+ Post New Thread
Results 1 to 2 of 2
  1. #1
    Newbie level 4
    Points: 192, Level: 2

    Join Date
    Sep 2016
    0 / 0

    [Moved]: Noise Analysis of Pixel Circuit in Cadence

    Hello all,

    I am currently designing some pixel circuits for my bachelor thesis and am trying to analyse the noise of such a pixel using Cadence. I have a couple of questions regarding this aspect as I am new to this..

    Also please consider that the pixel is a black box (made up of a photodiode and transistors) whose input ports are some logic signals used for selection and a photodiode input light intensity. For the output I have a voltage signal, say Vout. I am using ADE and selecting a noise simulation sweeping the frequency to obtain the power spectral density function and then integrating the range I require to determine the total noise output.

    1) Is the correct way of doing noise analysis of a circuit?
    2) I am assuming that any dynamic noise measurements for example kTC noise and such are not included in the simulation. Is it right to assume this?
    3) Moreover I get a plot which seems to include 1/f noise (the noise power increases and then drops proportionally to 1/f at around 22Hz).. Are the standard models such as NMOS4 and such valid for >20Hz or so? (sorry this might be a daunting question since I have not seen any information regarding the noise model for NMOS4)
    4) I also get a total noise voltage of around 0.5V which to me seems a bit high.. so again I ask if this methodology correct so that I can try to perform further tests in order to justify this value

    Thank you for help,

    •   AltAdvertisement


  2. #2
    Advanced Member level 5
    Points: 39,802, Level: 48

    Join Date
    Mar 2008
    1878 / 1878

    Re: Noise Analysis of Pixel Circuit in Cadence

    Not much of a "noise guy" myself but working around
    the edges of readouts (ROICs) I see a lot of the noise
    can be RTN / RTS which is not well modeled in SPICE
    devices, sub-Hz and critically, history-sensitive with
    respect to charge pumping that populates or depopulates
    the traps. The good news is, clocked circuits tend to
    "reset" RTN (but continuous-time micropower / nanopower
    circuits suffer badly and randomly). Bad news is, if RTN
    is "lumped" into 1/f noise models you may have a poor
    fidelity (or not) depending on topology and exercise and
    not to mention large process variability.

    A pixel by itself may not be loaded like one in a camera
    signal chain, which could affect noise voltage amplitude.

    1 members found this post helpful.

--[[ ]]--