I’m not a professional, but as a hobbyist, I repair various audio and radio equipment from time to time. I don’t have the money to spare for expensive equipment, so I bought a cheap Chinese portable oscilloscope/DMM/signal generator 3-in-1 ZOYI ZT-706S and I was surprised at how much you can get these days for $85.
Its parameters are quite good considering the price and you can make all the necessary measurements of current, voltage, and resistivity with sufficient accuracy. But the main reason I bought it is, of course, the oscilloscope. You can see the waveform, the device will show the voltage and frequency of the signal. It even has a basic spectrum analyzer function called FFT. But the measurement results cannot even be transferred to a computer. It only allows you to save screenshots of the screen in size 320×240. Therefore, I decided to try using AI for analysis. I uploaded the picture and asked: Please analyze the quality of the 12-volt DC power supply that produces this signal at its output. The oscilloscope readings are shown in the image below. Additionally, AC coupling, probe is 1X.
The AI’s response made me think:
Quality Assessment: Poor — significant switching noise present
Readings:
- Vpp (ripple) = 12.0 mV, AC-coupled, 1X probe, 20 mV/div
- Dominant noise frequency = 53.19 kHz (typical SMPS switching frequency)
- Timebase = 10 µs/div
Analysis:
- Ripple magnitude (12 mVpp on 12V) — this is ~0.1% of the DC level, which by itself is acceptable for most electronics (good supplies are typically <1% ripple, often <50 mV on 12V rails).
- However, the FFT (red trace) shows a serious problem. The fundamental at 53.19 kHz and its harmonics (visible as tall spikes at regular intervals) have very large amplitudes relative to the noise floor — the spikes reach into the -20 to -30 dB region while the floor sits around -70 to -80 dB. This indicates strong, narrowband switching-frequency interference riding on the output, consistent with a switch-mode (buck/flyback) supply with insufficient output filtering.
- The Y-T waveform (yellow) shows the ripple isn’t a clean sinusoid — it has a sawtooth/spiky character with sharp transitions, suggesting switching transients (spikes) are coupling onto the output, not just smooth low-frequency ripple from the rectifier/capacitor stage.
Verdict: The 12 mVpp absolute number looks small, but the spectral content reveals dominant switching-frequency noise with strong harmonics — a sign of marginal output filtering (LC stage may need a larger inductor, additional output capacitance, or a ferrite bead/π-filter to suppress conducted EMI). For sensitive analog/audio loads this level of switching noise could be problematic; for digital loads it’s likely tolerable but not ideal.
Recommendations:
- Add/increase output filter capacitance (low-ESR electrolytic + ceramic combo)
- Add a small output inductor or ferrite bead to attenuate the 53 kHz switching harmonics
- Check that the switching converter’s compensation/loop is stable (sharp spikes can indicate marginal stability or poor snubbing on the switch node)
So, it is impossible to agree with the categorical conclusion about the poor quality of the power supply, which the AI itself refutes in the next paragraph, and it is also unclear where it got the numbers “spikes reach into the -20 to -30 dB region while the floor sits around -70 to -80 dB”, but as a basis for further research it looks quite interesting.