Five tips for accurate power integrity measurements using an oscilloscope

As power rail voltages and tolerances get smaller, it becomes increasingly difficult to accurately measure power integrity. In the past, any oscilloscope could measure a 10% tolerance ripple on a 5V rail because the 500mV requirement was much higher than the oscilloscope's noise level; but now, it is difficult to measure a 2% tolerance ripple voltage on a 1V rail, regardless of the oscilloscope used. Here are five techniques for accurately measuring power integrity using an oscilloscope.

Figure 1: Power rail DC voltage and its tolerance

Tip 1: Adjust display characteristics

Waveform intensity

Measuring the DC voltage tolerance of a power rail requires measuring the worst-case peak-to-peak voltage (Vpp), which is best accomplished with automated measurements; sometimes visual judgment is also helpful, and all oscilloscopes have a display setting that allows the user to change the intensity of the waveform. This intensity value is usually set to about 50%, and setting the intensity to a higher value makes it easier to see which oscilloscope pixels correspond to waveforms that occur less frequently. The downside of increasing the waveform intensity is that it makes it more difficult to judge how often the waveform displayed on any particular pixel occurs; while this is important for observing modulated signals, this level of resolution is generally not important for power integrity measurements.

Infinite persistence

Turning on infinite duration mode allows continuously acquired waveforms to be displayed in an overlapping manner; infinite duration is also very useful for archiving, and the oscilloscope can display the DC voltage tolerance range over a longer period of time.

Color grading

Turning on color grading mode generates a 3D plot of the power rail waveform; color grading combined with infinite duration display helps provide deeper insight into the power rail signal.

Tip 2: Reduce Noise

Choosing a Low-Noise Oscilloscope

If the signal strength is less than the noise of the oscilloscope and probe/cable system, you will never measure the signal. The signal will have front-end noise added to it after it enters the oscilloscope and before it enters the analog-to-digital converter (ADC); each stored sample will then contain the original signal value plus some offset, the amount of which depends on the amount of noise present when the sample was acquired. The user will see a thicker waveform on the oscilloscope display, which should not be confused with a fast update rate. Peak-to-peak values ​​greater than the actual signal will be displayed and measured.

The best approach is to use an oscilloscope with lower noise. How do you determine the noise level of an oscilloscope? Most oscilloscope manufacturers provide product specification sheets that list the typical root mean square (RMS) noise values ​​for that particular oscilloscope; these noise values ​​are characterized based on a large number of oscilloscope samples. Noise is a characteristic rather than a specification, and manufacturers only provide typical values ​​for RMS noise, but the peak-to-peak value of noise is actually an important factor that affects the accurate measurement of ripple.

Figure 2: Noise is the primary cause of inaccurate power rail DC ripple measurements

A simple method is to measure it yourself. Rapid characterization takes only a few minutes and does not require the use of external equipment. Disconnect all inputs to the oscilloscope, turn on the Vpp measurement, set the vertical scale and sampling rate for noise measurement, and let the oscilloscope run until a stable and consistent Vpp noise value is obtained. The noise level depends on the vertical sensitivity setting, bandwidth setting, and impedance selection (50Ω or 1MΩ), and there will be slight differences on different channels on the same oscilloscope.

The noise levels of oscilloscopes from different manufacturers can vary by as much as 100%. If you need to accurately measure ripple, make sure you choose an oscilloscope with lower noise levels.

Choose the signal path impedance with the lowest noise

Oscilloscopes used to measure power integrity typically have two signal path impedances: 50Ω and 1MΩ; users can use probes that support either impedance, or use cables to complete power rail measurements.

For oscilloscopes with the above two impedances, the 50Ω impedance is usually less noisy and supports the full bandwidth of the oscilloscope. The noise on the 1MΩ path may be two to three times the noise on the 50Ω path, and the bandwidth on the 1MΩ path is usually limited to 500MHz, so the 50Ω path is the best choice for measuring power integrity.

The output impedance of the power rail is usually in the mΩ range. For cable measurement equipment without any probe, a 50Ω path has a 50Ω DC input impedance, which will cause some loading effect and reduce the DC amplitude value of the power rail. Using a dedicated power integrity probe, such as the R&S RT-ZPR20 with a 50kΩ input impedance, can minimize the impact of this problem.

It is not recommended to connect a 50Ω cable (such as a 50Ω pigtailed coaxial line) directly to the 1MΩ input of an oscilloscope due to reflections that occur between unmatched 1MΩ and 50Ω transmission lines.

Use the most sensitive vertical scale

The oscilloscope noise level is related to the oscilloscope's full-screen vertical scale value. Therefore, using a more sensitive vertical resolution will reduce the total amount of noise measured. In addition, when the signal is amplified so that it covers most of the vertical range, the oscilloscope will make better use of the ADC resolution, and the Vpp measurement value will be more accurate.

Limit bandwidth

The noise is broadband in nature, and by turning on the FFT function with no inputs connected to the oscilloscope, the noise can be seen across the entire bandwidth of the oscilloscope. Turning on the bandwidth limit filter can reduce broadband noise and help to more accurately measure the power rail, but the disadvantage is that if the bandwidth limit is set too low, higher frequency anomalies will not be shown.

How much bandwidth should be used? The answer is that it depends on the specific signal. Although the switching speed may be in the kHz range, fast edges will generate harmonics in the MHz range. For higher frequency coupled signals, including frequency harmonics, a larger bandwidth is required to capture these signals. The R&S RTO and R&S RTE digital oscilloscopes are equipped with bandwidth limiting filters. In addition, the HD mode can further reduce the broadband noise and increase the vertical resolution to 16 bits.

Selecting the appropriate probe (attenuation, bandwidth, and connections)

Using a probe with a 1:1 attenuation ratio can significantly improve the accuracy of power integrity measurements; probes with higher attenuation ratios will amplify noise, and higher attenuation ratios will limit the vertical sensitivity that can be used. For example, using a probe with a 1:1 attenuation ratio on an oscilloscope with an input as low as 1 mV/div can reduce the sensitivity to 1mV/div, while using a probe with a 10:1 attenuation ratio can only be set to 10 mV/div.

How you probe the power rail signals is just as important as any other technique. Some users connect the rails to SMA connectors for high signal quality and ease of connection; others choose to solder the connections, others use a fixture on the bypass capacitors as an easy contact point; and still others use a handheld probe. Each technique has its pros and cons in terms of ease of use, required upfront planning, and signal quality.

Figure 3: For small signals, using a probe with a 1:1 attenuation ratio can yield more accurate measurements.

To obtain high-precision measurements, Rohde & Schwarz recommends using the R&S RT-ZPR20 power integrity probe with an SMA connector or soldered 50Ω SMA pigtail coax (probe included). This probe provides a full 2GHz bandwidth with very low noise. Although the bandwidth of this probe is specified at 2.0GHz, its frequency response has a slow roll-off and can capture 2.4GHz Wi-Fi signals that may be coupled to the power rail. Although the 2.4 GHz amplitude value will be attenuated to about 3dB, the ability to capture these coupled signals is very important for finding the coupling source.

When using the R&S RT-ZPR20 probe with the R&S RT-ZA25 probe head (browser), the bandwidth will be reduced to 350MHz. Using a grounding device that can minimize the ground loop area, such as a ground spring, can effectively improve measurement accuracy.

Tip 3: Achieving Sufficient Offset

AC coupling and blocking capacitors

The offset built into an oscilloscope is usually not enough to allow the user to center the waveform on the display and zoom in, which has two negative consequences: the oscilloscope uses only a small portion of the ADC vertical resolution and uses a larger vertical scale, which introduces additional noise; this degrades the measurement quality.

If you use blocking caps on selected paths and probes or use the oscilloscope in AC coupling mode, the DC component of the signal will be removed; this can partially solve the problem, but the actual DC value and drift will not be visible.

Probe with built-in offset

Some probes have additional built-in offset, which has the advantage of allowing the user to get enough offset to see the true DC value and low-frequency characteristics such as drift and sag. The R&S RT-ZPR20 power integrity probe has a built-in offset of ±60V and a dynamic range of 850mV, which means that the user can view AC characteristics up to 850mV on the DC power rail within the range of -60V and +60V.

Tip 4: Evaluate Switching and EMI

Frequency Domain Plot

Characterizing a power rail usually requires ensuring that there are no interfering signals coupled to the rail, and users sometimes need to consider switching harmonics. These interfering factors cannot be determined by looking at the time domain waveform, but they can be seen in the frequency domain through the FFT function of the oscilloscope.