Systematic Wi-Fi Troubleshooting

Connect802 recently had the opportunity to troubleshoot a wireless network in which signal strength was reported as good, but a usable connection could not be established. Ping packets dropped more than 50% of the time and FTP file transfer throughput was below 1 Mbps. This month, we’ll walk you through the troubleshooting methodology that we used to identify the cause of this problem, passing along our tips as we go.

FTP throughput averages around 200-300 kbps—very low for 802.11b/g!

At first, it might seem incongruous that signal strength was good while link performance was bad. These two facts are not necessarily in conflict. Signal strength refers to the absolute amount of signal energy coming into the antenna. It’s true that signal strength needs to be above a certain minimum value in order for data to be received. This value is known as the receive sensitivity of the wireless card. But signal strength alone isn’t the only factor that needs to be considered: the signal must be a certain amount stronger than any noise or interference in the area of the receiver. This value is known as the signal-to-noise ratio or signal-to-interference ratio.

The fact that there were dropped Ping frames suggested packet corruption in the wireless network. This could also have explained the lower-than-usual FTP throughput. AirMagnet, a wireless protocol analysis tool, was used to measure the number of retransmitted 802.11 packets. In the presence of interference, the number of corrupted packets would be expected to be high, resulting in a high number of retransmissions.

If signal strength is high, but there are still corrupted packets, then poor signal-to-noise or signal-to-interference ratio are the next most likely culprits. The Cognio spectrum analyzer was used to search for sources of interference. Whereas AirMagnet can identify other 802.11 networks in the area, only a spectrum analyzer like Cognio can identify non-802.11 sources of interference. Cognio identified a large number of interfering devices, such as DECT cordless phone handsets, Bluetooth devices, and microwave ovens. The signal strength of these devices was very close to that of the 802.11 network, supporting the diagnosis of environmental interference.

Signal strength of interferers (“spikes” in the spectrum analysis trace)
is very close to that of 802.11 devices (“humps” in the spectrum analysis trace).

Many 802.11 networks continue to operate to an acceptable level even in the face of interference. Various steps were taken to determine whether this network could overcome this interference.

The network had originally been designed with a 12 dBi omnidirectional antenna. This antenna was swapped out for a 5 dBi omnidirectional and a 2 dBi omnidirectional to determine whether some quality of the high-gain antenna was decreasing performance. Connect802 has observed that having too much antenna gain sometimes decreases performance. The lower-gain antennas might not have been appropriate for the network’s day-to-day operations, but for testing purposes, they allowed the engineer to determine whether the high-gain antenna was related to the problem. Network performance did not improve even when antennas were swapped out, indicating that the antenna was not the problem.

The antenna was mounted on a light pole approximately 30 feet in the air. The possibility was considered that the antenna’s height was giving it more of a view of interferers in the area. The antenna was lowered to the ground using a length of LMR-400 coaxial cable. Performance did not improve, indicating that the height of the antenna was not the problem.

The client device that was being used for testing was moved to within a few inches of the transmitting antenna. Performance was still not adequate. Interpreting this test required careful judgment on the part of the RF engineer. If a client can’t get good performance just a few inches from the access point, then interference must be pretty bad. On the other hand, an extreme amount of interference would be required in order to produce results that severe. It’s unlikely (but not impossible) to have that much interference. Should this test be interpreted as supporting or denying the diagnosis of interference?

An additional test was performed to answer that question. A Symbol PCMCIA card with external antenna connectors was installed in the test client device. A pigtail cable was used to convert the PCMCIA card’s MC connector to a standard RF N-connector. This N-connector was connected directly to the access point’s antenna connector with a length of LMR-400 coaxial cable. The result was that the 802.11g radio in the client device was wired directly to the 802.11g radio in the access point.

Some people have expressed surprise that this technique is even possible, but it is quite feasible. When an 802.11 radio transmits, electrical current flows down a wire to the antenna, which converts it into RF energy. At the receiving antenna, the RF energy is converted back to electrical current. Wiring the radios directly to each other just cuts the antennas and the environment out of the chain. The only difference is that, with directly-wired radios, the incoming signal strength is much higher than it would be if the signal were traveling through the air. An attenuator was installed between the two radios to bring the signal strength down to a normal level.

This setup allowed a definitive identification of whether environmental interference was the problem. There is only one difference between the direct-wired setup and the one where the two stations had antennas that were a few inches apart. The direct-wired setup totally protects the signal from the environment. If directly wiring the radios together improved performance, then the environment was conclusively the problem. Performance did improve dramatically, confirming the diagnosis of environmental interference.

A final test was performed in order to rule out the possibility of bad hardware or a failing access point. The access point was taken down from the pole and moved to a different location on site where performance had been reported to be fine. When the access point was moved to the different location, its performance was good. When it was moved back to its original location, its performance was bad again. This test and the previous one confirmed that the access point was capable of good performance when the environment wasn’t interfering with it.

Conclusions

AirMagnet is invaluable for measuring signal strength, detecting 802.11 networks in the area, and producing various charts and graphs that help determine whether a wireless network is operating properly, however it cannot detect non-802.11 sources of interference. The Cognio spectrum analyzer detects and identifies non-802.11 sources of interference such as cordless phones, Bluetooth devices, and microwave ovens. Together, these two tools allow comprehensive evaluation of the 2.4 GHz and 5 GHz RF frequency bands. Both of these tools can be purchased from Connect802 in our online store.

External antenna connectors can aid in troubleshooting issues that are difficult to troubleshoot any other way. Antennas can be swapped out to directly measure the effect of their differing gain and coverage patterns. Additionally, the client device can be directly attached to the access point to eliminate the possibility of environmental interference. The RF engineer is encouraged to have a wireless card with external antenna connectors to make this possible. If the client device will be directly attached to an access point, the RF engineer is also encouraged to have an adjustable attenuator in order to bring signal strength down to normal levels. For more information on these products and our consulting services, contact Connect802’s sales department.