Yxie,
You are correct in saying that Spice only computes V_n when computing the input referred noise to a voltage source. Similarly, it only computes I_n when computing the input referred noise to a current source. If you want both, you can simulate your circuit twice, once with a voltage source at the input and once with a current source at the input. If you do this, you must be very careful not to disturb the bias conditions of your circuit when changing the input source. Alternatively, you can use a port for the input source and perform an S-parameter noise analysis. It computes 4 noise parameters from which you can extract V_n, I_n, and their complex correlation coefficient. To do so, you will need to consult a book on S-parameter noise analysis. I believe Microwave Transistor Amplifiers: Analysis and Design
by Guillermo Gonzalez  would have what you need. If not try Introduction to Computer Methods for Microwave Circuit Analysis and Design by Janusz A. Dobrowolski.

I must add that your initial comments are misleading. You seem to suggest that somehow Spice is giving you the wrong answer when computing the input referred noise because it does not include the shot noise components. This is wrong. The shot noise components are included, and you can see this from the output noise summary.

Spice is giving you the right answer to possibly the wrong question. When you drive your circuit with a voltage source and ask Spice to compute the input referred noise, it is telling you how much noise would have to be added to the input to compenstate for turning off all the noise sources in the circuit. What this is not telling you is how this equivalent input noise would change if the source impedance changed. If you know both V_n and I_n (and their correlation coefficient) you can compute this. But Spice does not give you this information, nor can you get it from Spice. Above, I did tell you how to compute V_n and I_n, but there is no way to compute the correlation coefficient. Instead, you must use the S-parameter noise analysis, which is available in Spectre.

-August

[quote author=yxie  link=1074355991/0#0 date=1074355991]There are two input referred noise components in the case of a bipolar circuit, namely the input referred noise voltage and current.  For a single bipolar transistor, the noise voltage is related to the base resistance and gm (thermal noise), and the noise current is related to the base current and collector current (shot noise) plus 1/f noise. [/quote]

Actually, the input referred voltage noise is the combination of the base resistance, and the shot noise current of the collector.  The collector current noise becomes an input voltage by dividing by the transconductance squared,  gm^2.  The input current noise source  is a combination of the shot noise and flicker noise of the base current.  In addition there is the collector current noise referred back to the input current through Beta^2.

The correlation between the equivalent noise generators is the presence of the collector current in both.

Chapter 11 of Gray and Meyer, "Analysis and Design of Analog Integrated Circuits" gives an excellent discussion of this."

It is important to point out that at high frequencies, there are high frequency effects on the base and collector shot noise (related to high frequency conductance) that Gray and Meyer don't account for.  These affects are much more significant in long devices.  I don't know if they are represented in SPICE models.