ADS124S08: RTD, Thermocouples and Thermistors References and Errors

Hi Darren,

See my comments below.

Best regards,

Bob B

Darren (FAE) said:

Part Number: ADS124S08

I am trying to understand the proper reference setups for various topologies with the ADS124S08. Using the "Basic Guide to..." documentation, datasheets, etc, I have kind of summarized the main design points below. Is there anything else you could add, or correct if I am mistaken?

For an RTD setup...
- A high-side reference can eliminate the IDAC mismatch issue, but because IDAC1 has to go through the reference resistor then to the RTD element, it means you need a MUX if you want to measure more than a single RTD (or else you need multiple expensive reference resistors) [Bob] The big advantage with the high-side reference is the reference is excited by the same IDAC that is also exciting the RTD for 3-wire RTD sensors.  There is no advantage when using 2-wire or 4-wire RTDs.  When the high-side reference method is used there is no issue with the drift of the excitation source as the measurement of the RTD is ratiometric.  However if the second IDAC is used for lead cancelation, there could still be some error associated with the lead cancelation within the measurement.  To use the single reference source as a part of the high-side reference you would need to be able to steer the current to the desired RTD which would require a mux for the the current.  Another downside to the high-side reference measurement is the compliance voltage must be met for the IDAC and PGA input range.  This normally would require a bias resistor in addition to the reference resistor.  The low-side reference would only need the reference resistor as the reference resistor would also be the biasing resistor.

- A low-side reference solves the above problem, but IDAC mismatch errors could be a concern (gain error). Swapping the IDAC sources and taking two measurements, and averaging the codes, could help cancel that mismatch, but at the cost of more processing time. [Bob] Yes, you can route more RTDs to the reference resistor without adding an external mux.  And yes, IDAC chopping will eliminate the issue with the IDAC mismatch.

For Thermocouple setup:
- If using a reference resistor for a thermocouple setup, the IDAC accuracy results in a pretty large gain error. This is because the measurement is not ratiometric.
So for a thermocouple, it is not recommended to use a reference resistor? Instead, it is recommended to use the internal, or external, voltage reference? [Bob] The TC can be considered a self-powered device in that you are measuring the voltage created by the junction of the two dissimilar metals.  There is no way to easily make this into a ratiometric measurement so in this case it would be better to use a low-noise, low-drift reference source.  This can be accomplished by either the internal reference or an external reference. 

With TC measurements, it is important to have a stable cold-junction and a precision measurement of the cold-junction temperature.  The measurement of the cold-junction temperature can be done in a variety of ways such as a chip RTD, an IC temperature sensor, a thermistor, a diode, etc..  This cold-junction measurement could be accomplished ratiometrically if the sensor used requires excitation.

- What kind of typical reference voltage accuracy needs do you see for thermocouple applications? Do you recommend the internal reference voltage? (soldering can create slight deformations in the silicon (metal expanding/contracting) which can shift the internal reference voltage, whereas external references are centered in the center of a package and are more immune to this) [Bob] As I said, the TC measurement is only as good as the cold-junction measurement.  Using the internal reference is adequate in most applications.  Using an external reference will increase the part count and board space requirements.  Also, depending on the reference choice, the internal reference may outperform some external references.

For Thermistor setup:
A thermistor setup is similar to a two-wire RTD setup, so do the above high/low side reference concerns apply? I believe they do.  [Bob] You could voltage excite or current excite the thermistor.  So it really depends on the application.

When calculating the error, is there a recommended way to calculate the max/min/typical errors?
If we take the RTD for example, if you calculate the worst-case error from the following:
- Worst case reference resistor ΔΩ over accuracy/drift
- Worst case IDAC mismatch over accuracy/drift
- Worst case lead-wire resistance over accuracy/drift
- Worst case input bias currents
- Worst case ADC internal errors (gain / offset / INL) over accuracy/drift
You end up with a compounded, unrealistic, completely hypothetical worst-case result.
Where could you give-in and use the typical error values in the calculation, and still see most production units passing customer ship-out testing?
I mean, could you use the max/min IDAC, and max/min reference resistor, coupled with the typical ADC internal errors, and be relatively well-off? [Bob] I don't know that I will give a satisfactory answer here.  What depends on a hypothetical result will be the factors you mention and ones you don't.  Let's say I have a designed system where the ADC error results in 0.5 deg error worst case, but then I end up using a IEC Class 3 RTD that could have an error of +/- 1.6 degrees at 100 deg C.  Which typical error(s) could I potentially use for the ADC with respect to the system error?  In other words which error will dominate?  Usually gain error is the most significant, but here you must also be careful because the gain error will have a more significant impact at full-scale.  So the entire system must be taken into account and the design parameters that must be met.  There could also be error relative to calculation.  Let's say you use a lookup table and interpolation.  What kind of potential error will this add to the total error.

Some of the ADC error can be reduced.  Using the self-offset command will help to make the offset error insignificant.  Using global chop and/or IDAC chopping can also significantly reduce error. 

In an uncalibrated system, the reference resistor used will impact the gain error.  In the calculation the ADC output code is a ratio of the RTD value to the reference resistor value.  The reference resistor is not exact and often times the resistor initial value and drift will be the dominant factor if the resistor value used is not a precision resistor.  For example, you may use a 0.1% resistor where the initial resistance is within a specific tolerance range but the drift of 200ppm as compared to 10ppm may be a much more significant issue than the ADC min/max error.

Thanks,
Darren