Re: [ntp:questions] Trying to use Dimension 4 time keeper
On 14/09/2013 21:39, W. eWatson wrote: [] A little bit of data from the two XP PCs: Older and more troublesome PC Jitter high 300-650 Offset around 685 Delay about 100.xxx Reach 80 jumped to 377 an hour later One * for the single pool Newer XP Jitter low, 11-8 Reach 377 * + - + - on five servers Delay 100-150 Those are fairly rough, as seen in a five minute span. To really make any progress I would think a monitor of the data would be more helpful. I noticed Broken Clock in the documentation mentioned above. It looks tricky to evaluate. In my situation, I would be happy if one of the important parameters would sound an alarm in my den when one went out of bounds. Fat chance of that. How are you connected to the Internet? Let's start with your newer PC. The delay of 100-150 ms is much higher than I would expect for a direct "fast" Internet connection, it sounds more like dial-up. If you are using the "pool" directive as I suggested (NTP 4.2.6p5 required), I would expect to see more than five servers listed. You can monitor by enabling the statistics collection within NTP, and by using MRTG as explained on my Web site: http://www.satsignal.eu/ntp/NTPandMRTG.html If statistics are enabled, both Meinberg and I offer plotting programs: http://www.meinbergglobal.com/english/sw/ntp-server-monitor.htm http://www.satsignal.eu/software/net.htm#NTPplotter The "Broken clock" problem may manifest itself by NTP reporting that it is stepping the clock, or even giving up. Look for NTP entries in the event log to monitor this. To get alarm monitoring, you could use SNMP (although that's not yet in the Windows port of NTP), or you could write a small Perl script to be run once every few minutes by the Task Scheduler, which checked the parameters you want to monitor by running the appropriate ntpq command, and which then sounded the alarm as required. The alarm could be a wave file which you play - perhaps even inside the Perl script. I have Perl script examples on my Web site at the first URL I gave, and there is information here: http://www.xav.com/perl/site/lib/Win32/Sound.html on playing .WAV files with Perl in Windows. Sounds like a good Autumn project! Having said all of that, my situation a decade back was similar to yours - wanting to have several Windows PCs well synced. Were I to be tackling the same problem today, I would spend less than US $50 on a Linux PC (Raspberry Pi) and sync it to my Internet sources, and then sync the Windows PCs to the Raspberry Pi (using a much shorter poll interval, 32 seconds, maxpoll 5). If the location had GPS available (e.g. near a window with a sky view) I would spend approx. US $100 instead and add an Adafruit GPS to the Raspberry Pi, making a unit known to correct to a few microseconds, and sync my Windows PCs to that (with a "pool" directive for backup). -- Cheers, David Web: http://www.satsignal.eu ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Trying to use Dimension 4 time keeper
On 2013-09-15, David Taylor wrote: > On 15/09/2013 00:06, unruh wrote: > [] > >> I was probably refering to the Sure unit which is more difficult. >> I have never looked at the Adafruit module. Also doesn't the Adafruit >> also require an extra antenna to work properly? > > Yes, the Sure unit, another low-cost option, does require a little > modification to extract the PPS signal at RS-232 levels, although based That would completely burn out the RPi GPIO ports. > on my current experiences likely the more readily available TTL-level > PPS signal would actually be satisfactory unless very long cable tuns > were involved. I have detailed the modification required here: > >http://www.satsignal.eu/ntp/Sure-GPS.htm > > With the Adafruit unit I'm not using an external antenna in my "lab", on > the second floor of a two-storey building. The GPS signal creeping in > via walls and roof is enough. The Adafruit module has both a built-in ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Trying to use Dimension 4 time keeper
On 15/09/2013 08:09, unruh wrote: On 2013-09-15, David Taylor wrote: [] Yes, the Sure unit, another low-cost option, does require a little modification to extract the PPS signal at RS-232 levels, although based That would completely burn out the RPi GPIO ports. Which is why my own recommendation has been to use the Adafruit GPS module, which is 3.3V level compatible with the Raspberry Pi. Here's the URL - again! http://www.adafruit.com/products/746 -- Cheers, David Web: http://www.satsignal.eu ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Trying to use Dimension 4 time keeper
On 9/14/2013 11:53 PM, David Taylor wrote: On 14/09/2013 21:39, W. eWatson wrote: [] A little bit of data from the two XP PCs: Older and more troublesome PC Jitter high 300-650 Offset around 685 Delay about 100.xxx Reach 80 jumped to 377 an hour later One * for the single pool Newer XP Jitter low, 11-8 Reach 377 * + - + - on five servers Delay 100-150 Those are fairly rough, as seen in a five minute span. To really make any progress I would think a monitor of the data would be more helpful. I noticed Broken Clock in the documentation mentioned above. It looks tricky to evaluate. In my situation, I would be happy if one of the important parameters would sound an alarm in my den when one went out of bounds. Fat chance of that. How are you connected to the Internet? Let's start with your newer PC. The delay of 100-150 ms is much higher than I would expect for a direct "fast" Internet connection, it sounds more like dial-up. If you We use AT&T DSL. My wife has noticed at time, the speed has been slow; however, I would expect the newer PC to have time failures. As I wrote somewhere, the newer PC is mostly off, so I'll keep it on for several weeks to see if the problem shows there. are using the "pool" directive as I suggested (NTP 4.2.6p5 required), I would expect to see more than five servers listed. I have 4.2.7p5 on the older PC. The five are on the newer PC. Since it seems to hold time OK, I've left it that way. The pooled idea is on the older PC. You can monitor by enabling the statistics collection within NTP, and by using MRTG as explained on my Web site: http://www.satsignal.eu/ntp/NTPandMRTG.html If statistics are enabled, both Meinberg and I offer plotting programs: http://www.meinbergglobal.com/english/sw/ntp-server-monitor.htm http://www.satsignal.eu/software/net.htm#NTPplotter The "Broken clock" problem may manifest itself by NTP reporting that it is stepping the clock, or even giving up. Look for NTP entries in the event log to monitor this. To get alarm monitoring, you could use SNMP (although that's not yet in the Windows port of NTP), or you could write a small Perl script to be run once every few minutes by the Task Scheduler, which checked the parameters you want to monitor by running the appropriate ntpq command, and which then sounded the alarm as required. The alarm could be a wave file which you play - perhaps even inside the Perl script. I have Perl script examples on my Web site at the first URL I gave, and there is information here: http://www.xav.com/perl/site/lib/Win32/Sound.html on playing .WAV files with Perl in Windows. Sounds like a good Autumn project! Having said all of that, my situation a decade back was similar to yours - wanting to have several Windows PCs well synced. Were I to be tackling the same problem today, I would spend less than US $50 on a Linux PC (Raspberry Pi) and sync it to my Internet sources, and then sync the Windows PCs to the Raspberry Pi (using a much shorter poll interval, 32 seconds, maxpoll 5). If the location had GPS available (e.g. near a window with a sky view) I would spend approx. US $100 instead and add an Adafruit GPS to the Raspberry Pi, making a unit known to correct to a few microseconds, and sync my Windows PCs to that (with a "pool" directive for backup). Well, thanks for the help, but this problem is beyond me. The organization that sponsors the science effort is remiss in providing some solution. Unfortunately, the problematic PC has produced a new problem. A few weeks ago I decided to check the PC battery. In doing so I had to take out a graphics card and a video grabber card to get to the battery. I found the battery in good shape, and put everything back together. Several days later I found the PC beeping. Subsequently, I found that the beeping began from around five minutes to maybe an hour after I had logged in, possibly as long as a day. The science s/w was put into play immediately after logging in. I figured it was heat, but that idea pretty much fizzled. I need to check the power supply to see if it somehow gone bad. As a last resort to the timing problem, I might bring the PC into my den and see if the timing is still flaky. This is no longer a high priority issue. I'll pick at it as I can. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Trying to use Dimension 4 time keeper
On 2013-09-15, David Taylor wrote: > On 15/09/2013 08:09, unruh wrote: >> On 2013-09-15, David Taylor wrote: > [] >>> Yes, the Sure unit, another low-cost option, does require a little >>> modification to extract the PPS signal at RS-232 levels, although based >> >> That would completely burn out the RPi GPIO ports. > > Which is why my own recommendation has been to use the Adafruit GPS > module, which is 3.3V level compatible with the Raspberry Pi. Here's > the URL - again! > >http://www.adafruit.com/products/746 Yes, I saw it. Not sure what the voltages are on the Sure unit -- the direct output of the PPS but I think 5V which is still a bit high. I guess one could put a diode in series with the line to bring it down to 3.3V. > ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Trying to use Dimension 4 time keeper
On 15/09/2013 18:08, W. eWatson wrote: [] As a last resort to the timing problem, I might bring the PC into my den and see if the timing is still flaky. This is no longer a high priority issue. I'll pick at it as I can. Just pop back if and when you need further help. -- Cheers, David Web: http://www.satsignal.eu ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Trying to use Dimension 4 time keeper
On 15/09/2013 18:37, unruh wrote: [] Yes, I saw it. Not sure what the voltages are on the Sure unit -- the direct output of the PPS but I think 5V which is still a bit high. I guess one could put a diode in series with the line to bring it down to 3.3V. I would recommend using a restive divider - perhaps 10K and 4.7K to start with - and monitoring the voltage into the Raspberry Pi. Using resistors provides a current limit, and allows the voltage to get down to zero. If the voltage is well below 3.3V, the upper 4.7 K resistor could be reduced. Best to ask on one of the Raspberry Pi support groups as someone has likely already done the measurements. But the Sure unit is so big, physically, that I would still prefer to use the Adafruit instead. -- Cheers, David Web: http://www.satsignal.eu ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Trying to use Dimension 4 time keeper
On 2013-09-15, David Taylor wrote: > On 15/09/2013 18:37, unruh wrote: > [] >> Yes, I saw it. Not sure what the voltages are on the Sure unit -- the >> direct output of the PPS but I think 5V which is still a bit high. I >> guess one could put a diode in series with the line to bring it down to >> 3.3V. > > I would recommend using a restive divider - perhaps 10K and 4.7K to > start with - and monitoring the voltage into the Raspberry Pi. Using > resistors provides a current limit, and allows the voltage to get down > to zero. If the voltage is well below 3.3V, the upper 4.7 K resistor > could be reduced. Best to ask on one of the Raspberry Pi support groups > as someone has likely already done the measurements. Not sure why you would want a resistive divider. The diode would do it just as well and you would not have to worry about the input impedence of the GPIO. The GPIO already have an input impedence and you should not have to worry about current limits as long as the voltage was low enough. > > But the Sure unit is so big, physically, that I would still prefer to > use the Adafruit instead. It does have a usb for both power and putting out the nmea sentences. But it certainly is bigger. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Trying to use Dimension 4 time keeper
David Taylor wrote: On 15/09/2013 08:09, unruh wrote: On 2013-09-15, David Taylor wrote: [] Yes, the Sure unit, another low-cost option, does require a little modification to extract the PPS signal at RS-232 levels, although based That would completely burn out the RPi GPIO ports. Which is why my own recommendation has been to use the Adafruit GPS module, which is 3.3V level compatible with the Raspberry Pi. Here's the URL - again! http://www.adafruit.com/products/746 My Sure gps has both 5V ttl and 3.3V levels available by soldering wires or fitting a pin header. I had tried both the 5V ttl to parallel port and currently using your rs232 modification. I didn't note any significant difference in parallel port vs serial port. David ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Trying to use Dimension 4 time keeper
On Sun, 15 Sep 2013 18:56:26 +0100, David Taylor wrote: > On 15/09/2013 18:37, unruh wrote: > [] >> Yes, I saw it. Not sure what the voltages are on the Sure unit -- the >> direct output of the PPS but I think 5V which is still a bit high. I >> guess one could put a diode in series with the line to bring it down to >> 3.3V. > > I would recommend using a restive divider - perhaps 10K and 4.7K to start > with - and monitoring the voltage into the Raspberry Pi. Using resistors > provides a current limit, and allows the voltage to get down to zero. If > the voltage is well below 3.3V, the upper 4.7 K resistor could be reduced. > Best to ask on one of the Raspberry Pi support groups as someone has > likely already done the measurements. http://jamesreubenknowles.com/level-shifting-stragety-experments-1741 might be of interest. For converting serial signals it works fine, for PPS where one worries about being 100ns off maybe not so much. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Trying to use Dimension 4 time keeper
On 15/09/2013 20:08, David Lord wrote: [] My Sure gps has both 5V ttl and 3.3V levels available by soldering wires or fitting a pin header. I had tried both the 5V ttl to parallel port and currently using your rs232 modification. I didn't note any significant difference in parallel port vs serial port. David Thanks for the reminder, David. Yes, both the serial and the PPS are available on the headers at 3.3V levels. Should completely suit those who wish to run the Sure GPS with the Raspberry Pi. Good to have the comparison report between serial/RS-232 and parallel/TTL levels. I suspect that for NTP use the difference is negligible, although "time-nuts" measuring to the tens of nanosecond level would see a small propagation delay offset. -- Cheers, David Web: http://www.satsignal.eu ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Trying to use Dimension 4 time keeper
Not sure why you would want a resistive divider. The diode would do it just as well and you would not have to worry about the input impedence of the GPIO. The GPIO already have an input impedence and you should not have to worry about current limits as long as the voltage was low enough. Because: (a) the diode would prevent the source from sinking current. Sometimes the inputs to digital circuits will have a pull-up resistor to +Vcc (i.e. +5 or +3.3 V), and the diode would prevent the input being pulled down to zero. (b) having a suitably high resistance in circuit will limit the current, and may help prevent damage if the voltage levels weren't quite as expected. As David Lord helpfully pointed out, the Sure board also has the signals you need at 3.3 V level, so a direct connection should be possible. But the Sure unit is so big, physically, that I would still prefer to use the Adafruit instead. It does have a usb for both power and putting out the nmea sentences. But it certainly is bigger. The Adafruit unit also offers NMEA sentences over its serial line. With the Raspberry Pi, there are only two USB ports, and many GPIO pins, so you may prefer to use a serial rather than a USB feed for the NMEA data, but both do work (tested here using gpsd). Please bear in mind that I'm not suggesting that a Raspberry Pi and Adafruit GPS is the ultimate NTP machine - far from it - but it may well be a cost-effective, low-power add-on to an existing network, particularly for the OP who requires to sync Windows PCs. Having a local stratum-1 server (or more) is a significant improvement over Internet sync alone. -- Cheers, David Web: http://www.satsignal.eu ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Trying to use Dimension 4 time keeper
On 2013-09-16, detha wrote: > On Sun, 15 Sep 2013 18:56:26 +0100, David Taylor wrote: > >> On 15/09/2013 18:37, unruh wrote: >> [] >>> Yes, I saw it. Not sure what the voltages are on the Sure unit -- the >>> direct output of the PPS but I think 5V which is still a bit high. I >>> guess one could put a diode in series with the line to bring it down to >>> 3.3V. >> >> I would recommend using a restive divider - perhaps 10K and 4.7K to start >> with - and monitoring the voltage into the Raspberry Pi. Using resistors >> provides a current limit, and allows the voltage to get down to zero. If >> the voltage is well below 3.3V, the upper 4.7 K resistor could be reduced. >> Best to ask on one of the Raspberry Pi support groups as someone has >> likely already done the measurements. > > http://jamesreubenknowles.com/level-shifting-stragety-experments-1741 > might be of interest. For converting serial signals it works fine, > for PPS where one worries about being 100ns off maybe not so much. > His input impedence has a capacitor in it (about 100pf it looks like). 5V__\|__3.3V /| Z Z Input impedence Z Z ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions