Measure 0.010ohm step change, ADS1248

Vindhyachal,


I think the circuit may or may not work - I can think of one bit problem that I've outlined later in my post.

As you've noted, the noise with 5SPS at a gain of 128V/V should only be 0.29uV peak-to-peak. With 0.01Ω as the step change, you should be able to resolve the 15uV of voltage change on the input. If you're concerned about the small change, you could route both excitation currents to the resistor to get 3 mA of total excitation, doubling the step voltage seen at the inputs. This would increase the common-mode input voltage to 3mA x 1000Ω or 3V, but with a max range of 2Ω, you don't push the input outside the common-mode input range.

However, I don't have any idea about how much noise there is on the excitation current. Since you measure this against the internal reference, this might add significant noise to the measurement. I don't think there's any data on this and and you would absolutely need a clean current source to make the measurement. I can look to see if there has been any past work in characterizing this.

As an alternative, you could make this a ratiometric measurement, similar to the way RTDs are measured. I would use a precision resistor for R6, and attach that to the reference input of REF0. This way the excitation flows from R8 to R6 without going anywhere else. The output code would be proportional to R8/R6. This would eliminate the current noise from the measurement since the same current goes through both resistors. I think that would be the easiest (and most accurate) way to make the measurement.

There is one other note I would like to add. I wouldn't separate the grounds, generally, we suggest to layout a single ground plane, considering the return path currents for both the analog and digital supply.

If you have any other questions let me know.


Joseph Wu

Vindhyachal,


With this ratiometric method, I think this should be more accurate. Again, this is similar to the way RTDs are measured and it should give moderately good results.

I think it is possible to measure 4 (or more) resistors. Since the resistances that you measure are so small, you could stack them one on top of another so that you have one current source and then a stack of 2Ω resistors. Even with 7 resistors, the total is 14Ω. With a reference resistor of 1000Ω the voltage will not exceed the maximum input voltage of either the reference or the analog input.

The output of the ratiometric measurement will be (2/1000)*(2^23-1). Note that the output is in a twos complement notation for negative values. In this system, you won't get negative values (unless your offset is large), so you'll lose the some of your resolution because half of your full scale range is not used. You could also use the PGA gain, but note that it also supports negative values and make sure that the input signal stays within the common-mode input range of the device. There is a section in the new datasheet describing the PGA and what limits the input range.

I would also note that R6 should be a precision resistor (your 1000Ω resistor). This is not just for the accuracy, but for the drift as well. If you care about the precision in the measurement, you'll want the reference to stay as stable as possible.


Joseph Wu