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.

ADS1262: TI REF3012 vs TI INA125: Who is the best for my ADS1262

Part Number: ADS1262
Other Parts Discussed in Thread: REF3012, INA125, ADS1235, OPA388, ADS1261

Hello people!

With ADS1262 PGA set on Gain 32, the full scale is 78mV, so my load cell -with its 20mV max output- isn't able to exploit ADS capabilities.
I have two solutions:
I could set an external Vref by a REF3012 so 1.25V/32=37mv (fine!).
I can use a load cell amplifier as TI INA125, using internal Vref as INA125 ref.
Each solution is quite affordable (I need only 3-4 components for a year, so a money difference is not significant) and seems to goal my needings but... what do you suggest? Does using an external amplifier make harder system tare?
Thanks very much and have a nice day!
Mario

  • Hi Mario,

    Noise and drift are going to be your biggest issues. Making the measurement ratiometric will give you the best performance. I'm assuming that you have chosen the ADS1262 because it has 32-bit resolution. Because of noise, you will not likely come even close to 24-bit noise-free bits, especially if you attempt to add an external amplifier or use a low performing reference.

    I would suggest taking a look at the ADS1235 which is similar to the ADS1262 in a number of ways, but is meant to be used specifically in bridge applications. Take at the datasheet on page 63 for a typical application and the performance you can expect.

    If the ADS1235 does not meet your requirements, get back to use with the specific requirements you need to meet your goals.

    Best regards,
    Bob B
  • Hello Bob,

    I have chosen the ADS1262/63 because we need a 32bit class resolution with 100-150SPS (to have at least a 20bit ENOB).

    By using the INA125 I can use a 10V for power and reference (pin 16), but I'll insert a new amp that can make harder the tare.
    A REF3012 is simpler but probably less effective...

    Now, I set my ADS @1200SPS, a 32 gain and I made an averaging to have 100 points per second ( I sampled 10 points, then I removed max/min and I made an average) and it's not too bad but a 100 Gain would be better!

     

  • Hi Mario,

    If you use the INA in front of the ADC with a gain of 100 V/V, you will be adding the INA's noise to your signal and gaining up this noise by 100. Notice that the input signal and the INA noise will both be gained up by the same amount; therefore, the signal-to-noise (or resolution) is already limited just by the choice of amplifier and not by the ADC.

    NOTE: The benefit of using gain ONLY comes from the case where you have multiple active stages in your circuit and first stage has lower noise (and higher gain) than the second active stage in your system. In this case you can gain up the signal and the noise of the first, while keeping the noise of the second stage constant. The net effect is that the signal is amplified more so than the overall noise. However, as soon as the gain increases to the point of causing the amplified noise of the first stage to exceed the noise of the (non-amplified) second stage then the signal-to-noise ratio no longer shows any significant improvements. Here is a numerical example, notice that above a gain of 32 V/V the SNR no longer increases with additional gain.

    Here is the Excel file you can reference to see the computations: SNR-Example.xlsx

    The ADS1262's internal amplifier is a very low noise amplifier, the noise spectral density is around 7 nVrms / sqrt[Hz]. Compare this to the INA125's noise spectral density of 38 nVrms / sqrt[Hz]. Even though you can only achieve a gain of 32 V/V, you will still get better resolution with the ADS1262. Besides, you'll also save on the cost, PCB area, and improved accuracy (lower system offset and gain error) by removing the INA125.

     

  • Hello Christopher.
    I'm agreed with you about noise amplification by INA125, but I have this problem:
    I have a 2000Kg load cell with 2mV/V, so if I supply it with 10V, I will have 20mV full scale.
    ADS1262 has 78mv full scale (32V/V) with its Vref (2.5V).
    Using 32V/V, 1200SPS I will reach a 20.9bit resolution (18 cont.) that is better than a 24bit ADC, but (seems) it is obtained in ratiometric usage (5Vavdd, 5Vref): in these conditions my load cell will give me 10mV, with an ADC full scale of 156mV (@32V/V) so I'll lose approx 4 bit due the full scale (is it correct?).
    So my possibilities:
    a) Using an external signal amplifier (as INA) to have a low performance at light loads but a stronger signal and (I hope) better performance at heavy loads.
    I can try using a 5V supply for my load cell, 5Vref, and total (ampli+ADS) gain of 256V/V (full-scale 5V:256V/V=19mV, max cell output 10mV).
    b) Opposite way, using a lower external REF to increase effective signal amplification, I think I'll have less signal noise (the signal doesn't pass through INA and ADS but will be more direct). Problems would be caused by noise on Vref, thermal aspects, different supplies for cell and ADC (non-ratiometric).
    c) Using a different ADC, 32bit and 1200-2400sps capable more "tailored" for my load cell...

    Thanks a lot,

    Mario

  • Hi Mario,

    Sorry for the delayed response.

    You are correct that not utilizing the full input range of the ADC will reduce the ENOB; however, keep in mind that ENOB is a measure of "dynamic range" (calculated as the ratio of the full-scale input voltage to the input-referred noise).  "Resolution" on the other hand is limited by the input-referred noise, and while gaining up a signal will improve the ADC's input range-utilization, it does not necessarily improve the input-referred noise of the system.

     

    Here are my calculations for what you could achieve using the ADS1262 by itself (plus the performance after post-process averaging):

    Source File: ADS1262-Load-Cell-Resolution.xlsx

     

    Going through these same calculations with the INA125 + ADS1262 is a lot more complicated but there are a few things to be aware of...
     

    • The INA125 noise is much much higher than the ADS1262 PGA noise, so input-referred noise will be much higher and resolution will be worse, even with the higher gains. All you have to do is compare your 10mV input voltage to the input-referred noise of the complete system to see what kind of resolution you can achieve (this is not too bad to compute when the ADC + PGA are integrated, with multiple ICs there is a lot more math involved).
       
    • The INA125 noise is not flat, as it is with the ADS1262. Rather the INA125 has a significant 1/f contribution (which also gets amplified by the INA's gain). 1/f noise is not Gaussian, so averaging 100 samples with this type of noise will not result in the same 10x noise reduction that I showed above. This another reason that I would point to for discouraging the use of an external amplifier circuit, unless you can find something with lower noise and no 1/f noise! In my search for such a device, the OPA388 comes the closest to such a thing but it is not the best op-amp for driving ADC inputs directly, so you'd still need to enable the ADS1262's internal buffer which adds to the noise.
       

    • In response to question #1 above...as you increase the input voltage to the ADC, the ratio of "VIN / VREF" increases. Therefore, the system will be more sensitive to reference noise (to illustrate: when VIN is close to 0, the ratio of VIN/VREF does not change significantly as VREF fluctuates. Refer to The impact of voltage reference noise on delta-sigma ADC resolution). This means that you will see more reference voltage noise in the measurement, which will degrade some of the resolution you'd expect to gain from amplifying the signal.
       

    • In response to question #2...

      Reducing the reference voltage to increase the ADC's input range (improving utilization) is only effective when the ADC behaves like an ideal 12- or 16-bit noise free ADC (such devices do exist). In these cases, reducing the reference voltage reduces the ADC's LSB size and improves resolution (at the cost of a smaller input range, so ENOB remains constant).

      However, at 24-bit resolution you will not find an ideal ADC with 24-bits noise-free performance; all 24-bit ADCs are limited by thermal noise (instead of quantization noise). In this case, when you reduce the reference voltage you will see a thermal noise amplitude that remains the same; the LSB size (i.e. quantization noise) will get smaller, but the number of ADC codes that are flickering will increase in response (so that you measure the same thermal noise amplitude). This results in a smaller input range and unaffected noise amplitude, which means the ENOB (ratio of input range to noise) will be smaller.
       

    • Regarding question #3... If you really need the higher gain, take a look at the ADS1235 that Bob recommended, and also the ADS1261. Both of these devices do have gains up to 128 V/V and you'll see some small improvement in resolution at these higher gains.

      ...In the case of the ADS1262, the resolution did not improve with higher gains so it didn't need to provide for them.
       

    I hope you find this information helpful!

  • Ok!
    Thanks to Bob and Christopher for your help!