1
\$\begingroup\$

I've designed a synchronous buck converter using a gate driver IC and two MOSFETs. I have 10 mH and 6.6 uF capacitor at the output as described in the attachment. I am providing 15 V as an input to the converter through connector P1. I am also running at 20 kHz of switching frequency. How accurate is the open loop measurement supposed to be? I expected about 7.5 V when I configure duty cycle wit 50% and ripple current & voltage to be around 1%. Load is fixed as 75 Ω. But the actual measurement of output voltage of the converter is around 6 V which is about 40% of input. I've tried to vary duty cycle but similar result. I don't get any closer value compared to my calculations and this gets worse when I increase duty cycle up to 70%. I do have a closed-loop with ADC then feed back to MCU later but I came across about this when I was testing an open-loop configuration. Is this expected behavior? Also, how should I remove the spike appearing at the switching node when turned on? Please see below for scope images.

Expected (calculated) results with 20 kHz and 50% duty cycle :

  • I_out = 100 mA
  • V_out = 7.5 V

Open loop testing with no load result:

  • Yellow = switching node before inductor

  • Blue = Vgs upper MOSFET

  • Green = Vgs lower MOSFET

  • No issue at the output. I see what I expected when there is no load. enter image description here

With 75 Ω load with 50% duty cycle:

  • Yellow = node at output capacitor (after inductor)
  • Red = inductor current enter image description here

With 75 Ω load with increased duty cycle (62%):

  • Yellow = node at output capacitor (after inductor)
  • Red = inductor current enter image description here

Spikes at switching node before inductor:

  • Red = inductor current

  • Yellow = voltage at switching node (before inductor)

enter image description here enter image description here

\$\endgroup\$
11
  • \$\begingroup\$ If possible try to verify operation in CCM and not DCM. Can you add a sense resistor (say 10-100 m\$Omega\$ in series with your inductor, and take differential measurement across iit. In CCM, current through inductor will never fall to 0. \$\endgroup\$ Oct 27, 2023 at 4:14
  • \$\begingroup\$ I will verify that. Thanks \$\endgroup\$
    – Alia Sana
    Oct 27, 2023 at 4:17
  • 1
    \$\begingroup\$ 10 mH inductor @ 20 kHz sounds odd for 25 ohm load. \$\endgroup\$
    – Jens
    Oct 27, 2023 at 4:42
  • \$\begingroup\$ @MathKeepsMeBusy I have some scope measurements and updated in the post! It's clear that it is operating in CCM. So to my original question, am I supposed to see that amount of discrepancies in open-loop testing? I had to increase dome duty cycle to get desired output. Also, how should I remove spikes when switching node is turned on? \$\endgroup\$
    – Alia Sana
    Oct 31, 2023 at 23:16
  • 1
    \$\begingroup\$ I've reverted the last edit as it deleted essential context for understanding the question and answers. \$\endgroup\$
    – Hearth
    Nov 3, 2023 at 16:16

2 Answers 2

0
\$\begingroup\$

Open loop and accuracy don't really go together.

With inductors often having 20% tolerances, on-resistances and switching speeds of transistors bring very temperature dependent and the properties of output capacitors not being much more precise, if say your 70% error is still much, but not unexpected. I would take a guess and assume you're using a model that neglects parasitics in the inductor and MOSFETs.

From the schematic, and in and output specs, I'd say designing this as open loop system is a mistake, even if you don't need accuracy. Feedback is essentially free in terms of circuit complexity and solves a lot of issues at once.

\$\endgroup\$
2
  • \$\begingroup\$ Hi Marcus Müller. Thanks for your reply. Yes, I don't have an intention to operate the system as an open-loop. I forgot to mention and attach rest of circuits but I do have mcu connected and read current through ADC then control the PWM.I just didn't have an experience in this design before and I was curious about the accuracy as an open-loop. I haven't tested along with mcu yet. Thanks again tho. But would you be able to elaborate what you mean by "I would take a guess and assume you're using a model that neglects parasitics in the inductor and MOSFETs"? \$\endgroup\$
    – Alia Sana
    Oct 27, 2023 at 2:25
  • \$\begingroup\$ Such a large error between what you expect and what you measure can only be explained if your model of how your components behave is not modeling reality well enough. Typically, in such switcher applications, people neglect parasitic effects making MOSFETs switch slower and making inductors store less energy. \$\endgroup\$ Oct 27, 2023 at 2:50
1
\$\begingroup\$

Yes it should be better than that.

  • L value doesn’t matter
  • C value doesn’t matter
  • Mosfet/L/trace R does matter but should be low (verify)
  • Mosfet switching times do matter but should be low (verify)
  • Duty cycle obviously matters. Verify this first

Most likely duty cycle isn’t actually 50% and/or your switching waveform is messed up for some reason. Also possible you have R in the design you didn’t account for.

Should be easy to follow this circuit around and find the problem. Switching waveforms should be nice and square (some overshoot allowed).

For context note that class-d audio amplifiers have the same topology (essentially) and no feedback and produce extremely linear response from duty cycle to output.

\$\endgroup\$
6
  • \$\begingroup\$ I tried all of your suggestion, specially checked all waveforms (Vgs of two mosfets and etc) and they seemed very clean waveform with the correct duty cycle of 50%. I tried with all different duty cycle. I can show you waveforms tomorrow. \$\endgroup\$
    – Alia Sana
    Oct 27, 2023 at 4:01
  • \$\begingroup\$ Can you verify that it is actually in CCM? Also, you could be losing voltage from your MOSFETs if for some reason they are not turning completely on. \$\endgroup\$ Oct 27, 2023 at 4:19
  • \$\begingroup\$ @MathKeepsMeBusy it's a synchronous converter so it never enters DCM. \$\endgroup\$
    – Andy aka
    Oct 27, 2023 at 8:30
  • \$\begingroup\$ @Andyaka that assumes the bottom transistor is working properly. IMHO, when troubleshooting it is always a good idea to verify that your assumptions are confirmed in reality. \$\endgroup\$ Oct 27, 2023 at 10:46
  • \$\begingroup\$ @MathKeepsMeBusy updated in the post! please have a look and let me know. thank you \$\endgroup\$
    – Alia Sana
    Oct 31, 2023 at 23:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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