|
Due to the trends of higher switching speeds, faster slew rates, more active pins per package, and smaller signal swings, power supply noise has risen to a very high level of concern in new digital designs. Real-time oscilloscopes are commonly used to measure power supply noise. This article illustrates techniques for analyzing power supply noise by means of an example, and discusses selection and evaluation of tools for power supply noise measurements.
The problem
Three trends conspire to elevate power supply noise to a very high level of concern in new digital designs. Higher switching speeds and slew rates, more active pins per package, and more total gates on each IC all lead to larger and faster switching current demands on power supplies. Increased active pin counts are often the worst offender, as pin drivers have much higher current demands than internal gates.
At the same time circuits are becoming more susceptible to power supply noise. Decreased unit intervals mean shrinking timing margins. Reduced signal amplitudes translate to reduced noise margins.
As with all engineering problems, understanding the problem and having accurate and precise measurement data to characterize the problem are essential to solving the problem.
Insights to "noise"
Before we go any further, let's consider where noise comes from. Ideally there wouldn't be any noise on your power supplies. How did it get there?
In addition to simple Gaussian noise that arises due to unavoidable thermal processes, which by the way is usually NOT the dominant source of noise, almost all noise on power supplies comes from one of two sources.
Switching power supplies create their own undesired noise, usually at harmonics of the switching frequency or coherent to the switching frequency.
When gates, and especially output pin drivers, switch, this creates transient current demands on the power supplies. This is usually the primary source of noise in most digital circuits. These events may appear random in time, however they tend to be coherent with clocks in the system.
Once we realize we can think about these influences as "signals" superimposed on the power supply instead of thinking about them as "noise," the analysis gets a lot simpler and more powerful.
Measurement challenges
Due to the wide bandwidth of power supply noise, an oscilloscope tends to be the preferred measurement tool. Oscilloscopes can also provide unique insights into the cause of noise, as I will illustrate later in an example.
Real-time, wideband digitizing oscilloscopes and wideband scope probes unfortunately have their own noise, which must be taken into account. If the noise you're trying to measure on your power supply is of the same order as the noise floor of the scope and probe, you are challenged to measure it accurately. I will illustrate later some techniques you can use to extract information out of the scope's noise floor, however.
The second problem can best be described as dynamic range. Your power supply is at some dc voltage. The small ac noise riding on it is usually a tiny fraction of the dc level. With some scopes and/or probes there may be a challenge in offsetting the scope and probe sufficiently to allow you to use a more sensitive range to get a better view of the noise, and at a lower scope noise level. See "A brief lesson in scope noise" on page 2.
|
