Concept and Overview
The "Midwest Quantar Bridge" or MQB for short, was developed to allow a small group of amateur radio operators in the Milwaukee/Chicago area that were P25 enthuasists communicate during the morning commute.
Their individual stations stretch about 65 miles apart and on UHF was difficult to stay in touch. Having read a discussion on a then popular website (www.P25.ca - now gone), they set out to better understand the backbone linking required and setup a small private network for experimentation. Limited online documentation and a quest for knowledge only fueled the project; there are some published works in U.S. Patent Searches and by Daniels Corporation (PDF) on how P25 CAI works, but nothing regarding the private Motorola V.24 protocol used.
Some Hawaiian and South Pacific radio enthuasists were able to uncover many of the necessary data fields to make back-to-back or RT/RT linking possible using off the shelf Cisco router gear (explained on the RT/RT page) across the public internet (see the contribution by John Yaldwyn, ZL4JY here). Traditionally, P25 linking is done over a closed network, typically private telco wireline or microwave, to ensure reliability, security and reduce latency as much as possible. Using the public internet for this project introduces unknown variables, but since the Cisco SerialTUNnel (STUN) protocol uses TCP, transmission delivery is guaranteed over UDP fire-and-forget methods.
Once two nodes were connected "back-to-back", the system was then extended to allow for a one-to-many approach. Instead of a flat network design, the designers opted to "Zone" individual systems together in a logical fashion to allow for segregation as desired, starting with LOCAL only and SYSTEMWIDE. Those approaches are controlled by the subscriber radios' transmitted TalkGroup, as part of the P25 data stream and selected by the user. Subscribers selecting TalkGroups 1 or 2 (default on some radios) will keep their traffic local to that system (STUN ID); subscribers using more well-known Talkgroups like 10 or 10100 will send to all others in their Zone to cover a greater area.
The V.24 data stream provides the following information (that we have identified and are capturing into a database):
The capture of the above V.24 data as transmitted is collected in both the MQB program memory hashes for fast internal operations, plus a SQL database for later analysis and website updating (see MQB Layout Diagram).
Capturing of this data, and the discovery of verified calibrated DBM signal levels as seen by the FNE, we were able to take two or more input sources and "Vote" the best input source for retransmission (a software comparator). This is a simple RSSI high-watermark design, and has proven reliable and seamless to the end listener. There is still much that can be discovered and improved upon for voting with the introduced variance of the public internet latency/interruptions, and those improvements are ongoing.
Given that we are working with an ever-changing RF signal level, it was determined by using real data captured above that the -115dbm signal level produced a minimal respectable signal level. Anything below that resulted in broken "artifacts", difficult to understand words and increased decoding errors as the signal level continued to decline. We devised a "brick wall filter" to overwrite the voice bytes in each V.24 data output with known good "silent" bytes as long as the RSSI was below -115dbm. This made listening to a marginal signal easier on both the ear and transmitters by staying up between input drops.
More to come...