Ripple voltage measuring: confusion with using of capacitor along with oscilloscope probe

[NigthMoth]

Hello all!

I have some questions regarding usage of filter capacitor along with oscilloscope probe to measure voltage ripple, that is described as correct ripple measurement method in this aticle (above Figure 5) : How do you reduce voltage ripple?

My questions at the end of this post.

First, here are pictures and citation from above mentoned article i'm confusing of:

Figure 5 shows the correct ripple measurement method. It can be seen from the figure that the output of the converter is connected with a filter capacitor. The purpose is to suppress noise, so the capacitor value is usually not too large, mostly at 0.1uF to 1uF. And the probe should use a short grounding method for measurement. The measure point should change from the load to the output capacitor. The purpose is to avoid measuring noise. Figure 6 shows the difference between the ripples of short ground and no short ground. The ripple voltage of the converter can be measured correctly if used the right methods.

I tried to make test as on Figure 5:

• 8.3V AC-DC wallplug adapter (used for toy car battery charging);
• 0.33uF film capacitor;
• 1.2k resistor as load;

I had measured voltage waveforms across capacitor (as reccomended in abovementioned article) AND across load simultaneously and found indeed that the noise across capacitor is very low, but it reappears again on the load (the longer distance between capacitor and load, the higher level of re-appeared noise)

I got:

• Without capacitor: Noise+Ripple 1.050Vpp;
• With capacitor: Noise 0.110Vpp across capacitor;
• With capacitor: Noise 0.508Vpp across load;

My questions
1) Am I correct: Connected capacitor is just an "upgrade" to oscilloscope probe and I should think of this capacitor as it is attached between oscillocope probe's tip and probe's ground, think of it as it is part of probe, but not think of it as it is an external noise filter for power source? (Reason: Measure point is across connected capacitor, noise reappears in some distance from capacitor).
2) What is actual noise level of power source output: that measured with capacitor or that measured without capacitor?
3) Or abovementioned method focused only on measuring of ripple but not noise, so purpose of connected capacitor is just to filter some noise, just in order to make ripple waveform more clear on oscilloscope screen?

 

[D.A.(Tony)Stewart]

Merry Xmas from Canada.

Yes now both methods give accurate underdamped 650kHz RLC pulse response 150 mV pk overshoot without parasitic probe error.
Perfect measurement. Bravo
Now how to adjust overshoot knowing series and parallel RLC damping factors from every component that has influence from low series R or with light load. Delicate balance between efficiency and overshoot.. If there was no ripple, there would be no feedback and it could be an unstable loop so limits on Vin:Vout range:ratio or Iout max:min ratio or ESR+DC+RdsOn ratio must be examined along with sensitivity or each damping R. Some like to use Root Locus method. I like to determine sensitivity of each component to a parameter over-spec using a process called DoE (design of experiments) Taguchi Method with +/-x% dither for fine tuning after getting it close. Others use Monti Carlo methods.

But adding R reduces overshoot and increases %loss and temperature so measure this too in simulation (alt+clk?) until it matches your design, then you can optimize faster. test for surge start, L overcurrent and +/- step load response


Alternate test methods. Often DSO has 50 ohm option term but can burn out with DC or in a pinch use a BNC T adapter and stick in a leaded resistor in T and even 75 ohms is close enough.

Or using 20 MHz filter on DSO is another method but sometimes hides real issues.
or add 500 ohms to ground alligator clip(??) to dampen current impulse

Adding in low R across L for light load overshoot often helps
Using low ESL ground and V+ traces and low ESR parallel ceramic caps helps for high load currents or reducing 1/LC product to raise fo can help overshoot. Then a tradeoff for loop stability and overshoot must be examined.

A perfect SMPS is hard to meet best-in-class balance.

 

[JDW_]

Noise is generally measured as RMS, not peak-to-Peak.

The following Keysight FAQ answer puts the above remark in perspective:

Peak-to-peak noise is... particularly important for applications where noise spikes could degrade accurate measurements on a sensitive load, such as RF circuitry.

[RMS] is not an ideal representation of DC power supply noise because fairly high output noise spikes of short duration could be present in the ripple and not appreciably increase the rms value.

 

[BigBoss]

Peak-to-peak noise is... particularly important for applications where noise spikes could degrade accurate measurements on a sensitive load, such as RF circuitry.

[RMS] is not an ideal representation of DC power supply noise because fairly high output noise spikes of short duration could be present in the ripple and not appreciably increase the rms value.

I talk about Noise, NOT Ripple or Spikes. I know very well what the Noise is.
Noise is a Stochastic Process so it's non-casual, non-deterministic. So the Noise should be measured and modelled as Stochastic process as other non-casual processes.
So you cannot know where and what time the noise variable is. RMS is used in these cases.