a few questions about folded cascode design

[airboss]

I'm designing a 2 stage op amp and I use folded-cascode as my 1st stage.

1. I think output resistance is

[gm4 * ro4 * (ro7//ro2)] // [gm9 * ro9 * ro11]

so, I want to match [gm4 * ro4 * (ro7//ro2)] up with [gm9 * ro9 * ro11] as
much as I can. However, I saw from another tutorial that I should make the
ro of M4 and M7 match with M9 and M11. So which one is correct?

2. In the same tutorial, the author says it's a rule of thumb that I_2 = 4 * I_4. Is
that true? Why?

3. I have a question about sizing M1 & M2, the input fets. The spec I got is an input
signal with 0.9 as CM V and signal swing p-p 0.5.
So I was thinking that, I have to make sure when input is 0.9 - 0.25, M1, M2 and
M5 are all in saturation region, which means 0.9 - 0.25 - Vgs1 >= Vov5. Then I
size M1 and M2 to get a small Vgs1 and then size M5.

On the other hand, an example on Razavi's textbook doesn't have a requirement
for input swing. The way he calculated the size of the input is,

connecting input to output and find the minimum size of M1 & M2. Do I need to
take this into consideration?



Thanks!

 

[engrvip]

in this ckt. ro7 and ro2 are not in parallel as the res. we r calculating is small signal and hence othere node of ro2 is not at analog ground. so resistance of upper branch is gm4.ro4.ro7 and we do need to macth these two parallel branches only from where signal is being taken to second stage.

 

[airboss]

Hi,

I think we need to take ro2 into account. There's a virtual ground (correct?). So looking at the drain of M2 and M7, we actually see ro2 to a virtual ground and ro7 to ground, which means ro2 is parallel with ro7.

Can you explain why they're not in parallel?

 

[rambus_ddr]

when you design a high BW opamp, the parasitic capacitor need be considered, such as cgs of M3. and the Cgs and gmbs of M9 when the compensation cap connected the source of M9 with the second stage output, because there is a zero at right plane, which affects your phase magin.