Making ripple measurements has traditionally been using an oscilloscope. For a number of technical reasons, an oscilloscope, in general, is not necessarily the best choice for measuring ripple. These include relatively poor sensitivity, poor signal-to-noise performance (SNR), low resolution (generally 8 bits), and insufficient bandwidth. Bandwidth is generally reduced for low amplitude AC coupled signals.
There are several common issues. The first issue occurs when we try to measure ripple with a 1X AC coupled scope probe. A typical 1X scope probe only provides a bandwidth of 8-10MHz, while an uncompensated probe can show peaking at higher frequencies. Second, many scopes reduce the available bandwidth when used at the most sensitive settings or when AC coupled is enabled. Scopes often have a higher noise floor when set to high input impedance mode. Fig. 1 shows the typical frequency response of a 1X/10X scope probe in both the 1X and 10X positions.
A 50 Ω coaxial cable, on the other hand, terminated into 50 Ω, offers a very flat response and, with unity gain, maximizes sensitivity of a passive connection. If the circuit being tested can tolerate the low AC impedance presented by the coax cable and the 50 Ω termination, this is a better solution than the 1X or 10X probe.
Couple Coaxial Cable
In order to make the best use of the scope’s sensitivity and bandwidth it is necessary to AC couple the coax cable to the 50 Ω scope input using something for DC voltage blocking other than the scope’s AC coupled mode. Here, we use the Picotest J2130A DC Bias injector. Keeping the scope in DC mode with a 50 Ω termination lets us obtain the maximum measurement bandwidth.
Fig. 2 is a sample measurement of a point of load regulator, Texas Instrument’s LMR10515Y 5.5V, 1.5A Simple Switcher that is part of the Picotest VRTS2 demo board.
Fig. 3 shows the measurement setup for the ripple measurement using the LeCroy 10X scope probe connected to a Picotest J2180A preamplifier (adds 20dB of gain). The VRTS2 demo board is part of a power supply measurement learning kit that was created to help engineers come up to speed on using a VNA to perform all types of power supply measurements. The output of the POL is coupled to the scope using a preamplifier and a bias injector to remove any DC component from the signal. This allows setting the scope to its highest sensitivity without over-ranging. Some high frequency is evident in the left side of Fig. 4 as a result of the inductance of the ground clip resonating with the capacitance of the probe. On the right side of Fig. 4, a 50 Ω coaxial cable with very short leads is connected to the header close to the POL output capacitors. The coax is AC coupled to the scope using the bias injector and the scope is set to DC 50 Ω input and full bandwidth. The measurement is much cleaner, though the amplitude and general wave shape are approximately the same. The resonant ringing is eliminated due to the very short leads.
While we have successfully measured the ripple and noise using two configurations, the time domain results still offer little information about the spectral content. The spectral content is often more important than the peak-to-peak result, since the spectral content provides insight into the source of the ripple and noise, as well as information about how to filter out the noise if the signal levels are too high.
We can conveniently perform frequency domain measurements using the spectrum analyzer built into the Lecroy 640zi. This measurement uses the same test setup used for the right side of Fig. 4. The spectrum analyzer is set to measure from 100kHz to 50MHz, though the instrument capability is much wider, both lower and higher in frequency. The results of this measurement, shown in Fig. 5, provide insight into the sources of the noise. We can clearly see the switching frequency (and the harmonics) from the 2.8MHz point-of-load regulator. We can also see signals related to a 10MHz clock which is also on this demo board and powered by the point-of-load (POL). We can now see that there are two sources of noise and the levels of each harmonic of each of the sources.
Converting the 2.8MHz spectrum measurement (-37.81dBm) to a peak-to-peak amplitude we have:
The majority of the 14mVpp signal is due to this fundamental frequency, which is also evident in the nearly sinusoidal shape seen in Fig. 4.
The ripple and noise from the regulator is not uniform across the printed circuit board (PCB). It varies with location due to the finite impedances of the printed circuit board traces and the proximity to the decoupling capacitors.
In the top image of Fig. 6, the measurement is repeated using the 1X 60MHz scope probe to show how significantly the measurement is degraded. In the bottom image of Fig. 6, the measurement is made at one of the PDN SMA connectors as shown in Fig. 2. This location is further from the point-of-load output capacitors and closer to the clock, which does not have a local decoupling capacitor. The decoupling capacitor is purposely left out in order to display this point more clearly. Note that in this measurement the point-of-load signal is much smaller than the clock signal.
The FFT spectrum measurements in Fig. 5 and Fig. 6 provide much more detailed information on frequency and harmonics that can be used to determine the source of the noise and improve the decoupling performance to better filter the ripple.
We have shown that the traditional method of using an oscilloscope to make ripple measurements does not always produce the high fidelity and bandwidth we expect or need. In most cases, engineers need to make both time domain and spectrum domain measurements. We have also shown how to use the time domain and spectrum domain information to determine the sources of the ripple. This ripple is not always due to the switching regulator, but sometimes is due to what the power supply is driving.