How you measure your ripple can make you or break you


Testing switching power supplies includes many different tests, one of them being the output-voltage peak-to-peak ripple. Output-voltage ripple is the alternating current (AC) component of the direct current (DC) output voltage. It’s generated by a combination of factors, including the output capacitor’s equivalent series resistance (ESR), the voltage drop across the output capacitance, duty cycle and switching frequency.

Because it impacts the overall voltage tolerance of the rail, the peak-to-peak output-voltage ripple is a target specification in many processor, field-programmable gate array (FPGA), application-specific integrated circuit (ASIC) and system-on-chip (SoC) data sheets and design guides.

How you measure the ripple can affect your ability to meet the design requirements.

Figure 1 shows a typical output-voltage ripple probe setup.

Figure 1: Output-voltage ripple probe setup

Probing using a clip-on probe shows an increased ripple that may be partly the result of the ground-wire loop picking up noise, as shown in Figure 2.

Figure 2: Output voltage ripple probe with a clip-on probe and ground wire fully extended, picking up noise from the nearby switch node

Probing using the pigtail method improves ripple, even though the tip is again near the switch node, as Figure 3 shows. The ground loop is much shorter; thus the noise pickup is less severe.

Figure 3: Output-voltage ripple probe using the pigtail method; the probe ground is in contact with the pigtail, which is connected to the board ground

Using a coaxial cable method improves results even more, as Figure 4 shows. Directly soldering the woven copper shield on the board ground minimizes the ground loop further.

Figure 4: Output-voltage ripple probe using coaxial method

Figure 5 shows a close-up of the coaxial cable.

Figure 5: Coaxial cable close-up: the outer plastic sheath (a); woven copper shield (ground) (b); inner dielectric insulator (c); and copper core (VOUT) (d)

Another similar measurement method is to use a probe jack like that shown in Figure 6. The outside jacket is the ground connected directly on the board while allowing the probe tip to connect to the voltage test point.

Figure 6: Probe jack

Of all of these methods, using a differential probe is probably the best way to measure ripple accurately. It can eliminate the ground-loop noise pickup error, especially when connecting other electronic equipment to the same board ground (such as electronic loads and multimeters).

Figure 7 shows the two test points for differential probe connections on the board.

Figure 7: Differential probe test points CPU_VSEN+ and CPU_VSEN-

If you are trying to meet tight output-voltage regulation requirements and have a low peak-to-peak voltage-ripple target, how you measure the ripple on your board can make you or break you. Optimizing your probe method will help you with your measurements and meet the specifications. Practice and compare any of the probe methods discussed in this post on a switching regulator evaluation module (EVM) like the TPS40304EVM-353. As well, read the application report, “Output Ripple Voltage for Buck Switching Regulator” and understand how ripple voltage is calculated and reported in WEBENCH® Power Designer.

  • George, In the coaxial method how to connect the oscilloscope probe to the cable?

    Thanks.

  • Haroldo,

    you use a standard 50 Ohm termination coaxial cable like the ones you use to connect your cable TV STB.

    The drawings show the low-cost method; people also put a co-ax connector on the board and use the standard 50 Ohm coaxial cables without the stripping and soldering at all.

    Please make sure you set the oscilloscope to 50 Ohm termination when you use the coaxial cable.

    Please note that this method does not provide ground isolation, as a differential probe does.  

    The coaxial cable technique is better for power measurements than using a probe with long ground leads, but not as good as differential probe.  For 50Ω signal measurements without power supply switching noise, this is a very good method.

  • Hello George.  Appreciate this note.  Quite a big deal to remove parasitics of the probing method in order to observe the correct signal produced.  However, you state below that one should set the scope input to 50 ohms.  Although RG-58 coax is a 50 ohm transmission line, you will load your power supply with the 50 ohm termination.  The output impedance of the converter power supply is very low, much lower than a 50 ohm source as in a sign gen.  Any reason why you would not wish to use the 1 Mohm scope input setting?

    Thank you.

  • It’s true, using the 50 Ohm termination does essentially add a load of 50 Ohms to the output of the supply, but it also keeps the whole transmission path relatively low impedance. With the 1 Meg setting, you’re a lot more likely to be picking up noise with the cable itself with the antenna effect.

    If you’re confident the 50 Ohm impedance won’t affect the signal (which you definitely should be for a power supply output), then you’ll get much cleaner waveforms with a 50 ohm termination.

    However, you do need to use a high impedance probe in order to measure signals with only a small drive capability (like a FB or COMP node for example), because the extra 50 Ohms to ground will change the answer.

    PS > For higher output voltages or switch nodes, you may need to look at the full scale range of your oscilloscope, since usually, the 50 Ohm setting means 1:1 conversion (most commercial probes include a 10:1 or 20:1 divider so it’s not an issue).

    If you need to measure a higher voltage than this allows, you can either go to the 1 Meg setting, and accept that you’re more likely to get a little bit of noise pickup, or you still can keep the path relatively low impedance by making a 10:1 divider with the internal 50 Ohms, by putting a 450 ohm resistor in series with your signal.

  • Thanks for the article. We use RF-174 50 Ohms co-axial cable for power noise measurements. Because we can't afford Differential probes we went with this method. Our regulators are all High input to low voltage (high current) Buck converters (12 to 0.9/1.0/1.2 etc). Setup will be RF174 Co-axial cable will be directly connected across regulator's Output bulk capacitor and DSO is 50 Ohms input. We are seeing good results. I have 2 questions on this. 1. you said differential measurement is more accurate than 50 Ohms co-axial measurement - How it is, any results (can be approximate) to know how much will be the error in measurement (coaxial to Differential probe) 2. As noise spec near ASIC is < 2%, whether we need to measure near to load (near to ASIC) or at the regulator output Bulk capacitor

  • We don’t have any real apples-to-apples comparison of differential vs. single ended.

    The differential probe is (as the name implies) a true differential measurement, so it will report one capacitor terminal minus the other, so it accounts for any common mode “bouncing” that might be occurring.

    This sort of thing would show up in a single ended measurement, but not a differential one.

    Also, with a differential probe, the amplifier element is right at the tip of the probe, so you can get some very small lead lengths (avoids noise pickup via antenna effect). Everything after the amplifier is actively driven by the amplifier, and is relatively low impedance.  

    The input impedance of the differential probe is greater, so it will not load the input signal, this matters more for sensitive signals than a power supply output, but it’s worth noting.

    To get an idea of the voltage ripple at the ASIC, it’s best to measure at the ASIC, across the closest high frequency capacitor to the device pins.

    It’s possible the bulk capacitor will appear noisier, because it’s closer to the circulating currents, and fast switching of the SMPS, where remotely at the ASIC, you’ve got better HF bypassing, and board impedance in between to squash really high frequency components.