1
\$\begingroup\$

A servo accelerometer when powered by batteries has a clean output, but when used with this SMPS supply its output seems like have switching artifacts from the power supply(the datasheet says its switching freq is 100kHz). Now since the desired range is very low I can easily filter out such noise at the sensor output.

But having said that, I want to know what we can say about the noise at the accelerometer output. So by using a digital scope(at AC coupling mode) I made two measurements and logged the time series data for the standing accelerometer output. The first one is for longer period, here is the screenshot:

enter image description here

Then I plotted the time series and FFT of this as follows:

enter image description here

enter image description here

Then I did the same thing for a smaller time scale:

enter image description here

enter image description here

enter image description here

So by having all these information what can we say about the frequency of the switching noise? I'm trying to learn in such a situation how to quantify the switching noise. Should we look at the time series data or the spectrum for that? And for this particular case what steps should one follow?

EDIT:

For a much longer acquisition I observe the following:

enter image description here

enter image description here

enter image description here

MEASURED DIRECTLY AT THE POWER SUPPLY OUTPUT:

enter image description here

enter image description here

Closer view:

enter image description here

And spectrum:

enter image description here

Supply Output at 50us scale:

enter image description here

Time series and FFT:

enter image description here

enter image description here

The PSU noise(yellow) directly appears at the sensor output(blue):

enter image description here

\$\endgroup\$
  • \$\begingroup\$ If that timebase is 250 nanoSeconds per division, then with 2 cycles per division you have 8MHz ringing. And some 2MHz (2 divisions period). To filter that, use 1uHenry and 1uFarad, which resonate at 160KHz, giving 100:1 attenuation to the 2MHz and another 16:1 attenuation to 8MHz.. \$\endgroup\$ – analogsystemsrf Sep 25 at 9:28
  • \$\begingroup\$ Well just by judging the time between the pulses, they are not with a fixed frequency, there is quite noticeable jitter on it. And by counting the number of events over the time you plotted you get an average frequency of 15-16 kHz. So I'm not really convinced that it is the switching noise of 100 kHz doing that. \$\endgroup\$ – Arsenal Sep 25 at 9:35
0
\$\begingroup\$

If I'm interpreting your first scope plot correctly it's showing switching noise from a SMPS running just under 100KHz. The apparent jitter is due to the fact that noise occurs on both edges of the switching waveform, so analysis of this waveform will show a fundamental at just under 200KHz, spread somewhat by the jitter on the switching edge. You can see this nearly buried in the noise on your first frequency plot.

However, there's very little fundamental in this waveform. Most of the energy is in the ringing, which you can see in the higher order spectral content.

\$\endgroup\$
  • \$\begingroup\$ How did you quantify the switching noise close to 100kHz? Between each spike I see like 50us to 60us that makes below 20kHz isnt it? \$\endgroup\$ – cm64 Sep 25 at 13:23
  • \$\begingroup\$ You're right, I misread the scale. I'll give it some thought and come back with an edit. \$\endgroup\$ – Cristobol Polychronopolis Sep 25 at 13:53
  • \$\begingroup\$ Please also see my edit meanwhile. \$\endgroup\$ – cm64 Sep 25 at 13:58
  • \$\begingroup\$ I will also add info from directly supply output. \$\endgroup\$ – cm64 Sep 25 at 14:00
  • \$\begingroup\$ Can you post the supply output on a 50us scale? \$\endgroup\$ – Cristobol Polychronopolis Sep 25 at 15:26

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.