Post Go back to editing

LT1464 is there a flaw in the LTspice model?

I am comparing Analog/Linear op amps (along with some by Brand X) using the NOISE analysis of LTspice. The context is: single supply via 9V battery, high source impedance, low bandwidth needed (-3dB at 20KHz would be fine). 

When I test the LT1464 against other models such as LT1006 and LT1112 I find that its noise at 1K is higher than either part and higher than the specs suggest to me. (Yes I know the LT1464 is dual and the others single).

Compared to LT1006, LT1464 voltage noise density is similar and current noise density is MUCH lower in the LT1464 (at 0.4fA), yet noise is slightly higher as shown below.

Based on my reading of several excellent Analog papers on op amp noise I expected that current noise density would dominate because of the 200K source resistance, at least until current noise density of the amp is VERY low, as with the LT1464. And if current noise is sufficiently low then resistor noise would dominate, which I have eliminated by the 'noiseless' spec. Yet if op amp voltage noise becomes the dominant factor then why are LT1006 and LT1112 very close in LTspice-reported noise when LT1112 voltage noise density at 1K is almost half that of LT1006 (22nV vs 14nV)?

These are the analyses for the tests:

https://musicsystems.com/images/Analog_LT1464.jpg

and

https://musicsystems.com/images/Analog_LT1006.jpg

I am very eager to understand these apparent contradictions.

Thanks,
JayGunn

  • I don't see any way to cancel or delete this, even though I am signed in as the guy who submitted it.

    I looked into the noise levels and found that the results from LTspice are reasonable in proportion to each other and to the values in the analysis at the 1K point.

    Rin and Rf in parallel (net 100K) give 40.5nV/rootHz of Johnson noise, which I left out of the analysis since this noise is constant for every op amp. 

    When I calculate noise voltage attributable to the op amp current noise density, in the case of most op amps I am evaluating, it hardly matters what the number is when the op amp has JFET input transistors. It is swamped by the dominant noise which (in my analysis) is the op amp voltage noise.

    Since uncorrelated noises are combined via root sums of squares, small noise values (any value 1/4 or smaller than the largest value) have almost no effect.

    In the case of the LT1006, which does have a non-negligible current noise (giving 12nV/rootHz with a 200K ohm source impedance) the op amp voltage noise of 22nV and the voltage of the current noise of 12nV sum to 25nV.

    In the case of the LT1464, the op amp voltage noise is 24nV and the voltage of the (tiny) current noise is 0.08nV, so the 2 noises sum to 24nV, virtually the same.

    Just bothering to do the math would have answered my questions, so I feel embarrassed that I asked before I checked my assumptions.

    Best regards and PLEASE DELETE THIS!

    John