Loading ...
Sorry, an error occurred while loading the content.

1584[dxatlas] Re: Timing Question

Expand Messages
  • vk4iu
    Nov 18, 2009
      Hi Dave,

      Accurate time is a difficult thing. As the other messages explain - a serial port/USB based "consumer" GPS just does not deliver time accurately enough for use by Faros, even if the routines were included inside Faros - they were never designed to do so.

      You may be confused by the use of the serial port to interface the PPS - pulse per second - industrial GPSs to a PC - the ones G0WBX and I intend using.

      The technique makes use of the CHIP that implements the serial port - it does NOT use data sent serially from the GPS. The GPS has a wire that connects directly to the CD - carrier detect - bit of the CHIP that implements the serial port on the PC. Special, real time, high performance, sometimes kernel mode, "driver software", is needed in the NTP software on the PC to respond to the "pulse" every second on the carrier detect line and from that, keep accurate time. The pulse is generated by the GPS chip from signals received from the GPS satellites. Fundamentally, the GPS system uses time to calculate position, and knows time accurate to microseconds.

      In a serious implementation they even worry about the temperature of the CPU chip in the PC causing instability of the "PC clock" that starts and stops - schedules - the "driver software" that responds to the "pulse" on the carrier detect - causing errors in time in the NTP software. They try to ensure that the intensity of activity on the PC is smooth and regular to maintain an even temperature.

      Given the intense, real time nature of the Faros activity with the sound card and the "DSP style" analysis of the signal, the two requirements - signal analysis, and PPS time calculations - could very well interfere with each other if implemented on the ONE PC. Only real practical experiments will determine that. Clearly the raw speed of the PC CPU would be a factor. Alex's use of a Kalman filter, with very low CPU load, in Faros, against ntp based internet time, ensures there is plenty of CPU power for the intense analysis of the beacon signal even on the low power PCs common in amateur radio.

      Have you found a suitable list of time servers for your system?

      Peter VK4IU

      --- In dxatlas_group@yahoogroups.com, "Dave" <dave@...> wrote:
      > Hi Dave Baxter- That's the issue and that's the solution. The closer we
      > get to a true timing signal the better the app will work. Has Alex
      > responded? I haven't any idea how difficult adding a timing routine to
      > extract time from a GPS would be, but the true value in the app has to be
      > rooted in timing. I prefer the GPS serial-port/USB solution as opposed to a
      > receiver into a sound card channel if I had a voice. Some GPS receivers
      > boards are fairly inexpensive, but, if one has a GPS already, why not use
      > that device and not buy more *stuff*?
      > 2c worth,
      > Dave
      > /++++++++++++++++++++++++++++++++/
      > -----Original Message-----
      > From: dxatlas_group@yahoogroups.com [mailto:dxatlas_group@yahoogroups.com]
      > On Behalf Of Dave Baxter
      > Sent: Wednesday, November 18, 2009 10:19
      > To: dxatlas_group@yahoogroups.com
      > Subject: [dxatlas] Re: Timing Question
      > Hi Peter (David)
      > Just got back from a frustrating on site service visit.
      > Yes, I know the Satsignal site, and have had a good exchange with the David
      > Taylor there and have successfully got a FreeBSD machine suposedly
      > configured to work as needed, from the same distro and version as he used,
      > but I've yet to get it to successfully see and work with a Garmin GPS16LVC,
      > that has a PPS output (yes, I buffered it to RS232 levels.) The problem I
      > have with all that, is I know next to nothing about that level of tweakage
      > at the kernel level of any 'nix, that and the 'nix community seems intent on
      > flaming anyone who asks questions that they do not see as relavant to their
      > latest and greatest distro. (or more relevantly, they do not themselves
      > understand.)
      > You also it seems found out the hard way what I was told, that the timing
      > accuracy of the NMEA sentence delivery from any GPS receiver currently
      > available, is not good enough for time keeping to the level we
      > need. The eTrex (I have one of the "Yellow" ones too) is otherwise a
      > good bit of kit, but useless for any accurate timekeeping.
      > Also, like you, I've found that protocols like NTP don't play well with
      > mobile internet services. 3G or otherwise.
      > VLF Radio time keeping broadcasts could be another way to go, but again, the
      > hardware to do the job is prohibitivey expensive, and none of the PC
      > software I've found so far that can decode signals like MSF, DCF, WWV etc,
      > provide a true full NTP server function. (Some do SNTP, that's about the
      > best it gets, and Faros doesnt do SNTP, only NTP it seems.)
      > My ISP issues. Well, (long story short) during the business day, there
      > seems to be a drastic increase in the network latency getting to anything,
      > evne in their own domain/network! If that is not bad enough, it's variable
      > one ping to the next, by a wide margin (10's to 100's of mS variation, but
      > no packet loss, that is their goal it seems.) Evenings, early mornings and
      > weekends, Latency is down to typicaly 25mS.
      > During the day, it can go to as much as 500mS, but routinely stretches out
      > to 150 to 300mS, depending on the exact time of day, day of week
      > etc. Some days are better, some worse. Watching the time accuracy
      > plot in Faros, at certain times of day, you can see what looks like the
      > effect of someone throwing a switch somewhere. I did some checks, and
      > it's not just me, other users of the same ISP are experiencing similar
      > effects, though they don't use NTP as we do. It is however messing with
      > lots of other systems it seems too.
      > As a result, it totaly screws up NTP for any sort of accurate time
      > synching, like Faros needs. Hence why I was messing with FreeBSD
      > trying to setup a local Stratum 1 server etc. It's getting to the
      > state at home now, where I am going to take another serious stab at it, and
      > see if I can get it going on an old early pentium laptop, as that is much
      > smaller, quieter, and uses less power, than the current desktop I have sort
      > of working.
      > I'd idealy like to have a "Time Appliance", but commercial units are way
      > outside my pocket money range, and like hen's teeth to find too. The
      > people who port 'nix etc onto old router hardware are also uninterested in
      > helping, from the few exchanges I've had on other forums, and I don't know
      > enough to do it myself.
      > I did find this... http://scss.com.au/family/andrew/gps/ntp/ But...
      > When I asked if he would release a copy of the BINARY (not the propriatry
      > TCP stack 'C' code from the compiler) he's gone all silent
      > (like no reply, ever!) Pity, as that seems to be what we need. Maybe
      > someone else could persuade him to if not release the BINARY, then perhaps
      > sell pre-programed (and copy protected) PIC's for such use?
      > For now, Faros bumbles along otherwise OK, but when you look at the plots,
      > the path charts in the UK afternoon, are often showing a snow storm of white
      > "unknown" spots. http://g8kbv.homeip.net:8008/
      > As a reception report, most people are interested in the fact that a beacon
      > has been realiably heard, not what path it was heard by, I think.
      > What I'd like to see long term, is GPS PPS support built into Faros itself,
      > that's where it's needed, NTP could still be used if GPS
      > reception is not posible, or as a fall back. That alone would make
      > Faros most valuable when out portable, or for DXpeditions etc. But of
      > course, only Alex can implement that.
      > Another option, but would also no doubt be a major undertaking on Alex's
      > part, would be to use one channel of the soundcard for the beacons, and
      > one channel for a radio code input (WWV, MSF etc) No GPS needed, just
      > two RX's, and one of them could be a realy simple SDR like thing.
      > (Keeping the audio tones very different, so any crosstalk would be ignored.
      > Or, take a pulse train from any of the commonly available miniature VLF
      > clock receiver modules, into a com port line somewhere, and use that to set
      > Faros's time.
      > That's enough from me I suspect. Oh, one more thing. The Javascript I
      > use, is more or less that, that is cited as an example to use, if you look
      > on the main NCDXF IBP site links. http://faros.ve3sun.com/ is
      > where you need to go to start that learning exersize! I make no claims
      > for originality on that subject.
      > Cheers All.
      > Dave Baxter
      > G0WBX.
      > --- Original Message ---
      > My current plan.
      > I initially dismissed creating a stratum 0 reference time server - the
      > Meinberg PCI card is priced at $3500 in Australia. But ...
      > Alex VE3NEA alerted me to the following web site http://time.qnan.org I
      > have ordered a Garmin 18x LVS, $100 n Australia, and I will proceed to
      > integrate it into my Linux server. On that site there is a reference to
      > http://www.satsignal.eu/ntp/index.html which is a fantastic source of
      > information relevant to getting Faros a good time signal. I have high hopes
      > of having a fantastic Faros installation for propagation study, site and
      > antenna comparisons.
      > Peter VK4IU
      > ------------------------------------
      > Yahoo! Groups Links
      > ----------
      > No virus found in this outgoing message.
      > Checked by AVG - www.avg.com
      > Version: 8.5.425 / Virus Database: 270.14.72/2511 - Release Date: 11/18/09 07:50:00
      > [Non-text portions of this message have been removed]
    • Show all 29 messages in this topic