• State Resolved
  • Locked Locked
  • Replies 8 replies
  • Subscribers 49 subscribers
  • Views 327 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

OPA4192: Output noise from , PSRR graph

Yusuke Mukuno
Expert 7395 points
Part Number: OPA4192
Other Parts Discussed in Thread: OPA4387, OPA4376

Hi Expert,

Customer found output noise issue with OPA4192.

Circuit : voltage follower.

V+=5V, V-=0V

This noise seems to come from power supply noise(1MHz). It has same frequency with 5V rail switching ripple.

So it is depend on PSRR performance.

But, they didn't find this issue in case of OPA4387(same circuit)

They checked the PSRR graph in datasheet. PSRR of Both device is around 22dB at 1MHz.

Is there any spec to cause of the output noise that is same freqency with power supply ripple other than PSRR?

Also, PSRR graph have different test condition between OPA4192 and OPA4387.

OPA4192  : +/-18V

OPA4387 : +/-2.75V

Does it affect PSRR performance? Also do you have any PSRR data at V+=5V, V-=0V?

Thanks

Mukuno

  • 0
    Guru 140200 points

    Hi Yusuke,

    DC/DC switcher noise coming from the supply voltage can enter a circuit by many many ways. One hidden and nasty way is the injection of switcher noise via the decoupling caps directly into signal ground. And with an imporper signal ground (omitting of solid ground plane) the switcher noise will be everywhere in the circuit then. 

    It's no good idea to make the OPAmp responsible for this issue. The DC/DC switcher is the wrongdoer here. And you urgently need to remove the switcher noise from the supply voltage by using LRC-filtering at the output of switcher. Pi-filter are very effective.

    Keep in mind that the supply voltage of an OPAmp must always be clean, stable and noise-free, because even with a superb PSRR of OPAmp supply voltage ripple will always enter the circuit.

    Kai

  • +1 in reply to kai klaas69
    TI__Guru* 93105 points

    Mukuno-san,

    Kai has very nicely summarized the switcher noise problem and it is the issue that needs to be addressed in the application.

    22 dB of PSRR at 1 MHz fundamental frequency provides only about 1/100th reduction of the noise amplitude. However, a switching waveform contains many higher frequency harmonics, and the op amp PSRR falls off further with each successive harmonic. After a few harmonics there is no PSRR remaining and the noise couples directly through the op amp output without any attenuation. 

    Regards, Thomas

    Precision Amplifiers Applications Engineering

  • 0 in reply to Thomas Kuehl
    TI__Expert 7395 points

    Hi Thomas-san, Kai-san,

    I understand your comments.

    My customer question is,

    - OPA4192 have the issue.

    - OPA4387 don&t have the issue

    PSRR of OPA4387 is also around 22dB at 1MHz. (from datasheet)

    Is there any spec to cause of the output noise that is same freqency with power supply ripple other than PSRR?

    Also, PSRR graph have different test condition between OPA4192 and OPA4387.

    OPA4192  : +/-18V

    OPA4387 : +/-2.75V

    Does it affect PSRR performance? Also do you have any PSRR data at V+=5V, V-=0V?

    Thanks

    Muk

  • +1 in reply to Yusuke Mukuno
    Guru 140200 points

    Hi Yusuke,

    the OPA4192 is much faster than the OPA4387 and it does not have the PSRR specified at frequencies above 1MHz. So we do not know whether both OPAmp suppress the ripple at frequencies above 1MHz equally good. Thomas already mentioned that the harmonics of the 1MHz ripple can be the issue here.

    For a better understanding we would need to see a scope plot of the ripple and analyze whether there are far reaching harmonics. If this is the case then the slower OPA4387 may better "seem" to suppress the supply voltage ripple.

    But again, only because you see a lower supply voltage ripple at the output of OPA4387 doesn't mean that the supply voltage ripple is less harmful with the OPA4387. As I already told, the supply voltage must be clean, stable and noise-free. Supply voltage noise can cause many unwanted issues which can totally ruin the precision of a circuit. Why? Because at very high frequencies the ripple seen at the supply voltage pins is no longer treated linearily by the OPAmp but very unwanted things can happen, like demodulation of supply voltage ripple or intermodulation effects. Demodulation and intermodulation can generate low frequency artefacts like unwanted and unexplainable, slowly changing offset voltages. Believe me, you don't want to have this in a precision circuit.

    So, as I already told, the issue here is not the OPAmp but the supply voltage ripple. If you see supply voltage ripple at the output of an OPAmp in a precision application, you urgently need to suppress the supply voltage ripple appearing at the supply voltage pins of OPAmp by the help of LRC low pass filtering, best by the help of Pi-filters which are mounted directly at the output of supply oltage source.

    Changing the OPAmp will usually not cure the situation but only hide the impact of supply voltage ripple.

    Kai

  • +1 in reply to kai klaas69
    TI__Guru* 93105 points

    Mukuno-san,

    Attempting to use the op amp PSRR is not the solution to this issue because of its limited effective frequency range, and the wide frequency content of the switcher noise. The solution needs to be external to the op amp used.

    There are two external approaches that can be considered. One is adding an RF choke in series with each supply line to the op amp supply pins. The RF choke is added right at the op amp power supply pins. The choke in conjunction with the op amp power supply decoupling capacitor creates a second-order LC low-pass filter. The lower its cutoff frequency is made, the higher the attenuation of the power supply noise will be.

    Alternately, low-noise three-terminal linear voltage regulators can be added to each supply pin. Three-terminal linear voltage regulators can exhibit high PSRR at higher frequencies than that of the op amp being used used. Linear voltage regulators do require a higher input voltage than the output voltage and must be selected for minimum drop out voltage characteristics in addition to best PSRR performance. The Linear Voltage Regulator e2e forum could help you with the best suited product for the application.

    Regards, Thomas

    Precision Amplifiers Applications Engineering

  • 0 in reply to Thomas Kuehl
    TI__Expert 7395 points

    Hi Thomas-san, Kai-san,

    I understand your solution idea.

    <background>

    Customer understand, the root cause come from switching noise of power supply.

    Customer had already used  OPA4376 to released product. The circuit is fixed(they cannot add LDO or filter).

    We are proposing OPA4192 and OPA4387 etc as alternate part, due to supply issue of OPA4376.  Also they have to select alternate part urgently.

    They are evaluating our alternate part. Then, they found this issue.

    I understand, they shoud improve power supply noise for solved root cause.

    But, they would like to know about which parameter of opamp will be affect to this issue.

    Is it PSRR only?

    Also, will the PSRR performance change by V+/V- test condition?  (I understand, this issue is also depend on higher freqency performance than graph range, from your comment)

    It is OK, Yes or No answer. Then, I can discuss about it with customer.

    Thanks

    Muk

  • +1 in reply to Yusuke Mukuno
    TI__Guru* 93105 points

    Mukuno-san,

    Yes, PSRR.

    Precision Amplifiers uses a large supply voltage change when production testing our op amp PSRR. Using a large voltage change is a more rigorous test than when a small voltage change is used. It is essentially a DC test and that is what the datasheet PSRR specification is based upon. 

    Regards, Thomas

    Precision Amplifiers Applications Engineering

  • 0 in reply to Thomas Kuehl
    TI__Expert 7395 points

    Hi Thomas-san,

    Thank you for your answer.