Post Go back to editing

Noise performance comparison between the OP27 and OP470

Good afternoon,

Some background on why I care about this:  NASA GSFC launched a mission (ASTRO-H) a while back, and the front-end amplifier designs incorporated the OP470 in virtually all of the bias control.  However, the engineer at the time claimed that the OP470 spice model was not available, so he ran worst case analyses using the OPO-27 since they seem to be very close in performance to each other.  We are very interested in the low noise performance of this device for our application.

I am reviewing these analyses and managed to find a SPICE model for the 470.  When I run a noise analysis, I am seeing some different behavior between the OP27 and OP470.   

I've attached a brief write-up on what I've done with some results.  Are these results consistent with how the device actually works, or am I seeing a model issue here? 

Thank you very much for your time, please contact me if you have any questions on how I've done the analysis.  (I welcome all darts thrown my way...)

Dave Durachka

QuestionforAnalogDevices.docx
  • Hi Dave,

    This was very interesting to know . Let us verify what you're observing and we'll get back to you.

    Best Regards,

    Goz

  • Hi Dave, 

    Thanks for the information! It is always interesting to find where our parts go into, and especially interesting if it's somewhere as far as space

    You are correct in your assumption -- the OP470 does not have the noise modeled correctly. I will take a look at the model and see if I can fix that. However, as the model is very, very old, this may take some time. 

    I would like to point out that the thermal noise of the resistors will add to the total noise of your circuit. For example, the 301k resistor would give you about 69nV/rtHz of noise, so you might not be getting the advantage of the low noise of your amplifier. May I ask what sort of performance you want to simulate?

    Thanks, 

    Kris 

  • Dave:

    It looks as if your circuit actually uses the heavily decoupled (ft = 48Hz) "Bias_Out" as its output, not the buffer's "output" you appear to be simulating.  Why not simulate the output the circuit actually uses?

    Two features of your "noise" plots for the OP470 make me suspicious - the large shelf in noise below ~2kHz, and the I/F corner apparently below 1Hz.  I don't see anything in the circuit time constants around 2kHz, so where is that noise enhancement coming from?  Nothing like that shows up in the data sheet, and it would be anomalous if it did.  And from the data sheets, the OP470 I/F corner is at 5Hz, versus 2.7 Hz for the OP-27; the noise outputs of both are otherwise pretty similar, for example the 0.1Hz - 10Hz photographs. So the OP470 I/F corner should be at a higher frequency than that of the OP-27, not smaller.

    To me this looks like a simulation problem.

    I'm curious as to what signal you're actually buffering - DC, AC, frequency - and whether there is significant loading at "Bias_Out".  Using the dual feedback path of R5 to "Bias_Out" and R4, C2 directly to the buffer output implies there is loading; R5 assures accuracy at low frequencies, and R4, C2 allow the loop to close at higher frequencies so that the buffer does not oscillate.

    And what noise levels in what frequency band are troublesome?