Keysight Technologies: Understanding power integrity measurements

By Keysight Technologies
Download White Paper
Share:

Providing a clean power source has become increasingly important as the speed of microcontroller-based circuits has risen. Noise on DC power rails can cause clock and data jitter in digital systems, while a voltage drop can cut the propagation delay through gates, reducing circuit timing margins and even causing bit errors. Read More


Increasing clock speeds have made it more likely that switching noise will be induced into the power supply. Reducing signal amplitudes in digital systems demand narrower noise margins on their power supplies, sometime to as little as +/- 1 %, which also makes noise control increasingly important.

This combination of factors means that designers have to find ways to accurately measure ever smaller and faster AC noise signals on DC power supply rails, in order to find their sources and so reduce their impact. Here are some key issues of which designers need to be aware to control supply noise effectively.

DC power supply noise sources

The first step in controlling noise from a DC power supply is to understand where it comes from. The dominant sources of noise on a DC power rail are the supply’s own switching, and the transient current demands created by the switching currents of devices in the circuit. The noise created by switching events may appear random, but tends to align with clock signals in the system.

Thermal agitation of electrons in the supply’s circuitry will also create Gaussian noise. It’s helpful to remember that the noise you measure on the supply is the sum of several noise sources, since this makes it easier to work out the series of steps necessary to mitigate each source of noise.

Power delivery networks and power integrity

Power integrity (PI) checks involve analysing how effectively power is converted and delivered from source to load, through a power distribution network made up of passive components and interconnects. Checking PI usually involves making measurements from DC to multi-gigahertz.

Some common PI measurements include:

  • PARD — periodic and random deviation represents the deviation of the DC output from its average value, when other parameters are constant. This measurement represents the undesirable AC and noise components that remain in the DC output, even after any regulation and filtering circuitry. It is most often measured as an RMS value, at between 20 Hz and 20 MHz. Variations below 20 Hz are known as drift.
  • Load response — this measures a supply’s ability to remain within specified output limits given a predetermined static or transient load. Load response usually reflects the time it takes the supply to settle back into a predefined band after a load is applied.
  • Noise — deviations of the DC supply from its nominal value.

Oscilloscopes: the flexible measurement option

Oscilloscopes have the bandwidth necessary to match the spectrum of noise signals on many DC supplies. Working with them means understanding their limitations, such as the fact that they have internal noise sources which can be confused with the noise signal that you are trying to measure on the supply.

Dynamic range is another issue, since the AC noise signal under measurement is usually only a very small fraction of the DC level of the power supply. This makes it difficult to examine the details of the AC noise without turning up the sensitivity of the scope so high that it is swamped by the intrinsic noise of the instrument.

Scope noise issues

Where does this noise come from? Take a look at Figure 1:

Figure 1: Noise sources in a scope (Source: Keysight Technologies)

It shows two main noise sources: the input amplifier and buffer circuits in the scope, and the amplifier of any active probe that is being used with it.

Scopes use an attenuator to vary vertical scaling, and as the diagram shows, the scope’s noise arises after this attenuation occurs. When the attenuator is set to more than 1:1 (the most sensitive range), the scope’s noise will appear larger relative to the signal at the scope’s input.

For example, consider a scope with a sensitivity of 5 mV/division without attenuation (1:1), and a noise floor of 500 μVrms at 5 mV/division. To change the sensitivity to 50 mV/division, the scope puts a 10:1 attenuator in series with the input. The noise then appears as if it were 5 mVrms, relative to the input (500 μV x 10).

The same thing happens when an attenuating probe is used, since the scope noise will then appear larger relative to the signal at the probe’s input, by the amount of the attenuation.

 

Using a scope’s FFT functions

An oscilloscope’s FFT capabilities can be useful in showing signals in the frequency domain, which helps identify sources that contribute to noise on a supply.

This facility needs to be used with some care. An oscilloscope will capture a finite amount of time on each trigger, depending on its memory capacity and the sampling rate. The FFT cannot ‘see’ frequencies in the incoming signal that are less than the inverse of the scope’s time-capture window, that is 1/[1/(sampling rate x memory depth)].

Figure 2: Comparison of noise on 1:1 and 10:1 probes, measuring a 50 mVpp sine wave (Source: Keysight Technologies)

To see a suspect source in the FFT, therefore, designers must set the scope to capture enough samples. For example, if a switching supply operates at 33 kHz, users would need to capture 1/(33 kHz), or 30 μs, of signal activity to see it in the FFT. For a sampling rate of 20 GSample/s, this would require 600,000 memory points.

Conclusion

Measuring AC noise on a DC power supply is increasingly important, but not as straightforward as it may, at first, seem. Supply noise is created by multiple sources, and can easily be confused with the intrinsic noise of your measurement instrument.

To make useful power integrity measurements, therefore, designers should use the lowest noise input on their scope, and the lowest attenuation probes possible, after making a null measurement to set a baseline. The scope’s bandwidth should be set to be just enough to capture troublesome transients and noise but no more, to avoid adding noise to the measurement unnecessarily. FFT facilities can help understand the source of various aspects of the noise on the supply, but must be applied wisely to ensure they can capture enough of the signal to enable proper diagnostics.


Disclaimer: by clicking on this button, you accept that your data might be communicated to this company. If you do not want us to communicate your data, please update your details on your profile

Download White Paper
White Papers
10s