Noisy power-rail measurements can ruin your day and wreak havoc on system performance. Fortunately, you can mitigate noise from those measurements.
With today’s circuits and systems operating with power rails at 1.2 V and lower, even small variations from nominal can produce bit errors. Jitter, false switching, and transient-related problems can leave you struggling to resolve problems.
Noise measurements on power-distribution networks (PDNs) have become a focal point of debugging and troubleshooting system designs. But, the process in determining PDN integrity isn’t without “gotchas.” In this article, we’ll cover some challenges of PDN measurements and probing that could result in incorrect results and how to surmount them.
Beware RF pickup
Noise from EMI/RFI sits atop the list of challenges which becomes apparent in even a voltage measurement from a 1.5 V battery. Between the battery’s internal electrochemical reactions and a bit of current draw due to probing, we should expect some modest amount of noise on the voltage trace.
Try placing a battery in a holder and probing its terminals; you’ll be surprised at the amount of noise appears on your oscilloscope screen. The top trace in Figure 1 is the voltage trace (magenta, ch2) for the battery. For reference, the bottom trace (yellow, ch1) show’s the oscilloscope’s noise-floor measurement for reference. Both traces use the same vertical scale. The battery’s trace reveals a high level of noise on its voltage, much more than anticipated. The mean voltage measures 1.56 V with a 33 mVPK-PK of noise.
A useful consistency check is to look at this signal in the frequency domain (Figure 2). From the full-spectrum frequency plot (upper trace), we see that the noise is indeed wideband, reaching to the oscilloscope’s full bandwidth (1 GHz in this case) with no sign of attenuation.
The lower trace in Fig. 2 shows a zoomed view of the first 100 MHz of the noise spectrum. It reveals clear peaks in the noise, which, curiously, begin at almost exactly 15 MHz, followed by 30 MHz, 45 MHz, and so on. This is undoubtedly RF noise from an outside source.
The obvious remedy, then, is to properly shield the battery (Figure 3), ensuring that the shielding is connected to the probe’s return line.
The difference from adding the shield is dramatic in the lower trace of Figure 4. Shielding reduced the noise from roughly -60 dBm range to -100 dBm range, representing a 4x reduction, with an amplitude of around 45 nV.
As a final sanity check, let’s compare the battery noise with proper shielding to our oscilloscope’s noise-floor measurement (Figure 5). The oscilloscope noise floor is on Ch1 (yellow, lower trace) and the battery on Ch2 (magenta, upper trace), and they are virtually identical.
Thus, whenever you probe low-level signals with anything other than a well-shielded coax connection, you’ll get interference. Any exposed conductors that are separated from the DUT’s shielding will act like antennas.
EMI-RFI pickup will typically be of a broadband nature. To minimize this aspect, your probe’s tip should be designed as much like coax as possible. Any inductance in that tip will degrade your measurement bandwidth and probably cause some ringing in your measurements. Worse, you will get that “antenna effect” and the probe will be prone to EMI/RFI pickup. Ensure a connection between the oscilloscope and the DUT that looks as much like a coaxial connection as possible.
In terms of design for test, if you can add test points to your board in the form of micro-coax connectors and then connect your coax cable to those points, you will have gone a long way toward minimizing any potential for EMI/RFI in your power-rail measurements.
[Continue reading on EDN US: Know your 10X probes]
David Maliniak is Technical Marketing Communications Specialist at Teledyne LeCroy.