Hi,all
I designed a 600M~800M GBW two stage opamp.
the first stage is rail to rail input,gain boosting folded cascode structure,
the second stage is normaly class A output.
the compenstation method is miller compenstation, that is a resistor in series a cap.

After simulation,the main OTA GBW and Total opamp GBW is about 600M~800M,PM is about 40~70,which can be adjust by compensation.
The boosting opamp GBW is at the order of 1.4G, i didn't postion the boost opamp GBW near Total opamp GBW for not  degrade the settling,while the boosting  loop is stable,for a PM about 30~40.
gm2=3m,gm1=0.43m,Cl=2.5p,Cc=1p.

But now, i find that the settling is not as good as i expect,it seems there is a doubltes in total opamp GBW, or to say there is a zero in GBW,the settling waveform is in the attached fig.the green waveform is Rc=100ohm, while blue waveform with no resistor but Cc is 100f bigger.

we can see that point A, B are degrade by a slow settling, i am  confused by this degraed phenomena.
I think this transition point is  not the point from slew to linear settling,the waveform didn't seems that.

i think about the following questions,am i right?
1. because the GBW is big, about 600M~800M, so the RHP zero caused by zero resistor through miller compensation is easily enter the GBW region or 3x GBW region, and caused a degraded settling,even thouth the Rc is very small. But if we don't use a zero resistor,is PM is poor, and need a big Cc and big I2.
2. there is still a degraed settling even though i didn't use a zero resistor showed by the blue waveform at point A. What caused this and how to solve it?
3.it seems there is a zero is in GBW?
4.Is the Cgd of input differential pairs caused the zero?

i am confused much and didn't how to avoid this degrade.
thanks.

hi,
   i think you are following that paper completely....i feel these are assumption that pap
1.Boosted gain amplifier gain is more.
2. Amplifier which is used for the boosting is completely dominate pole compensated......they neglected pole zero doublet formed due to active pmos load.
3.Main amplifier don't have any zero

But in your design
Two stage opamp and miller compensated.....so there is zero but you have compensate that very carefully........so in this type situation better use indirect compensation  (check baker book). And loop PM is 30 means there is lot of chance for oscilations , so decrease boosted amp BW or increase 2nd dominate pole in the main amp.

Ans to your 4th question is it create very high frequency zero provided output pole is dominate...so nominally you can ignore that. Can you post schematic?

Thanks,
Rajasekhar.

hi,
   you see gain plot in the pap,there he gave ugb of boosted amp is grater than 3db of the main amp..this is needed only in case we have high gain in boosted amp..but any how this is not needed in this discussion
  its two pole system.... to get feel break the loop and see..
  ya thats what i am telling you will zero's in simulation because they are coming because of the active load
 indirect compensation eliminates  zero, better use cascode
 regarding miller effect if o/p pole is dominate then by the time cgd is active o/p impedance would have reduced.........but   better to investigate this through simulation....

Bottom line is try to increase loop PM, i think problem may solved.

Thanks,
Rajasekhar.

why only high gain boosted amp need its gbw is bigger than 3db or colsed loop gbw?

active load,do you mean a mosfet biased by a dc voltage or a regulated cascode mosfet? i biased the cascode amp by dc voltage,
but the zero seems remain the same.

To increase PM, the settling is becoming worse,for the ringing and overshoot is very small, linear settling is almost flatten,but the slow settling(which showed degraed in the waveform) is remains the same.

raja.cedt wrote on Aug 10^th, 2009, 3:39am:

I calculate the GBW based on gm1/(2pi*Cc) derived from a two-stage amplifier with a miiler compensation and assume Cc=1.2p to include parasitic caps from mosfets. I will get GBW=57MHz if it is a single pole system. I am not convinced by your gm1.

Maybe now it is more clear. GBW=0.43m/(6.28*1.2p)=57MHz