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.

OPA828: Gain error of the voltage follower with OPA828

Part Number: OPA828
Other Parts Discussed in Thread: OPA827, TINA-TI

Dear Ladies and Gentlemen,

What is the expected gain error at 100 kHZ and 1 MHz of the simple voltage follower with OPA828 op-amp? I simulated it with Spice and got 60 ppm at 100 kHz and 5700 ppm at 1 MHz. Very high...

For comparison, for the voltage follower with OPA827 I got 3 ppm and 285 ppm, respectively.

Is really the gain error of the  OPA828 worse than OPA827 in this application?

The gain error is defined as 1E^6*(Uout-Uin)/Uin, where Uout and Uin are magnitudes of the output and input voltage of the follower, respectively. The supply voltage was -15V/+15V.

Thank you and best regards,

Marian

  • I simulated in Spice the complex open loop gain of both amps. At 10 MHz the phase shift is ca. -98 deg for OPA827 and -220 deg  (!) for OPA828.

    The open loop gain characteristics presented on the 1st page of the OPA828 datasheet show approximately -95 deg at 10MHz.

    Something is wrong with the Spice model or I am doing something wrong?

    Regards,

    Marian

  • Hi Marian,

    TINA-TI shows a deviation of only 44ppm at 1MHz:

    marian_opa828.TSC

    Kai

  • Hi Kai,

    That's a good news.Thank you!

    I am using LTSpice and your Spice model. Could you advice how to correct this model?

    Best regards,

    Marian

  • Hi Marian,

    I am using LTSpice and your Spice model. Could you advice how to correct this model?

    I do not see issues with OPAx828.LIB that I downloaded from the following website. 

    https://www.ti.com/product/OPA828#design-development##design-tools-simulation

    I duplicated the same simulation as Kai's schematic in LTSpice, the measurement is exactly identical at 1.0MHz. 

    At 1.0MHz, LTSpice is measured -380.78144udB or 10^(-380.78144udB/20)=0.999956161 in gain, which is 43.84ppm in deviation. 

    If you need further assistant, please let me know.

    Best,

    Raymond

  • Dear Raymond,

    Thank you!

    I found the error in my simulation: the .options cshunt = 1e-14 was the reason.

    But I see another problem: what gain error at low frequencies (<100 Hz) do you get? I am getting 1.7 ppm even when using the alternate solver.

    Best regards,

    Marian

  • Hi Marian,

    I run the same simulation from 1Hz to 10MHz (LTSpice), the frequency response is "flat" from 1Hz to 400kHz. 

    For example, at 10Hz, the gain is measured at 14.570343udB or 1.000001677 or 1.678ppm, which is the same as yours. 

    If you have additional questions, please let us know.

    Best,

    Raymond

  • Hi Raymond,

    Thank you.

    Where does this ca. 1. 7 ppm come from? It should be zero. I played with various controls of the LTSpice simulator and it is still there.

    We are designing very precise measuring instruments so this 1.7 ppm is high for us.

    Best regards,

    Marian

  • Hi Marian,

    It is probably related to Aol and internal bias current inside of op amp design. What are your gain errors requirements per your application? Please provide us the frequency operating range, gains, load requirements and operating temperature that are used in the application.  

    Best,

    Raymond

  • Hi Raymond,

    The gain error of the follower should be within +-1 ppm from DC to 100 kHz at ambient temperature 20 - 40 deg C and 10 kohm load.

    Best regards,

    Marian

  • Hi Marian,

    1.7ppm is too high an error for you? :-)

    1ppm is 0.0001%. Even ultra-ultra-precision foil resistors offer a TCR of more than 0.14 ppm/°C, a tolerance of more than ±0.005% and a long-term stability (1 year) more than 25 ppm.

    Kai

  • Dear Kai,

    AC voltage standards, which we had developed have AC-DC transfer difference below 1 ppm between ca. 100 Hz - 10 kHz.

    We are able to compare them with ca. 0.3 ppm standard uncertainty,

    Please take a look at my publications in IEEE Transactions on Instrumentation and Measurement: ieeexplore.ieee.org/.../searchresult.jsp

    Anyway, the voltage follower circuit does not contain any external resistor.

    Moreoover, with ca. 140 dB open loop gain at DC the OPA828 should in theory have sub-ppm gain error at frequency well below the first pole, but it shows there the 1.7 ppm.

    If the reason of the 1.7 ppm is the CMRR then it should somehow depend on frequency, but it does not.

    And it is AC analysis, so the op-amp offset voltage and other DC parameters should not affect the AC gain error.

    Maybe LTSpice is giving bad results. Maybe...

    Regards,

    Marian

  • Hi Marian,

    Q: Where does this ca. 1. 7 ppm come from? It should be zero.

    Here is the answer to the prior question. 

    Why Gain Error flattens around 50Hz and stays at 1.7uV (-115dB) despite the fact that the OPA828 DC AOL goes all the way to 140dB (Gain Error of 100nV).  The answer lies in the fact that in a buffer configuration the gain error will be controlled by worst of the AOL or CMRR responses – see below red composite of AOL and CMRR.  For higher frequency above 50Hz the error is controlled by the AOL roll-off but for frequency below 50Hz lower DC CMRR of 115dB takes over limiting the gain error to said 1.7ppm value.

    Best,

    Raymond

  • Hi Raymond,

    This explains the error. Indeed, the simulation model includes the CMRR vs. frequency  effect.

    The cure could be bootstrapping the power supply of the op-amp so the ground would follow the input signal. I will check this.

    Thank you very much.

    Best regards,

    Marian

  • Anyway, the voltage follower circuit does not contain any external resistor.

    :-)