# Spice model for analog devices instrumentation amplifier giving unexpected outputs

I'm trying to do noise analysis with the AD8421 (although the specific in-amp doesn't matter). Right now I have two time dependent voltage sources on the inputs of the amplifier and the output seems to only follow one of the inputs. Also this signal is centered at the top voltage rail and seems to move to the bottom when the inverting input is around (~0.9 V). Here is what I am using for my test.

* pins for clarification
*.SUBCKT AD8421  IN-  RG-  RG+  IN+  -Vs   REF  VOUT  +Vs
xinamp 1 0 5 2 7 0 6 8 AD8421
rg 5 0 1000k
rl 6 0 1000k
vsr1 7 0 -20
vsr2 8 0 20

vss1 2 0 sin(0 7 1k)
vss2 1 0 sin(0 0.9 1k)

.probe
.op
.tran .1m 4m
.end


Shouldn't I be getting the difference of the signals? I'm not sure what I'm doing wrong here. I've tried different Spice models from analog devices and they seem to do the same thing.

Here is another attempt by grounding the inverting input and applying a 4mV source on the amp input. This time different gain resistors were used. At a very high resistance the output is at the top rail while lower resistances will move it down. Instead of increasing the gain the gain resistors increase it to a limit then decrease it. The image below display resistance 5 50 and 1k. After 1k the gain drops again and the output shifts to a higher voltage.