[time-nuts] Long period variation of GPS PPS timing?
I am using the PPS from a Garmin GPS16 to timestamp a VLF signal received directly via a PC soundcard. Signal into one channel, PPS into the other. Using pulse centroid timing, I'm seeing about 0.5uS jitter of the pulse-to-pulse interval, and an exponential moving average with time constant 100 seconds is applied to smooth this out. What I have left is a slow cyclic variation of the signal timestamping. It shows up when I measure the phase of an off-air signal, eg here is MSF 60kHz from Anthorn. http://abelian.org/vlf/live/pp100701msfb.png The same variation appears (in unison) on other off-air timing signals, eg DCF and HBG both show the same cycle, so I guess this is variation of the GPS PPS itself, not the VLF signal phase. The period is about 3000 seconds and the amplitude corresponds to about +/- 1.4uS. Having no prior experience of using GPS PPS timing, I was expecting some pulse-to-pulse jitter, but this slow variation has caught me by surprise. Question: is this typical of GPS timing signals, or have I got a problem in my software? -- Paul Nicholson http://abelian.org -- ___ 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.
Re: [time-nuts] Long period variation of GPS PPS timing?
Thanks all for the various replies, on and off list. John WA4WDL wrote: > Perhaps you should use only the leading edges for the > pulse-to-pulse interval measurement. Yes, I'm just using the leading edge, turning it into a pulse using a couple of RC networks, slowing the rise and fall enough to give a nice shape for accurate locating of the centroid. Antonio CT1TE wrote: > I can send you the yesterday record of my GPS TBolt #2 against > DCF77, which apparently doesn't show any periodic phase > variation. Thanks for the offer. Hold on that, I'll assume for now that this slow cycle is not ionospheric. Rob Kimberley wrote: > I'm assuming your location is fixed, and you are tracking > sufficient SVs, Fixed. Maybe I can relocate the GPS antenna for a better view of the sky. Peter Vince wrote: > I don't believe that cyclic variability is GPS - I've not > noticed such a thing on any of the systems I monitor. Thanks, that's very helpful. Knowing that this is not normal behaviour for GPS timing means that I can set about tracking down a system fault. There are a few things I can think of to check. The cycle must correlate with something - soundcard rate, temperature, CPU clock, etc. It is pretty consistent, here is a bit more of MSF 60kHz and also DCF at 77.5kHz, http://abelian.org/vlf/live/pp100701msfc.png http://abelian.org/vlf/live/pp100701dcfc.png Thanks again, I'll report back when I find the problem. -- Paul Nicholson http://abelian.org -- ___ 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.
Re: [time-nuts] Long period variation of GPS PPS timing?
Bob Camp wrote: > I'd put a heat gun on your R/C networks and see what happens... Yes, good point. The RCs must have a temperature coefficient, maybe not something I can neglect, I'd better look at that. In fact, now that I know it's not inherent to GPS, I have quite a long list of things to eliminate. Will get busy... thanks again. -- Paul Nicholson http://abelian.org/ -- ___ 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.
Re: [time-nuts] Long period variation of GPS PPS timing?
Matthew Kaufman wrote: > What is the soundcard sample rate? This feels like an artifact > of sample phase vs. pulse phase to me. There are actually two soundcards, running at 192003.8285 and 192002.1638 samples/sec, varying a little. They are on different computers. One takes an East/West signal, the other a North/South signal. Each is timed and resampled to 192000.0 based on the PPS supplied to both soundcards. The two timestamped data streams are then brought together for bearing and phase extraction. The plotted phase is actually the mean (weighted by signal strength) of the phases separately determined for each signal. Tests with a signal generator show that the signals are being correctly resampled, only a small and fairly constant phase error between the two signals. Perhaps there is some subtle aliasing between the two signals and/or soundcards. I have altered things now to only look at the signal from one antenna, one soundcard. Will know in an hour or so if this has made a difference to the slow cycle. Joe Gwinn wrote: > The centroid of pulses triggered by the leading edge of the > 1PPS pulse will vary with the width of the pulses. I see. It is a passive network but unprotected from temperature changes. It is shut in a box with 2 PCs and a bunch of other stuff. Various fans on and off. Temperature varies 26-30C in the box. I need to look at temperature, but gut feeling tells me the cycle is a little too regular and is some subtle software defect, aliasing or something. I must plot soundcard drift rate against the slow cycle, etc. I can tackle this with confidence now that I know it is not a GPS limitation. Thanks for that. -- Paul Nicholson http://abelian.org -- ___ 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.
Re: [time-nuts] Long period variation of GPS PPS timing?
FYI, some time has elapsed with one soundcard instead of two and the slow cycle is still present, same period, and the phase of the slow cycle did not make a step change. Now I must take a careful look at how the centroid is being determined, the resampling, RC temperature, etc. Maybe the 'sawtoothness' of the cycle should be telling me something. If nothing jumps out at me, I'll change the rise/fall time constants of the RC network and see if that alters the slow cycle. -- Paul Nicholson http://abelian.org -- ___ 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.
Re: [time-nuts] Long period variation of GPS PPS timing?
This morning I waited for the VLF timing signals to settle into their steady daytime phase - still showing the slow cycle of phase variation. Then I power-cycled the GPS, leaving everything else running. During the 25 seconds it takes the GPS to begin sending PPS again, my software continues to time the VLF signal using the average PPS pulse interval measured over the previous 100 or so seconds. See the phase plot of DCF 77.5kHz for the result, http://abelian.org/vlf/live/pp100702dcfa.png GPS rebooted at 16686 seconds, just after the slow cycle started its upward ramp. It seems to have caused a phase jump in the slow cycle - it switched to starting a downward ramp. To confirm, I rebooted the GPS again at 24915 seconds, as it was about half way down a downward ramp. Again a step change of phase, this time it restarted at the bottom of an upward ramp. Just to eliminate any effect in my software, I unplugged the PPS signal from the soundcard for 30 seconds, leaving the GPS running. My software is starved of PPS for the same length of time as during the GPS reboots. This occurred at 27440 seconds, halfway along a downward ramp. The ramp continued down. Just to 'shake the bag' of temperature effects, I opened the door of the cabinet containing the PCs and electronics, this quickly dropped the temperature by 10 C. Also took out the unit containing the RC pulse forming, took its lid off and left it in the open air. Result: no noticeable change to the slow cycle. Well now, this sure looks to me like a GPS effect. I think I'd better order another GPS, a different type, maybe a GlobalSat MR-350P, or something, for comparison. Recommendations? Will also try to move the GPS16 so that it can see more of the sky - only about 50% visible at the moment, some trees block the other half. -- Paul Nicholson http://abelian.org -- ___ 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.
Re: [time-nuts] Long period variation of GPS PPS timing?
David Partridge wrote: > Is this a timing GPS receiver, or a generic fast start > navigation receiver? Just an ordinary nav GPS, Garmin model GPS16HVS intended I think for marine/vehicle applications. Spec says cold boot 45 seconds, warm boot (position unchanged) 38 seconds. Measured warm boot is about 25 seconds. I haven't sent any setup commands to the thing, it is just running with out-of-the-box defaults. The 1PPS is specified at +/-1uS accuracy, which is fine for this application if the error averaged to zero (which it probably does) and was randomly distributed (which it appears not to be). My target for VLF phase error is +/- a few degrees at 77.5kHz, ie about +/- 0.1uS accuracy, and naively I thought I could reach this by a moving average of 100 seconds to smooth out the quoted 1uS jitter of the PPS. But that assumes random jitter! But this 50 minute cycle of about +/- 1.5uS is a bit of a show stopper. Pity, the GPS16 is a nice unit in every other respect. I'm trying to avoid using any specialised, hard to obtain, GPS. Will have to order some other types and hope they have different, better behaved, chip sets. Recommendations welcome. -- Paul Nicholson http://abelian.org -- ___ 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.
Re: [time-nuts] Long period variation of GPS PPS timing?
Peter Vince wrote: > I have just received an ebay notification of new > items for sale by a "favourite seller" - list member > Bob Mokai, aka fluke.l in China: Thanks for the heads-up Peter. That makes three recommendations I've had for this seller. I expect I'll order one or two of these 'timing grade' GPS, likely a Thunderbolt. I still want to see what can be done with 'commodity' GPS devices, but having a proper one as a reference standard would be nice. > I think it is fair to say that several list members have > bought things from Bob, and found him friendly, helpful, > and very trustworthy. Reassuring, thanks. Although I see he's advertised a rubidium clock to 'Enhance the sound quality' of CD, so have doubts now :)) PS I spoke to a chap at Garmin support about the slow cyclic drift of the GPS16X and he asked me to send details. We'll see what comes from that. -- Paul Nicholson -- ___ 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.
Re: [time-nuts] GPS Timing Source -- looking at buying
Chris H wrote: > Would be good as a timing source for GPS > ... make a PC into a Stratum 0 timing device? I think it would count as stratum 1, since you are 1 hop away from a master time source. This page may be of interest, http://www.satsignal.eu/ntp/FreeBSD-GPS-PPS.htm Trimble Thunderbolts have been recommended on this list as a proper timing grade GPS, superior to these EOM Garmin GPS units. There's a reliable source of Thunderbolts on E-bay. I currently have one of these on order, http://cgi.ebay.co.uk/290322053618 -- Paul Nicholson -- ___ 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.
Re: [time-nuts] Long period variation of GPS PPS timing?
Just a follow-up on this. Yesterday I received a Thunderbolt from fluke.l, http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=290322053618 This worked straight away and has cured the 50 minute drift cycle. I'm getting a standard deviation of pulse interval of about 0.25uS, as measured against the soundcard sample clock. It ought to be a bit better than this, but the 10uS PPS width is a bit too short for a steady timing of centroid through a soundcard at 192k samples/sec, I need to stretch the pulse a factor of 10 or so. The Thunderbolt seems to take about 45 mins to settle down to a steady rate, which is fine. Meanwhile, Garmin wrote to say they will try to reproduce the 50 minute drift on the GPS16X. So that problem is fixed. Thanks for the advice and suggestions. -- Paul Nicholson http://abelian.org -- ___ 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.
Re: [time-nuts] Long period variation of GPS PPS timing?
Hal Murray wrote: > From the TBolt PPS pin: > resistor (current limiter) > diode > cap to ground > resistor across cap (decay) A passive solution along those lines would be ideal, will try that before anything more elaborate. Doesn't matter if the resulting pulse isn't symmetric, I just need it to have stable shape and timing, and to span several soundcard samples, maybe 10 or 20. If I can stretch the decay to 100uS, that will do. The T'Bolt seems to have a setup command for everything - except PPS width! Thanks, -- Paul Nicholson http://abelian.org -- ___ 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.
Re: [time-nuts] Pulsars
Hal Murray wrote: > What sort of gear does it take to hear a pulsar? > ... Is there a convenient one up near the north pole? A potential target is PSR0329+54, 1.5 Jansky at 400Mhz, which corresponds to about 1.5E-26 W/m^2/Hz. I did some rough calculations and concluded that with a 16dB gain beam, a 120 hour epoch-folding integration might be enough for it to show up on a 'periodgram'. This was with 150kHz bandwidth, 40K front-end temperature. I tried this about 10 years ago, with no success but I didn't have very good antenna steering - I used a bank of four phased high-gain UHF TV antennas, polar mount clamped at +54 declination. It was necessary to write software to continuously re-time the received signal to barycentric coordinates before the integration. I injected a 1PPS from MSF 60kHz as the timing reference. A fun project, one that I will return to one day. -- Paul Nicholson -- ___ 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.
Re: [time-nuts] Trimble Tbolt kits on eBay
In July I bought this kit from a reliable supplier in China, http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&item=290322053618 It arrived in 6 days and I had it up and running within an hour, and it's been doing a solid job since. -- Paul Nicholson -- ___ 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.