Interference in the 900 MHz band appears to produce random errors in a gas station price sign. Spectrum analyzers help track it down.

I received a call recently from a former client who manufactures electronic signs. A gas station was experiencing intermittent loss of wireless control of its signs, which displayed the price of various grades of gasoline. This case study demonstrates the equipment and techniques used to track down this potential interference.

The displayed gasoline prices (Figure 1) are controlled from inside the station kiosk from a remote terminal that controls a Digi International XBee-Pro S3B transceiver module. It uses frequency-hopping spread spectrum transmission within the license-free 915 MHz ISM band (902 to 928 MHz). The station has two signs: one on the kiosk canopy and one on the adjacent street. Occasionally, the gas station would lose the ability to control the sign prices; the system would shut down and require rebooting.

Intermittent problems can present a real challenge, so I immediately thought my Tektronix RSA306B real-time spectrum analyzer with real-time capture would be the tool to use. Real-time analyzers can detect very short bursts of RF energy down to as fast as 1 µs.

I also used a TinySA Ultra pocket spectrum analyzer in Max Hold mode to help pinpoint possible interfering sources.

Note that the specific names of companies and locations have been “sanitized” to protect client confidentiality.

Start with measurements

The gas station and surrounding shopping center were bounded on two sides by busy streets. I took measurements at the four locations shown in Figure 2.

Test results

Here are the results taken with the spectrum analyzers. In each of the dual-screen images, the left screen is the real-time capture (you can see signals almost instantly), while the right hand is the same capture, except using swept mode (slower signal captures). All the dual-screen captures were scans from 900 MHz to 930 MHz, leaving 2 MHz on each end of the 915 MHz ISM band as a buffer zone. Used together with a handheld directional 700 to 6 GHz log-periodic antenna from Aaronia Figure 3, we can roughly locate the source of these larger signals through triangulation.

Note that in every case, there were multiple signals throughout the 915 MHz ISM band the gas station was using for the sign control. The real-time mode captures many more of these than the swept mode, making it better for capturing the intermittent larger signals.

Figures 4 and 5, show swept mode with max hold, which lets the spectral envelope fill in. Doing so shows the full spectrum usage.

In Figure 4, note the occasional signal (+70 dBµV) well over the average 30 to 40 dBµV “clutter.” This was strong enough to be suspicious (middle of screen). Recall in Figure 2 that the drug store was directly behind the station kiosk.

In Figure 5, the directional antenna was oriented in the vertical polarization (for maximized signals) and was pointed at the service station’s kiosk. Several large signals (+70 dBµV) appear that are well over the average 50 dBµV “clutter.” These signals were strong enough to raise suspicion. Later measurements at the kiosk did not, however, reveal any of these large signals. Recall that the drug store was directly behind the station kiosk. The aqua trace is the measurement noise floor.

In Figure 6, the directional antenna was oriented vertically. You can see the usual clutter across the whole band, although no large signals appear. Later, I used the TinySA Ultra spectrum analyzer with short antenna and swept all the electronic equipment inside the kiosk. Again, no large signals appeared.

The spectrum in Figure 7, from just outside the kiosk, shows the usual clutter, but with the larger (~50 dBµV) and wider signals. These don’t appear to originate from the kiosk, but possibly from an antenna tower on a hill about a half-mile away.

The measurement in Figure 8 occurred with the directional antenna in vertical polarization and pointing at the nearby drug store. You can see a much larger signal (at 60 dBµV), which was definitely in the direction of the store.

Figure 9 provides a clue. I pointed the Aaronia directional antenna at the nearby drug store, this time using horizontal polarization. The plot shows a much larger signal (at 60 dBµV), which was definitely in the direction of the store.

Polarization hints

Because most transmitters in this 915 MHz ISM band use vertical polarization (Aaronia antenna is held sideways), it’s expected to observe maximized signal levels using vertical polarization. Rotating the antenna to horizontal polarization substantially reduces the usual clutter of signals, as you can see by comparing this and the previous figures. The strong signal at 60 dBµV was, however, unchanged. This indicates that the transmission of these signals is likely using multiple antennas and that the signals could be reflecting in different polarizations. My suspicion is that the drug store may be the largest signal source in the area, so I walked into the store and made spectral measurements with the TinySA Ultra.

Using my small TinySA Ultra spectrum analyzer with a telescoping antenna set to resonate at 915 MHz (center of the ISM band) and in Max Hold mode, I walked around inside the store capturing the ISM spectrum over a period of about five minutes. This allowed the frequency hopping signals to “fill in”, showing the overall spectrum usage within the building. Maximum signal levels were 80 dBµV, about 10 to 20 dB higher than what we measured outdoors. The building has a long row of windows on the south side that could allow some of this transmitted energy out towards the station and kiosk.

Summary of findings

First thing to note is that the 915 MHz ISM band (902 to 928 MHz) has a high usage by many sources in this area. I imagine you’d find the same high spectral usage in other larger metropolitan environments. The use of spread-spectrum or frequency hopping transmitters for the sign control system is an advantage in such an environment. However, large signals can possibly disrupt receiving the control transmissions.

There were no interfering signals observed at or inside the kiosk.

The nearby drug store is highly suspicious in that there appears to be multiple high-powered transmitters within the building that use nearly the entire 915 MHz ISM band. As I walked throughout the store, the signal strength appeared to be nearly constant, which probably eliminates their point of sale terminals. I’m guessing it may have to do with their video surveillance system. The closeness of the building and the fact there are windows stretching about 50 feet long on the south side, may allow interfering signals that could affect the signage receivers.

Recommendations

These were the recommendations I made to the client:

• Try using frequency spectrum masking. Examining Figure 10, we see areas of non-use in between the major signal bands. Each division is 2 MHz wide, so adding up the unused spectrum provides about 10 MHz of usable bandwidth. The disadvantage would be that every geographical location experiencing interference would have to have a spectrum usage study and different spectrum masking (assuming you’d have the minimum bandwidth required for your particular system).
• Use software solutions. The system probably uses some level of error correction now, but using some bootstrap code to restart might be a consideration. One thing to try in-house would be to duplicate the problem. This would assist in troubleshooting with either hardware or software solutions.
• I would recommend purchasing a handheld real-time spectrum analyzer to assist in performing these field studies in the future. Several companies make these: Aaronia, Anritsu, Keysight, Rigol, Rohde & Schwarz, and Siglent.