Timing Errors of the Garmin GPS 35 HVS
by Oliver Kloes
PLEASE NOTE:
These errors are caused by the GPS 35 only. The STVASTRO is a
valuable tool for occultation observers and works fine.
INTRODUCTION:
I have been using the Garmin 35 together with the STVASTRO video
time inserter for measuring occultations of stars by the moon
and asteroids. I then received a warning to be careful using a
GPS receiver for timing. There could be "jumping seconds" and
the GPS could give the wrong time. So I contacted other
observers, who were using the "35" with the STVASTRO. None of
them had ever observed a "jumping second". One of the
observers, Steve Preston from Seattle, U.S.A., suggested looking
for these errors with the help of the "KIWI" timing software.
So Geoff Hitchcox from Christchurch, New Zealand, joined the
hunt for timing errors. During the next five months, with more
than 300 e-mails and hundreds of hours testing, we really found
"jumping seconds" and solutions for this problem. The following
information will be helpful for users of the "35" and the
STVASTRO. Even "35" users with other timing purposes will find
the information valuable.
To provide an easy overview, I have created the following Q & A:
1. How was testing done?
2. What kind of timing error can happen?
3. What causes these errors?
4. How often will these errors appear?
5. Will there be a fix from Garmin?
6. Can I predict timing errors?
7. What can I do about these timing errors?
8. LINKS:
9. Thanks to:
HOW WAS TESTING DONE?
Steve Preston used a GPS 35 HVS with software version 2.5 with good sky coverage.
Oliver Kloes used a GPS 35 HVS with software version 2.21 with limited sky coverage (30% open sky)
Geoff Hitchcox, using the data from S.P. and O.K., developed
KIWI according to the test results. New versions were then
tested by O.K. and S.P. During most of the tests, the display
of STVASTRO was recorded on a video recorder. Test runs had
durations of hours up to days.
WHAT KIND OF TIMING ERROR CAN HAPPEN?
We found different errors:
1. STVASTRO display stuck for one second. Example: Display of seconds reading:
11 12 13 13 15 16
Second "13" is double, second "14" missing.
2. STVASTRO display reading (seconds): 39, 40, 41, 42, 42, 43, 44, 46, 47
Second "42" is double, the seconds "43" and "44" are late for
one second, second "45“ is missing.
WHAT CAUSES THESE ERRORS?
We know two sources of error:
1.SLOW CALCULATION OF THE GPS 35
According to our tests, it is the "slow" CPU of the GPS 35 that
causes these errors. KIWI can now show the time of transmission
of the NMEA data (At option#8 KIWI 1.07). Sampling the NMEA
data for a period of time, can show the transmission of the data
is sometimes very late.
Example:
KIWI screen:
13-03-2003 23:30:47, 479, 485, 627, 629, 772
Meaning:
The NMEA data of second 47 was transmitted the following way:
479 ms from 1PPS, "$" start of the RMC
485 ms from 1PPS, "R" 4th character of RMC
627 ms from 1PPS, LF END of the RMC
629 ms from 1PPS, "$" start of the GGA
772 ms from 1PPS, LF END of the GGA
(To operate correctly with the STVASTRO, the GPS 35 must be
programmed to transmit only the RMC and GGA sentences.)
But sometimes you can see this:
13-03-2003 23:29:20, 685, 691, 833, 835
The GGA sentence is completed in the next second AFTER the following 1PPS.
In some cases, the NMEA data overflowed to the following
second. The data of second A is shifted to following second B
and the GPS transmits the wrong time to the STVASTRO causing
jumping seconds.
2. MISSING 1PPS
In rare cases the GPS 35 "forgets" the 1PPS signal, so there are
2 or more seconds between pulses. Each 1PPS signal is timed by
the PC, if it is outside a window of plus or minus 50
microseconds (that's a 50 millionths of a second), KIWI gives an
alert and logs the details.
If a 1PPS signal is missing for one whole second, jumping
seconds will occur.
HOW OFTEN WILL THESE ERRORS APPEAR?
All these errors are very rare!!!
You will get about five in a 12 hour run. So less than 0.02 % of
all timings of the GPS are wrong. But you have to be aware that
they can happen!
WILL THERE BE A FIX FROM GARMIN?
Unfortunately we don't think so. Garmin was informed about our
results, but it seems the limits of the hardware can't be
corrected by a new software update. The "35" belongs to an older
generation of Garmin receiver, and they may not spend any more
research on this unit.
CAN I PREDICT TIMING ERRORS?
We found only one error that shows a pattern.
At a time of XX:00:06, we sometimes found the display (seconds):
04 05 06 06 08 09
We called this specific error "T06" (Triple zero six), because
it appears at a time when the display reads three zeros and one
six.
This "T06" appears more at even hours than at uneven hours, but
you can't count on it. In a typical 12 hour test run, you get
two or three of these errors - sometimes none!
Something happens at the sixth second after a full hour. It
seems that the "T06" is a timed housekeeping task in the "35"
that sometimes cannot be completed in time, like us trying to do
65 minutes work in an hour. It is because someone has programmed
too much work for a "small brain" (CPU) used in the Garmin 35.
That is why it cannot be fixed by a software upgrade.
Other errors are random.
WHAT CAN I DO ABOUT THESE TIMING ERRORS?
With good sky coverage for the GPS receiver, these errors appear
less often than with poor sky coverage. But you can't eliminate
these errors completely with an open sky.
You have four options:
1. Be aware timing errors can occur
The easiest and cheapest way. You know now that there can be
errors in the timing display. So check your recordings around an
occultation for "jumping seconds" carefully, and correct the
time in your observation report if an error appears. It's an
easy task and should be no problem for any observer. As the
errors appear not very often, it is possible you will never get
one during an occultation ever. The longest timing error we
ever found was about 3 seconds. According to our knowledge
today, there is no danger that the STVASTRO will show the wrong
time for longer.
2. Use KIWI 1.07 for monitoring GPS errors
You can run KIWI in parallel with the STVASTRO, where it can
monitor the output of the GPS 35, and write timing errors to a
log file. KIWI's filtering is now very good at hunting down
these errors. (This was the hardest task of our testing and
Geoff's software writing.)
You have to do some hardware modifications for this setup. Split
the data and 1PPS stream from the GPS 35, and send to the
STVASTRO and KIWI simultaneously. The 1PPS signal has to be
inverted. For more information on how to do this, read
wiring.txt. You'll find this text file at the zipped KIWI
package at the KIWI website.
3. Use KIWI 1.07 as a "Virtual GPS" (VGPS)
KIWI can now generate its own NMEA data from analysis of the GPS
information. The GPS 35 still provides the 1PPS signal, but the
NMEA data for the STVASTRO comes from KIWI which finds errors in
the "35" NMEA, and corrects if necessary. The GPS still beats
the drum of time, KIWI tells which time. Additional to the
hardware modifications above, you need to connect the NMEA data
from KIWI to the STVASTRO.
4. Buy a Garmin GPS 16
This 1PPS unit can be used with the STVASTRO too. Steve Preston
made a continuous test run of 6 (!) days!!! Using KIWI, he
detected no timing errors at all! The manufacturer of STVASTRO
now recommends the Garmin 16.
for the STVASTRO. So if you are considering buying the STVASTRO,
and have no GPS, get the "GPS-16".
LINKS:
KIWI timing software.
STVASTRO video time inserter.
Garmin 35 GPS receiver.
Garmin 16 GPS receiver.
If you want to have a closer look at our tests, read this
Fault Summary.
THANKS :-)
I want to thank...
Geoff Hitchcox (Christchurch, New Zealand)
for uncounted versions of testing software he provided just
hours after I have written a new request. Without him the
"timing errors" would still be a mystery. He has advanced KIWI
to a very useful timing tool even for users without a STVASTRO
video time inserter.
Steve Preston (Seattle, USA)
for running hundreds of hours of tests with his "35“ and with
the "16" and for contacting Garmin. Without his ideas and his
contact to Geoff, this project would never had started.
Stefan Messer (Hofheim am Taunus, Germany)
for helping me constructing the cables and wiring between
STVASTRO, Garmin 35 and KIWI. If I got lost in the wiring
diagram, he showed me the way back. He was the valuable "third"
hand, while I was soldering the connections.
Don Oliver (Houston Texas, USA)
who joined us with a version of the STVASTRO for the TRIMBLE GPS
unit. Without Don, "Virtual GPS" would still be a dream.
Oliver Kloes
Eppstein-Bremthal
Germany
2003 August 28