2

I've purchased multiple 3V switching power regulators form Amazon and they all have spikes in the ripple. Here's an O-scope capture: enter image description here

The spikes increase frequency depending on the load. Are these spikes a fact of life or are there better products out there?

I need something basic like this:enter image description here

peteey
  • 411
  • 1
  • 4
  • 10
  • Did you use any caps to reduce the spike ? How did you hook them up. Can you show a schematic. – deathismyfriend Aug 20 '14 at 17:28
  • "Spikes in the ripple?" What is the time base of the scope capture? Is the SMPS unloaded in that capture? – JYelton Aug 20 '14 at 17:38
  • There are "better" products out there, but you'll need to decide exactly how much variation you can tolerate so you can decide how much extra $ to spend to reduce the spikes. Also consider adding filters at the load, like other comments have said. Or get a 4.5 or 5 V SMPS and use a linear regulator to get a really clean 3.3 V at the load if you are doing really sensitive analog work. – The Photon Aug 20 '14 at 17:39
  • 1
    The 'spikes' likely correspond to the switching frequency. The switching frequency may be variable and increase with the load to improve efficiency. I agree with the previous comments about filtering. But disagree with the comment about the linear regulator (partially). High frequency ripple may pass through a linear regulator with little attenuation - depends on its own feedback loop. You should consider getting a fixed frequency switcher (PWM) whose switching frequency is outside of any band you are interested in and then tune your filter's cutoff to be less than that frequency. – Nick Aug 20 '14 at 17:47
  • I agree with @Nick, a filter is needed in addition to a linear regulator. (And a linear regulator is only needed if your circuit is very sensitive). – The Photon Aug 20 '14 at 17:54
  • 1
    When using the oscilloscope for this, make sure you use a very short gnd lead. – Tut Aug 20 '14 at 18:11
  • 2
    If you rephrase your question to something like asking if it is normal for cheap switching regulators to have spikes on the output voltage, it would be more on-topic. This sounds more like a shopping question but is not uninteresting in general. – Rev Aug 20 '14 at 18:46
  • 1
    @Rev1.0 Agreed - It could then spawn a broader and more useful discussion about switching regulator designs that are geared towards high efficiency vs. designs whose goal was low noise performance or the topic could be "how to I mitigate the fact that I have to power my electronics from a noisy regulator" – Nick Aug 20 '14 at 20:05

1 Answers1

3

This looks like a type of switching regulator that has a constant "on" period and variable "off" period. Once the regulator has measured the output voltage to be correct (at its upper limit), and, in the absence of significant loading effects on the output, the switcher switches off and the output voltage decays until the lower threshold of operation is reached whereupon it repeats the cycle. Sometimes one "on" period isn't enough to restore the voltage to the correct top limit and you'll see several pulses all close to each other.

If you need a better waveform there are better switching topologies but, you are unlikely to be able to determine that from an amazon web-page.

If you load the output with more capacitance, the spikes will appear less frequent but there may be more "on" periods closely grouped in order to restore the voltage to the upper limit. It's called a bang-bang type regulator and relies on the output waveform ramping up and down a little bit. The spike is usually related to resonance of the switching inductor but can be due to bad PCB layout, poor choice of capacitor in the output stage or your o-scope telling you something that isn't actually there (magnetic induction).

A way around this might be to use an output resistor of a few ohms and another capacitor but obviously the "few ohms" are going to lose some voltage when a real load current is taken. An inductor might help but without knowing the timebase of the picture it's difficult to guess at what value.

Andy aka
  • 434,556
  • 28
  • 351
  • 777