[time-nuts] A quick introduction to computer time synchronisation for new list members.
willism at westnet.com.au
Wed Aug 22 04:43:36 EDT 2012
To new and old time-nuts,
With the influx of new list members, I thought that this recent
presentation on time synchronisation and GPS might be useful.
It is part of an effort to get people interested in precision
timing and increase the pool of stratum 1 NTP servers in Australia.
(Australia has disproportionately small number of pool NTP servers).
The talk is intended as an broad-ranging introductory presentation,
and was given the Sydney Linux user group. The aim was to familiarise
the audience with some the hardware used with LinuxPPS project, which
had recently been merged into the mainline Linux kernel.
For the more experienced members this may be some useful in convincing
their friends and colleagues that precision time keeping is important.
Some of the more advanced topics covered include:
- GLONASS, GALILEO, COMPASS and the Japanese QZSS.
- GPS antenna choices.
- The future GPS/GNSS spectrum: L1, L2C, L5, L1C.
- GPSDO construction and applications.
- Indoor GPS, multiple antennas and splitters.
- Details of the Garmin and Sure Electronic receivers.
- Sawtooth errors with GPS receivers.
- Further information on leap seconds.
-------------- next part --------------
// This work is licensed under a Creative Commons
// Attribution-ShareAlike 3.0 Unported License
// Author: Mark Willis 2012.
// Filename: SLUG_2011Nov_GPS_timesync_notes.txt
This is a 30 minute presentation given at the:
Australian Sydney Linux Users Group (SLUG)
on the 25 November 2011.
GPS for computer time synchronisation.
The slide-set for the talk is:
This document is based on the script for the
talk, along with some background notes.
A large amount of material that was not presented
at the meeting, has been included in the
slide-set and the accompanying notes.
- A copy of the text in each side has included
between the tags [SLIDE:] ... [:SLIDE].
- Some slides were skipped in the presentation,
these have been included and a [SKIP] tag has
been added to these slides.
- Some paragraphs were skipped, they have been
included to expand on the areas covered.
These paragraphs have been enclosed in tags
[SKIP:] ... [:SKIP],
OR [SKIP; "line of text" ]
---- [SLIDE 1:][TITLE PAGE] ----
GNSS, GPS, GLONASS,
Galileo, Beidou/COMPASS, QZSS,
L1, L2C, L5, L1C,
SA Selective Availability,
DGPS, SBAS, WAAS, WAGE, EGNOS, MSAS,
Speed of light, nanosecond,
sawtooth, 0d-mode, GPSDO, GHz,
TIME TRANSFER SYSTEMS.
SETTING YOUR CLOCK BY GPS.
"The times they are are changing"
RS-232, PPS, USB, NMEA, EXTINT,
LinuxPPS, NTP, gpsd, PTP/IEEE1588, IRIG,
ACTS, WWV, LORAN, DCF77, CDMA, GSM NITZ,
Femto-Cell, GR-1244, IEC-61850, PCI-DSS,
SOX, GLBA, HIPAA, Securities Trading.
---- [:SLIDE 1] ----
---- [SLIDE 2:] ---
This work is licensed under a
Creative Commons Attribution?
ShareAlike 3.0 Unported License
Author: Mark Willis 2012.
Presented at: Sydney Linux User Group(SLUG)
---- [:SLIDE 2] ----
MD5: 7619ea9881e16da7121c2fded7f221ed SLUG_2011Nov_GPS_timesync_slides.pdf
---- [SLIDE 3:][TOPIC OUTLINE] ----
Timing Talks: Series Aims.
* To provide a quick background on
precision time synchronisation.
* To allow people to choose a level of
time synchronisation they require.
* From simply setting the time
periodically to high precision timing.
These talks are not intended to be a
how-to, these are available online.
---- [:SLIDE 3] ----
This is the first talk of series. The aim of
these talks is to give people an quick
introduction precision timing.
---- [SLIDE 4:][TOPIC OUTLINE] ----
Three parts of this talk series:
A. Time transfer systems and
hardware: choices, limitations
and installation. (Today's talk)
b. Linux kernel PPS support and
computer hardware. (next talk)
c. NTP software and hardware
monitoring (in the future)
---- [:SLIDE 4] ----
I have divided the TOPIC into three areas:
Hopefully each talk will be of interest to a
different group of people.
- Todays talk is a high level introduction to
GPS and time transfer
- The second talk will focus on computer
time keeping, specifically with Linux.
- The third talk is about timing software
choices and monitoring.
---- [SLIDE 5:][DISCLAIMER] ----
Disclaimer: Firstly I should say
that GPS and radio frequency
engineering, along with time
synchronisation and are highly
technical and specialised fields.
My interest is in precision time
synchronisation with a high degree
of assurance. Currently GPS is the
best method to achieve this goal.
---- [:SLIDE 5] ----
This is the standard disclaimer.
This is particularly important for time keeping
systems "as bad things could happen".
[SKIP; In the modern world many industries
depend on accurate time synchronisation.
Inaccurate time can result in frustration,
possibly financial losses or even dangerous
---- [SLIDE 6:][DISCLAIMER] ----
What this talk will NOT cover.
* Navigational GPS use is NOT the
focus of this talk.
* However much of the information
that I will cover about future of GPS
and systematic limits, is applicable.
* In particular precision navigation &
reception in challenging situations.
* In fact I have little knowledge about
mapping and user interfaces.
---- [:SLIDE 6] ----
The focus of today's talk is using GPS for time
transfer NOT consumer navigation.
However a lot of the technical information is
very much applicable to accurate and precise
---- [SLIDE 7:][TALK OUTLINE] ----
The aim of today's talk.
This talk is intended as an introduction
to NTP and GPS time transfer:
* To examine why you might need
precision time synchronisation.
* How in Australia, GPS is the only
real option for precision time transfer.
* Then to look at the current state and
the future of the GPS system.
---- [:SLIDE 7] ----
The aim of today's talk is to examine.
A. Why you might need Precision time
B. Show that in Australia, GPS is really
the only option (to achieve these needs).
---- [SLIDE 8:][TALK OUTLINE] ----
Contents of today's talk:
1. - Reasons for precision timing.
- NTP & synchronisation systems.
2. The future of GPS/GNSS.
3. Choosing a receiver & PPS output.
4. Additional hardware & antennas.
5. Connecting hardware together
& reception problems.
6. Chasing the nanoseconds.
---- [:SLIDE 8] ----
1. I will start off by covering common timing
2. Then look at the future of GPS and other
3. Next look at the options when choosing a GPS
receiver and why you NEED a Pulse Per Second.
4. I will quickly cover any additional hardware,
5. along with connecting all this new hardware
6. Then finish off by covering traceability to
UTC and deal with accuracy requirements.
---- [SLIDE 9:][SECTION OUTLINE] ----
Part 1: Introduction to time keeping.
1.1 Reasons for precision timing.
1.2 NTP "the default choice".
1.3 Reference Clocks and NTP.
1.4 Choice of time synchronisation
systems (GPS, Radio Clock).
1.5 Other indirect sources (CDMA,
NITZ, SONET/SDH, RDS etc.)
---- [:SLIDE 9] ----
The first part (of this talk) is about convincing
everyone that PRECISION TIMING IS IMPORTANT, (and
an external reference clock may be neccessary).
---- [SLIDE 10 - 11:][HUMOUR] ----
Why should I care?
"Accurate to nearest minute is
good enough for anybody."
Current date is Thu 01-03-1980
Enter new date (mm-dd-yy):
Enter new time:
---- [:SLIDE 10 - 11] ----
Time on Apple phones and mobile devices is only
guaranteed to be accurate to the nearest minute
Due to the vagaries of leap seconds, on Microsoft
platforms time is only guaranteed to within a
Some Android phones have a bug where they do not
apply leap seconds correctly, so the time is
wrong by 15 seconds.
[SKIP; This will become 16 seconds after the
next leap second insertion on 30 June 2012]
---- [SLIDE 12:] ----
Reasons for accurate and precise
* Legal/Regulatory compliance -
SOX, PCI-DSS, System log files.
* Transaction time stamping.
* Network performance monitoring..
. pool.ntp.org is a public service.
. eBay auctions.
. Because you can ;-)
---- [:SLIDE 12] ----
"So Why is sub-second time keeping important?"
Many modern regulatory frameworks require
accurate time as part of audit and accounting
Similarly for system administrators if your logs
are to be legally useful then they need to
contain accurate time-stamps.
[SKIP; For network performance monitoring such
as the network latency measurements, high
precision time keeping can be necessary
Then we get into the the vanity reasons...
Precision time keeping can be addictive.
---- [SLIDE 13:] ----
What level of Precision?
* With increasing work volumes -
each transaction occupies less time,
therefore higher precision is needed.
* Network performance: high speed &
multiple locations are commonplace.
* Computers are replacing specialised
electronics in time critical systems.
* Public expectations have increased.
---- [:SLIDE 13] ----
In the modern world expectations have increased.
---- [SLIDE 14:][HUMOUR] ----
[Cartoon: He won't be back for a few
nanoseconds -- care for a game of chess?]
---- [:SLIDE 14] ----
---- [SLIDE 15:] ----
Regulatory compliance. (IANAL)
* If it's the law: comply or don't exist.
* International regulation: achieve the
highest common denominator.
SOX, HIPAA, PCI-DSS, ETSI, OATS etc.
* To avoid negligence claims, you need
to demonstrate accepted practice.
* Proving time is expensive/impossible
to do after an event.
---- [:SLIDE 15] ----
To avoid litigation, I am not giving SLUG advice on
how to satisfy legal requirements.
BUT from an engineering perspective it is often
easiest to identify the most stringent applicable
requirements and implement systems that achieve
---- [SLIDE 16:] ----
List of Common Timing Requirements
3 secs (US OATS)
1 sec. (PCI/DSS)
[Trading: 1ms speed improvement
~= $120M per year
LSE <= 126us latency per transaction]
Timing Requirement: 200 ms
Timing Requirements: 10 us
Frequency Requirements: 5E-8
Timing Requirements: 0.864 us
Frequency Requirements: 1E-11
Timing Requirements: 1 us
Frequency Requirements: 1E-7
Timing Requirements: 5 us
Frequency Requirements: 1.5E-9
---- [:SLIDE 16] ----
Here is a list of some the common timing
As you can see there is a great variation in the
[SKIP; So when designing systems always check
what is REALLY needed. ]
---- [SLIDE 17:][SECTION OUTLINE] ----
RECAP: Part 1: General timing topics
1.2 NTP the default choice.
1.3 Reference Clocks and NTP.
1.4 Choice of time synchronisation
system (GPS, Radio Clock).
1.5 Other indirect sources(CDMA,
NITZ, SONET/SDH RDS etc.).
---- [:SLIDE 17] ----
OK enough of the reasons for precision timing.
I will now focus on the mechanics.
---- [SLIDE 18:] ----
NTP the "Hoover" of time sync.
* NTP is the default method of time
distribution over the Internet.
* Everything from national timing
labs to DSL routers use NTP.
* If all you want is accuracy of
1-100 milliseconds, NTP over the
Internet will do.
---- [:SLIDE 18] ----
NTP is everywhere from the clocks on our railway
stations to precision frequency transfer for
But with NTP over the INTERNET your accuracy
and precision are limited.
---- [SLIDE 19:] ----
NTP: Time Sync Choices.
1. Workstations: periodic SNTP
(ntpdate) queries (cron/if-up).
2. Servers: run ntpd: "more important"
and run 24/7 (ntpd can take up to a
couple of hours to settle down)
2b. Servers: run ntpd with an external
Reference Clock(PPS) - additional
hardware and time to settle down.
---- [:SLIDE 19] ----
There are two main approaches for using NTP:
1. The the simple and most common approach is
where your workstation periodically queries a
time server(s), and sets the local clock (then
free wheels until the next synchronisation).
2. The alternative more accurate approach is to
run NTP continuously and use Internet queries
to determine if the local clock is running
fast or slow and to allow ntpd to compensate.
2b. An additional step is to add a local external
reference clock (such as GPS).
The key advantages are:
A. GPS is far more accurate than time over
B. GPS receivers can be used once a second.
Rather than an Internet query a couple of
times and hour.
---- [SLIDE 20:] ----
Why an external Reference Clock?
* Higher precision/accuracy.
- Compensating asymmetric delays.
- High jitter Internet connections (3G).
* Higher assurance.
- Diversity/direct access.
- Potential NTP security problems.
- No Internet connection.
* Need precision frequency(GPSDO).
. The mountain is there to climb ;-)
---- [:SLIDE 20] ----
An external clock source is an out-of-band method
of time transfer (as opposed to in-band Internet
NTP), and this can have security/assurance
Additional reasons for using an external clock
Measuring asymmetric delays.
Coping with the extraordinary jitter on 3G
Internet connections. Latencies of around 0.1
and up to 2 seconds is not good for NTP.
The remaining reasons are about NOT depending on
somebody ELSE'S GPS receiver.
The majority of stratum one NTP servers are
synchronised by GPS. So by using a local GPS
receiver you are "cutting out the middle man".
---- [SLIDE 21-22:] ----
Alternatives to time over the Internet.
Internet NTP Query
Assurance ~ 3.5
Dial-Up Modem (ACTS)
Assurance ~ 4
HF Radio (WWVH Hawaii)
Assurance ~ 3
(WWVB, DCF77, LORAN)
Assurance ~ 3.75
(GLONASS, Galileo, COMPASS, SBAS)
Assurance ~ 4.5
< 10 ns
---- [:SLIDE 21-22] ----
So here is a list of all the possible, direct
[NEXT SLIDE] [NEXT SLIDE] [NEXT SLIDE]
[SKIP; ACTS costs a phone call per
synchronisation and can have a lower precision
that NTP over the Internet.]
Shortwave and Longwave are not available.
So in Australia GPS is ONLY high precision
---- [SLIDE 23:] ----
[SKIP] Indirect Synchronisation.
Another option is to use an indirect
method - usually to a GPS receiver.
* CDMA(10us), GSM NITZ(whole
seconds), ISDN, E1/T1, SONET/
SDH, TV signals, FM RDS(100ms)
(phone AGPS uses other methods)
* Before the "Internet everywhere"
world, these were a good solution to
the blinking 12:00 problem.
---- [:SLIDE 23] ----
It is worth mentioning indirect synchronisation,
these were great for synchronising the clock on
your video recorder.
---- [SLIDE 24:] ----
[SKIP] [Time sync is not a core function]
The biggest problem you are relying
on organisations (such a TV stations)
for which time synchronisation is
NOT a core a function.
(Also many of these schemes have a
lower precision than Internet NTP)
---- [:SLIDE 24] ----
There are questions regarding trust in these
systems and they often have a much lower level of
precision than can be achieved by simple Internet
---- [SLIDE 25:][SECTION OUTLINE] ----
Part 2: The Future of GPS/GNSS.
2.1 GPS time too good to be true.
2.2 GPS and the military.
2.3 Why should I trust GPS?
2.4 GPS constellation status.
2.5 Modernised GPS (L2C, L5, L1C).
2.6 Other constellations (GLONASS,
Galileo, COMPASS et al.)
---- [:SLIDE 25] ----
OK, that is long winded but necessary part done.
Here I will look at the "wonders" of satellite
navigation systems and GPS, and how it may
change in the near future.
---- [SLIDE 26:] ----
GPS/GNSS is The Best Option.
(even if there was a choice).
* GPS is the international de facto
standard for precision time transfer.
* GPS is no longer primarily a US
military system - GPS/GNSS is an
international critical civilian service.
* GPS synchronisation is now cheap.
($10-35 for a suitable GPS receiver)
---- [:SLIDE 26] ----
So GPS provides the highest precision and is now
---- [SLIDE 27:] ----
[SKIP] Why should I trust GPS?
* GPS is operated by the US military
& relies very accurate time signals.
* Differences between GPS and
UTC(USNO) are published.
* Precise time synchronisation has
been important to the navy after
Harrison's Marine Chronometer
solved the Longitude Problem.
---- [:SLIDE 27] ----
USNO = US Naval Observatory - which together
with US NIST are responsible for the
determination of UTC is the USA.
SEE: "Longitude" Dava Sobel 1995, 2007
The basis of GPS is precise time keeping and
radio signals traveling at the speed of light.
GPS is now so critical to the US military
functioning, that system outages are
Differences between GPS and UTC are available
on the website of the USNO and other standards
---- [SLIDE 28:] ----
The military gets the "cool" stuff?
* People can confuse errors from SA
with dual frequency improvements.
* L2 has 65% more ionic dispersion.
* L2 can be broken into with L1 info.
* RTK/Carrier phase (L1, codeless L2)
give the highest resolution (sub cm).
* Multi constellation beats dual
frequency, (n.b. urban canyons).
---- [:SLIDE 28] ----
[SKIP; There has long been a feeling that
military version of GPS is far superior to the
system that civilians get to use. ]
In reality high precision GPS is all about COST.
You can have most of the military features
If you are prepared to pay...
[SKIP; Land survey dual constellation and
carrier phase receivers and antennas can
easily COST more that a new car! ]
But like Moore's Law - multi constellation
receivers have recently become affordable.
---- [SLIDE 29:] ----
Stuff that ONLY the military gets.
Important differences between the
civilian and military GPS signals:
* Anti Spoofing: Out of band key
transfers needed/SAASM. (However
GPS simulators cost >$10k) *NOTE
* Anti Jamming: But due to the weak
GPS signal, a cheap wide band
jammer will obliterate BOTH signals.
---- [:SLIDE 29] ----
The most important advantage of the military only
system is cryptographic protection from spoofed
But GPS signals are so weak - a simple cheap
jammer off eBay, will overwhelm both (the
military and civilian) signals.
GPS signals are extremely weak:
30 Watts and 20 000 km away
A 1 watt "cigarette lighter jammer" can easily
overwhelm the satellite signals for up to 10km.
[*NOTE:] Simple "delayed replay spoofing"
(with a receiving antenna, a length of coax
cable, and a re-radiating antenna) is
However this attack is of most harm to
navigational GPS users, where the precision
requirements are higher (1us = 300m).
The latest generation of consumer GPS
receivers include systems to detect spoofed
signals (and implement countermeasures).
So most attackers don't bother spoofing GPS
signals they just use a wideband jammer that
covers all GPS and other satellite navigation
signals (GPS, GLONASS, Galileo & COMPASS).
---- [SLIDE 30:] ----
(civilian users are now important)
* No Selective Availability(SA).
* SBAS(WAAS, EGNOS, MSAS).
* 24+3 configuration (currently
32 satellites - original design: 24).
* L2C (1.2 GHz) civilian L2 signal.
* L5 (1.1 GHz) Safety Of Life signal.
* L1C (1.5 GHz) signal improvements.
---- [:SLIDE 30] ----
As a critical civilian service, it is no longer
acceptable for the military to provide a
deliberately inaccurate and restricted service.
Selective Availability(SA) has gone for ever, and
access to the formerly military only L2 frequency
has been introduced, along with the new L5
Permanent removal of Selective Availability was
a requirement for the US FAA to approve GPS
usage. Recently launched satellites do not
carry the hardware for SA.
The L5 frequency band is in the internationally
protected aviation band. This is also further
from the L1 frequency (than L2) which is better
for ionospheric corrections.
---- [SLIDE 31:] ----
Current GPS Constellation status (Nov 2011).
Projected satellite lifetime: 7.5 years
Oldest(still in use): 21 years
Average Age: 11 years
Cost per sat. inc. launch: ~$125 million
Constellation size: 32 (designed 24)
First Launch: 1978 (33 years ago)
Declared operational: 1993 (15 years ago)
Satellite Capabilities: [No. of sats]
Block IIA (L1 1.5 GHz) 11
Block IIR (L1 1.5 GHz) 12
Block IIR-M(L2C 1.2 GHz) 7
Block IIF(L2C & L5 1.1 GHz) 1 (+1 in testing)
Block III(L2C, L5, L1C 1.5 GHz) UNDER CONSTRUCTION
Next IIF launch scheduled for Sep. 2012
---- [:SLIDE 31] ----
What I am trying to show on this slide is that
GPS satellites are very expensive and stay in
service for much longer than projected.
So at the bottom of the table you can see that
only eight Sats transmit L2C satellites, and one
This is far less that the 18 needed for minimum
Currently there are 32 operational GPS
satellites in orbit, this is the limit of the
CDMA gold code.
A proportion of these satellites are
flown in pairs, which does not
significantly improve accuracy.
Over the past couple of years (finishing 2011)
there has been a project to re-organise the
constellation for the notional best use of the
27 satellites in orbit (24+3 configuration).
---- [SLIDE 32:][HUMOUR] ----
[SKIP] GPS Satellite Retirement.
"There's a chance it'll come back,
It did go active (many) years ago... maybe
they just want cut the springs, bolt a wing
on the back and add some racing stripes!"
[XKCD cartoon of Spirit Mars Rover.]
FROM: http://xkcd.com/695/ (CC-NC)
---- [:SLIDE 32] ----
After a satellite is launched, physical
maintenance such as refueling and replacement of
failed parts is very difficult.
GPS satellites are very different from the
geostationary communications satellites, such as
those that deliver satellite TV.
These communications satellites are often just a
"dumb bent pipe", where they simply repeat the
signal that is feed from the ground station.
The Medium Earth Orbits (MEO) that GPS occupy
ensure that the satellite are constantly moving
in relation to the earth.
So it is difficult for an individual GPS
satellite to be in constant communication with
the ground stations. Each satellite is specified
to operate for at least a week without contact.
This additional complexity and the choice of
orbit can limit a satellite's lifespan. The
atomic clocks (up to four redundant systems in
each satellite) have higher failure rates than
many electronic components. In addition there
the is failure of moving parts or when the fuel
supply runs outs.
When there are a sufficient number of failures on
a satellite and it is unable to continue
operation, it is permanently withdrawn from
---- [SLIDE 33:] ----
GPS is not the only show in town.
Other satellite navigation systems:
* GLONASS (.RU)(operational)
* QZSS(.JP)(1/3 satellites launched)
. SBAS (WAAS, EGNOS, MSAS) used
directly rather just than for DGPS.
---- [:SLIDE 33] ----
"Here we move into the geopolitical territory."
Next year (2012) there will be big changes in
satellite navigation with the introduction of
The Russian GLONASS fell into disrepair after the
break up of the Soviet Union. Recently it has
been fully restored.
The first two Galileo validation satellites have
been launched and the Chinese intend to go ahead
The Japanese Quasi Zenith is a regional system
but it will also cover Australia.
So in a few years time there will be over a
hundred satellites in orbit.
---- [SLIDE 34:] ----
[SKIP] GLONASS(.RU) as Plan B.
* After cold war, fell into disrepair.
* Uses old style FDMA (1 CDMA sat.)
* Above many L1 GPS antennas bandwidth.
* Has RUBBER LEAP SECONDS problem.
* Few/poorly documented receivers.
* Timing receivers not yet available.
* Mainly useful for urban canyons.
* Still susceptible to GPS jamming.
---- [:SLIDE 34] ----
During the Cold War the Soviet Union developed
the GLONASS satellite navigation system in
parallel to GPS. Whilst in many ways it is
similar to GPS, there are implementation details
that result in the receivers being incompatible.
Until recently only very expensive land survey
receivers could use this alternative system.
The main advantage of using both GLONASS and GPS
is that it effectively "doubles" of the number of
This increases the probability that there are
sufficient satellites visible in urban canyons
such as in high rise central city areas.
These additional satellites are particularly
important for high resolution users such as Real
Time Kinematic (RTK) surveying, which requires
six satellites to be visible simultaneously.
---- [SLIDE 35:] ----
[SKIP] Galileo and COMPASS.
* Galileo (.EU) in development >15 years.
(before Selective Availability ended)
- In Orbit Validation sats launched.
* COMPASS (.CN) is newer project.
- First sats are in geosync orbits.
(some coverage of Australia)
* Both will transmit on L1C & L5
* First receivers appearing.
---- [:SLIDE 35] ----
Galileo and COMPASS are systems that may become
useful in the next few years.
---- [SLIDE 36:] ----
[SKIP]Japanese QZSS and Australia.
[PICTURE: QZSS orbit: figure of eight,
traveling around Japan and Australia]
* Transmits on GPS L1, L1C, L2C, L5.
* 3 Satellites in Tundra orbits (1 launched)
---- [:SLIDE 36] ----
Unlike the other systems the Japanese Quasi
Zenith Satellite System is a regional system.
It is designed to solve the problems caused by
Japan being relatively far from the equator and
that it has GPS difficulties in urban canyons.
The satellites are in Tundra - Highly Elliptical
geosynchronous Orbits(HEO). With these orbits,
it is possible to provide constant coverage
DIRECT OVERHEAD for a region, using only three
This orbital pattern is also used by the US
Sirius satellite radio system. The are also
orbits are similar to the old Russian Molniyna
orbit (but with twice the orbital period).
---- [SLIDE 37:] ----
[SKIP] Japanese MSAS SBAS.
* Australia: only visible DGPS sats
* 2 Sats in geostat. orbit over PNG.
* PRN129 & PRN134
* Transmits on L1, L5.
* Australian Gov. did not participate so
IONIC CORRECTIONS ARE NOT VALID.
- (even though there is a Japanese
MSAS monitoring site in Australia!)
---- [:SLIDE 37] ----
Satellite Bases Augmentation Systems (SBAS) are
the modern equivalent of terrestrial Differential
They were in part designed to overcome Selective
They are now used to transmit ionic corrections,
satellite clock errors and orbit deviations.
The Japanese MSAS are the only publicly available
SBAS satellites that can be easily received in
However the ionospheric corrections are not valid
for Australia, and clock errors and orbit
deviations are typically very small.
US/Canada: WAAS (US Coast Guard)
US Military: WAGE
Commercial: Starfire (John Deere)
---- [SLIDE 38:] ----
The problem with both multi-frequency GPS
(L2C, L5, L1C) and alternative constellations
(GLONASS, Galileo, COMPASS) - there no
affordable timing receivers that implement
***Currently GPS L1 is the only choice.***
(U-Blox - promises Galileo firmware upgrade)
(Galileo/COMPASS sats are yet to be launched)
---- [:SLIDE 38] ----
Receivers that support the alternative
constellations AND are affordable have have just
come onto the market, and many only support one
of the new systems.
So these receivers are
A. Not mature and
B. Poorly documented.
For (accurate) time synchronisation this is NOT
So at the moment the traditional GPS L1 is the
---- [SLIDE 39:][SECTION OUTLINE] ----
Part 3: Choice of GPS receiver
Holden vs Ford.
3.1 How many sats/channels?
3.2 Consumer receivers don't really
care about time.
3.3 You need a PPS.
3.4 Timing receivers are nice.
3.5 GPSDOs are better.
---- [:SLIDE 39] ----
Here I will cover the technicalities of choosing
a GPS receiver.
---- [SLIDE 40:] ----
How many sats/channels to I need?
By recording the time of arrival from
4 satellites it is possible to solve for
4 unknowns - 3 dimensions of space
and 1 for timing error.
* Additional satellites increase
accuracy, configuration matters.
* 24 sats design (12 visible).
* 99 Channel receivers available.
---- [:SLIDE 40] ----
You need 4 satellites for a 3D fix,
There are currently 32 satellites in the
constellation, only half of which can be seen at
one time (the rest are on one side of the earth).
** So why do we need 99 channel receivers??? **
In a couple years time with GLONASS, Galileo and
COMPASS this would make more sense, but most
current receivers can NOT track the new
[SKIP; GPS receivers seem to get status anxiety
with the number of satellites they can track...
---- [SLIDE 41:] ----
Non-military GPS Users.
1. Consumer Navigation - a NICE
INTERFACE is most important.
2. Professional Navigation - HIGH
ASSURANCE systems are essential.
3. Land Survey - HIGH PRECISION with
capital costs and post processing.
4. Timing Receivers - real time,
used in CRITICAL INFRASTRUCTURE.
---- [:SLIDE 41] ----
Each tribe of GPS users have very different
1. For consumer navigation - it is the user
interface that sells.
2. In aviation each receiver must undergo
extensive certification before it can be
3. Land surveyors need the highest precision so
carrier phase tracking, dual frequency and
multi constellation receivers are used
(and they COST).
4. Timing Receivers need to be as bug free as
possible (and provide the required precision).
---- [SLIDE 42:] ----
[SKIP] All animals are not created equal.
* Even though the GPS relies on
high precision timing, you need a
receiver that allows you to access
* The clock display on most
"navigation receivers" is a secondary
feature so manufacturers take little care
beyond avoiding gross errors.
---- [:SLIDE 42] ----
Mobile phones and car navigation systems, seldom
have accurate or precise time so are not useful
for high precision time synchronisation.
---- [SLIDE 43:] ----
You NEED a PPS signal.
* The most common method GPS
receivers use to convey precise time
is a Pulse Per Second(PPS) signal
over a RS232 serial line.
* Most consumer GPS receivers don't
export the PPS line and/or use USB.
(NMEA and USB messages
have a high jitter)
---- [:SLIDE 43] ----
GPS is based upon precise time, but you need to
get that time into your computer.
For this you *NEED* a Pulse Per Second (PPS)
---- [SLIDE 44:] ----
A few consumer receivers:
a. Garmin(integrated antenna) 18LVC,
18xLVC, 16LVC, 16HVS/17HVS.
b. Sure Electronics GPS-10 (US$34)
. Accuracy of 1 us is often quoted:
an old and very conservative spec.
more pessimistic than errors caused
by SA (1us = 300m accuracy).
. 18LVC has a 130 ns sawtooth.
---- [:SLIDE 44] ----
The vast majority of consumer GPS receivers are
not suitable for precision time transfer.
A few that are suitable include these Garmin
models and ONLY these particular models:
(16LVC, 16HVS, 17HVS, 18LVC, 18xLVC/18LVCx)
A new arrival is the Sure Electronics GPS kit,
available cheaply on eBay, delivery direct from
[SKIP; There are other OEM GPS boards but these
require additional support electronics (MAX232)
which scares most users.
---- [SLIDE 45:] ----
Garmin 18LVC PPS delta delta Histogram
[GRAPH: a histogram of the changes in the
difference between successive PPS signals in
(used with permission)
---- [:SLIDE 45] ----
A B C D
1.000 1.006 0.994
\ / \ /
0.006 -0.012 DELTA DELTA
These receivers perform far better than stated on
their spec. sheets (1 microsecond).
But as you can see here, systematic sawtooth
errors limit the precision achieved.
---- [SLIDE 46:][HUMOUR] ----
"Can someone tell me what this is?"
[PICTURE: 30cm length of wire.]
Answer: 1 nano second.
---- [:SLIDE 46] ----
Grace Hopper's demonstration
A point to remember when ever discussing GPS
1 nanosecond ~= 30cm(1 foot)
1 microsecond ~= 300m
(speed of light in a vacuum)
But the speed of light in a wire is less than
in a vacuum.
Thick Coax .77c
Thin Coax .65c
Twisted Pair .59c
c = the speed of light in a vacuum.
This speed difference is important when
compensating for antenna delays.
---- [SLIDE 47:] ----
Timing receivers advantages:
"a better class of receiver"
* Time generation is a core function.
* Position hold (zero-D) mode,
1 satellite required rather than 4.
* Sawtooth correction.
* Higher precision outputs. (<10 ns)
* Resistance to interference (cell.
base stations have lots of RFI).
---- [:SLIDE 47] ----
Rather than re-purpose a consumer navigation
receiver another option is to use a receiver
specifically designed for time transfer.
These are primarily designed for mobile phone
The have the option of using a position hold
mode, which greatly increase the precision of
the GPS time calculations.
Equally as important is Sawtooth Correction which
allows the removal of the systematic errors
shown in the Garmin histogram.
---- [SLIDE 48:] ----
[SKIP] Histogram U-Blox LEA-6T
[Histogram showing the noise on the
sawtooth corrected PPS signal from a U-Blox
GPS receiver. <6 nanoseconds 1 sigma]
---- [:SLIDE 48] ----
[NOTE; The horizontal scale on this graph is about
16x smaller than the on Garmin 18LVC]
Unlike the Garmin there is not the peak spreading
due a sawtooth errors.
---- [SLIDE 49:] ----
[SKIP] Timing receiver disadvantages:
* Higher cost (new-old stock $15-50)
* May need to purchase an antenna.
* Often don't get firmware updates.
(Get it right first time, outages cost)
* Usually TTL serial - need to use a
MAX232 level convert circuit. (~$4)
[PICTURE: MAX232 TTL to RS-232 D9 from eBay]
---- [:SLIDE 49] ----
Timing receivers are aimed at a technical market
so are often not the turn key solutions systems
of consumer market.
However these downsides are not particularly high
to someone who has a basic electronics knowledge.
---- [SLIDE 50:] ----
Timing Receiver Choices
Limited number of brands/models:
* Motorola: m12-t, m12+-t, m12m-t
(now iLotus & Symmetricom)
* Rockwell-Collins: JupiterT
(now Conexant -> SiRF -> NavMan)
* Trimble (receivers and GPSDOs)
* U-Blox: LEA-4T, LEA-5T, LEA-6T
---- [:SLIDE 50] ----
Each of these receivers has a long history,
and the first three are readily available on the
eBay for about 20 - 50 dollars.
[SKIP; and some companies have long history of
"financial musical chairs". ;-)
Other lesser known brands and models of timing
GPS receivers include:
* Symmetricom(Furuno) 58534A "Smart Antenna",
* Navsync/Connor-Winfield - CW12-TIM, CW25-TIM,
* Novatel/CMC - SuperstarII, OEMV-2.
---- [SLIDE 51:] ----
GPSDOs do it better.
* GPS signals have short term noise
& medium term wander (diurnal).
- Short term noise can be averaged
out by an ovenised oscillator (OCXO)
after the GPS receiver. Know as a
GPS Disciplined Oscillator(GPSDO)
. SBAS corrections may help with
diurnal and solar wander.
---- [:SLIDE 51] ----
For navigation you can average a static location
over time to get a more accurate result.
But "time does not stand still", so a GPSDO uses
a high spec oscillator to average each result.
This arrangement also gives you a high precision
---- [SLIDE 52:] ----
"A tale of two Oscillators"
[PICTURE: of a Motorola M12 GPS receiver and
a Trimble ThunderBolt GPSDO.]
---- [:SLIDE 52] ----
* On the left I have circled an un-temperature
compensated surface mount oscillator on a M12
* On the right is a disciplined ovenised
oscillator(OCXO) on a ThunderBolt GPSDO.
The picture is to illustrate the size and
quality differences between the two
Some plain GPS receivers can be used as a
frequency source. However the conversion
between the receiver's internal frequency and
the chosen output frequency has significant
short term jitter.
---- [SLIDE 53:] ----
GPSDOs without GPS!
* GPSDOs allow the PPS to continue when
the GPS (signal) is interrupted.
* Without GPS, an OCXO will slowly
drift, but information gathered with
GPS allows better hold over.
* OCXO can run for up to a day before
it is detectable by NTP.
---- [:SLIDE 53] ----
GPSDOs also help deal with signal outages.
During hold over/outages GPSDOs are still a
precision time source, so there is no need to
fall back on the...
"Crummy Crystal in your PC".
---- [SLIDE 54:] ----
[SKIP] Discipline yourself?
GPSDOs contains three main parts
1. GPS receiver.
2. OCXO (crystal oscillator)
3. Controller to glue these together
* First two are readily available, the
disciplining circuit can be a home
brew PIC or AVR microprocessor.
---- [:SLIDE 54] ----
It is worth mentioning that a GPSDO is
conceptually a relatively simple system,
consisting of just 3 portions.
Ham radio operators that have published various
designs that are readily available.
---- [SLIDE 55:] ----
[SKIP] GPSDOs: The China effect.
* The major use for GPSDOs are
mobile phone base stations (E911).
* Decommissioned base stations
are sent to China as scrap.
* Tested Trimble Thunderbolts kits
are now available for US$200.
---- [:SLIDE 55] ----
There are various brands and models for GPSDO.
The Trimble ThunderBolt GPSDO is readily
available at a reasonable price.
---- [SLIDE 56:] ----
[SKIP] Summary: GPSDO Advantages
* Lower noise & hold over.
* GPSDOs output frequency -
replace motherboard oscillators,
and supply other timing gear.
* GPSDOs can be turnkey solutions,
fewer interface needs - no
MAX232 and power supply.
---- [:SLIDE 56] ----
If you need precision frequency source,
then you need a GPSDO!
---- [SLIDE 57:] ----
[SKIP] Summary: GPSDO Disadvantages
* More expensive: ~10x the cost.
* 10MHz output needs be converted
to be useful e.g. 14.318 MHz.
* Thunderbolts need a pulse stretcher
for PC use. (TAPR FatPPS US$60)
- In the end GPS/NTP will always win
over OCXO wander. (NTP is a
cheap low resolution GPSDO**)
---- [:SLIDE 57] ----
The disadvantages of a GPSDO are centred on cost.
and whether a simple GPS receiver is sufficient.
[**NOTE:] NTP does not actually disciple the
oscillator on a computer. NTP only adjusts the
the system clock (which is driven by the "free
As stated before, most stratum one servers
use GPS. So as such, when you use ntpd the
frequency of your system clock is being
indirectly disciplined by GPS.
---- [SLIDE 58:][HUMOUR] ----
"If you are not accurate to the
nearest nano/micro second,
you are not really trying!" :-)
---- [:SLIDE 58] ----
---- [SLIDE 59:][SECTION OUTLINE] ----
Part 4: More Hardware !!!
4.1 One receiver is a single point of
failure (firmware bugs !#$!%).
4.2 Receivers are cheap, just throw
hardware at it (electronic failure).
4.3 Good antennas & positions.
4.4 Multiple antennas.
4.5 Splitters GPS and PPS.
4.6 GPS outages (solar, jamming)
---- [:SLIDE 59] ----
OK, you have decided which of GPS receiver
But is one receiver enough? (!!!)
---- [SLIDE 60:] ----
Embedded software Sox.
"Untested code is broken code."
There might be updates but this
like "polishing poop".
(SEE: Garmin 18x updates)
---- [:SLIDE 60] ----
***Please excuse the language on this slide.***
What I am trying to convey here is the fear,
of waking up one morning and discovering your
receivers have turned into paper weights.
Garmin 18x updates:
2009Feb Version 3.0: Randomly "bricked".
2009Aug Version 3.2: Usable.
Version 3.3: Random NMEA delays.
Version 3.5: Random NMEA delays.
Version 3.6: Random NMEA delays.
2011Jun Version 3.7: Usable.
---- [SLIDE 61:] ----
Black Box GPS
* GPS receivers are like mobile
phones, even if the interface is
Linux based, the interesting stuff is
done is closed source firmware.
* In designing a GPS receiver,
assumptions need to be made.
* We can only guess at firmware
---- [:SLIDE 61] ----
Most GPS firmware is written in the cheapest
And it is difficult for an end user to test what
will happen at a particular date or situation.
So every day has the magic of Y2K...
---- [SLIDE 62:] ----
"A person with one clock,
knows the time precisely.
A person with two clocks,
is never sure what the time is!"
* One GPS receiver is a
Single Point Of Failure (SPOF).
* With two receivers you can only ever
have a disagreement.
---- [:SLIDE 62] ----
If you are happy with your one GPS receiver,
check it constantly against at least two/three
Internet NTP servers.
---- [SLIDE 63:] ----
Have a back up plan or three!
* NTP can find "false tickers", so
"just throw (GPS) hardware at it".
* The magic numbers are 1,3/4,5,7...
* All we need are the majority of
the receivers to be correct.
- A usable PPS may continue even
when the serial string is incorrect.
---- [:SLIDE 63] ----
But if you want to keep the high precision, you
need at least 3 GPS receivers (from different
NTP is "Byzantine" and can detect failures,
AND after all hardware is cheap.
---- [SLIDE 64:] ----
[SKIP] Which Antenna?
* With a separate antenna the receiver
is inside out of the weather and heat.
* Timing grade antennas have good
RFI filtering / narrow bandwidth.
* VIC-100 is a good choice (new-old
stock sells for US$40) (L1 C/A only)
. Use 75 Ohm sat. TV RG6, even though
you "should" use 50 Ohm cable.
---- [:SLIDE 64] ----
Rather than using a cheap $5 GPS "mouse" antenna
- Questionable quality electronic amplifier.
- That may not be fully weather/waterproof.
- May not handle Australian summer sun,
(black bodies can easily reach 100 C).
- Small/No antenna ground plane.
A better choice could be a $40 antenna intended
for a mobile phone base stations.
The Aromat/Matsushita/Panasonic/NAiS VIC-100
is a high gain (38dB) patch antenna with very
strong filtering (60dB).
These antennas are fully weather proof and are
shaped (sharpened) to shed the dirt/snow and
discourage birds from resting on top.
---- [SLIDE 65:][HUMOUR] ----
[SKIP] GPS Radomes:
"The peak is to not a lighting rod,
it's to stop avian carriers
from dropping their packets"
[Prussian Army helmet + Pigeon =? GPS Radome]
---- [:SLIDE 65] ----
Image from: http://wpclipart.com/
---- [SLIDE 66:] ----
[SKIP]Future Proofing Your Antenna.
* GPS L1, L2 and L5 are different
frequencies so require multiple
antennas or a different design.
* Galileo & COMPASS L1(& L5) straddle
GPS so existing antennas may work.
* GLONASS L1 is just above the
bandwidth of most GPS antennas.
* SBAS/DGPS (WAAS, EGNOS, MSAS)
are on GPS L1.
---- [:SLIDE 66] ----
In the future it will be possible to purchase GPS
receivers that support all these new
constellations and frequencies.
But if your receiver uses an external antenna
then you may need to purchase a new antenna.
Currently the best GPS antennas have strong
filtering to reduce the risk of interference.
(such as the case of LightSquared transmissions
in neighboring bands).
So the best option is to buy a good antenna now
(for your GPS receivers). Then in the future buy
a new antenna when you have a multi-constellation
---- [SLIDE 67:] ----
[SKIP] First we had L1 C/A
[DIAGRAM: showing the GPS L1 C/A spectrum.]
---- [:SLIDE 67] ----
Most consumer GPS receivers only use a very
narrow 2.046 MHz slice of the radio spectrum.
---- [SLIDE 68:] ----
[SKIP] Now L1 is busy neighborhood.
[DIAGRAM: Spectrum: GPS L1 C/A, L1 P(Y),
L1M, L1C, Galileo L1, COMPASS B1.]
---- [:SLIDE 68] ----
With GLONASS and other constellations this
increases by over 25 times to over 56 MHz.
---- [SLIDE 69:] ----
[SKIP] It's not just L1 any more.
[DIAGRAM 1. Spectrum: GPS L1, L1C
and GLONASS L1]
[DIAGRAM 2. Spectrum: L5: GPS, Galileo,
COMPASS and L2: GPS, GLONASS ]
---- [:SLIDE 69] ----
---- [SLIDE 70:] ----
GPS signals require line of sight.
* For portable applications signal
outages are normally not a problem.
- When the vehicle/person moves the
interference may go away.
* Static locations may need height to
see over trees and local obstacles.
* In Australia northern sky visibility
(towards the equator) is more important.
---- [:SLIDE 70] ----
In navigation mode, when the GPS receiver moves
any interference may disappear.
However -- in static locations, interference can
be permanent, and even for temporary outages NTP
can take along time to recover.
---- [SLIDE 71:] ----
[SKIP] GPS works in buildings!
* Modern GPS receivers are very
sensitive - often work indoors.
* Depends on your roofing material.
* Usually track fewer satellites.
* May need the almanac via AGPS.
* North facing windows may provide
reasonable reception (in Australia).
. Lightning protection ;-)
---- [:SLIDE 71] ----
Indoor GPS reception is not advisable for
important applications, but if you have no other
option then a north facing window may work.
How modern energy efficient windows may have a
metal film that makes GPS reception difficult.
---- [SLIDE 72:] ----
[SKIP] Multiple GPS Antennas.
* GPS signals are weak and amplifier
leakage may interfere with other
antennas and receivers.
* Best practice is for antennas 10m
apart, (inverse square law applies).
* Multiple antennas give diversity
from lightning strikes, but keep
receivers electrically isolated.
---- [:SLIDE 72] ----
If you need to use two GPS receivers in your car,
try to place the antennas at least 1m apart.
This is particularly important (but difficult)
for receivers with integrated antennas.
---- [SLIDE 73:] ----
[SKIP] One antenna Multiple receivers.
* Receivers with external antennas
can use a antenna splitter.
* Passive satellite TV splitters will do:
divides the signal strength, receivers
may interfere with each other.
* Active (amplified) splitters tend to be
single frequency(L1 only), but often
have strong filtering and isolation.
---- [:SLIDE 73] ----
Antenna splitters are useful to allow two (or
more) receivers to share the same sky view and
Passive satellite TV splitters are cheap but are
un-amplified and without port to port isolation.
Amplifier and oscillator leakage can be a problem
with these splitters.
Active GPS splitter often cost more per port than
using multiple antennas!
---- [SLIDE 74:] ----
[SKIP] One antenna Multiple receivers.
[PICTURE: A passive satellite TV splitter
compared to a HP/Agilent/Symetricon GPS L1
amplified Smart Splitter.]
---- [:SLIDE 74] ----
---- [SLIDE 75:] ----
GPS signal outages.
* GPS is now so important it has
now become single point of failure.
* GPS outages have brought down
mobile phone & telco. networks.
- Geopolitical GPS Jamming
- Solar weather("sun spots")
---- [:SLIDE 75] ----
GPS is now an essential service, most mobile
phone and radio transmitters use GPS
Recent jamming in by North Korea caused the
mobile phone basestation outages in South Korea.
The effect of the impending peak of the solar
cycle, on GPS and other satellite systems is a
current area of research.
---- [SLIDE 76:][SECTION OUTLINE] ----
Part 5: Connecting your GPS
5.1 Level Converters
5.2 Power supply
5.3 Electronics requirements.
5.4 RF connectors.
5.5 Garmin Interface Connections.
5.6 Sure Electronics GPS.
5.7 PPS and Frequency Distribution.
---- [:SLIDE 76] ----
So now I will cover hooking up your new GPS
receiver and some modifications that may be
For the viewers at home: "Warming up your
soldering iron might be a good idea."
---- [SLIDE 77:] ----
Receiver Interface Electronics
* !!! Do NOT connect TTL to RS-232. !!!
* A TTL PPS signals may need a level shifter,
but most computers are OK.
* Power supply: often steal power
from the host computer (USB/12V).
* 3 volt boards may need 5 volt
supply for the antenna.
* Antenna/RF/Interface connectors.
---- [:SLIDE 77] ----
Most GPS receivers with a Pulse Per Second use
But remember that the transmit voltage on your
computers RS-232 serial port (24 volt swing) can
easily be high enough to fry these receivers.
However, when receiving signals on your PC, most
modern serial ports still trigger at 3 volts.
So a level converter may not be needed for the
PPS line (NTPd can handle both polarity signals).
"RF connectors ...
there are are so many to choose from"
[SKIP; If you can't work out which connector
you have take it down to JayCar (electronics
store) and ask to try out the connectors until
you find the right one.
---- [SLIDE 78:] ----
[SKIP] Garmin Interface 18LVC/16HVS
* Available for <= $100.
* MediaTek based chipset.
* Combined antenna and receiver.
* Serial/PPS output - no GHz coax
* Need to attach serial D9 plug
* 18LVC can be powered from USB.
* 18LVCX UPGRADE TO LATEST FIRMWARE.
---- [:SLIDE 78] ----
---- [SLIDE 79:] ----
[SKIP] Sure Electronics GPS kit.
* New on the market.
* SkyLab chipset (MediaTek based).
* Soldering of questionable quality.
* US$34 approx. delivered.
* Uses USB for power.
* A small modification to get PPS.
* Serial, USB & Bluetooth interfaces.
* Includes separate antenna.
---- [:SLIDE 79] ----
---- [SLIDE 80:] ----
PPS Distribution Amplifiers.
* Allows one GPS to be used by
* Alternative to NTP everywhere.
- Get serial data via network.
- Combined/Separate Serial MUX.
* Two way splitting may work with
simple electrical T junction.
---- [: SLIDE 80] ----
A distribution amplifier allows a number
computers to share a single GPS.
This is far more precise than NTP over
[NEXT TALK: will be all about using PPS signals]
---- [SLIDE 81:][SECTION OUTLINE] ----
Part 6: Chasing the nanoseconds
how far can and should you go?
6.1 Computers fumble nanoseconds.
6.2 Sawtooth errors.
6.3 Diurnal variations.
6.4 UTC traceability.
[END OF PRESENTATION]
6.5 !#@$ leap seconds.
---- [:SLIDE 81] ----
OK, to finish the talk off.
Here is both a reality check
and a look at how far
you can soup up your time sync.
---- [SLIDE 82:] ----
Precision Achievable by GPS.
* The best general purpose
computers can achieve 100 ns.
* Most modern GPS receivers are at
this level or lower 10-100 ns.
. Some network capture devices
can use the higher precision.
. Higher precision can be useful for
GPSDO OCXO characterisation.
---- [:SLIDE 82] ----
*IF* the time is accurate on your GPS receiver,
Then the PPS on a cheap consumer model is
unlikely to limit NTP's precision.
---- [ SLIDE 83:] ----
[SKIP] Sawtooth is not "real" noise.
* Most receivers use the closest zero
crossing of an internal free running
oscillator to produce their PPS.
* As the phase of the oscillator moves over
time then the distance to the UTC second
* Timing receivers include this difference
in their serial output.
---- [:SLIDE 83] ----
One of the major differences between GPS
receivers designed for time transfer and
navigational receivers, is the ability to
correct the sawtooth error on the PPS output.
---- [ SLIDE 84:] ----
[SKIP] Sawtooth is not "real" noise.
[DIAGRAM 1. Shows a small sawtooth error.]
[DIAGRAM 2. Shows a large sawtooth error.]
The phase/error changes with time.
---- [ :SLIDE 84] ----
The two diagrams show how the PPS signal is
produced by most GPS receivers.
The PPS signal is generated:
1. After the UTC second(dark blue vertical line),
2. When the internal square wave oscillator
(light green line) rises and hits zero volts.
3. The purple area at the bottom is the resulting
PPS signal (which is asserted for a 100-100000
The left hand diagram shows a situation where
there would be a small sawtooth error.
The right hand diagram is produced if the
internal oscillator runs slightly fast.
There is almost a full cycle wrap around with
the PPS signal is significantly delayed.
---- [ SLIDE 85:] ----
[SKIP] Sawtooth is not "real" noise".
[ GRAPH 1. Motorola M12+t short term
performance showing sawtooth errors]
[GRAPH 2. Showing the timing performance of:
1. M12+t: WITHOUT sawtooth correction.
2. M12+t: WITH sawtooth correction.
3. HP58503B GPSDO (Motorola VP8)
(used with permission)
The first graph shows the short term
sawtooth structure over the period of two
The second graph shows the peak to peak
variations over 27 hours.
---- [:SLIDE 85] ----
The key observation from these graphs is that:
A. Sawtooth errors are highly systematic and
are about 10x normal noise variations.
(Pink Vs Yellow series on second graph)
B. When sawtooth corrected GPS still has
significant short term noise.
(Yellow Vs Light Blue series on second graph)
C. Even with averaging the signal the GPSDO shows
diurnal wander. (Light Blue line "wiggles")
---- [SLIDE 86:] ----
GPS Noise Summary.
2-20 ns Short term GPS noise: can
NOT be corrected, only averaged.
10-200 ns Sawtooth errors: CAN be corrected
or offsets used in calculations.
10-100 ns Environmental wander. DGPS may
correct common mode errors.
1-5 us Computer PPS timestamp jitter.
0.1-5 ms Network & USB timestamp jitter.
1-10 ms Computer temperature effects
>100 ms Computer power saving.
---- [:SLIDE 86] ----
So here is summary of the error sources.
As you can see: all the GPS sources are way below
the computer measurement errors.
[SKIP; These computer time variations will be
the subject of my next talk. ]
---- [SLIDE 87:] ----
Traceability to UTC.
* UTC is the agreed international legal
time scale (love it or hate it).
* UTC is derived from EAL/TAI - a
collaboration of national timing labs
and is calculated retrospectively.
* UTC differs from TAI by an integer
number of leap seconds.
* Local time is an offset from UTC.
---- [:SLIDE 87] ----
If you need to prove the specific time an event
occurred, then your local time needs to track the
internationally legally agreed time scale: UTC.
But there is no "real time" version of UTC, only
versions for each contributing country.
In Australia we need to demonstrate traceability
to UTC(AUS) which is maintained by the National
Measurements Institute (of Australia).
---- [SLIDE 88:] ----
Leap seconds and GPS.
* GPS has no leap seconds so differs
from UTC by an integer number.
* This offset is given in the GPS
almanac, which can take up to
12.5 minutes to receive.
* Until the receiver has the almanac it
may give the wrong time.
* IF the offset is ever/correctly applied.
---- [:SLIDE 88] ----
The GPS timescale is based on UTC but without
leap seconds, so as such it tracks TAI but is
offset by a constant 19 seconds.
In order to give the time relative to UTC, a GPS
receiver must know the current leap second
This can take up to twelve and a half minutes
to receive, assuming you have reasonable GPS
[SKIP; This is the reason that the time on some
Android phones is wrong by 15 seconds - the
current offset between GPS and UTC.
In order to speed up this process, designers
often take short cuts such as initially using the
UTC offset at the time that firmware was built.
There are many examples GPS receivers with
firmware bugs not applying leap seconds
correctly: either early (up-to six months)
or not at all.
---- [SLIDE 89:] ----
Going from GPS to UTC.
* Multiple levels of indirection,
GPS + leap secs -> UTC(USNO) -> UTC
* Differences between UTC(UNSO)
and UTC(BIPM) are typically only
at the nanoseconds level.
* Other constellations are based on
their own UTC standard.
e.g. GLONASS -> UTC(SU)
(with rubber seconds)
---- [:SLIDE 89] ----
Aside from the integer differences due to
leap seconds. GPS is not directly UTC.
GPS time is controlled to converge with the
USNO version UTC(USNO).
UTC(BIPM) AKA "UTC" is produced retrospectively
at the end of each month.
However these differences are usually less that
can be measured by a computer.
---- [SLIDE 90:] ----
---- [:SLIDE 90] ----
---- [SLIDE 91:][DISCUSSION] ----
All I want for Christmas
are NO more leap seconds.
* Proposals to stop adding them have
been around for over 15 years.
* UN standards vote is in Jan 2012.
* Astronomers are fighting strongly.
* Like other global geopolitical issues
it will be agreed upon ... eventually.
---- [:SLIDE 91] ----
[SPOILER WARNING; Some of this will be covered in
the next talk.]
Why leap seconds are evil could be the focus of a
On 19 January 2012 there was a meeting to decide
on stopping the insertion of leap seconds. The
UN - International Telecommunications Union (ITU)
made an "important decision" (!) to defer the
vote until 2015.
Both sides agree that there are long term flaws
with the current system. The fundamental and and
currently inescapable problem is that over time
the rate leap seconds are added will increase.
The proposal to cease the insertion of leap
seconds would mean that they will accumulate.
However the relationship between solar and local
time is already influenced by many factors such
as geographical and seasonal variations. Which
creates over half an hours variation over the
course of a year.
Political decisions have even greater effects,
such as with daylight saving and the use of a
single country wide timezone such in China
(which would normal occupy four time zones).
Within a normal time zone there are variation
such as the 25 minute difference in solar time
between Sydney and Melbourne due to their
Under the proposal individual governments would
be free to adjust for these change, through their
timezone and daylight saving systems (leap
seconds would produce a daylight saving change
once every six centuries). The result would be
that UTC - the international legal time scale
would be free from "discontinuities".
Leap seconds arise from the fact that the
rotation of the earth is slowing down (due to
tidal breaking from the moon). For the last half
century the difference has been on average about
1 second every 1.5 years.
However this is effected by many factors such a
earth quakes, seasonal variations and unusual
weather patterns, melting icecaps/permafrost and
even the construction of hydroelectric dams.
In fact there is so much variation that between
1999 and 2012 there has been only two leap second
insertions (rather than expected eight at 1.5 year
In comparison the earth rotation around the sun
is much more constant and the current method
for calculating the length of a year has been
known for about 500 years.
Even so programmers still make mistakes such as
in Microsoft COM/Excel where they wrongly
assumed that 1900 was a leap year.
Another example is comparing the earths
rotation to a cheap $35 eBay rubidium standard
that will lose less than one second is a
It is their unpredictability that causes the
greatest problems. Leap second insertions are
only announced six months in advance. So there
is no way of knowing of the number of seconds to
any event beyond six months in the future.
It was announced in January this year, that a
leap second will be added on 30 June 2012 +0000.
(~10:00 1 July Australia EST).
This is the first time June has been used since
1997. Whilst 1 July is generally not a public
holiday (unlike 1 January), this year at least it
is a Sunday, and this should avoid some bugs in
time critical systems.
However 1 July could be a normal trading day, and
in the modern world with high frequency trading
and "flash crashes" this glitch could have major
and expensive side effects.
As stated in the previously the GLONASS time
scale does not ignores leap seconds, so uses
"rubber seconds" to stay synchronised to UTC.
In the past this only effected high end dual
constellation receivers used for geodetic
However with the new combined GPS/GLONASS
receivers such as the Apple iPhone 4GS and
Samsung Galaxy Note/Wave 3 these problems will
affect a much larger audience.
---- [SLIDE 92:] ----
Leap seconds are unpredictable!
[Graph of the Length Of Day(LOD) over
from 2010-2012 with periodic lunar
and tidal variations removed]
---- [:SLIDE 92] ----
There is a great deal of variation in the rate
that the earth spins. Which is superimposed on
long term drift caused by tidal breaking, and
periodic orbit variations.
This graph shows the Length Of a Day varying by
over 1.6 milliseconds.
(1.6 milliseconds per day => 0.6 seconds per year)
---- [SLIDE 93:][HUMOUR] ----
"In what month was the
It took 341 for the Gregorian
calendar to replace the Julian.
How long will it take before we get
leap seconds fixed?
---- [:SLIDE 93] ----
---- [SLIDE 94:] ----
Time-Nuts mailing list:
European Space Agency(ESA) wiki:
---- [:SLIDE 94] ----
[END OF PRESENTATION]
More information about the time-nuts