My local radio club has put together a D-Star repeater and my role in it has been to set up the gateway computer. The gateway software is proprietary to Icom (no source code) and installs fairly easily as long as you do everything exactly as the poorly translated documentation specifies.
The documentation recommends CentOS 5.1, although I used 5.8. I used the 32-bit version as the administrators of the US Root Trust server, which manages the linking between hundreds of repeaters together over the Internet caution against the 64-bit version, and although I’ve usually managed to get most things to work on 64-bit CentOS, there was no reason to swim upstream on this project. I’ve also heard cautions about CentOS version 6 (now up to at least 6.2). I imagine the main problem is the tightening down of SELinux, but again, not worth the aggravation.
US Trust also has documentation “Joining the Network6″ which clearly explains what is necessary. I didn’t have much trouble getting it to work, although the last step of registration was delayed, I think, by a typo in one of the conf files.
The most curious part of the installation is that the Icom script installs a whole bunch of open source software packages (like Apache httpd, Tomcat, Postgres, etc.) from source. I suppose it is a way to control which version is installed.
The system worked as advertised, although at this moment it is down because the laptop which the (prototype) gateway is running on is prone to crashing every now and then with a hardware related kernel panic. It being the weekend of the Dayton Hamvention, I can’t get at it to reset it. Hopefully we’ll find a suitable server to run it on soon.
It’s a fairly small setup, with just 2 repeater modules and a controller. We’ve connected to the REF001C reflector, which is apparently the busiest of them all, but never saw more than 5% CPU utilization even on a 5 yr vintage laptop. So CPU performance doesn’t seem to be required. The memory utilization started around 500MB, but crept up to 900MB over a week or so of operation. I’ve heard there’s a memory leak somewhere in the software but no confirmation yet.
So what do I think of D-Star? It’s OK I guess. I’m not sure I’ll be buying a D-Star radio. The linking is pretty straightforward, and as it’s embedded into the signal can be supported by many memory slots in a radio like the ID-880h. The fundamental D-Star linking method, which is “connectionless” in that every transmission in both directions has to identify the path, seems less popular than the more traditional linking method of creating a nailed-up connection.
The low-speed data transmission capability is really slow. In combination with D-Rats a small email takes a couple of minutes to transfer. A PDF or other large document would take a very long time indeed. Technologies like HSMM-MESH seem more promising for usable data networking.
I was also shocked when setting up an ID-880h borrowed from the club that the speaker is in the body of the radio, not the front panel and the microphone also plugs into it. It seems pointless to separate the front panel from the radio by 12 ‘ but have a 3′ microphone cable attachment. You can buy extensions and external speakers, of course, but why not have the panel handle everything with one attachment cable? It’s not a D-Star thing though, the Yaesu FT-857 has the same setup.
So I’m not sure what radio is going to end up in my new Highlander. The D-Star options are very limited, either a plain dual-bander (70cm, 2m), a 1.2GHz (the Icom ID-1) only radio, or a very expensive all band base station radio.