ADS1244: How to measure in the lab 50Hz/60Hz noise rejection feature of ADS1244 or similar Sigma/Delta ADCs

Alberto,


How you test the 50/60Hz rejection depends on what equipment you have available. The ADS1244 does have an EVM, but doesn't have any software or GUI to go with it. If you want to use the ADS1244 device, you'll need an SPI master to get the data out of the device.

If you're interested in a similar device with the EVM and software, I would probably start with something like the ADS1220EVM. This EVM uses ADCPro to communicate with the device. Just note that ADCPro was originally designed with Windows XP and may not work beyond Windows 7. The ADS1220 has two channels and a few more features that the ADS1244.

For a procedure to measure 50/60Hz rejection, I would look up the following post that talks about CMRR and the 50/60Hz rejection measurements:

e2e.ti.com/.../508732

Beyond the EVM and software, you'll need a function generator and the ability to add some DC input voltage so that the measurement is within the input range of the ADC. After that, the rejection is a matter of recording the peak-to-peak input voltage and peak-to-peak output data.

If you do build something with the ADS1244, just remember to use a 2.4576MHz external clock. The frequency response scales with the external clock frequency. If you don't use the proper frequency, you won't get the right 50/60Hz rejection.


Joseph Wu

Dear Joseph,

thank you for your prompt answer. The link to your previous post concerning CMMR answers to 99% of my question.
If I may, I just take a few moments of your time.

In order to make my test, I should AM modulate the input DC signal with a 50hz-to-60Hz signal, then take a large number of
samples. In presence of a full rejection, the histogram should be centered on the DC value, the sigma should be determined
only by the resolution of the ADC. In absence of rejection, the histogram would show a bimodal shape, the two peaks being the
lowest and highest voltage values modulated by a 50Hz sinusoid. In real life, in presence of a finite rejection, the histogram
should be centered on the DC value, while the two extreme tails should allow to calculate the rejection. A series of
measurements made with differente AM apmplitude should allow to extrapolate a statistically more significant rejection value.

Did I get the point ?

Just a last question: in my application, I sample the signal at about 0.5 -to - 1Hz (I measure a temperature). What about
the aliasing effects ? Line noise is well beyond half the sample rate and it will sneak through the base band.
Apparently, the rejection measurement should/will give different results at different sample rate.
Do you agree ?
If so, could you suggest a way to normalize the results ?

Thank you in advance for your time and support.
Regards,

Alberto

Alberto,


To make the test you would just change the frequency and grab a collection of data. You could just record the peak-to-peak values of the input to the data and record the ratio in dB. By repeating this over different frequencies, you would eventually get plots similar to figures 7 or 8 in the ADS1244 datasheet.

One thing that I'd point out is that in the absence of rejection, it's hard to say what data you would get. I'd say that you are unlikely to get something purely bimodal.

Lastly, you can insert some basic RC filtering for anti-aliasing. Note that this is a delta-sigma (or oversampling) ADC. This means that many samples are taken for a single data. In the case of the ADS1244, the ADC samples the input at 19.2kHz, and the aliasing doesn't occur at the data rate. Figure 6 in the ADS1244 datasheet shows this effect. With the frequency response, you could build a basic anti-aliasing filter out to 10kHz and it shouldn't affect your measurement.


Joseph Wu