We are developing a design that, among other things, needs to provide sensor input from two load cells. The load cells are 100 lb rated with 3mV/V output and rated 10V nominal, 15V max excitation.
We had decided on using the AD7190 because it appeared that it would be very straight forward to implement using 5V excitation, even though we were concerned that might not be enough. Looking through the app notes and circuit notes it seemed like we could at least get 1kHz and less than 0.5 ounce resolution, so we were reassured.
However, we wanted to try and provide conversions for two load cells. It looks like this setup is supported, but I see some issues I want to clarify.
First, based on the description of the channel enable bits in the configuration register, it looks like auto sequencing would be supported for two load cells (two differential input channels). I would like to verify this first, because the rest is moot if I'm missing any limitations related to the auto-sequencing.
Second, how would you hook up the load cells? To be more specific, how would you hook up the REFIN inputs on the ADC? At first I thought it would be great to use the REFIN2 pair for the second channel, but it doesn't look like it works that way - you manually select the reference input used for the conversion. I just want to verify that this is correct - and if so, am I limited to connecting REFIN1 to AVDD/AGND and losing some of the ratiometric performance? If this is the case, then there wouldn't be much point in getting load cells with SENSE outputs I would imagine.
Finally, this might be the showstopper. From what I can gather reading the AN-1084, our sample frequency is essentially cut by a factor of four if we want to auto-sequence the two differential channels and maintain our noise performance, correct? If I'm calculating correctly, to maintain sufficient noise-free resolution we'd need to set the filter value to 2 decimal yielding about 0.14 oz resolution (with sinc4 and chop disabled) but with two auto-sequenced channels our frequency would drop to just under 600 Hz.
I just want to be sure I'm understanding all of this correctly because I really want to avoid putting a second chip in the design (and all the supporting hardware) if I can help it. Cost is one issue, but board area and development time is probably a larger issue.