Noise floor of op-amp circuit

[Scotty1978]

I have built and tested a basic non-inverting op-amp circuit. I measured the noise floor of the circuit and plotted it in nVrms/rtHz. I'm trying to simulate the noise floor with Multisim and I want to make sure I did it right. When I physically measure the noise floor, I short the input impedance, which is a capacitor, to ground and measure the spectrum on an HP 35670A spectrum analyzer. In Multisim, do I short the input impedance to ground, or do I need to include an ac voltage source? If so, does it matter what the source is set to for the noise simulation?

I laid out the design with an ac source, selected Simulate/Analysis/Noise Analysis..., selected the input noise reference source to my ac source, selected the output node to be the output of the op-amp, and selected the reference node as ground. I chose to plot the inoise_spectrum. I saved the output to Excel and then actually imported it to Matlab, where I plotted my actual measurement results. To convert from V^2/Hz (as it is given in Multisim) to Vrms/rtHz, I took the square root of the output from Multisim and divided it by square root of 2. Is this right? Do I need to divide out a bandwidth correction factor or gain of op-amp?

Following the above procedure gave me something fairly close to what I measured, but not quite. As frequency decreases, the simulated result gets noisier than the measurement. Is there something that the model is not accounting for.

It would be great if I could accurately predict preamp noise floors using this software, as I need to do this a lot.

Thanks,
Scott