By using notes from C.Basso Bulk Capacitor Calculation, I managed to plot the ripple curve for bulk capacitors for offline converters.

This bulk capacitor location:


Assume that my system is AC/DC flyback converter, below is my curve for 150uF bulk capacitor calculated with the input specification.


From this curve, by choosing 150uF as my input bulk capacitor, I understand that at low input voltage, 85Vrms, the capacitor DC voltage will discharge from 120V to 68V (with bulk voltage of 52V) in order to maintain the peak power requirement, 94W. Please correct me if my understanding/statement is wrong.


  1. What is the maximum limit bulk ripple voltage that should we use for calculation? For my example, I'm using 50Vp-p as the ripple. Are this 50Vp-p value is logic for actual application?
  2. What are the impacts of having high ripple at this bulk capacitor during low input?
  3. Example with 150uF bulk capacitor, what will happen to the system if the minimum voltage suddenly goes lower than 68V during 94W peak power operation?
  • \$\begingroup\$ You haven't said anything about what system your rectified power is driving. So its reaction to low voltage input or excessive ripple is hard to say. \$\endgroup\$ – Simon B Jul 1 '20 at 10:56
  • \$\begingroup\$ @SimonB, oh, sorry for that. I just update that information in my question, Basically, the system is a AC/DC flyback converter. It will drive the transformer with specific turns to get desired output voltage. \$\endgroup\$ – Lutz Fi Jul 1 '20 at 11:13
  • \$\begingroup\$ It all depends on the specifications you need. How much is the minimum AC inlet voltage you want to operate, and what is the minimum voltage on the capacitor you want the PWM chip to still be able to provide output voltage within some specification of voltage tolerance or ripple or whatever? The minimum AC frequency might be 47 Hz, or something else. It's more a specification problem, than calculation problem. When you have a specification, you can then calculate things. And still allow for margin and tolerance of components. \$\endgroup\$ – Justme Jul 1 '20 at 11:50

Let me give you an answer since I am the author of the note you gave as a reference:

  1. What is the maximum limit for the bulk ripple: there is no real specification and a rule of thumbs given by Unitrode a while ago was 20-30% of the peak line voltage. Therefore, for a 85-V rms voltage (120 V peak roughly), you would size the capacitor value for a valley voltage of 85 dc for instance. I have seen many commercial designs in which the valley voltage was down to 60 V or so. The reason? A smaller bulk capacitor which drives size and cost down. Let's see a brief simulated example:

\$P_{out}=117\;W\$ , \$C_{bulk}=220\;µF\$ then \$I_{C,rms}=2.2\;A\$ and \$V_{valley}=86\;V\$

Now select a 100-µF capacitor:

\$P_{out}=117\;W\$ , \$C_{bulk}=100\;µF\$ then \$I_{C,rms}=1.9\;A\$ and \$V_{valley}=48\;V\$

So the rms current does not drastically change but the valley voltage does.

  1. What is the impact of this large ripple: there are many issues brought by a low-value bulk capacitor:
  • the valley voltage sets the minimum operating voltage at which the power supply must deliver 100% of its rated power. It is often overlooked by designers but if you plan on designing a 100-W power supply unit (PSU) at a 85-V rms input with a 50-V valley voltage, then the PSU must be dimensioned for an operation at 50 V dc with some margin. So you realize the burden on the power supply (rms currents, conduction losses etc.) for a design at 50 V.
  • with such a low input voltage, the controller will push the duty ratio quite far and you may need a strong compensation ramp in a current-mode-controlled design. This extra ramp component may affect the maximum peak current and require to tweak the sense resistance to still pass power.
  • the worse-case right-half-plane-zero (RHPZ) will be very low, severely hampering the possible crossover frequency which cannot exceed 20-30% of the lowest RHPZ position. This is really an impeding factor.
  • you'll have to really increase the operating peak current in the bulk valley voltage, bringing extra burden on the transformer with saturation risks at high temperature. The real plague is the over-power phenomenon (OPP) that needs to be harnessed for the mandatory limited power-source test (LPS). By increasing the power capability at 85-V rms because of a high ripple voltage, you will pass significantly more power at 265 V rms, forcing you to find a way to fold the peak current limit in high-line conditions. Furthermore, the compensated OPP curve might approach a bell curve with a peak of power in the middle of the input voltage range if you over compensate the converter.
  • there is less energy storage with a small bulk capacitor and any small power drop will pass through and be observed on \$V_{out}\$.
  1. What will happen if the voltage goes lower than 68 V in the example: well, either the PSU will go in protection mode because a) you've hit the maximum peak current limit too long or b) the power supply won't be able to regulate and a lot of ripple will be seen on \$V_{out}\$. If this lasts too long, temperature runaway may also happen with all consequences.

As a conclusion, do not neglect the selection of the bulk capacitor as it is one of the failing components in switching power supplies: too low a rms current capability, too hot, too close to a heatsink etc. Finally, select a serious brand for this capacitor.

  • \$\begingroup\$ I enjoy your thoughts from time to time. They offer unique, important perspectives to pass along --something I've learned to value. I just recently posted here a TL431 model you developed around 2005, along with a 1992 TI macromodel and a 2009 more detailed TI PSpice/AC model. Thanks for the time you offer and for what some of it means, at times, to my memory of Dr. Middlebrook. \$\endgroup\$ – jonk Jul 1 '20 at 13:00
  • \$\begingroup\$ Thank you Mr. C. Basso. Your answers, explanations & materials have never disappointed me. It was very fortunate for me to see you to answer almost all of my questions regarding power supply in this forum. Again, thanks. \$\endgroup\$ – Lutz Fi Jul 1 '20 at 15:34
  • 1
    \$\begingroup\$ @jonk, thank you for the comments on the TL431 model that I made some time ago already. I have a fond memory of Dr. Middlebrook. I met him several times, especially when he started a collaboration with my friends at Intusoft. It was during the time he worked on the general feedback theorem. A really kind person and what a contribution to our power electronics field! \$\endgroup\$ – Verbal Kint Jul 1 '20 at 20:35
  • \$\begingroup\$ @Lutz Fi, I am happy if I can modestly help with my contributions. I remember when I learned switching power supplies 30 years ago without internet and interactive forums like this one. The papers I wanted had to be ordered through the local library and they were costly. I would have liked to chat with experts at the time: all was done via faxes and it was the early beginning of emails. Keep the questions coming, I'll be glad to answer if I can. \$\endgroup\$ – Verbal Kint Jul 1 '20 at 20:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.