LTC2485 - Thermistor Measurement Circuit

Hello adamvs ,

If you are more concerned of the precision, you might want to omit the divider itself and connect IN+ directly to the 5V reference, IN- connected as is. This would have a ratiometric configuration since the reference of the ADC and the thermistor supply are one and the same.

For your common mode voltage setting, you should have a series resistance equal to the base resistance of the thermistor (R @ 25°C) i.e., if the thermistor has a base resistance of 5k, the series precision resistor should also be 5k. This way, the thermistor voltage reading at room temperature will be REF/2. It would be a good start, but you can always adjust the baseline at your liking.

Regards,

Yugel

Thank you much Yugel for your feedback.

I'll take your points into consideration (See below) but I was more worried about the differential input current concerns talked about in the datasheet and less about the measurement circuit itself.

---------------------------

The thermistor swings between 5KOhm and 200 Ohms over the system's operating range. This brings In+-In- between -1.4 to +2.2V; this seems good as it uses most of the available full scale. 

The divider should be a fairly negligible source of precision loss, at least if is use a  product like your MAX5492 (<2 ppm/C). It is a source of inaccuracy (<0.1% ratio accuracy), but this can be easily calibrated with a single measurement point.

If i eliminate the divider and set up like you suggest I would need to choose choose R such that In- does not go above 2.5V at the lowest temperature of concern, (~5K).  This would result in much lower precision at higher temps (dVdT is lower by about 4x).

Although I admit I'm am not sure what radiometric means in this context so I might be misunderstand your suggestion. Sorry if so.

Thank you much Yugel for your feedback.

I'll take your points into consideration (See below) but I was more worried about the differential input current concerns talked about in the datasheet and less about the measurement circuit itself.

---------------------------

The thermistor swings between 5KOhm and 200 Ohms over the system's operating range. This brings In+-In- between -1.4 to +2.2V; this seems good as it uses most of the available full scale. 

The divider should be a fairly negligible source of precision loss, at least if is use a  product like your MAX5492 (<2 ppm/C). It is a source of inaccuracy (<0.1% ratio accuracy), but this can be easily calibrated with a single measurement point.

If i eliminate the divider and set up like you suggest I would need to choose choose R such that In- does not go above 2.5V at the lowest temperature of concern, (~5K).  This would result in much lower precision at higher temps (dVdT is lower by about 4x).

Although I admit I'm am not sure what radiometric means in this context so I might be misunderstand your suggestion. Sorry if so.

Oh sorry, I have missed that part.

The differential input current errors in this device are effectively cancelled out due to the method of input sampling. This improves the DC accuracy of the ADC. Nothing to be really concerned about.

I see, you're correct about the range as it is still within +/-VREF/2. Although can you check if your thermistor will not be self-heating significantly given that your series resistor is only 200 ohms?  At minimum resistance, the thermistor will dissipate around 10.2 mW. This might affect your measurements greatly.

Ratiometric configuration only means that the ADC reference and the excitation source is the same. In this way, any variations on the source would be cancelled out as in Eq. 3 below.

Source: A Simple Thermistor Interface to an ADC | Analog Devices

Thanks and regards,

Yugel

Thanks for your feedback.

Yes I agree there might be some significant non-linearity due to dissipation. This would only be a hit to accuracy, but I still may halve the reference voltage to 2.5 V to reduce the effect.

Hi adamvs ,

I don't suggest halving your reference voltage as this will limit the discernible input to -1.25V to 1.25V typically.

Regards,

Yugel