• State Resolved
  • Locked Locked
  • Replies 6 replies
  • Answers 1 answer
  • Subscribers 47 subscribers
  • Views 2623 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

OPA827 CMRR test bench in Tina-TI - changing R values change CMRR BW?

RER_Consulting
Prodigy 140 points

I'm trying to do some sims of CMRR of the OPA827 on Tina-TI.  Below is the topology that I'm using to test the CMRR of the SPICE model:

I first do a sim with all R's set to 10K.  Looks nice, pretty close to datasheet:

Then, I change all R's to 100K.  Looks like this:

I was not expecting such a drastic change.  The BW is much different using 100k's instead of 10k's.

Is that right?  Or, could there be something wrong with the model?

If anything, I would have thought that changing from 10k's to 100k's would make things worse.  I'm thinking that because the larger R could create a bigger V difference because of input currents.  Or, if anything, I would think that larger R's would make the BW worse if there is any input capacitance in the model.  Regardless, I can't think of many practical situations where larger R's result in wider BW.

Thoughts?  Does this make sense?  Something wrong with the model?

I'd attach the Tina-TI file, but I can't figure out how to do that ...

  • 0
    TI__Mastermind 27715 points

    Hello,

    I've duplicated the issue you're observing in TINA. Interestingly, increasing or decreasing the value of the feedback resistors by a decade seems to move the dominant pole of the CMRR response by a decade as well. This is not intuitive behavior for a typical op amp. Depending on the input capacitance value, you'd expect a higher resistance to cause a pole at lower frequency, pulling the bandwidth of the CMRR measurement down.

    Unfortunately, this model was built by a third party contractor so I don't have access to the schematic-level design in order to debug and modify in detail. However, I've noted this issue in our internal tracking system so that when we upgrade the OPA827 SPICE model to our new architecture, the CMRR behavior will be corrected.

    Best regards,

    Ian Williams
    Linear Applications Engineer
    Precision Analog - Op Amps

  • 0 in reply to Ian Williams
    Prodigy 140 points
    Ian,

    Thank you for your response!

    I took a look at a collection of datasheets and models. This is what I've found so far:

    LF CMRR value, 3dB freq
    datasheet sim 10K sim 100K
    ua741 90dB, note 1 95dB, ~1.4kHz 94dB, ~14kHz
    LM7732 90dB, ~40kHz 87dB, ~37kHz 89dB, ~97kHz
    LM6172 116dB, ~1kHz 110dB, ~5.8KHz 110dB, ~6.3kHz

    note 1: uA741 fig 8-x axis markings are messed up


    So, the only sim between 10k/100k where the corner freq held was the LM6172.
    The ua741 and OPA827 sims had an order of magnitude difference.
    The LM7732 was not quite as different at 100K.


    Your answer implies that there should not be a difference of corner freq between 10k and 100k sims. But, the data is all over the map. So, please allow me to follow up.

    So, I guess the question is, should the corner freq be the same for both 10k/100k sims? Or, is there an actual physical difference in some amps where it is correct that there should be a difference (as shown by the sims). For those amps where it is justified to have that difference, I'd like to understand physically what causes that?


    Best regards,

    Robert
  • 0 in reply to RER_Consulting
    Prodigy 140 points

    The formatting on the data wasn't retained after posting.  Allow me to re-post to re-format the data

    ==============================

    Ian,

    Thank you for your response!

    I took a look at a collection of datasheets and models. This is what I've found so far:

    • LF CMRR value, 3dB freq
      • ua741
        • datasheet
          • Low-Freq CMRR 90dB
          • can't ID 3dB freq.  x-axis on fig 8 is messed up
        • 10k sim
          • 95dB LF CMRR
          • 3dB freq ~1.4kHz
        • 100k sim
          • LF CMRR 94dB
          • 3dB freq ~14kHz
      • LM7732
        • datasheet
          • LF CMRR 90dB
          • 3dB freq 40kHz
        • 10k sim
          • LF CMRR 87dB
          • 3dB freq ~37kHz
        • 100k sim
          • LF CMRR 89dB
          • 3dB freq ~97kHz
      • LM6172
        • datasheet
          • LF CMRR 116dB
          • 3dB freq ~1kHz
        • 10k sim
          • LF CMRR 110dB
          • 3dB freq ~5.8KHz
        • 100k sim
          • LF CMRR 110dB
          • 3dB freq ~6.3kHz

    So, the only sim between 10k/100k where the corner freq held was the LM6172.

    The ua741 and OPA827 sims had an order of magnitude difference.

    The LM7732 was not quite as different at 100K.

    Your answer implies that there should not be a difference of corner freq between 10k and 100k sims. But, the data is all over the map. So, please allow me to follow up.

    So, I guess the question is, should the corner freq be the same for both 10k/100k sims? Or, is there an actual physical difference in some amps where it is correct that there should be a difference (as shown by the sims). For those amps where it is justified to have that difference, I'd like to understand physically what causes that?

    Best regards,

    Robert

  • 0 in reply to RER_Consulting
    TI__Mastermind 27715 points

    Hi Robert,

    In a real op amp, there is a complex interaction between the common mode rejection, open loop gain and open loop output impedance, input impedance, and external components. A mismatch on the input capacitors or the external resistors will cause the cutoff of the CMRR response to shift, since the ratio of the offset generated to the common mode voltage applied will change over frequency since the effective input filters have changed.

    However, in our typical op amp model architecture the input capacitance is perfectly matched, so this effect would only happen if you forced a mismatch of your external resistors. You can try this on your simulations and see how the bandwidth of CMRR changes.

    The only time I can think of where you would see a huge change in CMRR with external resistor scaling is if the pole caused by the resistors and the input capacitor is well within the flat region of CMRR. The pole frequency is simply equal to 1/(2*pi*Rcm*Ccm), and usually is located in the hundreds of kHz and above. Most of the time this will just look like a second pole at high frequency that causes roll-off of the response to increase from -20dB/decade (single pole) to -40dB/decade (double pole). Sometimes the rolloff of CMRR is already -40dB/decade for certain amplifier designs and it will increase to -60dB/decade.

    After discussion with my colleagues, we think the behavior observed with the OPA827 model is simply an artifact of how the model was created. We're not familiar with that architecture and unfortunately cannot comment further. However, if you check devices with our architecture (either Green-Lis or Green-Williams-Lis, noted in the netlist inside the op amp symbol), the phenomenon should not occur.

    Best regards,

    Ian Williams

  • 0 in reply to Ian Williams
    TI__Mastermind 27715 points

    By the way,  the devices which typically include Green-Lis or Green-Williams-Lis SPICE models are the precision amplifiers supported on this forum and found on this overview page.

    The other products you've reviewed, such as the ua741, LMP7732, and LM6172, are either legacy products or inherited from National Semiconductor. An "LM" or "LMP" prefix is usually a good indication that the product is a former National Semi device.

    Best,

    Ian Williams

  • 0 in reply to Ian Williams
    Prodigy 140 points

    Ian,

    Thank you for your thoughtful replies!

    Please keep this thread in mind, in case you hear anything from the people who model the devices, and let me know if they have anything more to add ...