When probing power-distribution networks (PDNs) in embedded systems, many considerations come into play. Signals of interest are often very small signals riding atop larger voltages, so measurement system noise must be well understood and managed. Impedance mismatches in the signal path can cause reflections of higher-frequency content. Some probing options may not offer enough offset range to match rail voltages. Bandwidth limitations can impede capture and characterization of high-frequency noise. And, with PDN impedances often being as low as 1Ω and below at DC, low-impedance probing options can place unacceptable loading on the DUT. In this Application Note, we will discuss various power-rail probing options for Teledyne LeCroy’s 12-bit High Definition oscilloscopes and how they fare in terms of the above considerations.
Overviews of Probing Options and Noise Comparisons
Generally speaking, there are five approaches to probing power rails with High Definition oscilloscopes (Table 1):
As with most oscilloscopes, all of Teledyne LeCroy’s 12-bit instruments come with a set of 10-MΩ passive probes. Bandwidth for such probes is typically ~500 MHz. When connected to a 1-MΩ coupled oscilloscope input, the probe’s high impedance provides good DC loading properties.
Grounding must be a consideration with the 10-MΩ passive probe. Using the typical 3” ground lead can result in high levels of RF interference and high inductance and signal ringing. The shorter spring-type ground lead is generally preferable; it presents a smaller “antenna” and thus less RF pickup as well as a shorter inductance loop that reduces ringing (see Application Note titled, “Getting the Most Out of 10x Passive Probes”).
The 10-MΩ passive probe has 10:1 attenuation, which means that the signal is attenuated 10x but the noise in the measurement system is unaffected—thus, SNR is 20 dB lower than a 1:1 probe or direct connection. As can be seen from Figure 1, the signal is displayed as 80 mV full scale because of the 10:1 attenuation, which limits sensitivity to 10 mV/div or higher in this case.
(Note: To compare the noise performance of all five methods, we applied each to the same 900-mV rail. All signals were bandwidth-limited to 500 MHz for an equitable comparison, and input levels were adjusted to maximize SNR as available at the required offset. In each example, the lower traces are zooms at 5 mV/div.)
The second probing option for power rails with High Definition oscilloscopes is a coaxial connection to the oscilloscope’s 1-MΩ input. The chief advantages of coaxial connections to the DUT, whether designed into the DUT or achieved via a solder-in connection, are two: high bandwidth at the connection point and a small ground loop for low RF pickup.
Using the oscilloscope’s 1-MΩ input coupling brings the advantages of low loading and 1:1 attenuation for low noise. On the other hand, there will be limited bandwidth of 1 GHz or less at the oscilloscope end and possible reflections of high-frequency content.
Figure 2 depicts measurement of the 900-mV rail with a coaxial connection to the oscilloscope’s 1-MΩ input. With no attenuation, this approach allows high SNR, but the impedance discontinuity from the 50-Ω coaxial cable to the 1-MΩ input can cause reflections. The high offset capability of Teledyne LeCroy’s HD4096 oscilloscopes means we were able to offset the input sufficiently to match the 900-mV rail while maintaining maximum vertical sensitivity.
A third option, a coaxial connection to the oscilloscope’s 50-Ω input coupling, differs from the previous option in that it offers even higher bandwidth. However, the small offset range it affords can be an issue for some oscilloscopes. There may also be possible loading issues when a power rail with extremely low impedance (much less than 1 Ω) is loaded with the 50-Ω oscilloscope input.
Measuring the 900-mV rail with a coaxial connection to the oscilloscope’s 50-Ω input also allows high SNR, and Teledyne LeCroy HD4096 High Definition oscilloscopes have sufficient headroom to offset with the oscilloscope input without adverse sensitivity effects (Figure 3).
Because this rail presents a high impedance, the 50-Ω load is not an issue. However, were it a very low-impedance rail, the loading could be a significant problem.
Probing option #4 is a 10:1 coaxial probe. These can be homemade using coaxial cable and a 450-Ω resistor or purchased commercially. A 10:1 coaxial probe attenuates the input signal by 10:1 when connected to the oscilloscope’s 50-Ω coupled input. Effective bandwidth can be very high, being a function of the quality of connection to the DUT. However, the 10:1 attenuation exacts the same 20-dB noise penalty as the 10-MΩ passive probe (Figure 4). Loading at 450 Ω may be acceptable depending on the PDN impedance.
Lastly, the fifth option, the voltage-rail probe, is specifically designed for probing voltage rails (an example is Teledyne LeCroy’s RP4030 active voltage-rail probe). This probe offers high bandwidth of 4 GHz, low noise with attenuation of just 1.2x, and high offset capability of ±30 V DC.
When measuring the 900-mV rail with the RP4030 voltage-rail probe, the probe’s 1:1 attenuation delivers low noise, while its large DC offset range of up to 30 V is more than capable for the task at hand (Figure 5). It also presents 50-kΩ loading at DC, which means no significant current draw from a low-impedance rail.
Offset and Sensitivity Comparison: 3.3-V Rail
Meanwhile, the voltage-rail probe can achieve full sensitivity at a very large offset, which enables higher SNR and more accurate measurements (Figure 7). Measuring at a full-scale input range of 40 mV pk-pk means more ADC resolution being applied to digitize the signal, which accounts for the SNR improvement.
Bandwidth Comparison: Noisy 3.3-V Rail
To compare the bandwidths of the five approaches to voltage-rail probing, we deployed a 4-GHz oscilloscope to measure a 3.3-V rail with copious pollution from a high-frequency clock signal. Again, input levels were adjusted to maximize SNR as available at the required offset.
Being bandwidth-limited to 500 MHz, the 10-MΩ passive probe was incapable of capturing all of the signal’s high-frequency content (Figure 8, top left). Similarly, the coaxial connection to the oscilloscope’s 1-MΩ-coupled input has more bandwidth at 1 GHz, but still not enough to cover the full spectrum of the noisy clock signal (Figure 8, top right).
Using the coaxial cable to the oscilloscope’s 50-Ω input fares somewhat better, in that there is 4 GHz of bandwidth: enough to capture the full spectrum of the noisy signal. However, to achieve sufficient offset at this coupling and bandwidth condition, sensitivity is limited (Figure 8, center left).
For its part, the 10:1 coaxial probe also has 4 GHz of bandwidth for full-spectrum capture. But the probe’s 10:1 attenuation comes with a 20-dB reduction in SNR (Figure 8, center right).
When using the RP4030 voltage-rail probe to measure the noisy 3.3-V rail, not only does the probe have enough bandwidth for full-spectrum capture but it also has no attenuation and a very broad offset range (Figure 8, bottom). In an overlay comparison of the 10:1 coaxial probe and the RP4030 probe (Figure 9), note the difference in noise floor between the former (in green) and the latter (in grey).
The various options for probing power rails with Teledyne LeCroy’s 12-bit High Definition oscilloscopes present their respective strengths and weaknesses. Some, like the 10-MΩ passive probe, present low loading to the power rail but also offer limited bandwidth. Others, such as a coaxial connection to the oscilloscope’s 50-Ω input, are problematic in terms of loading if rail impedance is low but offer high bandwidth. The exception is the RP4030 voltage-rail probe, which is designed specifically for power-rail probing and is architected to eliminate these tradeoffs.