Phone chargers produce EMI: We compared four

The switching power supplies that regulate output voltage produce radiated emissions. Our tests compare waveforms that give clues as to how the chargers operate.

Figure 1. The four chargers (l-r): Shenzhen Fujia, JBL, Anker, and Apple.

While evaluating the Uni-T UPO1202 oscilloscope, I used the Uni-T UTG962E Function/Arbitrary Waveform Generator and noticed some noise in my measurements. Some came from the LED task lighting above my bench, so I turned the LEDs off. Still seeing some noise, I checked the function generator’s USB power supply, which is a 5 V, 2 A phone/tablet charger. Now I was curious as to how that charger’s radiated emissions compared to others of similar output power. Doing what any engineer would do, I compared the radiated emissions of the four chargers. Here’s what I found.

The four chargers in Figure 1 were:

Uni-T manufactured by Shenzhen Fujia Appliance Co., (2 A rated)
JBL (Harman International Japan Co.) from a JBL Bluetooth speaker (2.5 A)
Anker model A2620 (two USB-A ports, 2.4 A)
Apple model A1357 (came with an iPad 2, 2.1 A)

Test load

Figure 2. The load consisted of four 10 Ω, 10 W resistors in parallel.

Wanting a stable load, I used a set of four 10 Ω, 10 W power resistors in parallel shown in Figure 2. That produced a constant 2.5 Ω passive load. Measuring the resistor bank with an HP 34401A DMM using the four-wire resistance mode, the resistor bank measured 2.499 Ω.

To connect the load to each power source, I used a universal USB breakout board and connected it to the power sources through its mini-USB port. I chose the mini-USB cable simply because I had a one that’s no longer used. I was going to use a micro USB but if you look carefully at Figure 1, the micro-USB connector (upper left corner) was not aligned with the board’s edge. It’s not soldered very well.

USB DMM — Figure 3. A USB inline multimeter measures the DC output voltage and current from each charger.

At first, I connected the load to the board using clip leads but later changed to a direct-soldered connection with a short wire. When measuring the voltage at the load, I found a significant loss in the connections. The loss was approximately 2 V, leaving the load voltage at around 3 V. The shorter and heavier wires increased the voltage at the load to between 4.55 V and 4.75 V depending on the power source. Table 1 shows the load voltage compared to the source voltage for each charger, measured with a USB inline multimeter shown in Figure 3.

Charger	V_OUT	V_LOAD	I_OUT	W
Anker	5.24	4.75	1.91	9.95
Apple	5.02	4.58	1.83	9.18
Fujia (Uni-T)	5.08	4.55	1.81	9.30
Harmon (JBL)	5.18	4.68	1.88	9.75

Four USB chargers for EMI measurements — Figure 4. The four chargers connected to the DMM and load. Clockwise from upper left: Anker, Apple, JBL, and Shenzhen Fujia.

Figure 4 shows each of the chargers with the oscilloscope ground lead surrounding each.

Anker

Figures 5, 6, and 7 show the emissions from the Anker charger. Figure 5 shows time and frequency domains. The frequency of the pulses is around 190 kHz, which you can assume is the switching rate of the regulator.

Figure 6 removes the frequency plots and focuses on the time domain. As in Figure 4, Figure 5 shows the trigger point. I used an edge trigger to obtain this plot. You can see the ringing as the EMI decays following the switching circuit turning on or off. The switching signals away from the trigger point appear blurred, which could result from slight timing differences or from triggering given that I used edge triggering on the oscilloscope. Another theory is that it could be caused by the oscilloscope screen refresh rate, though Uni-T’s specifications say the refresh rate is 500,000 waveforms/sec. The plot shows a peak-to-peak measurement of just under 100 mV.

Figure 5. A time, frequency, and frequency over time (waterfall) display from the Anker charger shows EMI when the power transistors switch on and off.

Figure 6. A time-domain plot of EMI from the Anker charger shows ringing, something present to varying degrees in all four chargers.

In Figure 7, you can see several peaks around 1 MHz given the resolution of 1.98 MHz/div. These peaks occurred intermittently and were hard to capture. I managed to capture this plot by manually triggering the oscilloscope. The plot also shows a gradual rise of a much lower amplitude peaking around 7 MHz. The video below showes the appearance of the intermittent frequencies. Anker was the only charger to exhibit this condition. Why? Leave a comment.

Anker Charger intermittent EMI — Figure 7. An intermittent frequency peak appeared in the Anker charger. The video below shows it clearly.

Apple

The Apple 10 W charger came from an old iPad 2. As with the other chargers, you can see in Figure 8 the emissions resulting from switching. What’s different about the Apple charger shows up in the frequency domain plot. The frequency peaks occur at regular intervals of roughly 40 kHz. That’s consistent with the 39.6392 kHz measurement shown in the time-domain plot. The frequency of the intervals seems to spread out with each harmonic. Is that due to noise or another condition? None of the other chargers show this condition. Perhaps the Apple charger uses spread-spectrum clocking to modulate the switching frequency enough to lower the emissions and pass regulatory limits. Can you think of another reason? Leave a comment.

Figure 9 shows the enlarged EMI time-domain waveform from the Apple charger. As with the others, we see a double pulse ending with some ringing. This plot shows a maximum peak-to peak measurement of about 125 mV. The trace shows a gradual rise, most noticeable just before the trigger point. That could be caused by transformer saturation. What do you think?

Apple charger radiated emissions — Figure 9. A gradual rise in the Apple EMI precedes a switching event. One theory is that it’s due to transformer saturation.

Figure 10 shows an expanded view of the pulse at the trigger point along with a frequency plot. This emission results from a switcher turning off, as you can see by the initial downward movement. The peak in the frequency plot occurs around 850 kHz.

JBL Harmon charger EMI — Figure 10. The initial drop indicates that the switching transistor is turning off.

Harmon (JBL)

The JBL charger, which came with a combination Bluetooth speaker/power bank, has by far the cleanest emissions based on the frequency plots in Figure 11.

EMI time and frequency plots — Figure 11. The JBL Harmon charger’s radiated emissions has the cleanest frequency plot of the four chargers.

Figure 12 shows the emissions from the switching waveform plotted in the time domain. It displays far less ringing than the others, which accounts for the cleaner frequency plot in Figure 10 compared to those from other chargers. Here. You can clearly see how the left pulse starts with a downward spike while the pulse on the right starts by going upward. That shows the off-on activity in the switching circuit.

JBL Harmon charger low emissions — Figure 12. The initial downward spike indicates the power transistor turning off, the positive pulse shows it turning on. Note the lack of ringing, hence the low emissions.

Shenzhen Fujia (Uni-T)

The Fujia charger came with the Uni-T function generator. The time-domain plot in Figure 13 shows a gradual rise from the first pulse to the second, similar to that in the Apple charger. Have you seen this before? If we are to believe the frequency number in Figure 13’s time-domain plot, it shows a frequency far lower than the other chargers, just 7.63574 kHz. That apparent in the FFT plot, though it shows two small peaks and one larger peak, that being around 5.5 MHz. I attribute that to the ringing in the time-domain plot.

Figure 13. The Shenzhen Fujia charger shows the same gradual rise seen in the Apple charger.

The expanded frequency plot in Figure 14 shows several peaks at roughly 25 kHz intervals given the scale of 49.44 kHz/div. It’s interesting because the highest peak occurs at twice the frequency of the leftmost (lowest frequency) peak.

Figure 14. Expanding on the frequency plot shows three peaks at around 250 kHz intervals.

Figure 15 shows expanded time and frequency plots. In the time domain, we see a small drop followed by a larger one, followed by ringing. What might have caused the initial small drop?

Figure 15. The Shenzhen Fujia charger emits a small amount of energy before the main pulse indicating the switcher circuit turning off.

Further work

Where do we go from here? For starters, remember that EMI emissions are caused by changes in current. Further experiments might be to integrate these waveforms, which will reveal voltage.

Because I conducted these experiments by wrapping a wire around the cases, we’re seeing the results in the near field. What about the far field? That will require a larger antenna to capture the emissions. I have a combination log-periodic TV antenna that also has a pair of “rabbit ears.” That seems worth a try. I’ll have to set the oscilloscope’s impedance to 50 Ω and get a 75 Ω-to-50 Ω network.

A third experiment could involve an AM radio. Yes, I have one somewhere though I can’t recall the last time I used it.

Note: All frequency plots were made using a Hamming Window.