aaron_do
|
Hi,
I didn't exactly follow what your professor meant there. Maybe he misunderstood your question.
Anyway, for a NMOS diff pair, and a PMOS current mirror, in order to get the best possible matching between your diff pair devices, you would want the drain voltage of the two PMOS transistors (the current mirror) to be equal (you do that by properly sizing the PMOS transistors). A result of this is that when a PMOS is used in the subsequent stage, it is effectively "mirrored" over. It is not a true current mirror because it only works at the quiescent operating point, but the design becomes easier. For example, in your picture, VDQ4 = VGQ3 = VDQ3 to get the best matching in your diff pair. Therefore, VGQ6 = VGQ3 (only at the operating point).
Also, because a PMOS mirror is used, there is only one diode drop from the supply to the output of the diff pair. Therefore, the voltage at the output of the diff pair is a little high to connect an NMOS to. Your NMOS might end up in the triode region. Its just a little trickier to design. For example, one diode connected PMOS drop might be 0.8 V. For a 2.5-V supply, that would make the DC level at that point equal to 1.7 V. If you connect an NMOS gate directly to that, and your output DC level is 1.25 V, then your device is operating a little close to the triode region. It could show up in your common-mode range.
As I pointed out earlier, I believe you get better PSRR using a PMOS. Its difficult to explain, so I'll leave it to you to analyze if you're interested.
Aaron
|