• State TI Thinks Resolved
  • Locked Locked
  • Replies 9 replies
  • Answers 2 answers
  • Subscribers 46 subscribers
  • Views 2582 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

ADS1247: Measuring 6 PT100 using a single ADS1247

Noam Weidenfeld1
Intellectual 780 points
Part Number: ADS1247

Hi,

Attached is a schematic for measuring 6 PT100 using a single ADS1247 and an external MUX.

1. Will it do the job?

2. Any consideration on the reference resistor, the MUX represent a 125 OHM resistor which is not part of the reference path.

Thanks

Noam

PT100 In.pdf

  • 0
    TI__Guru* 97655 points


    Noam,


    This circuit might work, but I think you would be unhappy with the result. There are a few things that I think would seriously disrupt the performance. I'll explain what I think the problems will be and you can consider the impact to the measurement.

    1. By inserting the multiplexer between the RTD and the IDAC current sources, you've included the resistances as part of the measurement. If the mux resistance is exactly equal, then the resistances cancel out. If they are not equal, then the difference between the mux resistances becomes an error. Remember that the typical resistance of 125Ω. If the mismatch error is 10Ω - 20Ω then your accuracy may be very poor.

    To get around the mux mismatch resistance, you would need dedicated IDAC pins and a second multiplexer to route the current directly to the RTD.

    2. I assume that R296 is some sort of hardware compensation designed to center the measurement near 0V at a particular temperature. While we've written this into old datasheets and applications notes, we don't recommend doing this compensation anymore. Normally this may give you more resolution by including negative measurements. However, with the accuracy and drift of this resistor, it often adds more error than it is worth. Generally, you'll have better performance without adding hardware compensation.

    3. The addition of TVS diodes (DESD5V0S1BA-7) may take away current from the RTD and reference. This will also give you error in the measurement. I'd note that the typical leakage is 5nA, while the maximum is 100nA. This may appear as a gain error if the current is removed before the reference resistor.

    If you have any other questions let me know.


    Joseph Wu

  • 0 in reply to Joseph Wu
    Intellectual 780 points

    Joseph Hi,

    The ON resistance difference between channels for this MUX is 2.2 Ohm  typical and up to 6 Ohm MAX.

    I looked at AD multiplexer product and they have the  ADG1606_1607 which have a lower typical on resistance and lower difference between channels <~1 Ohm.

    Would you say that using the AD multiplexer will give better results?

    Does TI have a similar multiplexer with lower difference between channels?

    Or maybe 2.2 Ohm is good enough?

    Noam

  • 0 in reply to Noam Weidenfeld1
    TI__Guru* 97655 points
    Noam,


    I'm sure that the AD multiplexer is a fine part. However, the point is to remove the error contribution from the multiplexer, not reduce the error.

    Going with your current topology, if you have 2.2Ω of multiplexer resistance mismatch, that is the equivalent of about 5.7°C of error in your temperature measurement. Even if you were to use the AD mux, where the typical mismatch is 0.2Ω at 25°C. you are still over 0.5°C of error. That would just be the typical error. If you added the maximum resistance mismatch, and added the temperature drift, then you would be much larger error.

    By using a second multiplexer, you could remove the resistance mismatch error reacting with the IDAC current. There may be some error because of the mismatch reacting with the input bias current, but it would be significantly smaller.


    Joseph Wu
  • 0 in reply to Joseph Wu
    Intellectual 780 points

    Joseph Hi,

    I understand what you are saying however I want to keep the solution with minimal number of component as possible considering/compromising with the measurement accuracy.

    There is a anothert chip of AD AD707:

    - 2.5 to 5 Ohm resistance over temperature rang of -40 to 85. (The application is 0 -70).

    - 0.3 Ohm resistance match.

    - 0.5 Ohm resistance flatness.

    Taking this number what will be the error rate and how it will affect the bit resolution?

    If the results will not be good enough I will use two MUX chips.

    And one general question, if the resolution required is for 16 bits, is there an advantage of using a 24 bit resolution chip instead of 16?

    Noam

  • 0 in reply to Joseph Wu
    Intellectual 780 points

    Joseph Hi,

    I am adding the updated schematics with the Current MUX. Please let me know if this is what you had in mind.

    Regarding the TVS, it was suggested by one of TI's reference design (Also attached)

    BR

    NoamPT100 In 2 MUX.pdf

    tidu491 tidrc86.pdf

  • 0 in reply to Noam Weidenfeld1
    TI__Guru* 97655 points
    Noam,


    I'm sorry I'd missed this follow up post.

    First, your circuit with the two muxes is what I was describing. A first mux routes the ADC input signals, while a second one routes the IDAC currents. In this way, you bypass any mux resistance from IDAC current in the signal path that would add error. There is still input bias current, but that is many orders of magnitude down in the error.

    Second, I can see that TIDU491 does have these TVS diodes for circuit protection, but it still helps to know that these diodes will have a non-zero leakage current and that will cause some small amount of gain error in the measurement. For the DESD5V0S1BA, typ is 5nA, while the max is 100nA. If you're using small IDAC currents to drive the sensor element and reference resistor, the max might be an issue. I haven't completely read through the TI design, but don't think the diode leakage error is part of any calculations in the write up.

    Going back to the previous post, the resistance mismatch will directly impact the measurement of the RTD. A 0.3Ω mismatch of resistance is an error of 0.3Ω in the measurment of the RTD. The flatness and the absolute value of the resistance isn't really a factor. The mismatch is what matters. I would also note that the mismatch likely will change over temperature. This is why I think the two multiplexer solution is the correct one.

    As for the resolution, I would go with the 24-bit ADC. If you want 16-bit resolution, you have two things going against you with a 16-bit ADC. First, A 16-bit ADC is 16 bits of resolution for both a positive and negative full scale input. With an RTD measurement, you are typically only able to get positive inputs, so you're already down to 15-bit performance at best. Second, gain comes at finite steps, so you may not fill the complete full scale range of the input based on what you PGA is set to. What type of accuracy are you trying to get out of your system?


    Joseph Wu
  • 0 in reply to Joseph Wu
    Intellectual 780 points

    Joseph,

    The temperature range is 0 -150 C with resolution of 0.1 C.

    Going back to the resistance mismatch, isn't this error constant and can be compensate by calibration?

    I also so an article showing how to compensate for the current mismatch. You take two measurements and you switch between IDAC1 and IDAC2. then you sum the measurements and get read of current in the equation.

    I think that doing that and calibrating the measurement at start will give a fairly accurate measurement. Especially when the changes with temperature is very low and my environmental temperature is 0 -45 C.

    Any way I think I will layout two multiplexer in case the results will not be good enough.

    Noam

  • 0 in reply to Noam Weidenfeld1
    TI__Guru* 97655 points
    Noam,


    I would note that ±0.1°C accuracy is a rather demanding specification. A class A PT100 sensor has an accuracy of ±0.15°C, so, you'd already need something better than that. a 1/10 DIN will have a ±0.03°C accuracy over a limited range.

    Also, note that an error of 0.1°C goes back to an error of 0.0385Ω in the RTD. That means that you'll need to be able to calibrate out all lead and series resistances carefully.

    Returning to the resistance mismatch, the mismatch is listed as 0.3Ω (I'm not sure if this is max or typ - the AD707 is an older op-amp, not a mux so I'm not sure what part you refer to). Even if this is listed as a max, I don't think you could calibrate this out. Because the switch resistance likely comes from a CMOS transmission gate, the resistance will vary with temperature, supply voltage and DC voltage of the input.

    I'm aware of the technique of chopping the IDAC current sources by chopping them. You might be able to swap the current sources and averaging them. However, I don't know if this technique helps enough here. Normally you either assume that the lead resistances are equal and you cancel the IDAC mismatch with chopping, or you assume that the IDACs are equal and you cancel the lead resistance with chopping. In this case the lead resistances are mismatched and the IDAC has a typical mismatch of 0.15%. This might be ok for you, but that mismatch is the equivalent of a typical gain error of 0.075%.

    Again, I would encourage you to consider the leakage of the current from the TVS diodes (also consider the leakage current from the mux) as part of your error. Any leakage current that is taken away from the reference resistor is adding gain error.


    Joseph Wu
  • 0 in reply to Joseph Wu
    Intellectual 780 points

    Joseph Hi,

    Thank you very much for your support.

    I will adopt your recommendation and use two multiplexer as an assembly option.

    BR

    Noam