• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Answers 1 answer
  • Subscribers 40 subscribers
  • Views 669 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.

WEBENCH® Tools/LM2735: Suggested output capacitor gives excessive output ripple

Richard Price
Intellectual 280 points
Part Number: LM2735

Tool/software: WEBENCH® Design Tools

My shared design webench.ti.com/.../SDP.cgi uses the default components now chosen by WEBENCH for the specified input and output requirements. The tabulated results show an output ripple voltage of 11.06mV peak-to-peak. However, the steady-state simulation shows a much larger ripple:

This ripple is consistent with the voltage dropped across the ESR of Cout (250mΩ) by the capacitor current, which suggest two things:

  • The default 47μF tantalum capacitor suggested by WEBENCH is a really poor choice for this design.
  • The calculation of output ripple in the tabulated data is failing to take account of ESR.

In a previous thread "Inconsistent options for alternate parts" (https://e2e.ti.com/support/tools/sim-hw-system-design/f/234/t/779121) Amod explained that the tool imposes constraints on Cout to ensure stability, and said "With the added ESR of an electrolytic/tantalum cap, you might be able to do with smaller values of capacitor for Cout. If you are looking to use a ceramic cap, you may need to bump up the cap value to push the dominant pole lower in frequency". He shared a design webench.ti.com/.../SDP.cgi that makes Cout a 47μF ceramic, quantity 2 (derated to 39μF total with voltage bias) with ESR of 2.1mΩ each. For this design WEBENCH predicts an output ripple of 893.65μV peak-to-peak and the steady-state simulation looks like this:

This looks much more like the triangular wave ripple that I'd expect, and the amplitude is now about 1.7mV peak-to-peak which is much smaller than for the first example above but still almost twice the tabulated value. I think this suggests that ceramic capacitors are a much better choice, as suggested in the LM2735 datasheet and mentioned in the previous forum thread. Amod's reply there said "the algorithm picks a cap that can ensure stable design operation and keep ripple and transient behavior acceptable", but I believe that because WEBENCH is calculating ripple incorrectly (due to ignoring the ESR) the algorithm is picking the wrong capacitor, a tantalum by default.

  • 0
    TI__Mastermind 21755 points
    Hi Richard,

    Thanks for the detailed post. We are looking into an update in the tool (2.a and 2.b points from the other post's last reply by me - e2e.ti.com/.../779121) that will use ceramic capacitors as the default option. This will reduce the ripple values that you are observing due to a tantalum cap being used. And, in addition, we also are correcting the Vout-pp equation to be more accurate and in close alignment with the SPICE model values you observe. We will keep you updated on this fix.

    Regards,
    Amod
  • 0 in reply to Amod Vaze
    TI__Expert 3320 points
    Hi Richard,
    We are correcting the Vout ripple equation and will be available on ti.com by next week. We will update you once the change is on ti.com. Thanks for your patience.

    Regards,
    Bhushan
  • 0 in reply to Bhushan Waghmare
    TI__Expert 3320 points
    Hi Richard,
    We have corrected the Vout ripple equation for LM2735 and now fix is available on ti.com Thanks for your patience.

    Regards,
    Bhushan
  • 0 in reply to Bhushan Waghmare
    Intellectual 280 points

    Bhushan,

    Thanks for the update. I've tested this by opening my previously shared design from the link in the first post above (https://webench.ti.com/appinfo/webench/scripts/SDP.cgi?ID=8C3DB089644B56E2), which initially shows Vout p-p 11.06mV i.e. unchanged. Clicking on Recalculate changes this to 47.26mV, confirming that the ripple calculation has been updated. Changing Vin from 4.38V (the default, minimum value) to 4.815V (the average value, used by the simulations) changes this only slightly, to 42.68mV.

    I've run a new steady state simulation, which looks much the same as before:

    Including the spikes the total ripple is about 110mV peak to peak. This is still significantly different from the tabulated result, which seems to correspond reasonably well with the roughly rectangular waveform excluding all spikes, the largest of which (the negative-going one) appears to be due to the diode reverse recovery. That's a real effect shown up by simulation, but I guess we shouldn't expect the WEBENCH calculation for the tabulated data to take it into account?

    Incidentally, the above screenshot shows a new problem with the time axis, where the last label value of 0.012ms is rounded to 0.01ms and is therefore indistinguishable from the previous label. The chart in my original post was fine, because the labels were in µs. (See also another thread about a similar but different problem: Time axis labels misleading during simulation.)

    Regards,

    Richard

  • 0 in reply to Richard Price
    TI__Expert 3320 points

    Hi Richard,

    You are correct. Tabular calculations does not considers the spikes in Vout for ripple calculations. We are looking into time axis label issue. We will keep you updated about the same in below post. 

    https://e2e.ti.com/support/tools/sim-hw-system-design/f/234/t/782208

    Regards,

    Bhushan