In Reply to: About noise and gain posted by Jason on 08/16/03 at 11:46 AM:
Equation 1 just says that gain=gm*Rload. No problem there. gm4ro is the gain for Vn4 but Ro is not the series connection of ro2 and ro4, it is the parallel connection: Ro=ro2*ro4/(ro2+ro4). If Vno is the total noise output, then it should include the noise from Q4. The noise from Q3 can be neglected because the feedback gives Q3 unity gain. Q4 has no signal gain, but I don't see any thing to prevent noise gain.
The statment: "For Vn3, notice the curent through this voltage source must be zero because ONE SIDE is connected to the gate of Q3. Therefore, assuming all the sources are zero, the drain current is unaffected by Vn3.". Is faulty logic. The feedback from drain to gate reduces the noise gain.
The noise current from Q5 is split between Q1 and Q2. Since only the current thru Q2 appears at the output, Vno/Vn5=(1/2)*gm5*ro. I don't know how gm3 got into the equation, and what happened to ro?