Hello,

   I am simulating a traditional two-stage operational amplifier with miller and lead compensation. The DC gain is 70dB, and gain bandwidth is 580MHz, with the phase margin 115 deg.

   After I achieve the specifications above, I start to simulate the integrated noise. The opamp is fully differential, the load is pure capacitor in each branch with the value 350fF. According to the first-order filter theory, the final integrated noise should be KT/C. For each branch, the integrated noise is 1.08e-4 volt.

   When I use ADE L/print/noise summary to plot the integrated noise, the results are as follows,
 When integrated band is from 1KHz to 10MHz,
total summarized noise = 0.132026V, total input referred noise = 6.20415e-5V.

   When integrated band is from 1KHz to 100MHz,
total summarized noise = 0.132042V, total input referred noise = 1.10047e-4V.  According to the simulation results, it seems the total summarized noise is converged, and the input referred noise is increasing when the band is higher (not converged).

   When I show this results to my boss, he told me how the total summarized noise can be so large, it is around 132mV!!! For the opamp, the noise cannot be so high. Compared with the value of KT/C, it seems the integrated noise is huge.

   This is my first question, why the integrated noise of the opamp is so large. How does spectre calculate the input referred noise?

   According to Ken's view "the input-referred noise is the ratio of the output noise to the gain". When the integrated band is larger, the gain of the opamp is dropping, therefore the input referred noise should be worse. But my boss told me when the band is larger, the gain of the opamp is dropping, but the noise of the opamp at higher frequency is dropping as well, therefore the input referred noise should be converged too even for the infinite integrated band.

   My second question is how to get the converged integrated input referred noise if the integrated band is going to infinite.
 
   I am confused now. Anybody can tell me give me some help. Thanks!

Hello Stephan,

    Thank you for your reply.

    I think when using spectre noise analysis, I can get both the total summarized noise and input referred noise either in the form of volt or square(volt). I think the total summarized noise is the integrated output noise, and it is huge, around 132mV. Then divided by the gain, I get the input referred noise.

   I am confused why the total summarized noise is so large. Now I think the total summarized noise is output noise, not intput noise, If I understand wrong, please correct me.

   For the input referred noise, Ken said "the input-referred noise is the ratio of the output noise to the gain", but Ken didn't menthion if this gain is DC gain, or the gain should be varied as the frequency.

   Any comments are appreciated.

Hello,

   I turn off the flicker noise, then the integrated noise reduced to 12mV. Then I turn off thermal noise of the transistor, just keep the noise from parasitic resistance at gate, source, and drain, i get the integrated noise 400uV.

   I want to ask if this integrated noise value normal for opamp.

   And if I don't turn off the thermal noise of the transistors, in the Pnoise simulation, the thermal noise of opamp dominates the total integrated noise, around 94%, and the switched capacitor sampling noise only occupies 6%, is it normal for SC-amplifier?


   Thank you.