ADS1248 - How to test CMRR & NMRR

Dinu,

I think this question has been asked of one of the FAEs supporting you and it was asked in an internal forum. I gave an answer earlier today, which I'll duplicate here. I've done a few tests just to test this out and I've got a few results.

Originally, it was asked if there were any application notes on this subject. I don't know of any application notes on test circuits for our ADCs. However, most of the tests would be straight-forward, and not different than you might expect.

For common mode rejection, I would assume you would use the internal reference for the ADC. The positive and negative inputs would be tied together and measurements would be made as the input was moved through the input range. Since the device has a digital filter with a frequency response that varies with data rate, I would do this as a DC test at whatever data rate is desired.

As an example, for PGA=1, you would set AINP and AINN to 0.1V, 2.5V, and to 4.9V. This gives three measurements, one at the center of the common-mode input range, and the other two measurements at the edges of the common-mode input range. I would take 128 measurements at each input setting, and average each of the 128 measurement sets to separate the noise from the CMR measurement. In the end you'd have three sets of data converted to voltage: Data0.1, Data2.5, and Data 4.9.

To get the CMR, you would calculate

CMR = 20 * log ((2.4V)/|(Data2.5-Data0.1)|) and

CMR = 20 * log ((2.4V)/|(Data4.9-Data2.5)|)

and take the larger of the two as the CMR in dB. I used ADCPro with an ADS1248EVM and I got a value that was about 110dB. It's better than reported in the datasheet.

With a different PGA value, you need to incorporate the gain into the calculation. As a second example, for a PGA=32, the input range is reduced because of the PGA. In our own testing, we do use slightly larger input range considering the gain (data is taken at +2.4V and +2.6V. Regardless, if your PGA gain is 32, input will vary less, given the PGA entire input and output range.

It's also possible to take the endpoints of the CMR measurement in this example, I use endpoints, instead of taking the larger between variations from the midpoint.

CMR = 20 * log ((32*0.2V)/|(Data2.6-Data2.4)|)

With this measurement method, get something similar to that in the datasheet. In the end, it was close to 126dB

For the normal mode rejection, it's another straight-forward measurement. In this one, you'd need a function generator with some DC offset so that the input would be in the center of the common-mode input range.

I used an Agilent 33522A, which has two outputs, both are set to have a DC output of 2.5V. with 1Vpp and one 180° out of phase with each other. This gave me a ±1Vpp 50Hz signal. It also gave me a 1mV DC offset error, which I ignore for the AC measurement.

Using two inputs that are 1Vpp and 180° out of phase gives me an ADC input that is 2Vpp as the input. Using ADCPro, I get this result:

Which looks like I get a Vpp measurement of about 300uV. For this NMRR measurement the result is:

NMRR = 20 * log (2V/300uV) = 76.5dB

Which is a bit better than the expected value. Other measurements for 5Hz and 10Hz were a bit worse than expected. I got results that were about 95dB. Based on the output of the data, it looks as if the input has some extra low frequency noise. If I wanted to see the expected 110dB NMRR, I'd need to make sure the input signal noise (and input signal offset) would be less than the 40-50uV than I'm seeing. This might require extra input signal conditioning to clean up the signal source.

I'm not sure I understand the last question about the noise rejection. In these examples, I've allowed the noise in the measurement and most have worked out fine. For the NMRR, test at 5SPS and 10SPS, the result is limited by the signal source and not the ADS1248.

Regardless, I hope this is a good outline for the measurement. It's a standard set of measurements and it shouldn't be too difficult to test.

Joseph Wu