If bias current through LVDS Tx is 7mA ( shown in red line )
Then how can i calculate the bias current in CMFB feeedback circuit ( shown in blue line )
well depends a little bit - I would look at the whole thing as a two-stage opamp with miller compensation (forget about the nmos h-bridge and dthe common-mode-feedback-resistors) ...
so now you have to solve the problem how to design a two-stage opamp - and you should think first about the point imo - where there nondominant pole in the left hand side realy is lying - then you can go on to do the rest ...
by the - I think the circuit is good for setting the common-mode voltage in lvds transmitters - but I think the paper itself is not so smart -> why ?
because they compare their solution with bonis paper and arge that they use less current - but (!!!) I don't think that they were able to build with the 5.5 mA that they use a "two-stage-amp" that can reach enough gbw so that the feedback can compensate also for common-mode-distortions which get from the line into the transmitter in the relvant range of serveral MHZ. IMO Boni used in his approach so much current because he wanted to be able not only to regulate a "DC-voltage" ...
Anyway - to break it down - you could design a first stage of the opamp wth miller compensation that can have a very low bandwidth in my opinion - the only aspect that limits the bandwidth of the diffpair in my opinion is the "swing-in" time of the feedback loop ...
OK - my 2 cents - maybe I'm wrong - if I'm le me know ....