NEWBIE: Cheap GPS or WWVB time source?
Jonathan
I'm looking for a cheap WWVB radio or GPS, with serial interface, that can
be hooked to a Unix box running xntpd to create a time server. Considering
this is for a home network, I suppose accuracy is not the highest priority.
However, I want to know how much bang-for-the-buck I can get, ie, what's the
best cheap-but-accurate time source?
Please be specific if you have a model to recommend, as I'm really not
familiar with many GPS or radio-clock mfrs. & models.
thanks,
Jonathan
jonathanNO at FISHsprintmail dot com
(remove NO FISH to mail)
Jonathan Buzzard
"Jonathan" <jonat...@SPAMsprintmail.com> writes:
> Hi,
> I'm looking for a cheap WWVB radio or GPS, with serial interface, that can
> be hooked to a Unix box running xntpd to create a time server. Considering
> this is for a home network, I suppose accuracy is not the highest priority.
> However, I want to know how much bang-for-the-buck I can get, ie, what's the
> best cheap-but-accurate time source?
>
> Please be specific if you have a model to recommend, as I'm really not
> familiar with many GPS or radio-clock mfrs. & models.
Take a look at
http://www.buzzard.org.uk/jonathan/radioclock.html
Probably the cheapest option out there. The antenna and clock module
are not available in the U.S.A., but the main supplier in the U.K.
Maplin (a bit like a Radio Shack) will take orders and ship to the
U.S.A.
JAB.
--
Jonathan A. Buzzard Email: jona...@buzzard.org.uk
Northumberland, United Kingdom. Tel: +44(0)1661-832195
Jonathan
"Jonathan Buzzard" <jona...@buzzard.org.uk> wrote in message
news:j60qma...@192.168.42.254...
> In article <up0ogq3...@corp.supernews.com>,
> "Jonathan" <jonat...@SPAMsprintmail.com> writes:
> > Hi,
> > I'm looking for a cheap WWVB radio or GPS, with serial interface, that
can
> > be hooked to a Unix box running xntpd to create a time server.
Considering
> > this is for a home network, I suppose accuracy is not the highest
priority.
> > However, I want to know how much bang-for-the-buck I can get, ie, what's
the
> > best cheap-but-accurate time source?
> >
> > Please be specific if you have a model to recommend, as I'm really not
> > familiar with many GPS or radio-clock mfrs. & models.
>
> Take a look at
>
> http://www.buzzard.org.uk/jonathan/radioclock.html
>
> Probably the cheapest option out there. The antenna and clock module
> are not available in the U.S.A., but the main supplier in the U.K.
> Maplin (a bit like a Radio Shack) will take orders and ship to the
> U.S.A.
Thanks for the pointer Jonathan. I actually found your site via Google
yesterday and it looks like a really easy project. However to be complete
I'm also exploring all available options right now.
How accurate is your radioclock? Also, I may be using a Sun system running
Solaris as the host. Has your clock been tested on any non-PC hardware and
any commercial Unix flavors?
thanks,
Jonathan
joseph c lang
David Schwartz
> Take a look at ultralink (www.ulio.com)
I bought an ultralink and found the time reference unsuitable for NTP
use. Jitter was around 15 milliseconds and drift between
synchronizations was intolerable despite my heroic efforts to clean up
the data it was giving me with software filters. I returned it.
Unless you hear a few success reports from other people, I'd suggest
you stay away from these. If you're not talking about NTP
synchronization, though, these are perfectly fine. If you need, say, 1/4
second accuracy for a remote data logging or call center application,
you should be perfectly satisfied.
A GPS reference will cost you at least $100 to $150 more, but you
should get microsecond accuracy.
DS
Jonathan Buzzard
"Jonathan" <jonat...@SPAMsprintmail.com> writes:
[SNIP]
> How accurate is your radioclock? Also, I may be using a Sun system running
> Solaris as the host. Has your clock been tested on any non-PC hardware and
> any commercial Unix flavors?
Depends on the time source to some extent. As I have never been able
to test against the WWVB transmitter I cannot say for sure. With
MSF and DCF77 and Linux offset and dispersion would appear to be
just under one millisecond. I am experimenting with some patches that
improve this by 'averaging' the time of the pulses over the last thirty
seconds. This leads to nearly an order of magnitude improvement in
jitter, and a much more stable frequency correction.
The polling mode should work on any comercial Unix, though this
degrades accuracy significantly. It is also possible use the
PPS interface which I believe exists in Solaris, though I don't
support this myself, but see URL below for some software that does.
My software is fairly plain Ansi C and Posix. The only Linux specific
bit is the use of TIOCMIWAIT ioctl to wait for a status change on
a serial port DCD or CTS line. FreeBSD has some problems with a
mlockall call which attempts to lock the program in memory, and
a sched calls to set realtime scheduling, but these are all Posix,
it is just FreeBSD does not support them, though they are not critical.
http://www.jonatkins.com/radioclk2/
Basically this is the cheapest hardware you will get for NTP.
Any problems with WWVB decoding will get my highest attention,
because this mode of operation has not been tested yet, though
I believe it to be correct I would like to try and verify it.
Hal Murray
>be hooked to a Unix box running xntpd to create a time server. Considering
>this is for a home network, I suppose accuracy is not the highest priority.
>However, I want to know how much bang-for-the-buck I can get, ie, what's the
>best cheap-but-accurate time source?
[xntpd is pretty old.]
How much are you willing to spend and/or how much tinkering/soldering
are you willing to do?
For GPS to work well for timing, you need a 1 PPS signal. Generally,
it is connected to the DCD pin on the serial port.
Here is my list of low cost GPS units:
Garmin:
GPS16, GPS35 under $200
http://www.gps4fun.com/ http://www.navtechgps.com/
I haven't used them. The GPS35 is intended for automotive use.
GPSClock:
GPSClock200 $380
http://www.gpsclock.com
I have one at home. Almost works OK inside my house. I'll
mount it outside when I get around to it. I have one at work
with a good outside mount. Works fine.
The Garmin and GPSClock units are intended for mounting outside
where they can get a good look at the sky. You need to provide
power and wire things to a DB-9. (Wall bricks are cheap.)
Motorola and Trimble both make cards for OEM use. You will need
to add an antenna, power, RS232 level shifters, and a box if you want
it to look good.
Motorola:
http://www.motorola.com/ies/GPS/products/
http://www.synergy-gps.com/
Trimble:
http://www.trimble.com/timing.html
I haven't found a good web site that sells Trimble OEM cards.
I picked up a couple of recycled Trimble units for $25 each. I
think they were pulled from a truck fleet. I haven't had
time to see how well they work.
If you are a serious geek, feed Z3801A to ebay and/or google.
--
The suespammers.org mail server is located in California. So are all my
other mailboxes. Please do not send unsolicited bulk e-mail or unsolicited
commercial e-mail to my suespammers.org address or any of my other addresses.
These are my opinions, not necessarily my employer's. I hate spam.
Jonathan
"Hal Murray" <hmu...@suespammers.org> wrote in message
news:up2og91...@corp.supernews.com...
> >I want to know how much bang-for-the-buck I can get, ie, what's the
> >best cheap-but-accurate time source?
>
> [xntpd is pretty old.]
sorry, I (obviously) don't use ntpd much, but the old xntpd is what Sun
*still* ships with Solaris, at least up thru Solaris 8 and I'm a big Solaris
user, so that's what popped into my head. Maybe they've gotten more current
in Solaris 9 but I haven't checked.
> How much are you willing to spend and/or how much tinkering/soldering
> are you willing to do?
Love tinkering. Soldering is no problem. I would probably draw the line at
etching PC boards. :-)
> For GPS to work well for timing, you need a 1 PPS signal. Generally,
> it is connected to the DCD pin on the serial port.
Hmmm. Do most GPS units provide a PPS signal? Do other clock sources, like
WWVB radios (see Jonathan Buzzard's project pages) require PPS?
Sorry for this question, but I don't think I yet fully understand how the
PPS signal fits into the equation.
> Here is my list of low cost GPS units:
[SNIP]
Excellent, thanks for the list! Do most Garmin units have RS-232
interfaces? Ie, how hard is it to find a PC-compatible (or specifically,
ntpd compatible) Garmin GPS unit?
OTOH, I may just end up building the WWVB receiver, as I'd rather not have
to worry about stringing a GPS antenna outside, if avoidable. I already
have a WWVB desk clock in the same room as the computers, and it gets a
great WWVB signal.
> I picked up a couple of recycled Trimble units for $25 each. I
> think they were pulled from a truck fleet. I haven't had
> time to see how well they work.
>
>
> If you are a serious geek, feed Z3801A to ebay and/or google.
This is the kind of hardware-hacking I find challenging... finding
somebody's scrap and making use of it. :-)
I'm a big eBay fan and will see what I can find.
thanks,
Jonathan
Jon Atkins
Buzzard) wrote:
> The polling mode should work on any comercial Unix, though this
> degrades accuracy significantly. It is also possible use the
> PPS interface which I believe exists in Solaris, though I don't
> support this myself, but see URL below for some software that does.
[...]
> http://www.jonatkins.com/radioclk2/
The version that *was* there this morning wasn't intended for public release.
I've just uploaded the latest version, 0.04, which now includes some
(incomplete) documentation.
Can the two people who downloaded 0.03 please return for the latest version.
--
Jon Atkins
Tapio Sokura
> A GPS reference will cost you at least $100 to $150 more, but you
> should get microsecond accuracy.
It depends.. I bought a couple of "new" 6 channel Motorola Oncore VPs from
www.synergy-gps.com, $25 USD each. These are old tech OEM GPS units in
excess inventory sale and you need to do some extra tinkering to interface
them to a serial port and power (TTL level serial port and 5 volt regulated
power).
If I remember correctly, TAPR (www.tapr.org) sells a suitable interface
board kit for $40 USD. I built my own for a few euros and it certainly looks
like it: http://iki.fi/oh2kku/kello/ . Note the recycling of a floppy disk
as voltage regulator cooler :-) But it seems to work, keeps my Linux time
server within some microseconds of UTC.
You also need a separate antenna, I bought some ANPC-135 active antennas
($24 each) with the GPSes and they've been working fine (the cable length is
closer to 3 meters than 5 as is mentioned at synergy-gps.com).
So the total price of my GPS sync equipment is somewhere around 60 euros
including the GPS, antenna and home-made interface board.
Hal Murray
>WWVB radios (see Jonathan Buzzard's project pages) require PPS?
>Sorry for this question, but I don't think I yet fully understand how the
>PPS signal fits into the equation.
GPS units use mostly RS-232. The PPS signal is needed to get accuracy
down to the microsecond level. Some GPS units have it. The rest aren't
much good for accurate timekeeping.
I think Johathan's radio setup uses the DCD signal rather than the
normal RS-232 serial paths. (Same connector, different pin.) The
data rate is very slow. You can afford to take an interrupt on each
bit and the data on the air isn't in character format that RS-232
chips support. The message parsing is done in software.
>Excellent, thanks for the list! Do most Garmin units have RS-232
>interfaces? Ie, how hard is it to find a PC-compatible (or specifically,
>ntpd compatible) Garmin GPS unit?
Check the data sheets. I think most GPS units sold for timekeeping
have RS232 or RS442 type interfaces. The OEM units might have
(inverted) CMOS logic levels. (Use a MAX232 type chip to fix that.)
Jonathan Buzzard
hmu...@suespammers.org (Hal Murray) writes:
[SNIP]
> I think Johathan's radio setup uses the DCD signal rather than the
> normal RS-232 serial paths. (Same connector, different pin.) The
> data rate is very slow. You can afford to take an interrupt on each
> bit and the data on the air isn't in character format that RS-232
> chips support. The message parsing is done in software.
At the moment it uses either or both of DCD and CTS, though I
intend to extend this to include DSR as well. You are right though
that the software actually watches the radio pulses, you get two
interrupts per second. Because the software on the PC gets to
see these pulse edges directly it makes it more acurate than a
radio clock where the signal is decoded externally. It would
even appear to be more accurate than all those RawDCF clocks.
An accurate discription of the clock would be to call it a software
radioclock.
>
>
>>Excellent, thanks for the list! Do most Garmin units have RS-232
>>interfaces? Ie, how hard is it to find a PC-compatible (or specifically,
>>ntpd compatible) Garmin GPS unit?
>
> Check the data sheets. I think most GPS units sold for timekeeping
> have RS232 or RS442 type interfaces. The OEM units might have
> (inverted) CMOS logic levels. (Use a MAX232 type chip to fix that.)
>
Or for real cheapskates use a 741 opamp. In some respect these are better
as you can get proper RS232 signalling levels which a MAX232 will not
do. They are also faster, which is handy when you are doing timekeeping.