• State Resolved
  • Locked Locked
  • Replies 7 replies
  • Subscribers 47 subscribers
  • Views 1176 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.

OPA4379 introducing extra harmonic into feedback loop

CHEE H. HOW
Prodigy 90 points

Other Parts Discussed in Thread: OPA4379

1007.PB_FA1-001.pdf

In the circuit we are using OPA4379 to close the feedback path of a switching regulator. Among about 40 units one of them is suffering from pulse width jitter issue. Our debugging work confirmed that OPA4379 is the root cause to the problem.

The ultimate goal is to confirm whether OPA4379 is suitable for our application from mass production point of view, we are seeking answer for the following questions:

  1. Which parameters of the NG IC responsible to the malfunction of our circuit.
  2. How is this parameter of the NG IC difference from the Good samples? Is it still within specification?
  3. For item 2 above what is its max, average, and max values from specification point of view and how is the actual mass production distribution look like?

Please check the attached report (PB_FA1-001.pdf) for detailed experiment data and waveform.

  • 0
    Intellectual 860 points

    I'll go out on a limb and make a guess...

    I'm a little puzzled as to why one part oscillates and another doesn't, but looking at your circuit I would be concerned with the large resistor values interacting with the input capacitance of the amplifier.  Ignoring the differential mode input cap, 1 MEG feedback with 6 pF CM input cap yields a zero at ~26kHz.  This may be enough to reduce the phase margin at the intersection of the noise gain and open loop gain curves to where you have ringing or oscillation.  

    Try either shorting the feedback resistors R515 and R528 or cutting all of the resistor values by a factor of 10 and see if the problem goes away. 

    Nice report btw.

    Best of luck.

    Tim

  • 0 in reply to Tim Sobering
    Prodigy 90 points

    Tim,

    Thanks for your reply. The feedback resistor is set high to minimize the consumption therefore it need an extra buffer.

    Based on your suggestion I have tried the following combinations.

    (a) R513=1M, R514=120K, R528=1M    Good IC = Stable                   BAD IC = Jitter

    (b) R513=1M, R514=120k, R528=100k,     Good IC = Stable              BAD IC = Stable

    (c) R513=10M, R514=1.2M, R528=100k, Good IC = Stable               BAD IC = Stable.

    The conclusion is R528 value for being too large is the culprit.

    I will stick with option (c) with the slight increase of error voltage due to offset current.     

    Thanks.

    How

  • 0 in reply to CHEE H. HOW
    Intellectual 860 points

    You might want to look at Technotes 7 and 9 at http://www.k-state.edu/ksuedl/publications.htm.  The issue is most likely the zero in the noise gain resulting from the feedback resistor and the opamp input capacitance.  You can probably fix the problem with just a cap across the feedback resistor.  This does limit the bandwidth of the amplifier (you didn't state the bandwidth needed for your application).

    Glad it helped.

    Tim

  • 0 in reply to Tim Sobering
    Intellectual 860 points

    I went back and looked at your original document and the mystery to me is why R515 doesn't seem to play a role.  True, dual opamps are not necessarily matched in specifications, but are the loads different on U503A and U503B?  Another possibility is that there is something in the layout of the circuit that is resulting in a higher parasitic capacitance on pin 6 than on pin 2.  You might take a look at the comment regarding removal of the ground plane under the inverting node under:

    http://e2e.ti.com/support/amplifiers/precision_amplifiers/f/14/t/364481.aspx

    Tim

  • 0 in reply to Tim Sobering
    Prodigy 90 points

    Tim,

    Thanks for the introduction of the Technotes they are well written and I am sure it will be a good resource for me to gain extra op-amp’s knowledge.

    The system bandwidth is about 7kHz, it should have enough room for Cf.

    R515? I don’t really catch what you mean here but I try to answer:

    Yes U503A and U503B are supplying different loads. During the first prototype only one buffer amp was used its output was supplying two different loads one to error amplifier of PWM and the other to an ADC. U503B is more critical since it involve feedback loop whereas U503A is just for mainly for output voltage monitoring. But the S&H (sampling & hold) activity of ADC is injecting charge back during its operation this caused serious noise problem to the PWM controller, therefore two buffer amplifiers are used in current version.

    R515 is to match up with R513//R514 so that the result voltage due to input biasing current is balancing up for both inputs. 

    Ground Plane – Yes there is ground plane underneath the inverting node of both amps. Their inverting node’s trace is kept very short.

    The PCB layout can’t be changed at this moment therefore removing ground plane underneath inverting inputs and introduction of small feedback capacitor can’t be done in real application until there is opportunity for layout modification.

    Here is the test result with Cf

    (0) R513=10M, R514=1.2M, R515=1M (original condition)                              BAD IC = Jitter ,  Good IC = Stable

    (1) R513=10M, R514=1.2M, R515=1M, Cf=5pF (Cf in parallel with R515)         BAD IC = Stable, Good IC = Stable

    (2) R513=10M, R514=1.2M, R515=1M, Cf=10pF                                             BAD IC = Stable, Good IC = Stable

    (3) R513=10M, R514=1.2M, R515=1M, Cf=2.5pF                                            BAD IC = Stable, Good IC = Stable

    10pF is more common so preference is 10pF.  

    In here I attached U503 Op-amp’s (BAD IC) output voltage waveform with (Fig 1(b)) and without (Fig. 1(a)) Cf (10pF). Obviously the oscillation wave dismissed with the inclusion of Cf.

    The criticality of Cf with large value Rf – I learnt the lesson.

    Thanks again Tim.1323.U503B,7.pptx

  • 0 in reply to CHEE H. HOW
    Prodigy 90 points

    Sorry for item (0) ~ (3) it should be R528 instead of R515 here is the correction

    (0) R513=10M, R514=1.2M, R528=1M (original condition)                             BAD IC = Jitter ,  Good IC = Stable

    (1) R513=10M, R514=1.2M, R528=1M, Cf=5pF (Cf in parallel with R528)      BAD IC = Stable, Good IC = Stable

    (2) R513=10M, R514=1.2M, R528=1M, Cf=10pF                                            BAD IC = Stable, Good IC = Stable

    (3) R513=10M, R514=1.2M, R528=1M, Cf=2.5pF                                           BAD IC = Stable, Good IC = Stable

  • 0 in reply to CHEE H. HOW
    Intellectual 860 points

    How,

    I'm glad this was useful.  The reason I asked about the opamp loads is that load impedance can also cause a zero in the noise gain (by interacting with the opamp output impedance, which increases with frequency).  So with two identical circuits, one might oscillate if it is driving a heavier capacitive load.  

    If you think this is fixed, please mark the thread as solved.

    Best of luck!

    Tim