ADS122U04: ADS122U04 measurement Question

Jonathan Craig

Part Number: ADS122U04

Hello,

I am looking at using the ADS122U04 ADC in my measurement system and I had a few questions.

We are measuring a strain gauge sensor that outputs a 0-20mA signal.

For our application the sensor will output in the range of 3.5 to 7mA as our application does not utilize the full sensor range. The total weight measured across that 3.5 to 7mA is 18500kg.

We've calculated that per kg measured the current output from the sensor changes by approximately 0.187uA.

I wanted to measure the voltage drop across a 100 ohm 0.005% resistor and then have the ADC take the measurement. This means that the ADC would have to measure 18.71uV changes per kg.

We want to take 4x single ended measurements with 4 separate sensors using this ADC. A data rate of 20SPS would suffice.

Is it possible to measure a 18.71uV change in measurement using this ADC? Looking at table 2 in the data sheet, the ENOB for 20SPS at a gain of 2 is 19.65 bits, which gives an lsb of 2.48uV when using the internal v reference. To me this suggests that the ADC would be able to measure the 9.35uV change, but i could be wrong or overlooking something.

Any advice / help with this is really appreciated.

Thanks

over 2 years ago

0 Bob Benjamin over 2 years ago

TI__Guru** 113765 points

Hi Jonathan,

It might be more useful to look at the actual noise tables that show the voltage instead of bits of resolution. I think you were looking at Table 2 of the datasheet and you should look at Table 1 (assuming you can use the PGA). Also, you are looking at the Effective Resolution and most likely will want to look at the Noise-free Resolution. The noise-free values will give a stable value in the measurement result with respect to the ADC itself. External noise can make the measurement less stable.

From the tables at a gain of 2 and 20sps we find that the noise voltage is 2.49uV RMS and 10.71uV Vpp. As I said, this requires that the PGA is used and the input range is not exceeded while using the PGA. You cannot measure a ground referenced signal using a unipolar analog supply with the PGA enabled. If you bypass the PGA Table 3 would be appropriate and the actual noise voltages are quite similar.

You can improve the overall response with averaging. However, averaging can become a bit more difficult to accomplish when cycling through the mux.

Best regards,

Bob B

0 Jonathan Craig over 2 years ago in reply to Bob Benjamin

Prodigy 110 points

Hi Bob,

Thanks for your response.

So going by Table 3 noise will be 2.53uV RMS and 10.06uV PP.

Table 4 suggests 17.64 noise free bits, using the internal reference gives me an LSB of 10uV.

Once i apply a gain of 2 to my measurement, 1kg works out to be 37uV . Does that mean that the ADC has enough resolution to take this measurement when we include noise? Would using an external reference be more beneficial?

Thanks for your help

0 Bob Benjamin over 2 years ago in reply to Jonathan Craig

TI__Guru** 113765 points

Hi Jonathan,

The noise numbers shown in the datasheet tables are taken with shorted inputs where the reference has essentially no effect. These numbers are the absolute best the ADC can do with respect to noise. The reference plays a larger factor with respect to the input signal. Think of the noise of the ADC similar to a bathtub shape where the noise of the reference plays a larger impact on the conversion result as the input voltage approaches full-scale.

It is difficult to say how much noise you will encounter with respect to the sensor output and any external influences such as EMI/RFI, but theoretically you should be within 1 code of the measurement step.

As far as your calculations, the noise tables are input referred. You will notice in the tables that for each step in gain the noise decreases by almost 1/2. So it is better to think in terms of input referred instead of output referred. In other words, think of the full-scale range of the ADC reducing with applied gain instead of the input being gained to the reference voltage. The actual ADC conversion code is based on the formula shown in the datasheet as equation 8 on page 35 of the datasheet. So the input signal stays the same, but the full-scale range changes. The noise values shown in the datasheet tables are what you would use for the calculation with respect to the actual input voltage.

Best regards,

Bob B

0 Jonathan Craig over 2 years ago in reply to Bob Benjamin

Prodigy 110 points

Hi Bob,

Thanks for your response.

I've setup the evaluation kit for this ADC with the Delta sigma software from TI. After taking measurements the software reports back 13.3 Noise free bits. I feel like I may have to improve on that. Is there anything we can do to improve this? Or is there another part you could suggest which would perform better?

Kind regards

Jon

0 Bob Benjamin over 2 years ago in reply to Jonathan Craig

TI__Guru** 113765 points

Hi Jon,

The 13.3 noise-free bits is with your sensor connected I assume. Is that correct? This is quite a bit different than the noise-free of the ADC by itself, so I would venture to say that you will not see any additional improvement from another ADC as the noise source is either from the sensor itself, the excitation source, or from external noise pickup from EMI/RFI in the wiring.

One option would be to attempt to lower the overall noise by using the same source for excitation as you also use for the reference. This will make the measurement ratiometric so that any noise from the excitation source would also cancel in the ADC measurement result as the voltage would be common to both the sensor and the ADC reference. You can use up to 5V (for a 5V analog supply) for the ADC reference.

Best regards,

Bob B

0 Jonathan Craig over 2 years ago in reply to Bob Benjamin

Prodigy 110 points

Hi Bob,

Yeah 13.3 bits with the sensor connected on the bench. You are right that it is probably noise.

I've a question regarding filtering, if i want to sample 4 single ended inputs referenced to AVSS (as opposed to AINN referenced to ground), what is the process for calculating the capacitor value for the low pass filter? Is it still as per this thread: [FAQ] Delta-sigma ADC anti-aliasing filter component selection - Data converters forum - Data converters - TI E2E support forums where they use equation: C_DIFF = 1 / [2*π*f_C*(2*R_FILTER)] ?

Furthermore, is there any difference between using 3.3V and 5V supply voltages if i plan on using the internal 2.048V reference?

Thanks for your help,

Jon

0 Bob Benjamin over 2 years ago in reply to Jonathan Craig

TI__Guru** 113765 points

Hi Jon,

For the filter equation you would use just Rfilter as opposed to 2*Rfilter.

As to the need to use 5V as opposed to 3.3V for the analog supply, the extended range is only required if the input voltage should exceed 3.3V. In the case of the single-ended ground referenced input less than 2.048V, there is no advantage to using 5V for AVDD. If you use a differential input where the sensor may have a common-mode voltage that is greater than 1.25V, then you would need an AVDD supply greater than 3.3V. Another way to say this is the absolute input voltage on the AIN pins cannot exceed the supply.

Best regards,

Bob B