LT1028 noise modelling in LTSPICE

[splin]

I was just having a go at noise modelling in LTSPICE but am rather confused by the results. The circuit and noise results are attached. LTSPICE calculates the .1 to 10Hz noise to be 873,43nV rms which, as expected, is almost alll due to the input noise current in the 10k resistor. It gives the same result if the 1 ohm and 10k are swapped.

However if both resistors are set to 10k the output noise drops to 613.45nV. Can anyone explain why - I was expecting it to increase by approx sqrt(2)?

[Marco]

I suspect the modeled current noise for the two inputs has high correlation (presumably caused by the current source for the differential pair input stage) and thus sums out when the source resistances are equal.

[splin]

Thank you both - I just assumed that they would be uncorrelated.

Ironically I had (skim) read that note on page 11, but it was a long time ago and I didn't really understand it at the time - it now makes sense! Very clever of LTSPICE to model it correctly.

[T3sl4co1l]

A "gotcha for young players", amps with input bias current cancellation.  There's a paragraph about it in AoE3: the input current noise is much higher than the shot noise from the input bias current.  For these, obviously you need to keep Rsrc under 10k or so (preferably under 1k), but much more importantly is to keep the resistances matched.

Tim