Oscilloscope Basics - 10 Power Integrity Measurement Tips

Let's take the measurement results shown in Figure 3 as an example. The same signal, a 20 MHz 50 mV pp sine wave signal, was measured using a 10:1 probe and a 1:1 probe. The only difference between the two measurements is the attenuation ratio. The value measured by the 1:1 probe is 52 mV pp, while the value measured by the 10:1 probe is 65 mV pp. The higher attenuation ratio reduces the signal-to-noise ratio, so the measured value exceeds the actual value by at least 25%. It can be seen that when measuring small signals, oscilloscope and probe noise may have a significant impact on the measurement results, and it is best to use the smallest possible attenuation ratio.

A Little Lesson About Oscilloscope Noise

See the block diagram below. Noise in an oscilloscope and probe system comes from two main sources. The input amplifier and buffer circuits in the oscilloscope contribute some noise, and the probe amplifier of the active probe also contributes noise.

Oscilloscopes use attenuators to change the vertical scale factor. When a signal is attenuated, the noise of the oscilloscope becomes larger relative to the input. When the attenuator is set greater than 1:1 (the most sensitive oscilloscope hardware range), the noise appears to be proportionally amplified relative to the signal at the oscilloscope input port. For example, the basic sensitivity of the oscilloscope is 5 mV/div when no attenuation is inserted (1:1). In this example, we assume that the noise floor of the oscilloscope at 5 mV/div is 500 µVrms. If we change the sensitivity to 50 mV/div, the oscilloscope will have a 10:1 attenuator in series with the input. The noise now appears to be 5 mVrms relative to the input (500 μV*10). The same thing happens if you connect a probe equipped with an attenuator to the oscilloscope. The noise of the oscilloscope appears to be proportionally amplified relative to the signal at the probe input, and the amplification factor is equal to the attenuation ratio.

Figure 3. Noise comparison when measuring a 50 mVpp sine wave using a 1:1 and 10:1 probe.

Tip 4. Use probe offset to increase dynamic range

Probe offset is a feature of active probes that allows the user to remove DC content from the signal being measured. Probe offset is especially important when there is a small AC signal on top of the DC signal, such as when measuring power supply noise. Figure 4 shows the noise on a 1.5 V power supply measured with and without probe offset. The difference is due to the attenuation applied by the oscilloscope at the larger V/div setting.

Note: Most active probes that provide offset also have an attenuation ratio greater than 1:1, which defeats the purpose of reducing the noise in the oscilloscope measurement system. Some probes, such as the Keysight N7020A power supply probe, provide both the offset function and the 1:1 attenuation ratio. The N7020A has an offset range of ±24 V.

Figure 4. Measuring noise on a 1.5 V DC supply with and without probe bias.

Tip 5. Understand the Disadvantages of DC Blockers

A DC block is a specialized large capacitor that can be inserted between the signal and the oscilloscope input. The purpose of a DC block is to block or remove the larger DC component of the signal so that the oscilloscope can be set to a more sensitive range, which is the same measurement principle as described earlier with respect to using probe offset. The disadvantage of a DC block is that it blocks low-frequency AC content, such as drift or power supply compression, in addition to DC content. Figure 5 shows a comparison of a 5 V DC power supply measured using a DC block and an N7020A power supply probe with probe offset. From this example, you can see that the measurement made with a DC block removes low-frequency power supply drift and can be misleading. In addition, because the DC information is blocked, it is not included in the measurement results, so the DC value of the power supply noise cannot be determined with an oscilloscope. To obtain this information, an additional DMM or similar measurement is required. To illustrate this point, Figure 6 compares the results of a 1.5 V DDR3 power supply measured using a DC block and an N7020A power supply probe with offset.

Figure 5. Loss of low-frequency content (such as compressed supply drift) when measuring DC power supply noise using a DC block.

Figure 6. Oscilloscope measurements do not include DC content when using a DC block. Therefore, to know what DC value the noise is at, additional steps are required (such as measuring the DC value with a DMM).

Tip 6. Minimize the loading effect of the oscilloscope and probes on the power supply

When an oscilloscope probes a system, it becomes part of the system because it is in electrical contact, and thus changes the characteristics of the system being measured. This effect is called loading, and the goal is to minimize it. When measuring a DC power supply, it is easy to create excessive loading if you connect a 50 Ω coaxial cable to the power supply and the 50 Ω input of the oscilloscope. You have good intentions in doing this, choosing a 50 Ω oscilloscope input for its low noise and a coaxial cable for its shielding and low ground inductance, but the 50 Ω termination of the oscilloscope will load the power supply by 20 mA/V. For example, probing a 3.3 V supply in this way will load it with 66 mA from the oscilloscope. A better approach is to use a probe with a 50 kΩ DC input impedance, such as the N7020A power supply probe. Figure 7 shows a comparison of these two methods. First, the power supply is measured with a DMM, which gives a result of 3.31 V. The power supply was then measured with an N7020A power supply probe and the measurement did not change, it was still 3.31 V.

Finally, the power supply was probed with an oscilloscope by connecting it directly to the 50 Ω input, and the supply voltage was measured to drop from 3.31 V to 3.25 V. Not all power supplies will be adversely affected in this way. Some power supplies may be strong enough to drive this additional load, while others may not, or this additional load may affect the characteristics of the PMIC (power management IC) in the system, so extra attention should be paid.

Figure 7. Probe loading effects on power supplies. A 3.3 V supply is probed using a 50 kΩ DC impedance N7020A power supply probe and the same supply is probed directly to the 50 Ω input of an oscilloscope.

Tip 7. Use frequency domain for analysis

Using an oscilloscope's FFT function to view the signal in the frequency domain can be helpful in identifying the source of the power supply noise.

In this example, we have a switching DC/DC converter that converts 5 V to 3.3 V. The converter operates at 2.8 MHz. There is also a 10 MHz clock and a 125 MHz clock running on the PCB. Using the previous technique, we measure the noise on the 3.3 V supply using a Keysight N7020A power supply probe with a 1:1 attenuation ratio, applying a 3.3 V probe offset, and limiting the bandwidth to 500 MHz. The probe is connected to a Keysight S-Series oscilloscope. Figure 8 shows the results of this measurement in the time domain. From the time domain view, we can see a signal with a period of ~360 ns, which is a remnant of the 2.8 MHz, but the noise added to the 3.3 V supply by the 10 MHz clock and the 125 MHz clock is not noticeable.

Figure 8. Time domain view of a 3.3 V DC supply. The remnants of the 2.8 MHz converter can be seen in the zoomed-in middle trace. In the zoomed-in bottom trace, the 10 MHz and 125 MHz clocks are the main sources of noise, though less obvious.

Figure 9 shows a frequency domain view of the same data. In this view using FFT, we have set two different windows, covering two different frequency ranges. We can clearly see a peak at 2.8 MHz, which is related to the frequency of the switching converter, and a spike at both 10 MHz and 125 MHz, indicating that the noise is coupled to both clocks. In addition to the time domain, viewing the noise in the frequency domain also gives us more insight into where the noise is coming from.

Figure 9. Using FFT to determine that the noise from the 2.8 MHz converter and the 10 MHz and 125 MHz clocks is located on the 3.3 V supply.

Notes on FFT in Oscilloscopes

An oscilloscope will capture a waveform for a limited amount of time on each trigger, depending on the memory depth and sample rate. The FFT cannot "see" frequencies in the input signal that are below the inverse of the scope's time capture window; the lowest frequency it can analyze is 1/[1/(sample rate) X (memory depth)]. To see suspected noise sources in the FFT, the memory depth should be set correctly so that enough samples are captured. For example, if your switching power supply is running at 33 kHz, you would need to capture 1/(33 kHz) or 30 microseconds of signal activity to see it in the FFT. At a sample rate of 20 GSa/s, this would require a memory depth of 600,000 points. The FFT typically processes only the data displayed on the screen.

Tip 8. Use triggering to view and measure signal components in power supply noise

Triggering can help visualize and measure components of the power supply noise that couple into the power supplies of other components in the system and are phase-coherent with those components. To demonstrate this, we will use the same measurement system as described in Tip 5 (N7020A probe with offset and bandwidth limiting connected to an S-Series oscilloscope) and use a 2.8 MHz switching regulator to provide 3.3 V and 10 MHz clocks in the system. Figure 10 shows the measurement results. We can see the 2.8 MHz clock and its harmonics in the FFT, as well as the 10 MHz spike that represents the clock. From this, we know that the clock is coupling noise into the 3.3 V supply. Now we will trigger on the clock and turn on averaging. Averaging will remove all random noise and other signal components that are not coherent with the clock. The final result is the portion of the power supply noise that is associated with the 10 MHz clock. Figure 11 illustrates this problem.