Joseph Gwinn kirjoitti:
I'm studying up on how IRIG-B decoder circuits work. What are the good approaches, the bad approaches, especially in the presence of noise? (I asked on the NTP group, with little result beyond the C/C++ decoder software written for the audio channel of a 1990s Sun workstation, which it ate alive: 50% cpu load.)
I was also searching chips for IRIG-B decoding lately, but didn't find any. Then I decided to create my own, mostly just for fun but there's also some uses for it. So I ended up to use 8-bit PIC16F873 and do IRIG-B DCLS decoding with it. DCLS means logic level IRIG-B signal from Symmetricom TS2100. So it's not 1 kHz modulated.
At a start it was only a time code decoder... Then, maybe because very rainy weather in the Finland, new features was added daily. For now, it calculates local time (calendar date and weekday) for IRIG-B day number and year and supports european daylight saving time. Leap second is also supported, if it's encoded in the IEEE1344 control bits (Tymserve TS2100 encodes this, leap seconds are flagged one minute before the actual leap second). IRIG-B timecode is also verified by checking it's continuity. If there's momentary errors or total loss of timecode, timing continues in freerun mode. PPS is also generated from IRIG-B with about +-100 ns. maximum jitter (it's one instruction cycle of 'F873, so it cannot be done better with this MCU). If the whole system is rebooted due to long blackout with UPS batteries runout, TS2100 will jump back to January first of current year. And because TS2100 GPS functionality is now dead, it means that it will also continue with wrong time until it's manually set. Because of that, support for TS2100 resets was also added. Now it keeps record of passed dates on the EEPROM... :)
Now I have also DCF77 version of this, where PPS signal is replaced with unmodulated DCF77 timecode with same accuracy than PPS has. I'm planning to use this at least to build some kind of IRIG-B wallclock and possibly to syncronize radio controlled clocks locally with close field magnetic coupling, because actual DCF77 does not work here. However, when testing this I was little bit disappointed when noticed that radio controlled clocks doesn't seem to support leap secods at all. Also their time setting accuracy is not millisecond grade, so the +-100 ns. accurate DCF77 output is little bit overkill when the final setting error can easily seen by eyes...
Code is 100% assembler and full version with DCF77 encoder included (and of course with debug LCD drivers) takes only about 1,5 kilowords and needs only 66 bytes of RAM when running.
And it's still raining in Finland.. Have to see, what features will be added next.. :)
-- 73s! Esa OH4KJU _______________________________________________ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
