Most bench power supplies and main power regulators used in electronics are switching regulators, which may be further isolated or non-isolated depending on safety concerns. Power supplies or power regulators can exhibit varying levels of noise given a required grounding strategy and connections to loads. When you’re testing a bench power supply or an on-board power regulator, it’s important to properly measure the noise being generated.
The main tool for measuring power regulator noise is an oscilloscope, and possibly a spectrum analyzer if you want to look at the noise power spectrum. To help newer engineers get a better handle on the noise they can expect in their power electronics, the guide below outlines how to capture accurate power supply noise measurements with an oscilloscope.
Power Supply Noise on an Oscilloscope
An oscilloscope can be used to directly sample a waveform in the time domain and display the measurements to a user. For a power supply, this requires adding a load to the output and connecting to the circuit with a standard high bandwidth probe. The complete list of components required can be found below:
- Probe with required bandwidth and attenuation ratio (10:1 is appropriate)
- If available, use a differential probe for highly precise measurements, otherwise single-ended is appropriate
- A resistive load, preferably high enough resistance to stabilize the regulator
The connections for these measurements are quite simple: place the load on the output of the power supply and measure with the oscilloscope. A precision resistor that can handle high power dissipation (order of ~1 W) is required for this measurement, particularly if you will run this measurement at higher voltages. The output is then measured directly across the terminals of the power regulator.
With this configuration, we intend to be able to measure the primary switching waveform associated with the inductor in the primary switching regulator circuit. In a more advanced isolated system, such as an LLC resonant converter, the primary side inductor, transformer, and parallel capacitor will determine the ripple frequency and intensity seen on the secondary side. The output capacitor bank (and the ESR values of caps) will determine the magnitude of the ripple measured at the output port.
Upon initially powering up the power supply and bringing the time-domain waveform onto the oscilloscope, the waveform will appear very low level with some DC offset due to the voltage division setting. Decrease the volts per division and time per division settings to bring the ripple into the viewing area and to resolve the main ripple component associated with the inductor. The waveform should appear as noise superimposed on a constant voltage. Large peaks may be present, indicating switching events, similar to what is measured below.
Large peaks observed on the output from a power regulator.
From here, there are three possible noise metrics that we will care about:
- Major frequency components in the time-domain waveform
- Peak-to-peak voltage, or peak voltage excursion from nominal DC
- RMS voltage, or the standard deviation from the nominal DC value
The peak-to-peak voltage or the RMS voltage can be extracted from an automated measurement in most oscilloscopes. If your scope does not contain an automated measurement, you will need to use a set of cursors to mark measurements or you will need to change the horizontal/vertical divisions to take measurements.
Generally, when running in the continuous conduction mode, the ripple on the output will be a semi-smooth waveform that has a somewhat triangular shape. It will not be perfectly triangular for two reasons: high-frequency rolloff in the probe and oscilloscope input, and any low-pass filtering on the supply output. However, it is common to see additional frequency components superimposed on top of the main ripple associated with the inductor.
Typical switching frequencies for the switching elements in a power supply (either synchronous or asynchronous) could be anywhere from 100 kHz to 10 MHz. The probe being used for the measurement will add some capacitance to the output of the switching regulator, and this will modify the appearance of the waveform that is seen on the oscilloscope. Ideally, the probe should have bandwidth reaching around 100 MHz with minimal probe capacitance (<100 pF) to get the most accurate measurements of the switching waveform.
Why is a Load Needed?
There is sometimes a perception that a load should not be included when taking a power supply ripple measurement using an oscilloscope. Inclusion of a load is quite important for characterizing noise on the output of the power supply. This is because the operating characteristics, and thus the noise waveform that you will measure, depend on the load that is connected to the power supply.
Generally, bench power supplies and power regulators are designed to run in the continuous conduction mode so that the inductor current (the current leaving the switching node) is always greater than zero. A power supply could also be intentionally designed to run discontinuously. In either case, the load needs to match the intended operating range for the
If the load is too small, then spiking (ringing) may be seen on the output of the regulator. If there is ringing on the output, then parasitics may be creating underdamped oscillations when the inductor current drops to zero, followed by a switch back on into forward conduction. The SPICE results below show an example of the ringing that could be observed due to running in the discontinuous mode, as well as excessive inductance on the regulator’s output.
Periodically excited ringing seen on the output of a switching regulator circuit. Learn more about these glitches in this article.
In this case, it is important to probe the inductor region with a near-field probe to see if the artifact can be seen in the switching noise, or use a larger load and see if the output stabilizes. At some minimum load resistance, the switching regulator circuit’s operation will transition to continuous and the ringing events should cease.
What About Power Rail Ripple in Digital Systems?
The technique used here can be used to measure ripple along a power rail, either with direct probing of the power rails or with a near-field probe. A near-field probe can pick up fields from other areas of the board, so a direct contact probe measure along the rail would be preferable. This would typically be done with a coaxial probe connected to an SMA, and this would ideally be impedance matched to the input impedance of the oscilloscope in order to correctly determine the magnitude of the high frequency noise components.
The other factor in these measurements is the load characteristics of digital components. The power rail in a digital system is connected to a switching element, which switches from high to low impedance depending on an input data signal in a CMOS buffer circuit. So, you are effectively measuring the voltage on a very low impedance circuit that switches very quickly, rather than measuring a load with fixed DC resistance.
The connections for a PDN measurement are also not the same connections as those used in a power regulator output noise measurement. In a power supply output noise test, the output is being measured from the output terminal of the regulator circuit or voltage regulator module (VRM). In the power rail feeding a large digital processor, the ripple due to power rail collapse is seen at the Vcc/Vdd input, which could have any number of moderate frequency oscillations that are not associated with power regulator switching noise.
You may see strong spikes like this on a power rail in a digital system.
To measure the power rail ripple due to switching of I/Os, the ideal measurement location would be at the output of a power rail. In reality, this is often difficult because of the structure of a PDN in a PCB, so some other point needs to be found to connect the probe. The PDN can span across multiple layers, and it can connect to a large processor at multiple points. As another example, the PDN can branch into different rails, each with their own power integrity problems depending on the transfer impedance between rails.
As these measurements are much more complex and they require more specialized equipment, this discussion will be left for discussion in a future article. However, in general, the measurements can be taken in the time domain or the frequency domain, the latter being taken with a vector network analyzer.
After you’ve identified any problems in your switching regulator circuits and power supply, you can solve these problems in your circuits and PCB layout using the complete set of design features in OrCAD from Cadence. OrCAD is the industry’s best PCB design and analysis software with utilities covering schematic capture, PCB layout and routing, and manufacturing. OrCAD users can access a complete set of schematic capture features, mixed-signal simulations in PSpice, and powerful CAD features, and much more.