2
\$\begingroup\$

I had a working design of a buck switching regulator based on a TPS62125. It worked fine, with minimal (<100mVpp) noise. I made some changes to my design, seemingly unrelated to the switcher. The result is audible noise from the switcher (a high pitch whine, that gets louder as draw increases), as well as this lovely waveform:

enter image description here

Here's my current layout:

enter image description here

The only changes I made to the switcher directly is adding room for a 250uF electrolytic cap (problem persists with and without it), adding a .2mF supercap elsewhere in the design, but off this power rail (problem also persists with and without), and adding a footprint for C33, but not populating. The amplitude of the noise does not seem to be affected by load.

Where do I go from here? How do I troubleshoot this? I thought I spotted a short between the two pins of the IC that lead to the two pads for the inductor. It doesn't look like there's an actual bridge while looking with a loupe, but could this be the source of the problem (pins are too close because of excess solder, creating some sort of stray capacitance)?

The only other thing I can think of is that it's possible that I installed a wrong capacitor for the input or output caps (did this by hand). Would that produce these results? Could it be a bad inductor?

\$\endgroup\$
  • \$\begingroup\$ You've got a lovely pickup loop across R16 - why are r's across other side from both Vout and FB pin? |C22 is in middle of that loop - noise filtered by C22 MAY be inductiely coupled into loop. Install C33 (small) . Temporarily move C22 physically and tack on in next best locn away from FB loop.|IP and OP caps could cause problems if well off values shown | Data sheet does NOT say so but possible it is unstable if output cap ESR does not fall in a certain range. Varies with load. Check. | What sort on in & out caps? Ceramic? C33 = transient response speedup says DNP. Is it populated? .... \$\endgroup\$ – Russell McMahon Oct 5 '14 at 6:41
  • \$\begingroup\$ I'd still try moving C22 - the work of moments and it does look like it could "do the job". \$\endgroup\$ – Russell McMahon Oct 5 '14 at 10:26
  • \$\begingroup\$ @RussellMcMahon are you saying to scrape off the soldermask and make new pads elsewhere? I did try installing the larger 250uF electrolytic (there are pads for that) and it decreased the noise somewhat, but not completely. \$\endgroup\$ – kolosy Oct 5 '14 at 16:13
  • \$\begingroup\$ Irrelevant now, but I meant just tack it on wherever it was possible such that there was no chance of it forming a one turn loop for coupling to the feedback resistor loop. As it was not the issue it's not the solution:-). It would be interesting to see if there is any measurable effect of this sort due to the positioning. \$\endgroup\$ – Russell McMahon Oct 6 '14 at 6:29
1
\$\begingroup\$

Just maybes:

If any of these prove useful I'll come back and tidy up the answer. Otherwise no great value.

You've got a lovely pickup loop across R16 - why are FB R's across other side from both Vout and FB pin? |

C22 is in middle of that loop - noise filtered by C22 MAY be inductively coupled into loop.
Install C33 (small).
Temporarily move C22 physically and tack on in next best locn away from FB loop.

IP and OP caps could cause problems if well off values shown.

Data sheet does NOT say so but possible it is unstable if output cap ESR does not fall in a certain range. Varies with load. Check.

What sort on in & out caps? Ceramic?

C33 = transient response speedup. Diagm says DNP. Is it populated? Try it.
What happens with small C33?

What is load.
How does op waveform and amplitude change with load?

\$\endgroup\$
  • \$\begingroup\$ R16 - this is per their recommended layout (see page 28 of the data sheet). All caps are MLCC "low esr". C33 is my addition to design, will try. Load starts off at ~100mA, then drops to about 20 as the supercap charges. No visible impact on waveform as load drops. \$\endgroup\$ – kolosy Oct 5 '14 at 6:58
  • \$\begingroup\$ The issue does appear to have been the shorted pins leading to the inductor. I assume it didn't just blow up because it was switching fast enough to prevent that (1MHz). I am surprised that it was hovering near the set voltage, but it's working well now, <100mV PP. \$\endgroup\$ – kolosy Oct 5 '14 at 17:21
  • 2
    \$\begingroup\$ @kolosy Yes, no inductor turns it into a PWM regulator :-). The small RC delay on the sense line plus internal switching delays give a finite time high or low. If you'd put some formal capacitance on the FB pin you'd probably have got a closer average output V. Definitely not recommended though :-). \$\endgroup\$ – Russell McMahon Oct 6 '14 at 6:27

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.