subgold wrote on Apr 23rd, 2009, 9:13am:gm/CL is a very rough 1st order estimation. Now, neglecting the impact from all the nondominat parasitic poles, you already have at least two poles in the loop, the pole from the boost amp will result in a pole-zero doublet, that's why the GBW of your main amp increases. you must be careful about the selection of your poles, otherwise you may get some ugly settling. and i doubt it is a good scheme to have GBW of the boost amp larger than that of the main amp.
there is a famous paper "A fast-settling CMOS op amp for SC circuits with 90-dB DC gain", in this paper the authors point out the fu1(fu1 is the GBW of the boost OTA) should be larger than β*fu2(fu2 is the GBW of main OTA, β is the feedback factor) and less than the p2(p2 is the second pole of main OTA). so i choose the fu2=200M and fu2=300M, did you mean 300M is not enough? so how much should be enough?