I am getting odd results for Eldo transient noise analysis on a fully differential OTA I designed.  I get 7 nV/rt.Hz input referred using AC noise analysis on both neg and pos signal paths.  When I run noisetran on the amplifier open loop with a uVolt differential sinusoidal input I get several millivolts of common mode noise with hundreds of uVolts riding on top  (the transient output looks great, no clipping or slewing).  I am not sure if this is statistically valid, but I diff the two rms_noise output signals and get about 600uV of zero mean noise, which itself has a 385uV rms value.  My amp has 90dB of gain and a dominant single pole roll off at 1.2kHz.  I would expect 2*((7nV/rt.Hz)^2*GBW*Pi/2)^1/2 = 108uV total integrated noise, but I am not even close with 600uV (or 385uV!)

I have run the same cross check on a single ended amp I built and everything ties out between AC Noise and noisetran using the GBW for that amp.  Am I doing something stupid with my test bench or committing some major statistical blunder?

Thanks in advance

May I know is the equation right?
2*((7nV/rt.Hz)^2*GBW*Pi/2)^1/2 = 108uV

I thought it should be *f(3dB BW) instead of *GBW.