exp
Full Member level 1
Hi,
Using the square law model, one can derive the distortion of a differential pair using NMOS as:
HD3 = 1/32 (vi/Vov)^2
I successfully used this approximation for an 180nm process and it was still reasonably accurate.
Now for a 28nm process, this formula seems to be unuseable. As example, I use vin=1mV. Vov is not very well defined but dc analysis in spectre shows me Vdsat=130mV. This yields -114dB.
Using the values vgd-vth from dcOpAnalysis I get 0.52489361-0.43463317=90mV, resulting in -108 dB. Close enough.
However, measuring the HD3 or THD with tran and dft gives me -76dB ... orders of magnitude different!
How do you deal with distortion in a modern process?
Using the square law model, one can derive the distortion of a differential pair using NMOS as:
HD3 = 1/32 (vi/Vov)^2
I successfully used this approximation for an 180nm process and it was still reasonably accurate.
Now for a 28nm process, this formula seems to be unuseable. As example, I use vin=1mV. Vov is not very well defined but dc analysis in spectre shows me Vdsat=130mV. This yields -114dB.
Using the values vgd-vth from dcOpAnalysis I get 0.52489361-0.43463317=90mV, resulting in -108 dB. Close enough.
However, measuring the HD3 or THD with tran and dft gives me -76dB ... orders of magnitude different!
How do you deal with distortion in a modern process?
