• State Resolved
  • Locked Locked
  • Replies 6 replies
  • Answers 1 answer
  • Subscribers 46 subscribers
  • Views 2265 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.

ADS1248: A question about Gain, Reference Resistor and Accuracy

Kemal ARIKAN
Prodigy 40 points
Part Number: ADS1248
Other Parts Discussed in Thread: ADS124S08

Dear Sirs,

I'm working on a four wired PT100 circuit which main goal is measure between -50, + 50 range with an stable resolution of 0,01C after a proper calibration.

PT100 will be in a circulating liquid bath so no self heating effect I guess.

I read all the forum, datasheet and application notes like


www.ti.com/.../tidu145.pdf

www.ti.com/.../sbaa180b.pdf

www.ti.com/.../sbaa201.pdf


But a little confused about Gain, Excitation Current and Reference Resistor.

Please look my schematic and correct me if I'm wrong about following comments:


Suggested Reference voltage is mid supply in my case its 2,5V(AVDD=5VDC)

If I use 3,3kohm reference resistor(for 2.5V reference IEXC will be 0.75mA) For this project PT100 will be around 120 ohm for 50C.

Max PT100 voltage will be around 90mV. So Max gain at this point is 2,5/90m=27,78 not even 32!


Basically we need higher gain for better noise performance.

With a higher reference resistor like 10kohm. Can it help us? For 2.5V reference voltage IEXC is 0.25mA.

Max PT100 voltage will be around 30mV so Gain can be 64.

With higher reference resistor also effects low pass filter values. Do you think it is applicaple? Because filter resistor values goes high.

Can we extend reference voltage higher than 2.5V instead of mid suplly like 3-3.5V? Does it help? Of course values should meet common mode input range

AVSS + 0.1 V + (VRTD MAX · Gain) / 2 ≤ VCM ≤ AVDD – 0.1 V – (VRTD MAX · Gain) / 2

Most of te examples uses 20SPS data rate(for 50/60Hz rejection) and it gives 250Hz frequency at 14.8Hz(several times more)

If we dont mind about conversion time and for a better performance if we use 5 or 10 SPS rates what will be frequencies at this values-more or less?


Could you evaluate our situation and suggest for a precise measuring?


Also I will take care following point
-Iexc current should be meet with compliance voltage
-Keep the analog devices close to ADCs pins
-Do not cross the digital lines and analog ones
-Leave unused analog pins floating
-Do not float unused digital pins
-Use LDO supplies
-Temp coeff of reference resistor shoul be low as 2ppm/C or better
-Inlet and reference filetrs should be well matched
-AGround and DGround should be equal

Best Regards.

Kemal

  • 0
    TI__Guru* 97775 points
    Kemal,


    I'd like to point out that achieving 0.01°C accuracy is very difficult. Even with the highest tolerance grade RTDs (1/10 DIN), the tolerance error of the RTD alone is limited to: ±(0.03 + 0.0005*|T|)°C.

    First, gain does not make your resolution better. If you increase gain, you decrease the noise, but you also decrease the range of your measurement. As you get higher and higher in gain, other contributors of noise become a factor, and you get less resolution in your measurement. It is not important to maximize the gain.

    What is important is to make sure the range of measurement takes up as much of your full scale measurement as possible. This way, the range of measurement compared to the noise gives you the best possible resolution. Changing the reference value to make it larger may help a small amount, but maximizing the usable range of measurement is the most important thing.

    In your case, based on the values that you give, it looks like the maximum RTD resistance that you expect is 120Ω. That is the maximum value that you care to read. If you choose a 1mA current and a 2kΩ precision reference resistor, this may be a ideal combination.

    First, 1mA * 2kΩ gives a good 2V reference. This puts the input voltage near mid scale so that it is within the range of the PGA. Second, the input voltage is 120mV. With a PGA gain of 16, the range of measurement is 1.92V. This makes up 96% of your positive full scale range for the ADC. This maximizes the input signal compared to the noise. Note that you could use a 1.92kΩ reference resistor, but it may be important to leave a margin in case there is some gain error.

    To get the best noise performance, generally you want to go with the lowest data rate. The bandwidth and filter plots for all data rates are listed on pages 30 to 32 of the ADS1248 datasheet. For the 5SPS data rate, the bandwidth is 2.26Hz. For the 10SPS data rate, the bandwidth is 4.76Hz.

    Hopefully this answers your questions. I would note several things about the design. First, the reference resistor is extremely important. The error in the reference resistor determines the error in the system. As an example a +1% error in the reference resistor becomes a -1% error in the ADC measurement. This resistor must be high accuracy and low drift. Second many of the reference material you cited made precision measurements with the ADC. Note that the ADC is kept at a constant temperature. There will be some amount of offset and gain drift that will contribute error. If you do calibrations, you'll want to calibrate the measurement and then keep the temperature of the ADC constant. Note that the reference resistor will have some drift. In many of the measurements that I made, the drift of the resistors were less than 5ppm/°C.


    Joseph Wu
  • 0 in reply to Joseph Wu
    Prodigy 40 points
    Hello Joseph,

    Thank you for your detailed explanations. I understand better now.

    I need to understand followning two points to make a road map:


    1-About filter corner frequency: I know the values from datasheet so wasn't asking bandwith for 5 and 10 SPS.
    Datasheet and app note says"–3-dB corner frequency at least 10 times larger than the bandwidth of the ADC"
    For 20 SPS its 14.8 and its set to 250HZ roughly. What should be for 5 and 10 SPS according to your experince?

    For example bandwitdh of 5SPS is 2.26 if we multiply it with 15 it will be 40Hz. Is it correct?
    There is built in 50-60 Hz rejection(for 5-10-20SPS) can be an effect around this band?


    2-According to "tidu145" application note they reached a better accuracy than 0.01C for 0-100C range.
    This applicaiton uses HW compansation and three wired. Also my goal is measuring 100 fold range(-50,+50) with four wired RTD.
    Thats why I choose the ADS1248 and already ordered some. Its a nice 24bit ADC with built in IDAC and low noise PGA. So I will go with ADS1248 now.
    Can you suggest another device or different RTD like PT500?

    Best Regards.

    Kemal
  • 0 in reply to Kemal ARIKAN
    TI__Guru* 97775 points
    Kemal,


    Ok, I think I understand. The -3-dB filters you refer to are the differential input filtering at the front end. These are used for cleaning up the signal and anti-aliasing. Generally, I would choose something that is about 20x of the bandwidth of the ADC. At 5SPS, the bandwidth would be 2.26, so you could design the filter to be around 22.6Hz.

    I would also note that you should try to keep the series resistance low. Any input current to the ADC reacts with the series resistance and adds error to the measurement. Also for 5, 10, and 20SPS the digital filtering will always have a large amount of 50-60 Hz line rejection.

    As for TIDU145, I now understand the maximizing of the gain. I have generally avoided HW compensation. The idea of it will work, but it is often difficult to achieve. As an example, the error of the compensating resistor directly affects measurement. An error of 1Ω in the HW compensating resistor is an error of 1Ω in the RTD measurement. If you want to get to ±0.01°C you need to be able to accurately measure the RTD to ±0.004Ω.

    For the reference resistor, an error in that measurement will translate to a gain error which may be compensated. However, for the HW compensating resistor, the error is direct, and you need to maintain that resistance through any future measurement that you make. You can get precision resistors that have drift of less than 5ppm/°C, but they are expensive. You can't have much drift for anything in the system, including the device.

    For TIDU145, the measurement was done with precision resistors that are likely 0.01% accuracy and 5ppm/°C or better. Resistors were measured directly for their resistance and all measurements were done at room temperature. I think that the resistances were first measured with an Agilent 3458A, and that you get 0.001Ω accuracy with the device. This way, you can calibrate your measurements, and remove errors associated with the resistors in the system. The remaining errors are the ADC gain error and offset in the system.

    With this method, it is possible to get great accuracy and precision, but it requires a lot of measurement and calibration.

    The ADS1248 is capable of this precision. I would also consider using the ADS124S08 also. It is considered the next generation of the ADS1248 and has better specifications in many cases. However, I don't think it doesn't make the calibration any easier. I don't think changing from a PT100 to PT500 will make much difference either. I suspect that with the lower current through the RTD, and the higher resistance, you might get more noise.


    Joseph Wu
  • 0 in reply to Joseph Wu
    Prodigy 40 points
    Hello Joseph,

    Thank you for your attention and detailed information. I will go with current design and try a noise free environment.

    Have a nice day.
    Best regards.
    Kemal
  • 0 in reply to Kemal ARIKAN
    TI__Guru* 97775 points
    Kemal,


    One other thing that I would recommend would be to get the EVM for the ADS1248 or the ADS124S08. You would need a different precision resistor for the reference, but you should be able to test the application with some real measurements. However, the ADS1248 is a slightly older software platform and would run better on Windows XP or 7. The measurements in the typical applications in the respective datasheets were done with the EVMs.


    Joseph Wu
  • 0 in reply to Joseph Wu
    Prodigy 40 points
    Hello Joseph,
    I consider about it, thank you.
    Best regards.
    Kemal