BQ76942: Accuracy of NTC

Hello Matt,

Thank you for your help.

What does the Adc0 parameter mean? Do I need to calibrate this parameter?

Hi Matt,

I have tried with the excel calculator and got the coef parameters, but I found the temperature measurement is very unaccurate.

Even at the temperature point of 15degrees, the error is 3degrees.  What should I set to narrow the error?

Hi Matt,

I have tried with the excel calculator and got the coef parameters, but I found the temperature measurement is very unaccurate.

Even at the temperature point of 15degrees, the error is 3degrees.  What should I set to narrow the error?

Hi Sean,

The calculator may be trying to optimize too much of the curve.  Is there a public data sheet for the thermistor you are using that you can share a link, and what temperature range are you interested in having good accuracy?

You asked about ADC0 the other day.  The technical reference manual section 13.2.8.10 Calibration:18K Temperature Model:Adc0 shows it as a gain term, you could calculate this for your thermistor after applying the coefficients.  Note there is also the offset calibration for each thermistor available in the calibration section of the data memory.

Hi Matt,

Thank you for you explanation.

What's the formula that calculate the temperature?

I have the data sheet for the thermistor, and the data is in the picture above. And the interested temperature is between -25 degrees to 65 degrees.

I have also put in the excel table. But with the coef parameters. the error of temperature measured is huge.

How to solve this issue?

Hi Sean,

Let us take a look at it, it could take a few days.

Thank you Matt. I have told this issue to Alex, and shared NTC datasheet to him.

Hi Sean,

I ran your thermistor through another internal tool and came up with a new set of coefficients, you can try these and see if they give better results:

A = 2978   -8937   -8937   26809    2978

B = -1160   3035  -3937   4383

For this, leave the adc0 at its default setting.  You can evaluate the results across temperature, then you can adjust an offset using the Data Memory parameter in Calibration:Temperature:TS1 Temp Offset for a thermistor on the TS1 pin, for example.

Hello Terry,

Very glad to get your answer. But with my test,  the temperature measurement is still not accurate.

I set the coef parameter as which you give me. And I set the input resistor 10K to calibrate at 25 degrees. Then I change the resistor value, and get the temperature report with reading the register.

When I set 12.1K in order to get 20 degrees, in fact I get 22.2 degrees.

When I set 27.6K in order to get 0 degrees, in fact I get 11.8 degrees.

When I set 68.97K in order to get -20 degrees, in fact I get 5.6 degrees.

When I set 4.9K in order to get 45 degrees, in fact I get 40.8 degrees.

When I set 3.0K in order to get 60 degrees, in fact I get 54.8 degrees.

Is there something wrong with my setting?

Hi Sean,

I think something else is awry here.   Can you export a gg file with all your settings and send that?  Also please capture the raw counts for the pin you are using (such as TS1 Counts if using TS1) for each value of thermistor resistance.

Thanks,

Terry

Hi Sean,

Also what is your thermistor circuit?  Excess capacitance on the pins can slow rise time and may result in a lower voltage reading, but that should be appear across temperature and your result does not indicate this. 

Hi Sean,

Also try removing the 0.1uF caps, since that generates a fairly large time constant (especially at cold, when the thermistor resistance increases).

Thanks,

Terry

Hi Terry,

I remove the 0.1uF, and use your parameter, it is very very accurate.

Could you pls help me to make another group of parameters? I use another 200K ntc, since the error distribution of 10k and 200k is different.

What's your mail adderess？I have sent the datasheet of the 200K NTC to Alex.

Thank you very much.

Hi Sean,

Ok, great to hear that cleaned it up.  When using the 18k pullup resistor, we recommend keeping the capacitance <4nF.  When using the 180k pullup, it needs to be <400pF.

I already received the thermistor datasheet in email from Yue.  I will work on coefficients for that one tomorrow.  I assume you will use the 180k pullup with that one, correct?

Thanks,

Terry

Hi Terry,

Many thanks for your kindly help.

Yes, the 200K NTC will use 180K pull up. But I have another question to ask you, do you think it is necessary to use two different NTC to measure temperature? Or the 10K one is enough?

Or I can use two 10K ones, if one fails, I can change to another one. Which choice makes more sense?

With my original thinking, I can get more accurate temperature value in low temperature range by using 10k one. While in high temperature range, the 200K one is more trustable. Am I right?

Hi Sean,

A NTC monitors temperature at one point, some designers want to have multiple sensors to check temperature of cells over a larger area (more cells).  The BQ76952 family can support multiple thermistors.   You could connect 2 or more 10k NTCs to the same TSn input of the BQ76952 and have the MCU control multiplexing, but it would not be able to synchronize to the BQ76952 measurement so there may be odd measurements taken during switching which would need to be filtered out.  It may be preferred to have the MCU measure additional temperatures if desired.

NTC curves generally have the same general shape, steeper at cold than at hot.  Accuracy should be similar if the appropriate parameter coefficients are prepared and loaded.  The higher impedance thermistor will draw less current, might be more easily influenced by interference.   The part can support the 2 pull up values for a variety of thermistors as desired for the design.  The part does not have a setting to use one thermistor for cold and another for hot, but the host can read the data and use it as desired.

It seems Terry is providing support through another path, I will close the post for now.