4
\$\begingroup\$

Well this is just a general question about switch mode power supplies, but i have recently built a buck converter, and read that it is very important where in the circuit you place your probes, to measure output ripple and noise.

As output capacitance i have a electrolytic cap, and a small 100n ceramic. When i measure directly above the ceramic capacitor, i can clearly see the ideal/textbook output ripple, with a small amount of high frequency ringing, which measures to about 50mVpp. But when i move my probes to measure above my resistive load, the output noise amplifies to around 200-300mVpp.

How come? I dont understand where this noise is coming from, when closer to the circuit im measuring way less noise? Is it radiated noise picked up from the inductor coil? Or am i missing something

\$\endgroup\$
8
  • 1
    \$\begingroup\$ In what conditions have you measured the ripple? You might be interested in seeing it explained by Dave Jones: How to measure power supply ripple noise. His explanations yet apply to measuring the internals of a power supply. \$\endgroup\$
    – user59864
    Jun 17, 2017 at 19:45
  • \$\begingroup\$ Hmm yea, well he is talking about common mode noise and very low voltage background noise. I dont think the 200-300mv noise im picking up is background noise in any way. When i probe the capacitor i use the small ground tip for the probe, but when i probe the resistor i use the big ground lead. But should that really matter, when we are talking this much noise? I mean, if i stick my probe into the lab bench PSU i am using with the ground lead, i measure almost no noise, i dont need to probe directly onto the output capacitor with a small ground tip and such \$\endgroup\$
    – Linkyyy
    Jun 17, 2017 at 20:20
  • \$\begingroup\$ What makes you so sure the 200-300 mV do come from the SMPS itself and not from the outside? Yes, the ground connection matters and what you must ensure is the shortest probe leads as possible to avoid picking external (i.e. common) noise from your environment. Otherwise all you have is an antenna, which is what Dave explains in his video. Do a differential measurement for example. \$\endgroup\$
    – user59864
    Jun 17, 2017 at 20:40
  • 2
    \$\begingroup\$ Can you put up images of the scope measurements? The waveforms might be enlightening. Also, try shielding the SMPS best you can purely to see the effect. Be rigorous with your methods of measurements in that it's easy to convince oneself of correctness, but adhering to objectivity etc always wins out. You have the power to be deductive - I for one am interested in your results! \$\endgroup\$
    – CL22
    Jun 18, 2017 at 5:05
  • 2
    \$\begingroup\$ Can you post a picture of the setup and the scope probes when you are measuring ripple? \$\endgroup\$
    – bobflux
    Jan 1, 2020 at 0:28

1 Answer 1

Reset to default
1
\$\begingroup\$

Consider your scope probe + its GND lead as a magnetic receiver. Use the formula $$Vinduce = MU0 * MUr * Area/(2 * Pi * Distance) * dI/dT$$

Let area be 4" by 4". Let distance between the probe/gnd and the SMPS be 4". Accept 4" is 0.1 meter.

Use dI/dT of 0.1 amp in 10 nanoseconds (the FET on and off time).

Evaluate the Vinduce (at this point, I cannot predict how small or big this will be, so we just do the math).

Vinduce = 2e-7 * Area/Distance * dI/dT

Vinduce = 2e-7 Henry/meter * 0.1 meter * 1amp/100nanoSec

Vinduce = 2e-7 * 0.1 * 1e+7 = 2e-8 * 1e+7 = 0.2 volts.

For fast edges, the planes (all your planes) will provide some attenuation; Skin Freq for 35micron foil (the standard thickness, 1.4mils, 1 ounce/foot^2) is 4MHz or 125 nanoseconds, thus a 10nS edge gets lots of attenuation.

If your scope probe + GND lead is up in the air, you'll have little or no attenuation from the planes.

============================================

This formula comes from combining Faraday law of induction with Biot-Savart law. We then assume source of magnetic field is a long straight wire that is coplanar with the Area, and that the ratio of minumum distance to maximum distance is close enough to One we can ignoe some minor natural-log terms.

\$\endgroup\$
1
  • \$\begingroup\$ Well that sounds pretty reasonably actually. Where does this formula come from? \$\endgroup\$
    – Linkyyy
    Jun 19, 2017 at 17:20

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.