Possible to change a thermistor resistance curve?

[Rynobomb]

TL;DR

Samsung fridge, cooling ok but maintains 3c over set point. No temp offset or calibration available. Thought was to adjust thermistor resistance to match actual temp.

Currently when set to 1C, the fridge cools to and maintains 4-5C(As a conversion when set to 34f it cools and maintains approx. 40f. ), all functions like fan/defrost are working correctly. Potentially the main board has an issue but all other zones are working correctly so I would prefer to not mess with that. Temp is verified with multiple thermometers. It looks like the main board puts out voltage and measures the voltage received back from what I have read. This is the table I found it is in f. Thought was maybe a combination of parallel and series resistors may work - The range is pretty small so it only needs to be accurate there.

https://postimg.cc/2Ld5Fdmn

Thanks for your time!

 

[jrmichler]

You could make a correction chart. For example, a label like this "To get 2 deg C, set to -1 deg C". On the other hand, if you are doing this for the fun of doing it, then read on.

This is what I recall from when I was tinkering with thermistors some years ago. A thermistor is connected in series with a resistor. The value of that resistor is selected so as to make the response as linear as possible over a temperature range. Typical NTC thermistors have lower resistance at higher temperatures.

Your next step is to find the temperature signal in your refrigerator. Is it taken across the resistor or across the thermistor? Is the thermistor NTC or PTC? What is the voltage across the resistor/thermistor pair? What is the voltage to the controller when the controller is set to the desired temperature, and when the controller is offset to get the desired temperature? Can you get the resistance of the resistor and the model number of the thermistor?

 

[hutchphd]

A thermistor is connected in series with a resistor. The value of that resistor is selected so as to make the response as linear as possible over a temperature range.

Sorry I don't see how a series resistor changes the linearity. But it is late by my new (old) brain standards.
Parallel maybe?

 

[jrmichler]

Found the spreadsheet. The calculations were for a Tewa TT02-10KC8-T105-1500 thermistor. That thermistor is rated 10K ohms at 25 deg C, and has a beta(25/85) constant 3435K. The thermistor temperature / resistance curve is calculated from those numbers. While the thermistor rated temperature range is -40 deg C to +105 deg C, I calculated only 0 to 30 deg C because the intended application was lake temperature monitoring. The temperature / resistance curve for that thermistor is:

The next step is to add a series resistor, calculate the resistor voltage vs temperature, then iterate the resistance to get maximum linearity as determined by minimum sum squared error. That turned out to be a 12K ohm resistor. The resulting plot of resistor voltage vs temperature is shown below for a 3.3 volt supply.

The maximum deviation from linearity is 0.4% and the sensitivity is 33 millivolts per deg C. Increased voltage increases the sensitivity and also increases the thermistor self heating. This thermistor was intended for use with an Arduino data acquisition system (EnviroDIY Mayfly) that had 3.3 volts readily available.

 

[Baluncore]

This is the table I found it is in f.

It seems the table is now in 1.8 °F steps, which was simply converted from 1 °C steps.
For 1 °C, it should have R = 12k7
But it is 3 °C high at 4 °C, so read down the table three rows to get R = 11k3
We need to convince the controller it is 3 °C too hot.
The thermistor is NTC.
To fake a higher temperature, we must lower the thermistor resistance.
To lower the resistance of the thermistor, we must use a parallel resistor.
1 / x + 1 / 12k7 = 1 / 11k3
x = 1 / ( 1 / 11k3 -1 / 12k7 ) ≈ 100k in parallel with the thermistor.

But first clean the electrical connectors, and check there are no cracked solder joints in the thermistor circuit, that might be raising the circuit resistance.

 

[DaveE]

Sorry I don't see how a series resistor changes the linearity. But it is late by my new (old) brain standards.
Parallel maybe?

I think he means the other part of a resistive divider from a reference. But I haven't actually paid enough attention to know. That would modify the linearity though. A schematic would be nice; it would only need, like, 3 or 4 things in it.

 

[hutchphd]

Thanks.

 

[JT Smith]

I don't think it's 100% linear but it's often good enough to do the job, at least over some range. NTC thermistors can be modeled with the Steinhart-Hart equation. If you put a fixed resistor in series then the voltage drop across the thermistor will be roughly linear with temperature. Usually a bit low at one end of the range and high on the other. Doing the math to show this would be an interesting exercise.

 

[DaveE]

NTC thermistors can be modeled with the Steinhart-Hart equation.

Everyone goes to this empirical equation, which is a good one. I found in a couple of normal designs that didn't have a huge temperature range (like this one) that a simpler polynomial expansion can fit well and be much easier to calculate without the logarithms. You can easily invent your own model. I suppose now it doesn't matter since uC have become so powerful.

 

[JT Smith]

I found in a couple of normal designs that didn't have a huge temperature range (like this one) that a simpler polynomial expansion can fit well and be much easier to calculate without the logarithms. You can easily invent your own model. I suppose now it doesn't matter since uC have become so powerful.

You'd think so. I did a project last year that used a pedestrian 8MHZ 8-bit processor and I simply made a lookup table of the values straight from the manufacturer's data sheet, converted to truncated ADC values.

But I also have an inexpensive probe thermometer that I'm pretty sure isn't doing that sort of thing. There's a kind of S shape to the temperature curve. The probe has a processor to run the LCD and read the buttons so I wonder why didn't they fix up the data a bit. It couldn't have been hard. Well, it did cost $10-15.

 

[Rynobomb]

It seems the table is now in 1.8 °F steps, which was simply converted from 1 °C steps.
For 1 °C, it should have R = 12k7
But it is 3 °C high at 4 °C, so read down the table three rows to get R = 11k3
We need to convince the controller it is 3 °C too hot.
The thermistor is NTC.
To fake a higher temperature, we must lower the thermistor resistance.
To lower the resistance of the thermistor, we must use a parallel resistor.
1 / x + 1 / 12k7 = 1 / 11k3
x = 1 / ( 1 / 11k3 -1 / 12k7 ) ≈ 100k in parallel with the thermistor.

But first clean the electrical connectors, and check there are no cracked solder joints in the thermistor circuit, that might be raising the circuit resistance.

This does seem to fit the bill at least reasonably from the range I need it to work. Is there any potential downside or reason that it would cause an issue? I believe the mainboard uses the return voltage to determine the temperature but that in my mind should still be accurate but its been many many years since I have built or played with circuits.

 

[hutchphd]

TL;DR Summary: Samsung fridge, cooling ok but maintains 3c over set point. No temp offset or calibration available. Thought was to adjust thermistor resistance to match actual temp.

Temp is verified with multiple thermometers.

My 30 yr old Amana, (after a few surgeries) still works fine. During my repairs I have noticed the internal temperature is not uniform and varies ~5C from door to center space. Does your Samsung not do the same? So why do you need an exact number and what does it actually measure?
The Amana allows me to choose "temperature" from 1 to 9 on the big dial. I set it on "5" 30 yrs ago and have not touched it since. The coldest measured temp is 30F in the meat drawer and 39 F in the butter tray on the door.

 

[Baluncore]

Is there any potential downside or reason that it would cause an issue?

If you cannot adjust the reference resistor, nor the reference voltage, then you can do no more than offset the thermistor resistance in the narrow operating range.

Is the fridge in a temperature regulated environment? If not, I would monitor the temperature error and look for a correlation with external temperature.

While the environment here wanders seasonally between 5 °C and 30 °C, I see a wandering 4 °C error in my old fridge, that is related to the thermostat being on a temperature gradient, between the inside and the outside of the enclosure. It is complicated by being a fridge-freezer with one sensor, that is also influenced by ice buildup.

 

[Rynobomb]

If you cannot adjust the reference resistor, nor the reference voltage, then you can do no more than offset the thermistor resistance in the narrow operating range.

Is the fridge in a temperature regulated environment? If not, I would monitor the temperature error and look for a correlation with external temperature.

While the environment here wanders seasonally between 5 °C and 30 °C, I see a wandering 4 °C error in my old fridge, that is related to the thermostat being on a temperature gradient, between the inside and the outside of the enclosure. It is complicated by being a fridge-freezer with one sensor, that is also influenced by ice buildup.

 

[Rynobomb]

The fridge is in a temperature controlled environment. I have moved the thermometer around and the 4 degree is the best case which hovers right on acceptable. Near the doors are a couple degrees higher. There are thermistors in each zone, at least 3 I know about. The freezer and mid drawer hold temp perfectly. Only the main fridge section is slightly warm. This is a pretty common issue with these fridges, unfortunately one of many. I was just hoping by tricking it to think it is slightly warmer it would improve the overall performance and extend the life until something else major fails.

 

[hutchphd]

I can make you a deal on a slightly used Amana!!
I am still confused as to the workings of your Samsung and your expectations thereof. My Amana (no microprocessor) is clearly mechanically designed to maintain different temperatures in different parts of the fridge (and the freezer!). Why do you expect the Samsung to produce only one Temperature? How does the Samsung manipulate multiple thermister inputs? If you eat meat put a thermometer in the meat drawer and turn the knob enough to keep it plenty cold. Put the fresh veggies in a non-freezing spot. The rest will be fine. How difficult is that?
I would tell you to go buy an Amana but alas Whirlpool bought them. I think they still make a pretty good machine that is very similar.

 

[Rynobomb]

I am not sure what you mean, it is all digital and controlled by a main board. I would love to turn the knob further but the maximum it can be set to is 1c and at that the coldest the main refrigerator gets is 4-5 with the door area being around 6. I wanted to trick the main board into thinking it was warmer to call for more cooling, effectively making it match the set temperature.

 

[Tom.G]

Your main concern seems to be that the thermistor and your thermometer disagree. Is the thermometer next to or in contact with the thermistor?

If not, then you are measuring the normal temperature variation in the box.

If you insist on uniform temperature you could keep a circulating fan running full time; perhaps the fan that comes on when the cooling cycle is running could be rewired to run continuously.

If your concern is that the temperature-set knob does not match the measured temperature at the thermistor, perhaps rotating the knob on its shaft would bring them into agreement -- or add a fixed resistor in series with one end of the potentiometer that the knob turns

Cheers,
Tom

 

[JT Smith]

Our 20+ year old refrigerator is old school. Analog electronics, dials with dimensionless numbers, no digital display of the supposed temperature. It works great. Never have to calibrate it, just adjust it sometimes. The temperature switch which controls the defrost cycle failed last year and the repair person told me to keep the refrigerator going as long as possible as the newer units aren't made as well.

We had our 30 year old washing machine replaced last year because parts for it weren't being made anymore. The guys installing the new one said the same thing: the new washing machines don't last as long. They told me our 30 year old dryer would probably last forever. And when they saw my 25 year old car they said the same thing about that: keep repairing it as long as you can; the newer cars aren't made as well. I've had several people tell me that about my car.

So go ahead and hack your refrigerator with a parallel resistor if it makes you feel better.

 

[jbw151]

I have the same problem and had the same idea. Just ordered 75kohm and 100kohm resistors to give it a shot.

Did you have any luck resolving?