with 1pF cap you have very good Phase margin (i guess this is what you want). Even if you don't want to add  any load, i would suggest reduce the input bias current. I thin you are getting zero at lower frequencies, may be it's because of m18.

In folded cascode better bias both legs with same current, in your case it's 4:1 ratio, some time you would end up with slew rate problem

Thanks,
raj.

raja.cedt wrote on Oct 15^th, 2011, 8:25am:

hi raja,
i have uploaded the circuit and its results.....!!
i have doubts such as
1)my theoretical calculated gain is about 8.5k but the result is only 42dB(139)
2) same for the UGB calculated is 2.8GHz   and  observed is 8 MHz...
why such a discrepancy in result occurs??
only for folded cascode opamp i get like that but for other configurations like CS amplifier, telescopic amplifier i get according to theory...

Hi,

Adc=gm2*Rout is not correct, since there is a current division on the current summing node given by the output resistance of the diff pair, active load and PMOS cascodes. Thus only a fraction of the current generated by gm2 will go to the load.
Assuming similar Rout for those three (although not the case according to the op points you showed)
Adc~gm2*Rout/3.

Your Adc (barely above 40dB) seems to be low for a typical folded cascode. Please check the op-point.

Minimizing drain and sources area in both PMOS and NMOS cascodes will push higher poles even further away, so your UGF will increase.

Increasing gm2 and better balancing the output resistances in the summing node will also help.

Best
Tosei

How can increasing gm2 help your phase margin?? your gm2 is in the numerator of your UGBW frequency eqtn, in other to improve PM, you want to reduce the UGBW frequency as a result, you must decrease gm2 and not increase it! I stand corrected.

Hi Raj,

that reference on folded cascode was a good read, thanks.

Hi Ashutosh,

if you follow the reference given by raj, i think you will understand that the PM of this circuit is only governed by the parasitic at the cascode node ( source of M18).

one suggestion, don't tie the body of M18 to source, instead tie it to VDD, this will reduce the gate to bulk cap. reduce the size of M18, as your Cgsm18 also dominates at this node.
by doing this, your compromising the output swing!!!

in the AC response you attached i think your are plotting the half circuit gain, that is why your gain is at 40dB.

k.praveen

@praveen: Connecting body to vdd is ok  from cap point of view, but also signal current from differential pair will divide at m1 drain  and m18 source, so now impedance looking into M18 source will be 1/(gm18+gmb18), hence improvement in the gain as well as UBG. But i didn't understand how this reduces the swing?

@thechopper: You have to improve gds2 to get better balance for Rout (i guess it's a typo)

Thanks,
raj.

Larry_80 wrote on Oct 17^th, 2011, 5:49pm:


True only under certain conditions. Your expression assumes gm5 >> gds2, gds4 and gds5, which is not necessarily true, especially when Id is high. A general expression for the folded cascode can be found if you consider the combo of M2, M4 and M5 as an equivalent transconductance amplifier with a corresponding output resistance, which will be loaded with M7 and M8.
If you run that analysis you will find that:

GMeq = gm2 * (gm5+gds5) / (gm5+gds2+gds4+gds5)   and
Go eq = gds5 * (gds2 + gds4) / gm5

Thus Adc = GM eq / (Goeq + gds7*(gds8/Gm8))

As you can see if gm5 is not >> gds2,4,5, then Adc cannot be simplified to your expression.
Since gm5 is proportional to √Id while gds5 is to Id, as you increase Id gm5 and gds5,4,2 will get closer in value to each other and the general formula should be used.

Note: in GMeq and Goeq you can see the current splitting effect I was talking about.

Best
Tosei

hello Tosei ,
your Gout is wrong seems...

a. Conductance looking down from output=gds8*gds7/gm8
b.Conducatnce looking up from output=gds5*(gds4+gds2)/gm5
so total Gout is sum of these two.

Thanks,
raj.