I'm also trying with an AD592, and i've got some data out of it.
The way to connect it, using a 10k resistor:
Connect +5v to ad592
Connect analog 0 to ad592
Connect 10k resistor to ad592
Connect resistor to ground.
(see examples for piazo or light-resistor, same principle)
(also see Tutorial/Calibration)
You mentioned the example in java, but I figured they use a different technique.
First of all they don't use arduino, and they don't measure resistance directly.
Instead they measure the time it takes to reach 2,5v (hence the capacitor? (again not very sure) ).
Then they convert this time to temp, (because it will take longer as the temp goes down ( again not very sure :)) )
I tried to mimic this with arduino, but I figured I need to set "HIGH" to 2.5V instead of 5V to be able to use the same "technique"
Returning to the "old" way I did get some "sensible" readings, 567 low, 633 high
(using ice and hot-water) but this posses a small problem:
Only a 66 (points) diff, which is not enough.
If the real temp-diff is about 80, I will still only be able to measure with more than 1 degree steps.
Does anyone have a idea how to fix this?
I would like 0.5 as a minimum step-size