I've made this SEPIC regulator based on this design, using a MT3608 boost regulator, which has the same oscillation (ripple) as this question:

SEPIC regulator

I've made these changes:

  • C1=100uF, C2=470uF
  • R2(bottom one which is 2K here) is changed to 1K
  • R2(top one which is 9.1K here) is changed to a 50K pot to give it a variable output
  • C5 is 2x10uF, 0805 caps in parallel
  • input is 5V from a USB wall charger

It oscillates too much. I tried to capture the waveform. I set the output to about 14 volts and used a 1K-3.3K resistor divider to bring it under 5V for my o'scope. The divider is considered a light load of 4.3KOhms then. Here is the waveform:

the output waveform

I marked one cycle which shows a frequency of about 15Hz ripple. The divider lines are 0.5V/div so the voltage shown is ~3.5V which ripples between 3V and 4V; multiplying the ripple by 4.3 (resistor divider) gives almost 4V ripple. (1V ignoring the brief spikes).

After connecting a 100 ohm resistor (the divider still exist for the sake of measuring,) the waveform changes as follows (notice the first half and the second half of waveform in the shown image):

After connecting a 100ohms load

Despite my PCB being worst case, in case of switching regulator designs, why is this ripple so high? Is there any theoretic problem with the regulator or design (e.g. the 0.6V Vref or 1.2MHz switching frequency)?

The PCB (red is top, blue is bottom and both planes are GND):



I changed the parallel 2x10uF caps back to a single 4.7uF. it got a lot more stable and now my multimeter, can show a precise voltage without much ripple under both 4K ohm and 100 ohm load. here is the waveform:

100 ohm load

the periodic noise seems to be from input. as I measured the input from my Wall charger it shows the periodic 100Hz noise. (also, my low-end USB oscilloscope has an 8 bit, ±10V ADC inside, so maybe this tiny ripple is lower than what it seems on the waveform):

5V input waveform

  • 1
    \$\begingroup\$ despite my PCB being worst case Are you saying that you didn't follow the layout suggestions in the datasheet (are there any?) of at least the "known good practices regarding layout of DCDC converters"? Also consider that how you do the grounding of your oscilloscope probe can have a HUGE difference on the measured amount of ripple. There are similar questions with suggestions and answers on this site, go search for those questions. \$\endgroup\$ – Bimpelrekkie Feb 4 at 19:46
  • 1
    \$\begingroup\$ You don't provide the information necessary for anyone to help you, first step would be to show your construction/PCB.. I am almost willing to bet, as @Bimpelrekkie already pointed out, that the way you constructed the circuit is to blame. Especially given your comment "despite my PCB being worst case".. \$\endgroup\$ – Vinzent Feb 4 at 19:58
  • 1
    \$\begingroup\$ A light load may be driving your converter into discontinuous conduction mode. Do you get better regulation when the load draws more current? Perhaps a 22\$\Omega\$ resistor? \$\endgroup\$ – Math Keeps Me Busy Feb 4 at 19:59
  • 1
    \$\begingroup\$ Very short ground lead on 10:1 probe is mandatory and ac couple into 50 Ohms for best noise results \$\endgroup\$ – Tony Stewart EE75 Feb 4 at 20:03
  • 1
    \$\begingroup\$ You could change your 50k pot to a dual 50k pot, and use the extra resistance to control a low-dropout linear regulator. Such that the SEPIC voltage is always above the drop-out voltage of the linear regulator. Good efficiency with very clean output. \$\endgroup\$ – rdtsc Feb 4 at 20:14

Here is a CircuitLab circuit which models what happens when the two inductors and coupling capacitor from your circuit are driven with a 5V supply and 1.2MHz, 50% constant duty cycle switch.


simulate this circuit – Schematic created using CircuitLab

Although this circuit shows a relatively stable Vout, (after it stabilizes when C2 is charged at around 2.2mSec) it can be seen that the voltage across the coupling capacitor has over 2V peak-to-peak oscillations, and the two inductors have nearly 2A peak-to-peak oscillations.

Output Voltage and Voltage across coupling capacitor

Current through inductors

Although by the output voltage everything seems OK, by the way these components are oscillating, it could be a problem. That is why I wrote:

By my analysis, the two inductors and sepic cap should be oscillating quite strongly around 5kHz. Could you post a scope shot of Vout for a 4 mSec total duration after the circuit has settled a bit (say after 100 mSec or even 1 sec)? – Math Keeps Me Busy

I will wait on the results, but I suspect that your sepic cap is too big. Do you need it that big for current? If not, I suggest that you remove your 20uF and replace it with the original 5.6 uF and test that. – Math Keeps Me Busy

You then changed the coupling capacitor to 4.7 uF. According to the circuit lab model, this reduced the oscillations to about a quarter of the value when the cap was 20 uF:

enter image description here

enter image description here

Since on your actual board:

@MathKeepsMeBusy I changed the cap back to 4.7uF, single cap. it got a lot more better both on 4K ohm load and 100 ohm load (set to 14V). I used 2 parallel caps for lower ESR. it seems that wasn't a good idea.

I think it is safe to conclude that much of your problem was related to excessive oscillations of the inductors and coupling capacitor.


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.