Search
Power supply: If several components are installed on one board, the direct currents must be measured exactly.
Power supply: If several components are installed on one board, the direct currents must be measured exactly.
( Source: gemeinfrei / Unsplash)

Embedded Systems Power-Rail: Measuring the DC power supply in embedded systems

| Author / Editor: Lee Morgan * / Jochen Schwab

Integrated systems (embedded systems) on chips as well as FPGAs and processors: If several components are installed, several supply voltages are required. Special high-impedance probes are required.

Many electronic components require several supply voltages to function correctly. This applies in particular to integrated systems such as microcontrollers, FPGAs and processors. Also, voltage levels have decreased over time, reducing noise tolerances on power rails. Besides, other factors increase the noise on the lines:

  • Performance efficiency functions: Power gating and dynamic voltage and frequency scaling as well as DVFS.
  • Dynamic loads with fast transients.
  • Greater crosstalk and coupling.
  • Switching regulators with faster rise times.

To accurately determine noise, the measurement engineer must measure very carefully. This is where specially developed power rail probes can help.

Gallery

Detailed view of the Power-Rail and its characteristics

Image 1: The noise components of a DC power supply.
Image 1: The noise components of a DC power supply.
(Bild: Tektronix)

The developer should have a close look at every DC line. This is the only way to determine whether the supplied voltage is within tolerance. This includes the DC value on the line and the AC noise. The AC noise in a supply voltage signal can be further divided into broadband noise, periodic events and transient events (Image 1).

Before these types of noise can be minimized, they must be detected and precisely measured. There are several measurement challenges when measuring busbars:

  • Bandwidth requirements.
  • Management of measurement system noise and sample noise.
  • Optimize the dynamic range of your measurement system.
  • Loading of the circuit by the sample (sample loading).

Image 2: Channel 3 (red curve) shows a measurement on a power rail that has an interference distance with a high-frequency signal. If the total energy is too high, such interference can affect the function and lead to defects.
Image 2: Channel 3 (red curve) shows a measurement on a power rail that has an interference distance with a high-frequency signal. If the total energy is too high, such interference can affect the function and lead to defects.
(Bild: Tektronix)

Looking at some designs of power supplies, it may seem that the bandwidths of the measuring systems used are sufficient for a few tens of megahertz. Most designs switch from hundreds of kHz to several MHz. Larger designs and devices operated with higher supply voltages were less susceptible to interference. Therefore, noise above 20 MHz was rarely a problem. For these reasons, current designs are very sensitive to higher frequencies and are also exposed to higher frequency noise sources (Image 2).

Causes of noise and harmonics

The design of the power supply has also changed. While the basic switching frequencies of power converters are still relatively slow, the rise times have increased considerably thanks to faster switching technologies. Fast switching of the power supply, cross-coupling, simultaneous switching to chips and other interfering factors can generate noise and harmonics at much higher frequencies in the power distribution network. Tektronix, the metrology provider, offers 1 GHz and 4 GHz probes specifically designed for power rail measurements.

To measure the noise of the power rail accurately, the measuring system has to bring in minimum noise contributions. The less noise added by the instrument, the better the signal integrity of the DUT under test. If you examine a baseline noise measurement of both the instrument and the connected sample, you will quickly get an idea of the overall noise performance of the system.

Measuring and estimating the additive noise

Image 3: Channel 1 (yellow curve) shows an oscilloscope signal without a signal. Channel 2 (blue curve) is a TPR1000 Power Rail Probe with short-circuited input. At a bandwidth of 1 GHz, the probe adds only 17 µV of additional noise.
Image 3: Channel 1 (yellow curve) shows an oscilloscope signal without a signal. Channel 2 (blue curve) is a TPR1000 Power Rail Probe with short-circuited input. At a bandwidth of 1 GHz, the probe adds only 17 µV of additional noise.
(Bild: Tektronix)

Simple measurements such as peak-to-peak and RMS value of the voltage applied to the inputs (without applying a signal) are a quick and easy way for the measurement technician to detect and estimate the additive noise of the measurement system (Image 3).

Image 4: Channel 2 (blue curve) shows the signal with 157.1 mV peak-to-peak noise, measured with a 10x passive sample. For comparison, the TPR1000 is shown on channel 1 (yellow curve), which shows a better performance with a peak-to-peak value of 38.7 mV.
Image 4: Channel 2 (blue curve) shows the signal with 157.1 mV peak-to-peak noise, measured with a 10x passive sample. For comparison, the TPR1000 is shown on channel 1 (yellow curve), which shows a better performance with a peak-to-peak value of 38.7 mV.
(Bild: Tektronix)

Probes with a large attenuation offer a wide dynamic range, but introduce additional noise into the measurement system. This is also because amplification in the oscilloscope is necessary to compensate for the attenuation (Image 4).

The reason for this is that the signal is divided by the damping factor, which brings it closer to the noise level of the measurement system.

This is calculated with the so-called signal-to-noise ratio or SNR according to Formula 1.

The reason for this is that the signal is divided by the damping factor, which brings it closer to the noise level of the measuring system. This can be illustrated by calculating the signal-to-noise ratio (SNR) in Formula 2.

A measuring head with a low attenuation like the Power Rail Probe with 1.25x has a signal-to-noise ratio like in Formula 3.

Set the oscilloscope as sensitive as possible

Image 5: The effect of vertical scaling on the measured noise. Both channels have nothing connected to the input. Channel 3 at 1 mV/div has 521.2 µv peak-to-peak noise compared to channel 4 at 100 mV/div with 8.953 mv. For channel 4, the value is 8.953 mV <1% of the full-scale voltage.
Image 5: The effect of vertical scaling on the measured noise. Both channels have nothing connected to the input. Channel 3 at 1 mV/div has 521.2 µv peak-to-peak noise compared to channel 4 at 100 mV/div with 8.953 mv. For channel 4, the value is 8.953 mV <1% of the full-scale voltage.
(Bild: Tektronix)

The noise behavior of an oscilloscope can be scaled via the vertical sensitivity. Settings with higher vertical sensitivity provide lower noise than settings with lower sensitivity. A maximized displayed signal on the screen provides higher resolution and a more accurate representation of the signal. Lower vertical sensitivities can cause signals to often appear to have more peak noise than they actually do (Image 5).

Image 7: Load changes can lead to low-frequency power failures on power rails. The AC coupling masks these low-frequency changes.
Image 7: Load changes can lead to low-frequency power failures on power rails. The AC coupling masks these low-frequency changes.
(Bild: Tektronix)

Features such as High Res on a Tektronix 4, 5 and 6 series mixed-signal oscilloscope further reduce noise. A higher sampling rate is used to generate samples with higher resolution. This function uses sample averaging and applies special hardware and FIR filters based on the current sampling rate. The filter allows the maximum bandwidth to be used for a given sampling rate and filters out any aliasing effects that may occur. .

How to establish the best measurement connection

The connection to the DUT is one of the most important factors influencing the measurement of quality. It provides low inductance paths to ground and minimum effective capacitance, noise is reduced and the highest possible bandwidth is provided. The best connections are made using solder adapters and high-performance connectors. Microcoax and Flex Solder Adapter provide a semi-permanent connection to the DUTs when repeated tests are performed on an unplanned test point. Small form factor RF connectors such as MMCX and SMA allow repeatable and reliable access to signals, but must be planned into the design.

Browsers and adapters can be used for a faster and more convenient connection. Tektronix offers a dedicated power rail browser that provides 1 GHz bandwidth. Note that accessories reduce the bandwidth of the system. For example, flying lead square pin adapters usually do not have more than a few hundred MHz effective bandwidth. The bandwidth is further reduced the more clips and other connection aids are used.

Equip measuring system with sufficient offset

Image 8: Example of a device that adjusts the input voltage when the frequency is increased. The frequency component with about 2 Hz between the steps will be omitted with many AC coupling filters
Image 8: Example of a device that adjusts the input voltage when the frequency is increased. The frequency component with about 2 Hz between the steps will be omitted with many AC coupling filters
(Bild: Tektronix)

Low sensitivities (high V/div settings) provide poor measurement results because the oscilloscope uses less of its dynamic range for measurements. For high sensitivity, the DC component of the power rail voltage must be removed. The DC offset can be removed using an AC coupling on the oscilloscope. However, the DC offset masks possible low-frequency events: Voltage dips (Fig. 7) from load changes and dynamic frequency and voltage scaling (Image. 8).

A DC offset to the input signal and a DC coupling on the oscilloscope provide a more complete picture of the device behavior. Oscilloscopes and differential probes often support a certain DC offset. However, the front end of some oscilloscopes limits the available offset together with the selected vertical sensitivity. At lower V/Div settings, the instrument has less offset. The power rail probes provide the measurement system with sufficient offset to support DC coupling on most power rails. The TPR4000 and TPR1000 probes offer ±60 V DC offset, covering the most common standards in automotive, industrial and data center applications.

High-impedance probes with direct current

If power rails are to be measured, the chosen approach must be able to display the high-frequency AC component on the DC supply without overloading the DC component of the signal so that it is not inaccurate or interferes with device operation. High impedance probes offer low DC loading, but cause significant interference and may not have sufficient bandwidth to capture important high-frequency signal content with the oscilloscope.

The 50-ohm signal path on an oscilloscope provides the lowest noise contribution, but 50 ohms would put considerably more strain on the power rail with a direct current. An optimal power rail measurement sample has a high DC resistance and a 50 ohm AC transmission line.

This article was first published in German by Elektronikpraxis.

* Lee Morgan is Technical Marketing Manager at Tektronix, specializing in embedded, power and automotive applications.