How big difference of noise between the realistic circuit and pspice simulation?

[goodnight521]

I am trying to build JFET preamplifier which is connected to a opm amplifier behind.After pspice simulation the equivalent input noise can achieve 20nv/sqrt Hz @5Hz which meets the need.However when I finish building the circuit with realistic componets the noise measured is far worse than it should be.I use DAQ to aquire output signal and measure it by labview.The equivalent input noise level is approaching 1uv/sqrt Hz at low frequency ranged from 1-100Hz.(btw I use 2sk170 input pair)

The tricky thing is that I build a simple common source amplifier with 2sk170 whose gain is 10 times.The equivalent input noise from pspice simulation and realistic measurement is 0.8nv/sqrt Hz and 100nv/sqrt Hz.I am so confused there will be such huge difference between them.

So my question is that how to build realistic circuit to reduce noise to achieve ideal result.Or do I measure the noise in a wrong way?What noise is realistic component supposed to be like 2sk170?

[h_a]

Since nobody bothers to reply, I give you my 2 cents.

My Spice-experience is limited, mostly because the quality of the results so strongly depend on the quality of the used transistor models.

Honestly I would not trust Spice for the calculation of more complex functions like distortion or noise in a real amp, especially if the circuit just barely meets the spec.

Of course there are exceptions like the proprietary Spice that IC-companies use (I was told they achieve excellent results!!).

However if you find out what the problem is, I would be very much interested in this as well.

Have fun, Hannes