Improving LDO phase margin for the unloaded current

[shooks]

I am designing an LDO that has an external load range from 1uF to 10uF. Under loaded conditons (current load), I have pretty good phase margin, however for the unloaded (current), my phase margin reduces significantly. I have tried incorporating a zero to offset the phase rolloff, however for the no current case (other than internal bias of 100uA) and a 10uF cap load this zero proves to be detrimental. By the way, I am incoroporating a pmos output driver - Any thoughts?

 

[dick_freebird]

LDO Compensation

This is one of the tougher stabilization jobs, because the "load
pole" is varying by decades and can go from secondary
at high load, to dominant at low load.

Bringing Zout down in the lightly loaded case may be what
solves it for you. But that will tend to cost efficiency / ground
current most likely.

But consider that if you had an output voltage image, offset
by (say) 100mV with a 1Kohm resistor, you'd have a Zout that
never goes below 1Kohm at the cost of 100uA wasted
current. Contrasted with (say) 2.5V out at 100uA, an operating-
point Rout of 25Kohms and a pole 2+ decades further to
the left.

Perhaps you could make the offset smaller and the resistor as
well, but this comes down to tolerances. Might also make the
bleed sensitive to some sensed current or gate voltage so
as to not pay the tax full time.