Actions
Forum Thread Details
  • State Verified Answer
  • Replies 4 replies
  • Answers 1 answer
  • Subscribers 301 subscribers
  • Views 1466 views
  • Users 0 members are here
Related
See a helpful answer?

Be sure to click 'more' and select 'suggest as answer'!

If you're the thread creator, be sure to click 'more' then 'Verify as Answer'!

Noise simulation with SPICE

jc2048

Does anyone have any experience of simulating noise performance with SPICE?

 

I'm playing around (in Tina-TI) with an old discrete audio preamp design that was originally intended for use with a pickup coil (moving-coil). If I ask for the input-referenced noise, it gives me this

 

image

 

That looks quite plausible, with the 1/f noise in place and a sensible level for the white noise (1 or 2 nV/rtHz - higher than a very good bipolar low-noise op-amp, but still quite respectable). One thing I'm unsure of is the low corner frequency. An OP27 would be around 2.5Hz, and that's very, very good, so a third of a Hertz looks too good to be true, although this design does use a medium-power transistor in a common-base arrangement at the input, which is obviously something the chip people can't do.

 

Would it be reasonable to expect transistors to have sensible noise figures in their models?

  • 0

    Hi Jon,

     

    Interesting question, I don't know the answer, I've never looked closely at BJT SPICE files, and don't know if the models have stayed the same or have improved for newer devices.

    I came across this file, maybe you have already seen it: http://www.eng.auburn.edu/~wilambm/pap/2011/K10147_C011.pdf

    It mentions on page 8 that the corner frequency can be as low as tenths of Hz.

    This thread seems useful:

    https://groups.google.com/forum/#!topic/sci.electronics.cad/NMHuu4KdWUg

    And this SPICE document:

    https://class.ece.uw.edu/cadta/hspice/chapter_14.pdf

    on page 46 indicates the SPICE parameters related to noise (I'm thinking if the SPICE file is available, it could be checked to see if these parameters are present).

  • 0

    If you have AOE (Art of Electronics) third edition there is a lot of good stuff about discrete device noise in there.

     

    If you plotted on a log scale I think the 1/f corner (3dB rise in noise) is nearer 3Hz than 0.33 Hz.

     

    1/f noise "definition" here:

    https://www.analog.com/en/analog-dialogue/articles/understanding-and-eliminating-1-f-noise.html

     

    Spice models are very variable, some good, some bad - no way of telling which you have, other than by comparison with reality.

     

    I've found discrete MOSFET and transistor models to be worse than IC models.

     

    My only real life use of small power transistors for low noise was very poor on 1/f corner - but OK for us at 10kHz.

     

    MK

  • 0 in reply to michaelkellett

    Thanks Michael, that's just what I was after. I did wonder if maybe, because the transistor models

    are a part of SPICE, manufacturers might feel obliged to populate the various parameters

    accurately, but it's like all the rest of simulation: only as good as your models, so beware!

     

    The Tina help warns that op amp models may be simplified or not have any noise modelling at all,

    but so far I'm finding that models I get from manufacturer's do look to have reasonable modelling.

     

    Here's a selection, along with the discrete design I started with. [This time with a log scale. Why

    did I think I could judge a corner frequency from a linear scale and a constantly changing curve?]

     

     

    The OP27 is a bit off from the datasheet, but maybe the simulation model is biased more towards

    worst-case rather than the 'headline' typical figures they like to show on the datasheet. The CMOS

    OPA192 would be expected to have worse low frequency noise than the bipolar parts, but they've

    obviously done a lot of work to get a good wideband figure and the OP27 is a good bit better than

    an OP177 [which is fair enough, given what they're being pitched for and the price difference].

     

    image

     

    Even without knowing much about all this, I suspect the discrete design is a bit optimistic when

    you see it against an OP27.

     

    Although simulation might not always be that useful for actual design, it's very good for

    educational purposes. There are two curves for the discrete design because there's a switch to

    change the gain. That changes the resistance seen by the base of the transistor and that's the

    result. Ouch!

     

    I might try building the discrete circuit as it's so simple, and do some measurements on it. I

    don't have the test equipment to do that well, but I could make a stab at the low frequency noise.

  • 0 in reply to shabaz

    Thanks Shabaz, some good, helpful material there. Back to school for me!