Here's some background: I am making a modular resistance tester for checking for opens & shorts in fabricated circuits. Right now I'm using one ADG731 to multiplex the positive load supply out to one of 32 channels, and another ADG731 to multiplex the return path through one of two sense resistors (one low-value for measuring low resistances [0-10ohm]; and one high-value for measuring high resistances [5-100Mohm]). Scope creep on the low-resistance measurement tolerances has dictated that we move to a 4-wire measurement. I'm planning on using another pair of ADG731s, connected up pin for pin to the same gates as the "2-wire" ADG731s, since we already have two contacts on each end of the load for doing contact checks before the measurement.
And on to the question: I'm using an AD8224 instrumentation amplifier to handle the differential input from the 4-wire sense multiplexers, as well as the single-ended measurement of the sense resistor voltage. The output goes into an AD7902 for conversion. The 4-wire setup is pictued below (excuse the poor hypersnap art). The green lines indicate that all the gates of the two "high side" ADG731s are connected, as well as the two "low side" 731s. The red lines show the current path, and the blue lines show the 4-wire sense voltage path.
Under these conditions, I get an always positive voltage coming out of the AD8224-B side. However, when it comes to doing a "contact check", where only two pins are used on one side of the load, I'd still like to be able to use the 4-wire sense path to measure rather than calibrating out the resistances of the entire current path and controlling the +5V tightly (as is needed for the 2-wire measurement method).
Here's how the schematic changes:
Admittedly, I will be measuring the lead + contact resistance here, but that should be easy enough to calibrate out w/ some gold-plated artifact. The problem is now my (admittedly small due to the supposedly low resistance) voltage is negative, which the AD8224+AD7902 combo doesn't like very well.
So my question is do I switch to the fully differential AD7903 & add an op-amp on the output of the 8224 to make the output differential?
This seems the most bullet proof situation as everyone touts the better SNR & CMRR of fully differential ADCs, but my ADC transfer function goes from 0 -> 5V to -5 -> +5V, halving my resolution. Would it be better, since I don't care about measuring "negative" resistance, to put something like an absolute value circuit after the 8224, like either of the following?
After doing more testing w/ the initial setup, I also have another problem. The 8224 is "rail-to-rail", but that's w/ a gain greater than 1. It's got a rail-2V limit on the input, so w/ 0 & 5V rails, even if the input goes to from 0-5V, the output goes 0-3V. I can design a higher voltage rail for the next version, but then I get into the possibility of generating high voltages for the 7902(or 7903), since the "A" side of the 8224 is G=1 (which needs rails > 5V to output up to 5V), and the "B" side is G=4 (which will take a 5V input on a 9V rail and output 9V, which will give the ADC a big headache). Do I go and put clamping diodes on the input of the ADC, or is there a design decision I'm just not considering?
Add to that the relatively poor linearity I'm getting currently w/ the 8224 & a 0V negative rail w/ inputs below 100mV...I'd like to go w/ something like a +/-9V rail going forward, but a -5V input on the G=4 side would likewise give a -9V input to the ADC.
On the surface, it seems like the only option to keeping the 5V rail on the 8224 is to reduce the system voltage to something like 2.5V & change the "A" side to a gain of 2. But then I increasae noise w/ a lower system voltage trying to measure high resistances. Any advice AD community?