Hunting noise sources in wireless embedded systems
Hunting down noise sources has never been easy. However, as embedded systems have become more complex with the addition of wireless, designers face bigger hurdles than ever before to track down noise sources. And let’s face it – wireless is everywhere. It’s estimated that there are more than 1 billion wireless devices in use today and more than 30 percent of embedded designs now include wireless, and that number continues to grow daily.
In adding wireless capability to embedded systems, there are a number of issues typically encountered in the integration. For battery powered systems, a switching regulator is typically used to have the highest practical efficiency at the lowest cost. The size of the power supply is also often an issue. This can lead to the use of high switching frequencies to minimize the size and requirements of output filtering. These power supplies often have ripple on the output voltage which can show up on the RF transmitter output, especially when under load or under low battery conditions. To avoid this, additional power supply filtering may be needed to avoid unwanted impairment of the radio signal, even though the cost or size is undesirable.
The hardware circuits and the software configuration of the radio chip or module can affect the quality of the transmitted signal. If not properly set up and filtered, the radio can cause interference to other radio systems and/or fail to conform to applicable agency regulations. Some radio systems will need channel filters, RF Surface Acoustic Wave, or other relatively expensive filters to meet agency regulations for out-of-channel and out-of-band emissions.
The tool of choice for the embedded designer, the oscilloscope, is optimized for making time domain measurements only. An MSO (Mixed Signal Oscilloscope) can measure both analogue and digital signals, but it remains difficult to effectively measure RF signals with an oscilloscope at the RF carrier. It is also quite difficult to adequately correlate events in the time and frequency domains – something critical for finding system-level problems. While spectrum analysers are available for making measurements in the frequency domain, these are not the tool of choice for most embedded designers. Using spectrum analysers to make time correlated measurements with the rest of the system is virtually impossible.
In this article, we will explore tips and techniques for hunting noise sources using a new type of instrument called the mixed domain oscilloscope, or MDO. Tektronix recently introduced the world’s first MDO and the examples provided here are based on the MDO4000 Series. The oscilloscope has the ability to simultaneously display four analogue signals, 16 digital waveforms, up to 4 decoded serial and/or parallel buses, and one RF signal. All of these signals are time correlated to show the effects of control signals on the analogue and RF domains.
Before diving into a hands-on example on the use of the MDO, it might be helpful to first review some of the key concepts behind this oscilloscope. The primary value of a mixed domain oscilloscope for hunting noise sources is its ability to make time-correlated measurements across two domains: the time domain and the frequency domain. In addition, it can make these measurements across multiple analogue, digital, and RF signals.
When we talk about time-correlation, what this means is the MDO can measure timing relationships between all of its inputs. It can, for instance, measure the time between a control signal and the beginning of a radio transmission, measure the rise time of a transmitted radio signal, or measure the time between symbols in a wireless data stream. A power supply voltage dip during a device state change can be analysed and correlated to the impact on the RF signal. Time correlation is important for understanding the complete system operation or cause and effect.
Time domain signals are signals that are best viewed as amplitude versus time. These are the signals traditionally measured with an oscilloscope. Viewing signals as amplitude versus time helps answer questions like; “is this power supply really DC,” “is there sufficient setup time on this digital signal,” “is my RF signal on,” or “what information is currently being sent over this wired bus?” Time domain signals are not limited to analogue inputs. Seeing RF amplitude, frequency, and phase versus time can enable a study of simple analogue modulations, turn-on, and settling behaviour of RF signals.
Frequency Domain signals on the other hand are signals that are best viewed as amplitude versus frequency. These are the signals traditionally measured with a spectrum analyser. Viewing signals as amplitude versus frequency helps answer questions like; “is this transmitted RF signal within its allocated spectrum,” “is the harmonic distortion on this signal causing problems in my device,” or “are there any signals present within this frequency band?”
Wireless-enabled embedded system with a switching power supply
For the following discussion, the device under test will use a flexible radio integrated circuit already integrated into a module for radio test, the Microchip Technologies MRF89XM8A. This module incorporates the MRF89XA integrated circuit radio along with filtering and antenna matching. For demonstration, this module is mounted to a Microchip Explorer 16 board and is used with a PC to program the setup of the radio. To illustrate the effects of powering the radio with a switching power supply, a boost converter IC, the Microchip MCP1640, incorporated into an MCP1640EV evaluation board, is used. This converter switches at about 500 kHz which is common for switching regulators. It can provide the 3.3 V output voltage needed by the radio module with an input voltage down to 0.8 V. This means the radio can be powered from a single cell, reducing the size of the battery for the product. Figure 1 shows the test setup.
For reference, measurements of the radio spectrum centred at 868 MHz with a fairly low data rate of 2 kbps of FSK modulation are taken. Figure 2 shows the reference spectrum. Notice that the MDO displays both the time and frequency domain views, and all signals are time correlated.
The lower half of the display shows the frequency domain view of the RF signal, in this case the radio transmitter output, while the upper half of the display is a traditional oscilloscope view of the time domain. The spectrum shown in the frequency domain view is taken from the period of time indicated by the short orange bar in the time domain view – known as the Spectrum Time.
Since the Horizontal Scale of the Time Domain Display is independent of the amount of time required to process a Fourier Transform (FFT) for the Frequency Domain Display, it is important to represent the actual time period that correlates to the RF acquisition. The MDO enables separate time-correlated acquisitions of all inputs (digital, analogue, and RF). Each input has separate memory, and depending on the horizontal acquisition time of the Time Domain Display, the RF signal acquired in memory allows the Spectrum Time and can be moved within the Analogue Time as shown in figure 3.
Spectrum Time can be moved through the acquisition to investigate how the RF spectrum changes over time. In figure 3, Spectrum Time is placed to show the spectrum of the transmitted signal during the several symbols of the preamble of the packet. Spectrum Time is the amount of time required to support the desired resolution bandwidth (RBW) of the spectrum display. It is equivalent to the Window Shaping factor divided by the RBW. The default Kaiser Window has a shaping factor of 2.23, so the spectrum time is 2.23/220 Hz or approximately 10 ms in this example.
While FSK modulation has only one frequency of the RF signal on at a time, a longer acquisition time of the preamble is used for the spectrum to enable measurements of the occupied bandwidth and the total power. To see the packet transmission from the radio, RF versus Time traces have been added to the time domain view. The orange trace marked with “A” shows the instantaneous RF amplitude versus time. While the orange trace marked with “f ” shows the instantaneous RF frequency versus time, relative to the centre frequency of the display. The green trace (Channel 4) shows the current into the module. As can be seen, the current rises from close to 0 between packets, to about 40 mA during transmission. The yellow trace (Channel 1) shows the AC ripple on the power supply voltage at the module. There is only a small dip in the voltage during transmission.
The previous capture was taken with the module powered by a clean laboratory power supply – hardly real world, but a useful reference nonetheless. In figure 4, the same RF signal is shown but with a boost type switching supply powering the radio module. Boost regulators are notorious for generating noise, but are valuable to allow the use of a battery with one or two alkaline or NiCad cells and relatively few components, lowering the cost. As shown, the noise has increased at the base of the modulated signal. Near the transmitted signal, there is noise at least 5 dB higher than with the clean power supply. The noise is also readily apparent in the current and voltage waveforms. The additional noise would also degrade the signal-to-noise ratio of the signal at the receiver used to gather the data from this transmitter, reducing the effective range of the radio system.
RF marker measurements
The noise from the power supply can be measured with a commercially available EMI current probe used to observe the noise from the switcher in figure 5. For this example, the switcher is being loaded by a resistor and small capacitor. The automated marker function of the MDO is used to show the frequency and amplitude of the most prominent seven signals radiating from the supply. The instrument can provide up to 11 automated markers and display the results in absolute values, or referenced to the largest signal as delta values. The highest value is represented as the red reference marker. The fundamental frequency and the second harmonic are about the same level at around 30 dBµA. The upper half of the screen shows the waveform at the switching transistor in the MCP1640 IC. The measurement function is used to show the switching frequency to confirm the RF marker measurement.
When the power supply drives the RF board, the time and frequency domain displays of the noise power are changed. Figure 6 shows the same power supply noise plus additional signals. Here the second harmonic is reduced, but there is a lot of other low level noise. Some of this noise may well interfere with the operation of the radio receiver and needs to be carefully evaluated.
The digital board can generate noise as seen in figure 7. A simple wire probe can be used to look for the source, amplitude and frequencies of such noise. Here there is significant noise in the range of 220 MHz. The automated markers show the 868 MHz transmitted signal as well as the highest level of unwanted signal.
Manual markers can be used to measure the frequency range of the highest level of noise. The displayed measurement in the manual markers also includes the noise density of the signal of interest. Understanding this type of noise power can be important because, depending on the radio receiver architecture, the receiver sensitivity can be impaired by noise at various frequencies.
Noise generated by the radio
In adding a radio system to an embedded system, there is also the potential problem of the radio generating noise and either interfering with other parts of the system, or of failing to meet regulatory limits on radio signals. Measurements such as occupied bandwidth and total power transmitted can also help evaluate regulatory compliance.
Figure 8 shows the spectrum of the desired signal as well as the spurious transmissions in neighbouring frequencies. It shows some spurious signals around 500 kHz either side of the fundamental, but they are about 40 dB below the fundamental so they would be acceptable under most regulations. This figure also shows the measured signal power of 1.4 dBm and occupied bandwidth of 94.5 kHz which fits within a typical 100 kHz acceptable bandwidth.
The next step is to look at the second harmonic with the same measurements as the fundamental frequency in figure 8. In this example, we found that the power level at the second harmonic was slightly less than 40 dB down from the fundamental and the occupied bandwidth was twice the bandwidth of the spectrum of the fundamental.
Figure 9 shows the third harmonic which is often the most troublesome in radio systems. However, at this frequency the signal is down to a very low level of noise power relative to the carrier (about -60 dBc). Moving on, it is possible to take measurements in this band up to the sixth harmonic. By this frequency, this radio has almost insignificant emissions at less than -80 dBm.
There are many new issues to watch for when including wireless communications in an embedded system. These include the effects of power supply switching noise, correctly setting up the operating parameters of the radio integrated circuit, and assuring that the transmitted output conforms to applicable radio regulations.
With its ability to see time-correlated signals, the MDO provides helps the designer efficiently diagnose and test for power supply and other noise effects. It has the ability to confirm that the data commands to the radio are being set correctly, and the ability to check for spurious emissions from the transmitter and other circuits. It can be used to measure RF signals up to 6 GHz, but is also valuable to look at lower frequency noise from switching power supplies and from digital circuits with time-correlated acquisitions.
Darren McCarthy is the Worldwide RF Technical Marketing Manager for Tektronix – www.tek.com.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
