Figure 1 shows the
test setup used to generate the data. The setup is composed of a
switching power supply, a (black) wire draped across the supply, an Agilent N9320B
spectrum analyzer, and an
Fischer F-33-1 current probe
The power supply is powering a 48 Watt, 12 Volt incandescent light bulb
through the yellow test leads. The black wire passing through the
F-33-1 current probe is about three meters long and its center has been
positioned over the power supply. The plot on the analyzer screen in
Figure 1 extends from zero Hz to 100 MHz at 10 MHz/div.
noise floor on the screen of the analyzer is about 30 dBuV, a level
that corresponds to enough current flowing in a one half wavelength
dipole antenna to just hit the CISPR/FCC class B emissions limits in
the 30 to 100 MHz frequency range. You can see the level of current
near 40 MHz is about 20 dB above that level and could cause a Class B
emissions problem. In reality, the power supply exhibits broad band
noise, but the current in the wire peaks at the frequency at which the
wire is a one half wavelength dipole. At that frequency, the driving
point impedance of a dipole is near 70 Ohms, its minimum value, and the
induced voltage from the power supply in the black wire is able to
produce a current maximum at the center the wire.
field from the power supply is probably not physically large enough to
radiate efficiently at 40 MHz, but once it induces enough current in
the nearby black wire, the wire itself is long enough to be an
efficient radiator. That is what has happened in this case.
know that the source is the power supply for this simple case, but in
some cases the source might not be so obvious. One way to discover if a
broadband noise, such as shown on the analyzer screen in Figure 1, is
due to a switching power supply is to adjust the analyzer as shown in
Figure 2. Spectrum Analyzer Screen Centered at 43.5 MHz with a 100 kHz/div Horixontal Scale and a Resolution Bandwidth of 1 kHz
analyzer has been adjusted in Figure 2 to focus the plot around a
center frequency of 43.5 MHz with a frequency span of only one MHz, 100
kHz/div. The resolution bandwidth has also been lowered from 100 kHz in
Figure 1 down to one kHz, a value consistent with the horizontal scale
of 100 kHz/div. With these analyzer parameters, you can clearly see
peaks in Figure 2 separated by about 120 kHz, the fundamental frequency
of the switching power supply used for this experiment.
originating in switching power supplies that operate at a
constant frequency will produce a spectrum plot similar to Figure 2,
although the spacing of the harmonics may be different. That the noise
is actually coming from a specific supply can be determined by holding
a small loop probe near the power supply to pick up induced noise.
Display the loop output on an oscilloscope or spectrum analyzer to
determine the fundamental frequency. If the observed noise has a
harmonic spacing the same as the fundamental switching frequency of the
power supply, the supply is the likely source of the noise unless there
are other supplies in the system with "the same" fundamental frequency.
For the case where there are several power supplies with "the same"
fundamental frequency, they are likely slightly different and this
information can be used to narrow the source of the noise down to one
supply out of several.