Testing AvaLAN's Wireless Video Surveillance

Published Jan 24, 2010 00:00 AM

Wireless IP video surveillance has plenty of hype (claims of 300 Mb/s city wide meshes) and also significant challenges (repeated deployment nightmares). Recently, we helped an integrator friend of ours who needed to deploy a wireless link. He has many years of experience deploying both analog and IP wireless. As such, he knows too well the risks and problems that wireless video surveillance brings.

This report shares test results from the AvaLAN AW5800HTP-PAIR [link no longer available]. This is a kit with built in antennas that offers a tightly integrated solution though with modest bandwidth.

Our key findings include:

  • Wireless connection was "plug n play" without any configuration, however some poorly documented steps should be taken for optimization
  • The system offers limited advanced features 
  • We achieved consistent 3.5 - 4 Mb/s IP video throughput
  • In our setup, with HD/megapixel cameras, using MJPEG was more stable than H.264

The AvaLAN AW5800HTP-PAIR [link no longer available] consists of 2 wireless radios, running in the 5 Ghz range with integrated 23 dBi antennas. The system is sold as a kit - pre-configured to run together. AvaLAN sells their products direct on-line. The AW5800HTP-PAIR [link no longer available] costs $2,800 from their website.

5 Mhz Channels

AvaLAN's use of 5 MHz RF channels is an important and somewhat controversial/different design decision. Most unlicensed IP wireless systems use 20 MHz RF channels (like your home WiFi connection) and newer 802.11n / MIMO systems are using 40 MHz RF channels (needed to achieve higher bit rates). A tradeoff of using wide RF channels is that limited unlicensed spectrum is available for both 2.4 and 5 GHz systems (usually between 60 MHz and 100 Mhz). 

Using 5 MHz channels ensures lower maximum theoretical bandwidth. While AvaLAN provides configuration options for data rates up to 13.5 Mb/s, they recommend using 6 Mb/s as the data rate. Using this, we achieved IP video throughput of 3.5 Mb/s to 4.0 Mb/s.

AvaLAN claims that the use of smaller RF channels allows the system to be more robust to interference and to achieve longer distances (because the RF power is more concentrated in a narrow area). We cannot test nor judge this. We do believe that these smaller RF channels would allow using multiple links in the same physical area (since, unlike more common 20 MHz links, each 5 MHz uses a much smaller portion of the overall spectrum). 

Only Point to Point

The AvaLAN AW5800HTP-PAIR [link no longer available] only supports point to point wireless links. The kit cannot be modified for point to multi-point or for mesh. As such, it's primarily useful for a single camera or for closely located cameras to be transmitted back to a main building/headend.

Distance

AvaLAN claims the system can achieve up to 15 miles wireless links. Since our test was over a 100 meter link across two adjacent buildings, we cannot comment on distances achievable. However, multiple mile links almost always require expert planning as the narrow beamwidth of these antennas, the curvature of the Earth, etc. make antenna alignment non-trivial.

Physical Overview

In the screencast below, the key point to emphasize is:

  • No physical adjustments can be made to the radios. The built in antenna cannot be removed nor can replacement antennas be added. It's a sealed unit meant to use the built in 23 dBi antenna.

Configuration and Optimization

As we mentioned in the summary findings, the system did not require configuration to establish a connection and transmit video. However, there are some important points to keep in mind:

 

  • The default data rate is set to auto. However, according to AvaLAN's technical support that is not recommend for optimal use - 6 Mb/s is. Furthermore, any other data rate setting but 6 Mb/s (like 3 Mb/s, 6 Mb/s, 12 Mb/s, etc.) acts as auto mode. Finally, the the traffic generator tool only works with 6 Mb/s. For all these reasons, the radios should be switched to 6 Mb/s.
  • The IP address of the units are hard coded at 192.168.1.66/67. If this is being deployed in an existing network, the IP addresses will need to be changed.
  • The system does not offer any QoS optimizations.

 

Bandwidth Consumption Analysis

With the data rate set at 6 Mb/s, we achieved 3.5 to 4 Mb/s IP video throughput (unidirectional). We tested the AvaLAN system using 3 cameras - 2 megapixel and 1 standard definition connected to Exacq's video management software.

Not surprising, in our tests of a standard definition camera at 1 to 2 Mb/s, we had no issues.

Challenges arose when we switched to 1 or more megapixel cameras and, surprisingly, when we used H.264 encoding.

The screencast below demonstrates that using MJPEG, frames gracefully dropped but the connection remained and full quality video was recorded over AvaLAN at a slow frame rate. However, when using H.264, the camera's connections to Exacq disconnected. With 1 megapixel camera at H.264, the camera disconnected intermittently. However, with 2 megapixel cameras at H.264, both connections dropped and we were unable to re-establish the connection.

IP Camera Use and Load Recommendations

IP video surveillance regularly exhibits strange problems with insufficient bandwidth is available. However, different IP cameras and VMS software respond differently to shortages (some drop connections, some display macro-blocking, some display green 'slime', some drop frames, etc., etc.).

While it is not possible to accurately predict what will happen with various combinations of cameras and VMS systems, we can provide guidelines of what to consider when faced with the limited bandwidth of a wireless system like AvaLAN:

 

  • Test exactly how much throughput the system can deliver in your setup. For instance, while we achieved 3.5 - 4 Mb/s, a longer link or one with more interference could deliver significantly less. We recognize that this is logistically difficult to do (as it requires testing on site), however, we think it's risker to assume manufacturer claimed throughputs.
  • Test the bit rate of the IP cameras you plan to use at both low light and high motion. In wired systems, manufacturer 'rough' estimates of bandwidth consumption are generally not a problem. However, in wireless, if the manufacturer estimates 1 Mb/s but you find that at high motion or low light, consumption is 2.5 Mb/s, you may have a serious problem with your wireless system. We see such significant variations in our IP camera tests repeatedly.
  • Test the specific combination of IP cameras and VMS system you plan to use. This will help you understand how the combination degrades with insufficient bandwidth.
  • Favor the use of constant bit rate over variable bit rate streams. While variable bit rates stream may reduce storage use, a spike in the bit stream could break the connection.
  • Choose lower frame rates. Lower frame rates (10 fps vs 30 fps) do not significantly reduce usability of surveillance video but can dramatically reduce bandwidth consumption.
  • Plan to use less bandwidth than available. Wireless links throughput can degrade over time (new systems create interference, vegetation growth obstructs link, antenna alignment shifts, etc.) 
For this AvaLAN system, I would not use IP camera streams over 3 Mb/s. I would also limit frame rate to no more than 10 fps (though I think 5 would be even better). I would not do more than 2 standard definition cameras at 4 CIF or 1 megapixel camera at 720p (1080p consumes significantly more bandwidth with little video quality improvement).

 

Application Recommendations

We think the big constraints on the system are (1) limited bandwidth and (2) lack of advanced features. However, the biggest advantage is that it is simple to setup. We could see traditional security integrators installing these, pointing them at each other and establishing a solid connection with minimal RF expertise.

The most common application is likely to connect a single camera back to a building. We could also see using this to connect building but the link could not realistically handle more than 2 or 3 cameras (even at lower quality settings).