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.

OPA847: recommended compensation circuit for OPA847 - low gain, non-inverting

Part Number: OPA847
Other Parts Discussed in Thread: THS3491, OPA695, LMH6702, OPA3695

Dear Sir or Madam,

 I would like to use the OPA847 in non-inverting configuration at a gain of 2. Is there a recommended external compensation circuit for this configuration? 

Best Regards


  • Hi Uros,

    it's not recommended to run the OPA847 with a gain of 2V/V (in non-inverting configuration). This may work in an inverting configuration because then the -input sits on virtual ground. But this is not the case in a non-inverting configuration.

    I would choose an OPAmp which is unity-gain stable or at least needing a minimum gain of 2V/V.


  • Dear Kai,

    Thank you for your answer. the problem is that an inverting configuration has a higher output total noise (assuming the 50 ohm input matching).

    The problem is finding an op amp which can do 4 Vpp at the output, has an output total noise (with all the resistors included) lower than 3 nV/sqrt(Hz) , works with 60 MHz sine input, is matched to 50 ohm and is stable. I could not find anything suitable in the TI catalogue. I realize this might go beyond your duties but any hint which op amp to choose would be greatly appreciated. 



  • Do you need to be DC coupled? That 4Vpp at 60Mhz will ask for 750V/usec slew rate, for any kind of HD needs, you would need much more than that in the op amp. 

  • Hi Michael. it would have to run in both modes. DC and AC coupled. yes, I expected a few thousands V/us in slew rate are needed. 

  • I am sure you are aware of how impossible your targets are - but, maybe can get close. 

    1. Your noise targets imply a decompensated VFA

    2. Your slew rate needs imply a current feedback output stage

    When I need this I go to a composite structure where I essentially replace the VFA output stage with a CFA stage. 

    1. To shif the slew rate burden to the CFA, I operate that new output stage (inside the loop) at some gain, -4V/V here. 

    2. To get the highest slew rate and shortest prop delay inside the loop, I operate the CFA inverting. 

    3. Since I now have an inverting output stage inside the loop, I flip the feedback connection around to the V+ node for the VFA

    4. I have essentially gained up the intrinsic GBP for the VFA by a factor 4 here - and, its min stable gain number. 

    The final trick is an inverting compenstation method I developed many years ago and recently republished on EDN. inserted here.

    This will give you the noise target at low F, but for stability these caps are shaping the noise gain up (I targeted 30V/V at high F) to get good phase margin at LG xover. Closer but probably not what you want. 

    The frequency response looks pretty good actually (and you cannot do this non-inverting - I was guessing a 50ohm source to a 50ohm inverting input Rg with 100ohm feedback). 

    The spot noise does what I would expect, but 3.3nV at low frequencies, 19nV at 60MHz

    The CFA is driving a pretty heavy load here, 100ohm feedback with 200ohm load. With +/-2V output, might be ok with the OPA695 - I chose it for the highest slew rate, not the highest output current - the THS3491 is the most lunatic fringe part these days on both specs, but cannot operate down to +/-5V supplies. 

    And this file, 

    OPA846 to OPA695 composite.TSC

  • Thank you Michael for all the material and for taking the time to explain it. Gives me quite some subjects to read about.

    I realize the specs are pushing it on all fronts but I have to see to which parameter do I have to give some slack and which can take it a bit more...  

    However the 19nV/sqrt(Hz) at 60 MHz is too much. Noise is a hard limit. 

    I guess a LMH6702 in non-inverting configuration would do a better job (app. 5 nV/sqrt(Hz)) and meeting all the specs or?  

  • the other important point (which I did not mention, sorry about this) is that I have to parallelize (drive multiple from a single 50 ohm source) many of these stages. With an non-inverting configuration I have the high Z input into each of them which helps (matching over an R to GND)..when using the inverting I have the low impedance (because of the low ohm resistors -> noise)  Rg which is a problem for parallelizing them...

  • Oh well yes Gruffalo, if you need to drive parallel V+ inputs need to go a different direction, 

    Long time ago, we were pursuing how to push CFA down in Noise and HD. OA-22 is an app note I wrote at Comlinear exploring one way to do that, here is that circuit, parallel feedback networks at low F, transitioned out by a coupling inductor, 

    Page 9 in this,

    The output noise has a large inverting current noise term that this reduces, here is that spot noise sim, about 7nV at 60Mhz, 

    The OPA695 has a triple version, OPA3695, and you can of course adapt this to the LMH6702. 

    Gain of 2 parallel feedback.TSC

  • And then of course if cost and complexity is no problem, paralleling stages will also reduce noise, 

    Here are 4 of those stages combined at the output through 50ohm R (you never said your load, I used 200ohm here), still a pretty good response shape, 

    And some improvement in noise at the load, 4.6nV midband, 4.9nV at 60Mhz. 

    Gain of 2 parallel feedback parallel stages.TSC

    And then if I drop the LMH6702 into this same design, there you go, 3nV noise at the load, So maybe not so impossible after all, 

    And the frequency response looks really good, 

    Gain of 2 parallel feedback parallel stages LMH6702.TSC

  • A brief explanation how paralleling of OPAmps can decrease the total noise can be seen in the below figure:


  • wow! thank you for all the infos to both of you Michael and Kai. thank you for proposing so many new ideas an approaches. I need some time to go over it. I will close the thread.  

    Many thanks for your help!