They cost about$10 - \$30 depending on features.

18. Oct 1, 2009

### kinglerch

I have to use this one thermistor because it's in an appliance. I can't change the thermistor but I can wire a display to it, if I can find one that is compatible with it's resistance curve.

19. Oct 1, 2009

### vk6kro

Inside the displays above, there is a computer chip and I would guess that it has a long list of conversion values in it that it can look up to convert thermistor readings to temperatures.

Or, it might have a formula in it for converting them.

Whichever way it works, no doubt many hours of work and a lot of skill went into making it a commercial product.

If you got one of these displays from somewhere with a good product return policy, you could measure the resistance of the plug-in probe. If it is different to yours, take it back to the store.
If it is identical to yours, you could plug yours into the socket. Once you do that you shouldn't ethically take it back to the store.
Maybe it will work. Maybe it won't. If it doesn't, you have a nice thermometer for the house.