Finding Noise Sources in Wireless Embedded Systems Using a Mixed Domain Oscilloscope

When integrating a radio chip or module into a typical embedded system, a common task that designers must face is tracking down and eliminating noise and spurious signals. Potential noise sources include: switching power supplies, digital noise from other parts of the system, and external noise sources. When considering noise, any possible interference generated by the radio should also be considered, which is an important consideration to avoid interference with other radios and meet regulatory requirements.

Finding the source of noise has always been a challenge. However, the addition of wireless technologies has further increased the complexity of embedded systems, and designers face greater obstacles in tracking down the source of noise. Let’s face it, wireless technology is everywhere. It is estimated that there are more than 1 billion wireless devices in use today, and 30% of embedded designs now include wireless capabilities, and this number continues to grow every day.

When adding wireless functionality to an embedded system, many issues are typically encountered during integration. For battery-powered systems, switching regulators are typically used to achieve the highest practical efficiency at the lowest cost. Power supply output power is also often an issue. This requires the use of high switching frequencies to minimize output filtering specifications and requirements. These power supplies often have ripple on the output voltage, which may show up on the RF transmitter output, especially under load or when the battery is low. To avoid this, additional power supply filtering may be required to avoid unwanted impairments of the radio signal, although this may result in undesirable cost or power.

The hardware circuitry and software configuration of a radio chip or module may affect the quality of the signal being sent. If not properly configured and filtered, the radio may cause interference to other radio systems or fail to meet regulatory standards. Some radio systems require channel filters, RF surface acoustic waves, and other relatively expensive filters to meet regulatory requirements for out-of-channel and out-of-band emissions.

As the tool of choice for embedded designers, the oscilloscope is optimized for making time domain measurements alone. An MSO (mixed signal oscilloscope) can measure both analog and digital signals, but it is difficult to effectively measure RF signals at the RF carrier using an oscilloscope. It is also difficult to fully correlate events in the time and frequency domains, which is critical to finding system-level problems.

Although spectrum analyzers can make measurements in the frequency domain, they are not the tool of choice for most embedded designers. Making time-correlated measurements in the rest of the system using a spectrum analyzer is nearly impossible.

In this article, we will examine tips and techniques for finding noise sources using a new type of instrument called a mixed domain oscilloscope, or MDO. Tektronix recently introduced the world's first MDO, and the examples in this article are based on the MDO4000 Series. This oscilloscope has the unique ability to display 4 analog signals, 16 digital waveforms, up to 4 decoded serial and/or parallel buses, and 1 RF signal simultaneously. All of these signals are time-correlated, showing the effects of control signals on both the analog and RF domains.

Before diving into an operational example of using an MDO, let’s first review some of the key concepts behind this oscilloscope. The main value of a mixed domain oscilloscope in finding noise sources is its ability to make time-correlated measurements in two domains: time and frequency. In addition, it can make these measurements on multiple analog, digital, and RF signals.

By time correlation, we mean the ability of an MDO to measure the timing relationship between all inputs. For example, it can measure the time between a control signal and the start of a radio transmission, measure the rise time of a transmitted radio signal, or measure the time between multiple symbols in a wireless data stream. A power supply voltage sag during a device state change can be analyzed and correlated with the effect on the RF signal. Time correlation is important to understanding overall system operation or cause and effect relationships.

Time domain signals are best viewed as amplitude versus time, and these signals are traditionally measured using an oscilloscope. Viewing signals as amplitude versus time helps answer questions like, “Is the power supply truly DC?” “Is the settling time of this digital signal adequate?” “Is my RF signal turned on?” “What information is currently being sent over this wired bus?” Time domain signals are not limited to analog inputs. Viewing RF amplitude, frequency, and phase versus time allows for the study of simple analog modulation characteristics, startup characteristics, and settling characteristics of RF signals.

Frequency domain signals are best viewed as amplitude versus frequency, and these signals are traditionally measured using spectrum analyzers. Viewing signals as amplitude versus frequency helps answer questions such as, "Is this RF signal being transmitted within the allocated spectrum?", "Will harmonic distortion on this signal cause equipment problems?", "Are there any signals in this band?"

Application example: Embedded system with switching power supply and wireless function

In the following discussion, the device under test will use a flexible radio IC that has been integrated into a radio test module, the Microchip Technologies MRF89XM8A. This module uses the MRF89XA integrated circuit radio along with filtering and antenna matching. For demonstration purposes, this module is mounted on a Microchip Explorer 16 board and used with a PC to program the radio settings.

To demonstrate the effect of using a switching power supply to power the radio, we used a boost converter integrated circuit, the Microchip MCP1640, which is integrated onto the MCP1640EV evaluation board. This converter switches at a frequency of approximately 500 kHz, which is quite common for switching regulators. It can provide the 3.3 V output voltage required by the radio module and supports input voltages as low as 0.8 V. This means that the radio can be powered from a single battery cell, reducing the battery size of the product. Figure 1 is a diagram of the test setup.

Figure 1. Test connections between the device under test (Microchip Technologies MRF89XA 868 MHz radio) and a mixed-domain oscilloscope.

For reference, we measured a radio spectrum centered at 868 MHz, which has a fairly low FSK modulated data rate of 2 kbps. The reference spectrum is shown in Figure 2. Note that the MDO displays both the time domain view and the frequency domain view, with all signals time-correlated.

The lower half of the display shows a frequency domain view of an RF signal, in this case a radio transmitter output, and the upper half of the display is a traditional oscilloscope view of the time domain. The spectrum displayed in the frequency domain view is from the time period indicated by the short orange bar in the time domain view, called the Spectrum Time.

Figure 2. Observing the time and frequency domains. [page]

Since the horizontal scale of the time domain display is independent of the amount of time required to process the Fourier transform (FFT) of the time domain display, it is very important to represent the actual time period associated with the RF acquisition. The unique architecture of the MDO Series oscilloscope allows all inputs (digital, analog, and RF) to be acquired separately in a time-related manner. Each input has a separate memory, and the RF signal acquired in the memory supports spectrum time and can be moved within analog time depending on the horizontal acquisition time of the time domain display, as shown in Figure 3.

Figure 3. Occupied power measurement during multiple packet preamble symbols shown using a clean lab power supply.

Spectrum Time can be moved through the acquisition to examine how the RF spectrum changes over time. In Figure 3, we have positioned Spectrum Time to show the spectrum of the signal transmitted during multiple symbols of the packet preamble. Spectrum Time is the amount of time required to support the resolution bandwidth (RBW) required for the spectrum display. It is equal to the window shaping factor divided by the RBW. The default Kaiser Window shaping factor is 2.23, and in this example, the Spectrum Time is 2.23/220 Hz, which is approximately 10 ms.

FSK modulation has only one RF signal frequency at a time, and we use a longer acquisition time for the spectrum to measure occupied bandwidth and total power.

To easily visualize the packet transmission in the radio, we have added RF vs. time plots in the time domain view. The orange curve labeled “A” shows the instantaneous RF amplitude vs. time. The orange curve labeled “f ” shows the instantaneous RF frequency vs. time relative to the center frequency of the frame. The green curve (channel 4) shows the current to the module. It can be seen that the current rises from nearly 0 between packets to about 40 mA during transmission. The yellow curve (channel 1) shows the AC ripple on the module supply voltage. Note that there is only a small voltage dip during transmission.

The above graph was obtained with the module powered from a clean lab supply, which is difficult to achieve in a real environment, but serves as a practical reference. Figure 4 shows the same RF signal, but using a boosted switching supply to power the radio module. Boost regulators are notorious for generating noise, but they allow the use of one or two alkaline or nickel-cadmium units and relatively few components, reducing cost. Note the increased noise at the bottom of the modulated signal. Near the transmitted signal, the noise is at least 5 dB higher than the clean supply. The noise already shows up easily in the current waveform and the voltage waveform. The additional noise also degrades the signal-to-noise ratio at the receiver used to collect this data from the transmitter, reducing the effective range of the radio system.

Figure 4. Spectrum and power supply measurement results for a switching power supply.

[page]

The noise from the power supply can be measured using a commercial EMI current probe, which is used to observe the noise from the switching device in Figure 5. In this example, the switching device is loaded by a resistor and a small capacitor. The Auto Marker function in the MDO is used to display the frequency and amplitude of the seven most significant signals from the power supply. The MDO4000 Series can provide up to 11 Auto Markers, which display the results as absolute values ​​or as relative values ​​referenced to the largest signal. The highest value is always indicated by the Red Reference marker. Note that the fundamental frequency and the second harmonic are at roughly the same level, approximately 30 dBuA. The upper half of the screen shows the waveform on the switching transistor of the MCP1640 IC. We use the measurement function to display the switching frequency to confirm the RF marker measurement.

Figure 5. Power supply switching noise to an equivalent load.

The time domain and frequency domain views of noise power change when the power supply is driving an RF board. Figure 6 shows the same power supply noise along with additional signals. Note that the second order harmonic is down, but there is a lot of other low level noise. Some of this noise can be very disruptive to the operation of a radio receiver and needs to be carefully evaluated.

Figure 6. Power supply and board noise using a boost converter.

[page]

A digital circuit board can generate noise, as shown in Figure 7. A single-wire probe can be used to find the source, amplitude, and frequency of the noise. Here, there is significant noise in the 220 MHz range. The automatic marker shows the highest level of the 868 MHz transmit signal and the unwanted signal. We use a manual marker to measure the frequency range of the highest level noise. The measurement data displayed in the manual marker also includes the noise density of the signal of interest. Knowing this type of noise power can be very important because, depending on the radio receiver structure, the receiver sensitivity can be affected by noise at various frequencies.

Figure 7. Broad spectrum noise from a digital board when using a boost converter.

Radio-Generated Noise
Another potential problem when adding a radio to an embedded system is the noise generated by the radio, which can interfere with the rest of the system or fail to meet regulatory limits for radio signals. Measurements such as occupied bandwidth and total transmitted power can also help assess whether regulatory requirements are met.

Figure 8 shows the spectrum of the wanted signal and the spurious transmissions in adjacent frequencies. It shows some spurious signals around 500 kHz on either side of the fundamental frequency, but they are about 40 dB below the fundamental frequency, which is generally acceptable. This figure also shows that the measured signal power is 1.4 dBm and the occupied bandwidth is 94.5 kHz, which falls within the acceptable typical bandwidth of 100 kHz.

Figure 8. Out-of-channel spectrum around the fundamental signal.

[page]

The next step is to look at the second order harmonic using the same measurement as in Figure 8 at the fundamental frequency. In this example, we see that the power level at the second order harmonic is slightly down by less than 40 dB from the fundamental harmonic, occupying a bandwidth that is twice the bandwidth of the fundamental harmonic spectrum. Figure 9 shows the third order harmonic, which is often the most troublesome part of a radio system. However, at this frequency, the signal has very low noise power relative to the carrier (~ -60dBc).

Measurements can be made in this frequency band up to the sixth harmonic. In this frequency, the radio has almost no appreciable radiation, below -80 dBm.

Figure 9. Spectrum at the third harmonic.

Summary
When including wireless communications in an embedded system, there are many key issues to consider, including the effects of power switching noise, correctly setting the operating parameters of the radio integrated circuit, and ensuring that the transmit output meets the applicable radio regulations.

The MDO can view time-correlated signals, helping designers to efficiently diagnose and test power and other noise effects. It can confirm that data commands sent to the radio are set correctly, and can check for spurious emissions from transmitters and other circuits. It can be used to measure RF signals up to 6 GHz, and low-frequency noise from switching power supplies and digital circuits can also be viewed through time-correlated acquisitions.

To learn more, visit http://www.tek.com/en/scoperevolution/index53.html and watch the webinar "Finding Noise Sources in Wireless Embedded Systems."