maneatis ring oscillator

How do I go about finding the Ceff for the oscillation frequency of this particular conifguration? I checked the original paper and it links the oscillation frequency to the effective capacitance of the delay buffer stage, but I don't understand how to go about finding this.

Also, with regards to designing this particular delay cell, it is the width of the transistors that controls the voltages and not the length. My reasoning is that since current is a function of (w/l) and cox is given as c'*(wL) where cox' is capacitance per unit area, the lengths cancel making the current in the delay cell dependent on the width^2. The only time the length comes into play is by scaling the capacitance and changing the center frequency, right? I mention this because most textbooks just insert the mobility and oxide capacitance as a constant parameter when the oxide cap is a function of the area of the fet, and now I have started reading foundry parameters and that is not the case anymore.

Can anyone help me out with this? Thanks

I am not very familiar but I think in this type of ring oscillator the oscillation frequency is given by the number of buffers per ring, due to changing the overall delay which affect the coupling factor C.
Less number of buffers increases the oscillation frequency when more buffers decrease the oscillation frequency.

http://www-vlsi.stanford.edu/papers/jgm_jssc_93.pdf

I am referring to a different paper which I attached. He states that the buffer delay can be written as Ceff/gm or Ceff/(k * (vctrl - vt)). I am not sure how to obtain the Ceff. Can somebody help me out on this or maybe point me to a good reference? Thanks.

I will try to read the paper i nthe meantime.