Post Go back to editing

AD7124 RTD measurement sequencing


I read circuit note CN-0383 and I plan on using AD7124-8 to interface 4 2-wires RTDs for ratio-metric measurement.

One thing I can't figure out is what conversion mode to use on AD7124-8, and what will be achievable sampling rate.

If I understand correctly, I'm going to have to enable IOUT current to the RTD I want to measure before each measure, since IOUT0 and IOUT1 attributions can't be part of an "ADC Setup".

Is it correct to assume that "continuous read mode" is out of the question in this configuration ?
In the case of "continuous conversion mode", shall I reconfigure IOUT0 or IOUT1 for the next RTD measurement before reading the data register ?
What is the best mode to use in order to achieve maximal SPS, and what will be that maximum SPS ?

Thanks for your help !


  • Hi, 

    Any chance you've seen this article? How to Select and Design the Best RTD Temperature Sensing System | Analog Devices

    Please have a read through as you may find this helpful. 

    With regards to your question about the maximum SPS. Since you are required to convert 4 RTDs meaning 4 channel conversion. When multiple channels are enabled, a settling time is required to generate a conversion every time the channel is switched. The maximum ODR that can achieve with full power mode are 4.876ksps at Sinc3 and 3.889ksps at Sinc4 per channel. Please take note also that since CN-0383 is sharing a single IOUT0, the IOUT0 is alternately switching(on/off) between channels, the IOUT can be powered on immediately but it depends on the external RC values connected to the ADC before it completely settled for conversion. It is outside the sequencer so the delay or timing for this needs to be considered before taking measurements. 

    With regards to your question about which mode, it depends on configuration and what would be your target throughput rate. For example you are correct that continuous read mode seems not applicable as you will need to switch the current outputs. In terms of using the single conversion mode, If you are tend to operate in single conversion mode, the part revert to standby mode after conversion. On the standard silicon (LFCSP and TSSOP), the excitation currents are disabled when the ADC is placed in standby mode. On the B-grade, the excitation currents can remain active in standby mode. So you can use a 4 channel B-grade option the excitation current should remain active. But like I said it depends on whether your okay to run a single measurement then wait for a while before taking new set of measurements or you can just continuously converting and repeat it through all enabled channels. 

    So in the CN-0383 a single IOUT 0 is only used for all RTDs but since the AD7124-8 has two excitation currents and given that you have extra pins for the other IOUT1 you can used both IOUTs so that any delayed or settling time introduced by switching on/off the current can be minimized. So for example used IOUT0 for RTD 1 and RTD3 and then used IOUT1 for RTD2 and RTD4, however the drawback in this configuration is additional Rref for RTD2 and RTD4 since you are using different IOUT it can no longer shared with RTD1 and RTD2. So I still recommend a simplified approach which was suggested in CN-0383 and just considered the timing when switching the IOUT, it should fall within the settling of the channel conversion to ensure that the IOUT is fully settled before the next conversion.  



  • Thanks a lot for the insight!

    First, I'm thinking of getting rid of the analog anti-aliasing filter used in CN-0383 and your article (very helpful article !), because I'm using PT1000 and PT10K and because we are using a current source instead of a voltage source to excite the RTDs, the settling time of this kind of filter becomes to long and have low impact anyway. Do you agree with that?

    Second, am I right to think that, in the conversion sequence, I should change the IOUT output just before reading the ADC value (assuming continuous conversion mode)? This way the signal gets time to stabilize (or at least starts to stabilize) before the next conversion is started (because reading ADC value starts the next conversion, right?). I don't know if I'm clear... What I mean is, would I gain a little time if instead of the typical procedure given on page 74 of the AD7124-8 datasheet, I did as follow:

    1.     Reset the ADC.
    2.     Select the power mode.
    3.     Set the CHANNEL_0 register analog input to AIN1/AIN2.
    Assign Setup 0 to this channel. Configure Setup 0 to have a
    gain of 16 and select the reference source REFIN2(±).
    Select the filter type and set the output data rate.
    4.     Program the excitation currents to 500 µA and output the
    currents on the AIN0 and AIN3 pins.
    5.     Wait until RDY goes low.
    6.     Program the excitation currents to 500 µA and output the
    currents on the AIN0 and AIN3 pins.
    7.     Read the conversion value.
    8.     Repeat Step 5.

    Is it worth it ?

    Thanks for your help.

  • Hi, 

    I agree with removing the RC filter on the IOUT pin. As you can see in the CN0383 or even in the article the IOUT does not have any RC filter. It is okay to use two IOUT pins but like I said it will cost you additional Rref and additional REFINs. So the simplest solution is still what is recommended in the CN0383. I think switching IOUT0 without external RC on the IOUT pin will not introduce significant delay. So the only delay is the time you switch the IOUT (i.e. the time when writing to the register) 

    I think the CN0383 also provide the steps needed to measure multiple RTDs.

    1. Direct IOUT0 to AIN0. The voltage is measured on Channel 0 (AIN2, AIN3). Therefore, Channel 0 must be enabled. All other channels are disabled for this measurement.

    2. Disable Channel 0, enable Channel 1, and direct the IOUT0 current to AIN1. The voltage is then measured on Channel 1 (AIN4 and AIN5). 

    3. Repeat this sequence until all RTDs are measured.

    In the steps I think it is suggested to just enable one channel at a time. So that it will avoid performing conversion at channel with disabled IOUT. So the approach will really depends on the required throughput rate per channel. But like I mentioned it depends on how important the delay when switching IOUT in your system. 

    We also have example code that will soon to be release to help you easily configure and measure multiple RTDs using mbed code.  



  • Hi,

    Sorry for the misunderstanding; I was not talking about an RC filter on IOUT, but about those:

    I ran a simulation with LTSpice, and it takes a few milliseconds (I'd say about 7-8 for this PT100, but it will be much more for a PT1000 or PT10K) for the ratio-metric measure to stabilized. Of course because the capacitors are charged with a constant current.

    Since my measurements won't be in a noisy environment, is it OK to remove those filters ?

  • Hi, 

    I am not sure how are you measuring the 7-8ms. With the given RC values would it be just supposed to be in us? For example I think the time you need to measure is the time the IOUT is high up until it gets stable at VIN. And looking at the plot it seems less than 1ms. The RC filter there serve as an anti-alias filter. Using smaller R and C values will mean that there is less rejection at multiples of the modulator frequency. So, the design is more prone to external interference (RF pick up for example). 

    May I know what throughput you required in your application and why? 



  • Thank you very much for this article.
    It is very helpful and confirms my final design.

    Have a nice day !


  • Thanks Julien for that feedback. Slight smile