ADS114S08: Internal temperature conversion formula

Hi Tim,

The LSB or value of 1 count = +/- Vref/gain/2^16 = 2*2.5/1/2^16 = 2.5V/2^15 which is the same as your formula as you would multiply the ADC code result times the value of 1 code.

The internal temperature sensor measures the temperature of the die and not necessarily the ambient or package temperature.  The typical coefficient is 403uV/deg C.  The typical value at 25 deg C is 129mV.  So depending on how accurate you need to measure and what you use the measurement for, you may need to calibrate.

Best regards,

Bob B

When reading the internal temperature, I am getting a value that I perceive to be off.  I know that the internal temperature sensor measures the temperature of the die and not necessarily the ambient or package temperature, but I am getting the following reading when compared to my K type thermocouple device placed on the board near the ADC:

Are these values within expected tolerance or do I have a problem with my device?

Thanks,

Tim Davis

Hi Tim,

A few of questions:

1. How are you heating/cooling your board?  When you move through the temperatures the ADC trends follow.
2. What do you mean by K-type TC measurement?  Is this a temperature probe connected to a meter?
3. Do you have any pictures of your setup and how the measurement is being conducted?
4. Which device package are you using?

Best regards,

Bob B

Hi Bob,

I am heating the board in a toaster over that has been modified to allow me to run wires out to a PC to give me communication access to the board.  Yes, I have a K type probe that is connected to my EX330 Multimeter.  I don't have it setup at the moment, but I can set it up again and send pictures if that would help.  My device package is 32TOFP and part number ADS114S08BIPBSR.  These results are similar for more than one board.

Thanks,

Tim Davis

Hi Tim,

How do you get the toaster oven below room temperature? 

There can be a difference between ambient and even PCB temperature with respect to the die (which has some self-heating).  There can also be some thermal drift within the oven itself.  A better comparison would be to attach the TC to the package in some way (like superglue to the top).  An even better way would be to use a thermal bath (which is the method we use) that is a non-conductive oil bath and very stable.  These baths are bit expensive so we also have ovens and thermostreams that isolate the ADC so that it can be measured.  This method is similar to what you are doing, but the temperature is well stabilized (long soak times) before the measurement is made.

Best regards,

Bob B

So for low temperatures we using a modified freezer that allows for wired communication to the board, and also dry ice.  Our test right now are in office and are preliminary in nature of course.  We test in one environment at a time, meaning one test in the freezer and then later another test in the oven.  It is not a continuous test.  Regarding your test, can you share your test data?  Something that we can use to verify our test or device.

Hi Tim,

Temperature gradients can vary within the testing environment.  It is incredible the amount of variance one can find due to circulating air currents.  Some bench methods that I have used is to put the test fixture in a metal box where the box limits the circulating air current.  I have also used a large ESD bag that has a zip-lock end and placed the test fixture inside it then zipped the seal around the wiring.  In the end you need to eliminate the convection currents as much as possible.

I would then take a large series of data to determine any trends to ensure that once the freezer or oven has stabilized that the temperature readings have stabilized as well.  You really need to make sure that the manual measurement of the TC is stable with respect to the temperature sensor.  If one or the other is drifting, then your results will be skewed.

I also point you to the datasheet graph in Figure 36. Here you see that the end result is quite linear.  As your measurements do not appear to be linear, I suspect that the temperature is not stable, or that there is significant convection currents affecting the results.

Other issues that can affect the results is making sure that the supply voltage remains stable as well as the reference voltage.  Drift with respect to the reference will have a direct relationship with the ADC result.

As far as characterization results, the measurements are taken at 3.3V AVDD with various temperatures take throughout the operating range of -50 to 125 deg C.  As I said, the outcome is quite linear, but the initial 25 deg C measurement could be off by +/- 1 deg (six sigma).  The typical response is given in the datasheet as 129mV.  Keep in mind these a typical specifications and not guaranteed.

Making sure that the temperature is stable is key.

Best regards,

Bob B

Ben,

We are using the internal reference voltage of 2.5V provided by the device.  How does that change your previous feedback.

Thanks,

Tim Davis

Hi Tim,

It doesn't change what I stated before.  The internal reference drift will be quite small across the full operational temperature range.  You can see the drift characteristics in Figure 21 of the datasheet.  The drift will be within the level of noise of the ADC.  This is assuming that you have correctly connected a minimum value of 1uF across REFOUT and REFCOM and REFCOM is also connected to AVSS.

I still think there is a thermal stability issue with respect to your probe relative to the ADC.  As an example, let's say you heat up the oven to 65 deg C.  The temperature probe most likely will see the shift in temperature faster than the PCB can absorb the heat.  Also, the oven will vary some with respect to thermal cycling as the element turns on/off.

Cooling will work in a similar way as the probe will react more quickly to the change in temperature compared to the PCB that still contains some value of heat.  So I still think the key here is to take long time data sets (continuously converting) to determine when the temperature is more or less steady state less any thermal gradients from air currents.

Best regards,

Bob B

Hi Bob,

Under normal conditions, and correct implementation, would it be normal to see an internal temperature reading +8 to 10 degrees above the temperature of the package on the circuit board?  For example, I am using an IRT207 Infrared Thermometer to read the temperature of the surface of the ADC on the board, and it reads 22 degrees C while the internal reading is 30 degrees C.  I have several boards and they are all performing about the same, so I definitely see consistency.  I just want to know that my readings aren't way off or in error.

Thanks,

Tim Davis

Hi Tim,

The internal temperature measurement is a calculation and you are measuring with an IR thermometer that has an accuracy of +/- 2 deg C.  Also, it is not a trivial task to measure such a small area using an IR thermometer with a ratio 8:1.  The IR will report all temperatures (averages) within the field of view and you are trying to capture a very small piece of the PCB where the ADS114S08 is located.  So the proximity and angle of the thermometer is also important.  So can you guarantee to me that the IR measurement is more accurate than the ADS114S08?

I think that the important thing at this point is not the accuracy or the difference but rather the consistency.  It is quite possible that the internal temperature is slightly elevated from the ambient.  It is also quite possible that the due to process variations the voltage value at 25 deg C is not exactly 129mV.  The value in the datasheet can vary as it is listed as a typical value and also we do not test the parameter in our final device testing.

For me to give additional guidance I need to make sure that you understand a couple of things.  First, although the internal temperature is quite linear it does not have an accuracy specification like our temperature sensor devices or even ADC devices like the ADS1118 and ADS1120.  Secondly, if you want improved accuracy you will need to accurately calibrate each device.  This is not a trivial task.

It would be very helpful for me to know how and why you intend to use the internal temperature measurement in your system.  Most applications that I see the internal temperature sensor of an ADC used is for cold-junction compensation for a thermocouple measurement.  In many cases, as there is no guaranteed specifications for the internal temperature sensor, I don't see the ADS114S08 internal temperature sensor used this way.  Instead most systems I have seen either use a different method for measuring cold-junction temperature with the ADS114S08 such as a thermistor at the cold-junction, or a temperature sensor device placed in the cold-junction area.  So if I know more about the intended application I can be of better help.

Best regards,

Bob B