LTSPICE noise calculation

Hello everyone,

what are the calculation that LTspice does in background during noise analysis?

I am wondering about this issue because, I am not sure if the results I found out on my simulation are correct or not. In two points of my circuit I have the following noise density

which are very similar, then I use a BV generator to calculate the difference between the voltage at these two nodes and the result is the following:

How this result is calculated? I am not expecting to see that the noise becomes smaller, because the noise density sum as square at each frequency.

so why the total noise is smaller

  • 0
    •  Analog Employees 
    •  Super User 
    on Jun 4, 2021 4:02 PM

    Hello master89,

    Regarding your question about the calculations that LTspice (or any other simulator) does with regard to noise at different nodes when you sum or subtract from each other:

    My thoughts are that the results you get depend on whether the two nodes in question have correlated or uncorrelated noises. With correlated noise, if you subtract them from each other, you will be get some cancellation. That would not be the case if the the noise at the nodes in question are uncorrelated. I believe LTspice takes this into account during noise simulation.

    Hope this helps answer your question.



  • Hi Hooman,

    That is exactly what I also suspected, the I wonder, how ltspice can take into account about correlation?. At this point I am so curious what are the operations that ltspice perform during noise analysis :) .

    In any case thank you for confirming my hypothesis. Your opinion provided me great help.

  • +1
    •  Analog Employees 
    •  Super User 
    on Jun 6, 2021 10:50 PM in reply to master89

    Hi master89,

    I'm not an LTspice specialist. So, I cannot comment on how LTspice keeps track of correlated vs uncorrelated noise, but it does :-) Maybe other people more knowledgeable on this forum about how simulators treat noise can comment?

    To convince myself, I used the LTC6228 standard demo circuit in LTspice and looked at what it shows as the noise when I subtract it from itself (E1 creates a copy of the output noise, and E2 subtracts this from the original output).

    The results is 0nV/RtHz noise! The only reason this would happen is that the "subtracted" noise is keeping track of noise correlation.

    Here is the LTspice simulation file for reference:

    4201.LTC6228 Noise Subtraction Demo EZ 6_6_21.asc

    You can see 0nV/RtHz is the "subtracted" (resultant) noise:

    Hope this helps.