Back in August, I posed about some internal changes to the XLX330 Megalink transcoding reflector. The problem attempting to be solved was that due to the addition of and subsequent interest in DMR TG 311070 and not really wanting to buy another DVstick, the system was changed to use software-based processes to transcode to YSF to and from DMR using MMDVM_Bridge. After about four months of spotty, unreliable behavior, I bought a DVstick30 and moved all digital modes back to the integrated xlxd process.
Since August, the performance of the system has been unreliable. At random times, xlxd would “hear” the D-STAR or DMR traffic, but not reliably transcode between those modes and YSF. Additionally, after some period of time being in that state, transcoded audio between D-STAR and DMR would start to fail with errors in the ambed process. These seem to be correlated to hiccups with Brandmeister DMR. However no one cause of the hiccups was identified.
In order to reduce the complexity, I purchased a DVstick30 (single channel), removed the MMDVM_Bridge and YSFReflector processes, and re-enabled xlxd to serve as the YSFReflector service. However the transcoder process still fails from time to time. So there is now a “watchdog” script that will watch for transcoder timeout errors and restart the process. This should result in the system not working no more than 63 seconds when the issue hands up (60 second wait time, 3 seconds to re-establish the ambed transcode connection).