hi all,

i have a confusion about noise summary of spectre.

i have an opamp used in a closed-loop. when i change the feedback factor (the system is still absolutely stable), and resize some transistors (only in the output stage), the noise at the output node has increased (the white noise).

the noise summary shows that, now some current sources in the bias network become the main noise contributors, whose white noise has increased significantly. (before the change, the main noise contributors are the input transistor, which is as expected)

to me this noise summary is really confusing and should not be the case, because i checked the dc operating point of those transistors, and they remain exactly the same. (of course they should remain same because they are just in the bias network). also, this shouldn't have something to do with the change of the system bandwidth, because i only check the spot white noise at one certain frequency.

so ultimately what does this noise summary tell me? how can i reduce the noise then?

In pnoise analysis, one can separate noise into sources and transfer functions by setting the parameter separatenoise=yes. Unfortunately, a similar parameter does not seem to exist for noise analysis.

hi,
 1.To reduce noise from bias circuit  use less mirroring ratio (even though it increase's power dissipation it really helps some times)
2. I didn't understand why M1,2,3 don't contribute noise because their noise is directly dummped into signal path transistors

Could you please share you W/L ratio's of transistors.

Thanks,
rajasekhar.

hi,
  still i didn't understand why they are contributing...ok let me give last trail could you please increase their overdrive and check..if entire noise decreases means there are some nodes which are low impedance nodes. Then find Signal path node impedance and try to increase

Thanks,
Rajasekhar.