problem with PSRR of bandgap

[banosey]

i'm designing a bandgap circuit using banba scheme , and i have some problems with PSRR

vdd= 1.8 V
vout= 1.2 V

I got PSRR= -43 dB at low freq. which is known as a bad value.

note : i increased the length of the current mirrors to 8u and still i got this bad value and any more increase in the transistors' length and width, the PSRR improves with very very small value.

what can i do?

 

[dick_freebird]

First, determine whether this is a high frequency or a low frequency
problem. They will have somewht different solutions.

DC, you probably want cascode (although you don't have much
headroom to throw away, maybe a folded cascode could help
you out, or in some processes low- or 0-VT devices are useful).

High frequency, you may prefer to filter the output to roll off the
AC currents that blow through the bias and output devices. Using
narrow and short devices in follower applications may bring
less HF blow-through than a larger device. Compensation
networks that are gate-shunt-to-rail are probably better for
PSRR than integrator-style or Miller-multiplied-style feedback,
which by "floating" the ballast make it potentially a "gain block"
for stray supply-signal couplings.

 

[liusupeng]

I guess this maybe due to some systemetic offset error in the design.

 

[saro_k_82]

Post your schematic

 

[love_analog]

What is banba schematic?

In general, if its low frequency you are worried about, increase the impedence to Vcc - cascode pmos current mirrors etc.