+ Post New Thread
Results 1 to 8 of 8
  1. #1
    Advanced Member level 3
    Points: 5,833, Level: 18

    Join Date
    Feb 2014
    Posts
    830
    Helped
    273 / 273
    Points
    5,833
    Level
    18

    LTSpice Noise Analyis

    Does anyone understand how LTSpice's noise simulation works under the hood?

    I've used LTSpice extensively but barely touched the noise mode (I'm trying to catch up on my understanding of noise calculations in general). In a couple toy examples it seems almost too good to be true.

    Do the model designers need to explicitly prepare the model for this mode or does it emerge from standard model design practices? My understanding is that spice models can vary a lot under the hood and it seems the quality of the noise model would vary accordingly? Thus I wonder if LT models may be better than the AD models that probably weren't designed with LTSpice in mind from the start?

    •   AltAdvertisement

        
       

  2. #2
    Super Moderator
    Points: 260,224, Level: 100
    Awards:
    1st Helpful Member

    Join Date
    Jan 2008
    Location
    Bochum, Germany
    Posts
    45,441
    Helped
    13826 / 13826
    Points
    260,224
    Level
    100

    Re: LTSpice Noise Analyis

    It's basic SPICE noise analysis considering resistor noise and semiconductor shot noise.

    Also flicker noise parameters for FET and bipolar devices, if used by the respective model.



  3. #3
    Advanced Member level 3
    Points: 5,833, Level: 18

    Join Date
    Feb 2014
    Posts
    830
    Helped
    273 / 273
    Points
    5,833
    Level
    18

    Re: LTSpice Noise Analyis

    Ok and part of my question is about realistic model expectations.

    For example TI's OPA140 has no noise when simulated in LTSpice's noise mode.

    Of course this is what I expected - performance depends on the quality and makeup of the model. I'm still trying to get a sense of how many models have accurate noise performance.



    •   AltAdvertisement

        
       

  4. #4
    Advanced Member level 5
    Points: 19,083, Level: 33
    Achievements:
    7 years registered
    crutschow's Avatar
    Join Date
    Feb 2012
    Location
    L.A. USA Zulu -8
    Posts
    3,675
    Helped
    873 / 873
    Points
    19,083
    Level
    33

    Re: LTSpice Noise Analyis

    I'm still trying to get a sense of how many models have accurate noise performance.
    One way is to simulate the noise of the model and see how it compares to the specs from the device data sheet.
    Zapper
    Curmudgeon Elektroniker



  5. #5
    Advanced Member level 5
    Points: 39,705, Level: 48

    Join Date
    Mar 2008
    Location
    USA
    Posts
    6,413
    Helped
    1872 / 1872
    Points
    39,705
    Level
    48

    Re: LTSpice Noise Analyis

    I haven't really experienced anything I'd call "standard
    model design practices" in a pretty long IC design career.
    IME it's always a negotiation and you can't count on the
    noise params being populated in device model statements,
    let alone fitted inside a behavioral / macromodel.

    LTSpice has a lot of tricky stuff below the surface. The
    Yahoo! Group for LTSpice is very active and has a deep
    pile of archived info that might help you out.

    http://ltwiki.org/index.php?title=Undocumented_LTspice

    has some discussion of how LTSpice behavioral OTA does
    noise model fitting (when and if); maybe look at the .model
    or .subckt's referenced models / primitives and see whether
    you have any right to expect noise characteristics out of it.


    1 members found this post helpful.

  6. #6
    Advanced Member level 3
    Points: 5,833, Level: 18

    Join Date
    Feb 2014
    Posts
    830
    Helped
    273 / 273
    Points
    5,833
    Level
    18

    Re: LTSpice Noise Analyis

    Wait I have to take my last statement back. I simply screwed up the importing of the OPA140 model (symbol was still referencing an ideal model). So on a second quick look the OPA140 and a handful of other imported TI models do have decent correlation between datasheet and simulation noise results.


    For what its worth TI has a paper describing their new model approach:
    http://www.ti.com/lit/an/sboa338/sboa338.pdf



  7. #7
    Advanced Member level 3
    Points: 5,833, Level: 18

    Join Date
    Feb 2014
    Posts
    830
    Helped
    273 / 273
    Points
    5,833
    Level
    18

    Re: LTSpice Noise Analyis

    On a sort of related note does anyone have an understanding of why instrumentation amplifiers have such high noise?

    For example AD8429 billed as 'ultralow noise' has 45 nV/√Hz output noise. So 45 is the noise at G=1 (input noise is negligible at 1 nV/√Hz) which is quite bad (tons of precision amplifiers are <5).

    Why have an output stage with that high noise when it seems 'easy' to make amplifiers with much less. Are they assuming I'm going to use it at high gains (I'm not, I want a high impedance differential buffer and inst amps with integrated resistors are a cost effective way to buy that)?

    Am I missing something dumb?



    •   AltAdvertisement

        
       

  8. #8
    Advanced Member level 5
    Points: 20,688, Level: 34
    Achievements:
    7 years registered

    Join Date
    Jan 2011
    Posts
    3,382
    Helped
    1208 / 1208
    Points
    20,688
    Level
    34

    Re: LTSpice Noise Analyis

    Quote Originally Posted by asdf44 View Post
    For example AD8429 billed as 'ultralow noise' has 45 nV/√Hz output noise. So 45 is the noise at G=1 (input noise is negligible at 1 nV/√Hz) which is quite bad (tons of precision amplifiers are <5).

    Why have an output stage with that high noise when it seems 'easy' to make amplifiers with much less. Are they assuming I'm going to use it at high gains (I'm not, I want a high impedance differential buffer and inst amps with integrated resistors are a cost effective way to buy that)?

    Am I missing something dumb?
    Just look at the datasheet. The AD8429 is based on the typical three-amp circuit, so it basically has two cascaded amp stages. When the gain of the first stage is high, its noise dominates the overall noise performance. But if it's gain is low, then the second stage's noise will degrade overall performance. Meeting your requirement (high input impedance, unity gain difference amp) is tricky with this circuit, since the resistors used in the second stage will add significant noise.



--[[ ]]--