From inspection, the dominant pole is Rds of M2 and Cload, Cl. You show gds as 28u (35 kohms) and Cl as 5p, this would be ~890kHz. So, I see your problem. I assume that you have just experimented with VIN DC to get the output DC voltage midway, otherwise two current sources in series are going to rail out one way or another. Are you sure you have the gds correct? Assuming a low early voltage, of say 10V at 5 ua still gets you 2 Mohm, which is way, way larger than 35 kohms. Even if you were using 45nm with say, 1V Early voltage you would be talking 200k for rds. You need to post more details on sizes, currents and processes for a more accurate diagnosis.

GaAs_si,

Try putting a 1A AC current source at the output to simulate the output impedance and compare that to your calculation.  You can also use large ideal inductors (100H for example) to block the 1A AC signal and find the impedance looking into just the drain of the cascode or the drain of the current source.  This method should not effect the DC operating points.

Also, please check that only 1 AC source is on at a time for your simulations.

Tim

Tim,

I could not understand the procedure that u mentioned to calculate output impedance.

Could you please explain the procedure in detail..???

Thanks.

To find the output impedance of this circuit, replace the load capacitor with a 1A AC current source pointing in to the circuit.  After an AC simulation plot the the voltage across this current source.  Since Z=V/I and I=1A, plotting the real and imaginary part of the voltage across the current source is the same as plotting the impedance.  Realize that this is the SERIES EQUIVALENT output impedance, in other words, R+j*X.  You can convert this to parallel equivalent impedance which is what I normally do since I find it easier to work with.  Also, always make sure you inject only 1 AC source at a time in your circuit.  You would be surprised how many times I have seen this mistake.

Now, the use of the large inductor can help in debugging this problem.  For example, place a large inductor (maybe 100H, large enough so the impedance of the inductor is extremely high at your frequency of interest) in series with the drain of M2.  Use the same procedure to find the output impedance as described above, and you should see that the impedance is now the output impedance of your cascode device only.  You can compare this to your calculation, then put the inductor in series with the drain of the cascode and see the output impedance of your PMOS current mirror only.

It seems you have solved the issue with the capacitor on the gate of the current mirror.  Reviewing your calculations, this makes sense since I see you add capacitances Cgd and Cdb of M2 which is incorrect unless the current mirror gate is at AC ground.  I will let Kevin elaborate since it was his suggestion  :)

Hi Tim,

Thanks a lot for the info. It helped me to verify the M2 impedance correctly. Which match  the calculated value.
Thank You.  

What bit is requiring elaboration?

Apply a voltage at M2 drain, Vo. The voltage across the gate source due to this voltage is given by the potential divider, V0.ZS/(ZS + ZF). ZS is the impedance across gate source, ZF is the feedback impedance from drain to gate. This voltage Vo X gm2 is the current, Io, in M2 drain due to this voltage. Vo/Io is therefore the impedance due to M2, which is (1+Zf/Zs)/gm2. The transistor sucks more current. This impedance is in parallel with the direct impedance of ZF + ZS.