In the daily work of power supply engineers, ripple testing is a core step in verifying the performance of power supplies. However, behind the seemingly simple operation of oscilloscope measurement, there are many technical traps: improper selection of oscilloscope bandwidth can lead to high-frequency noise being filtered out, the load effect of the probe may change the actual working state of the power supply, and the ground loop will superimpose common-mode noise in the test signal. The combination of these factors often causes the test results to deviate from the true values by more than 50%. This article will systematically analyze the three major pitfalls in ripple testing and provide practical solutions.
I. Concealment Techniques for High-Frequency Noise
The bandwidth of an oscilloscope is the first hurdle that determines the accuracy of a test. When the bandwidth is set lower than the signal frequency, high-frequency noise will be attenuated, resulting in a smaller ripple measurement value. When the bandwidth is set too high, environmental noise may be amplified, masking the true ripple. The actual measurement data of a certain communication power supply manufacturer shows that when testing a 12V/10A power supply with a 20MHz bandwidth, the peak value of the ripple peak is displayed as 48mV. After switching to a 200MHz bandwidth, the measured value jumped to 126mV, of which 80mV was high-frequency noise filtered due to bandwidth limitations.
The golden rule for bandwidth selection: Determine the test bandwidth based on the switching frequency. For traditional silicon-based power supplies (with switching frequencies ranging from 65kHz to 200kHz), it is recommended to choose a bandwidth five times the switching frequency (for example, 200kHz corresponds to a 1MHz bandwidth). For GaN/SiC high-frequency power supplies (200kHz-1MHz), a bandwidth of over 500MHz is required. The actual measurement of Tektronix MDO3000 series oscilloscopes shows that at a bandwidth of 500MHz, the ripple measurement error of a 650kHz switching power supply can be controlled within ±3%.
High-frequency compensation technique: When testing high-frequency signals with a low-bandwidth probe, the frequency response can be corrected through the "high-frequency compensation" function of the oscilloscope. The Agilent InfiniiVision 3000X series oscilloscopes are equipped with intelligent compensation algorithms that can automatically adjust the probe response curve within the range of 100kHz to 500MHz, reducing the measurement error from ±15% to ±5%.
Second, the test tools change the system under test
The interaction between the probe and the circuit under test is a often overlooked trap. The input capacitance of traditional passive probes ranges from 15pF to 30pF, which forms a low-pass filter in high-frequency tests and significantly attenuates high-frequency ripples. A server power supply test case shows that when a 15pF probe is used, the 1MHz ripple component is attenuated by 42%. After switching to a 3.5pF low-capacitance probe, the measured value was restored to 98% of the true value.
Three key elements for probe selection:
Input capacitance: Prefer low-capacitance probes with a capacitance of ≤5pF, such as Rohde & Schwarz RT-ZP03(3pF) or Tektronktronktronkus P7500(4pF).
Common-mode rejection ratio (CMRR) : In environments with strong interference, select a differential probe with a CMRR>60dB, such as Keysight N2790A(80dB@1MHz).
Attenuation ratio: For small-signal tests (<50mV), use a 1:1 probe to avoid signal attenuation. For large-signal tests (>10V), a 10:1 probe is required to protect the oscilloscope input.
Load effect compensation method: When high-capacitance probes must be used, a compensation network can be constructed by paralleling small capacitors (0.1μF-1μF). In the test of a certain DC-DC converter, after a 0.47μF film capacitor was connected in parallel to the ground terminal of the probe, the measurement error of the 100kHz ripple was reduced from 38% to 7%.
Iii. Amplifiers for Common-mode Noise
The ground loop is the most concealed trap in ripple testing. When there is a potential difference between the ground wire of the oscilloscope and the ground of the power supply under test, a loop current will be formed, superimposing common-mode noise on the test signal. The power supply test of a certain medical device shows that when using a common alligator clip ground wire, the 50Hz power frequency interference reaches 200mV. After switching to a spring-loaded grounding needle, the interference was reduced to 8mV.
Four-step method for Eliminating Grounding loops
Shorten the ground wire length: Replacing the traditional 15cm alligator clamp ground wire with a 3cm spring-type ground pin can reduce the loop area by 95%. The actual measurement of the grounding needle that comes with the Tektronix P6139A probe shows that when the ground wire length is reduced from 15cm to 3cm, the 50Hz interference drops from 120mV to 15mV.
Adopt differential measurement: Use differential probes (such as PICOTEST J2102A) to directly measure the signal difference, completely eliminating the influence of ground loops. In the 48V communication power supply test, the peak value of the ripple measured by the differential probe was 82mV, while the measured value by the single-ended probe was 156mV(including 74mV common-mode noise).
Isolating oscilloscope channels: For multi-channel testing, using an isolation transformer (such as TDK B64290L0001X) to isolate the ground wires of each channel can prevent crosstalk between channels. The actual measurement shows that the coupling noise between the channels has decreased from 45mV to 3mV after isolation.
Optimize the test layout: Place the power supply, oscilloscope and probe on the conductive rubber pad to form an equipotential plane, which can further suppress common-mode noise. In the OBC test of a certain new energy vehicle, this method increased the stability of 100kHz ripple measurement by 40%.
Four. From single-point testing to full-chain verification
True ripple testing requires the establishment of a closed-loop verification system:
Front-end filtering verification: Parallel connection of X/Y capacitors (such as 0.1μF+10μF) at the power input end can reduce conducted noise by 20dB.
Intermediate loop compensation: Optimize the power supply dynamic response by adjusting the error amplifier compensation network (such as the RCOMP/CCOMP parameters of TI TPS5430).
Back-end load matching: Adding magnetic beads (such as Murata BLM18PG121SN1) at the output end to suppress high-frequency oscillations can reduce the peak value of ripple by 35%.
The practice of a certain data center power supply manufacturer shows that after adopting the triple solution of "probe optimization + ground loop elimination + system compensation", the ripple test repeatability of its 48V/1200W power supply has increased from ±15% to ±2%, and the test efficiency has increased by three times.
V. Breakthroughs in Intelligent Testing Tools
With the evolution of power supply technology towards high frequency and digitalization, traditional testing methods are facing challenges. The "Power Analysis" software built into Keysight's InfiniiVision 1000X series oscilloscopes can automatically identify switching frequencies, calculate ripple parameters, and keep measurement errors within ±1%. The R&S RTO6 oscilloscope from Rohde & Schwarz can distinguish between real ripple and noise interference through machine learning algorithms, achieving a measurement accuracy of 98% at a 1GHz bandwidth.
Today, with the increasing precision in power supply design, ripple testing has evolved from a simple oscilloscope operation to a systems engineering involving materials science, electromagnetic compatibility, and signal processing. Engineers need to establish the concept of "testing as design", and consider the selection of testing tools, the optimization of testing methods and the design of power supply architecture simultaneously, so as to truly break through technical bottlenecks and achieve a qualitative leap in power supply performance.