Joerg Schilling wrote: > James Carlson <carls...@workingcode.com> wrote: > >>> But I also mentioned that is is simple to write a driver that supports such >>> information. As Solaris does not support DCF-77 receivers, I did e.g. write >>> a >>> small driver that installs itself in the related callback tables from the >>> "zs" >>> driver in order to fetch the time when a related line from the DCF-77 >>> receiver >>> indicated a second clock signal. This driver later allowed to read the time >>> stamp via ioctl. >> As long as your DCF-77 receiver has a PPS output, the system already >> supports it, and has supported it for ages. No special driver required. >> >> Look into the TIOCSPPS ioctl on the termio(7I) man page. Instead of >> relaying the information to user space (which involves latency), the >> mechanism automatically adjusts the system clock and latches the >> timestamp of the last tick. You can get the timestamp with TIOCGPPSEV. >> >> I'm pretty sure that xntpd supports this interface. I remember a >> co-working in Burlington setting up a test system with a GPS time source >> many years ago. > > I did implement my code between 1992 and 1994 and the way I interpret the > documentation, it seems that the Solaris driver does not deal with the fact > that the pulse at the 59. second is missing. > > DCF-77 uses cheap reaceiver hardware that has been developed by Telefunken > for > Junghans wrist watches around 1980. For decoding the 0s and 1s in the data > stream, you set the baudrate to 50 baud and then use a char table to detect > 100ms vs 200 ms pulses. > > For DCF-77, you then need to throw away the pulse start time samples that are > outside the standard deviation and you need to implement a low pass on the > rest of the samples. If you do it the right way, you get aprox. 200 usec > accuracy when you have a user space daemon that implements the filters and > that sets an adjust value for adjtime(). > > Is this implemented on Solaris now?
As I plainly said before, if your hardware supports standard PPS output -- a quick Google search tells me that the Meinberg COM52HS supports this, but I have no clue about the hardware you were using -- then the generic PPS feature in the Solaris serial drivers should do the trick. If not, then you'd probably need to modify the drivers or buy better hardware. I suspect, though, that going up to user space to do the work (rather than doing it in the driver) is a bit of a bummer, particularly on a busy system where dispatch delays could become unpredictable. That's why it's quite intentionally not implemented that way on Solaris or OpenSolaris. In any event, this entire discussion has not a whit to do with the original poster's question. So I'm done. -- James Carlson 42.703N 71.076W <carls...@workingcode.com> _______________________________________________ opensolaris-code mailing list opensolaris-code@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
