eng wrote on Jul 12th, 2007, 12:16pm:Hi all,
Is it necessary to use minimum length for an RF circuit. i.e. LNA. I doubled both length and width of my LNA (Lmin=0.18um) there is a .6 dB NF improvement and -5 dBm P1dB improvement. Gain dropped 4.5 dB. The transient results look good. I don't know if I'm missing a important point.
What are the consequences of increaing L in RF design?
My thoughts are as follows:
Increasing L will decrease ft of the device and increase capacitance, although cgs is usually matched out. This doesn't really matter if the input gm transistor is operating at 0.1*ft anyway and in cascode configuration. The cascode device however, which develops the voltage gain, may suffer from larger L as a consequence of lowered ft means less gain available.
Obviously if current and W are not changed, overdrive increases which means less gm, but more linearity.
But mostly your choice depends on your specifications as Aaron_do said. I must admit the CMOS LNAs i have designed (few hundred MHz to few GHz) have seldom need minimum geometry (expect at >~0.35um), also having larger L might help ESD robustnbess as well.
cheers
aw