Sure Electronics GPS board with PPS signal @ $35
Does anybody know anything about this product? It looks attractive as a reference time source, as it exposes the PPS signal via RS-232 as well as USB/Bluetooth For this price it seems like it could wear many hats.

I wish I had seen this two weeks ago, I already bought a USB GPS for this purpose that I'm hoping to modify..

Mini USB & Bluetooth Interface GPS Demo Board_GPS/GPRS/Ethernet, Wifi and Wireless Communication
  • Low Power consumption
  • Powered by USB port, no need to feed with DC power supply
  • RF Receiver of Noise Figure is at 2.5 dB
  • High Sensitivity -161 dBm( indoor)
  • Seamless Outdoor/Indoor Operation
  • Support standard NMEA-0183 V 3.01
  • Graphic GPS testing software is provided for free
  • Software for modifying baud rate is provided for free
  • Interfaces: USB Mini-B and USB B interface, RS232 Female Interface, bluetooth
Marvin Hlavac
-161 dBm is decent. Are you working on a Car PC, or is this something you would want to squeeze into a laptop/tablet/whatever?
No, I already have an auto/portable GPS, my recent purchase was to be to allow my home Linux system to get accurate time (within a second or so) without needing to depend on the network. I saw a dongle GPS on sale at Geeks.com for $19, a brand I had never heard of, Unitraq. Its turned out to be very good, as far as acquiring satellites, and accuracy, but I still have not been able to figure out what chipset is in it. It does work very well for setting the time using the SHM interface, using NTP. I need to learn more about gpsd, and its interface to the hardware. I think the chance is good that the chipset in the GPS is only supported in NMEA, the binary protocol may be unsupported by gpsd. Does anybody have any tips on figuring out what chipset is in a "no name" GPS dongle. So many Asian products downplay the sourcing of the hardware particulars, perhaps because the components change often. Its frustrating. This Sure Electronics board looks really useful, given that it exposes the PPS interface directly. Using that I think it would be easy to set up a stratum 1 NTP server. (!) Which is way overkill for me but the opportunity to learn something new is interesting. I do some database work and the applications for a super-accurate clock are so many that I think it might be worth it to just do it, and use the dongle for my laptop unmodified. The dongle does have an LED that appears to be showing PPS. I wonder if its really that PPS signal. I guess it would not be so hard to find out using a photocell or something. Many power meters on homes have a similar flashing LED and people commonly attach a photocell to the outside of them to monitor their power usage.. something like that.
I would definitely like to do the car pc thing at some point, but its a low priority compared to a lot of other things going on in my life.. lol. The accurate network clock is something I can use immediately as I can start correlating logging events on my different computers better that way.
Size is completely unimportant, the ability to use the unit on any kind of interface is cool and unusual and the price is right. Also they have what look like an excellent value in 1 wire temp/humidity sensors, which could be useful in the same project.
I'm surprised that there isn't any "identify your hardware and firmware revision" string in NMEA. Is that really true- that there isn't? almost every other protocol that I know of has something like that.
Ken in Regina
Garmin's proprietary protocol is essentially NMEA but with extensions. The ability to query the device for things like software/firmware revisions, Unit ID/serial no. and battery charge level are among the extensions.

I figured out what chip my dongle has by cracking it open and looking at the module in it, its a 524C chip - which is a Venus 5 by Skytraq. So, I got a good deal, its performance is quite good. But its USB, which now I understand will limit my accuracy for TIME uses substantially. Its good enough for now, but looking ahead I think I'm going to get one of these boards with the RS-232C mod described here. http://www.satsignal.eu/ntp/Sure-GPS.htm for that, and keep on using the $18 dongle for when I "need a fix" with my laptop, since it performs so well. That is unless I can figure out an easy way to extract PPS from it. The LED might be it. Still haven't had the time to measure that with anything handy. This Sure Electronics board appears to be quite accurate for use as a time source, and these people should know. It also uses the MTK chipset, which I think might be quite fast (>10Hz) to update data http://www.leapsecond.com/pages/MG1613S/ http://www.febo.com/pipermail/time-nuts/ http://wiederlinge.de/cgi-bin/lnwiki.cgi/147 also http://home.mira.net/%7Egnb/gps/index.html http://www.rigacci.org/wiki/lib/exe/fetch.php/doc/appunti/hardware/gps/mtk_packet_user_manual.pdf http://www.eecis.udel.edu/%7Emills/ntp.html Don't have the time to follow up on this for a few days though.
Ken in Regina
I understand "timing" so the idea of a cheap Stratum 1 timesource is intriguing. But I'm wondering what sorts of computer applications, especially "database" related applications, would benefit from anything more than, say, Stratum 3??

Specifically, the "accuracy" of Stratum 3 is pretty much the same as Stratum 1. The main difference is the drift, e.g. how long it stays accurate. And if you are linked to a Stratum 1 source (e.g. the GPS satellites) it's going to self-correct regularly anyway. Whether it's once per second, more than once per second, or every few minutes shouldn't make any practical difference. Should it?

Not a criticism ... genuine curiosity. My experience with timing comes from the timing (sync) for data networks in the telecommunications industry (telco engineering). I never really gave it much thought as it relates to computer database applications, even though my background is IT.

Ken, currently, absolutely nothing justifies going to all that effort to have such an accurate time source. However, in the future, who knows, I like the idea of total accuracy without depending on the fickle net. But basically I just think it would be cool to be able to do it. Perhaps secretly I resent having to look to remote time servers for "the" time, knowing that "its all relative" anyway.
Ken, Perhaps this is a better answer.. I have several home automation devices, a PBX and a bunch of VOIP stuff, along with computers, TV/DVD and two laptops on my home network. I'm also about to add a bunch of sensors connected via owfs or something similar (haven't decided what interface makes the most sense for that yet) If the time on all of them is synched closely, I guess I'm just expecting that I could reliably extract much more meaningful data from the logs when something doesn't work properly. Which begs the question, why don't I set up a syslog for all of them on the box that gets the accurate time whatever, and indeed, that's also in the plan. There isn't any argument for the extreme accuracy.. anything less than a second would be acceptable, as long as its consistent.
Ken in Regina
I understand the value of synced time logging of events. I've written and supported SCADA systems. As you say, usually accuracy to a second or so is sufficient. But for some issues, especially where they might cascade through a number of apps or devices, it's helpful to get somewhere into the subseconds. Otherwise you can't see the correct "direction" of the event as it moves through the network.

In your case you have a couple of options.

You can connect a GPS to one PC and make it the network time server. The GPS dongle and a utility that corrects the time on the time server PC from the GPS dongle should get you more than sufficient accuracy and stability. If you want, say, Stratum 2 or 3 stability, just check the spec and make sure you update the system time in slightly shorter frequency than the drift rate of the appropriate Stratum would take you outside the related accuracy (factoring in the measured drift rate of the PC's system clock).

The other option you've already mentioned: do the logs on the "time server" box so there is only one clock being referenced for the time stamps.

Or you can do a hybrid. Do time serving for the devices that can use a time server. There may be other reasons than logging to have them locked fairly close to each other. Shucks, just having them locked at an assured Stratum 2 or 3 would be a kick, if for no other reason than ego. And for the devices that can't use a time server, do their logging for them on the/a central box.

.............hmmmm.... Now you've got me thinking. I think I can get an NTP app for Windows. I've got a bunch of GPS receivers kicking around. It would be interesting to use one of them to create a time server and lock my five PCs to Stratum 2 or so.

My i.Trek M7 updates at 5Hz. I don't think even a crappy Wintel system clock can drift fast enough that it can't be forced to maintain a stable time with something like that. The real trick to this is to determine what the drift on a typical PC clock is and then figure out the least frequent updates necessary to keep the time accurate. The goal would be to reduce the amount of network traffic from NTP queries on the synced boxes while still maintaining the accuracy at the desired level.

Exactly. I can't think of many situations where accurate logging isn't extremely useful.. Its great that so many devices now do logging.. Its going to get better as IPv6 and its ability to self configure takes hold.. more and more devices will expose some of their functionality to net control.. And of course, it seems like its not the speed of the updates that's important. Its the stability of the source, specifically the rise time of the edge of the pulse that you get the fix on.. (since the instant in time is triggered by that) and since its coming from a GPS, the likelihood of it being excellent is good. But like you imply, I think it would be nuts to rely only on the GPS given the fact that sometimes they deliver abberant times - ir no time at all..(say, right after a reboot.)

There are some great NTP-capable apps for Windows. They are every bit as accurate as anything on *nix, although perhaps getting multi-year uptimes (between rebootings) is too much to ask for ... Yesterday I saw this web page from Japan that gives the plans to use a Garmin puck style GPS as a timing source. They also link to what looks like an *excellent* free NTP app for Windows. SATK http://www7.ocn.ne.jp/~set/Satk/SatkE.html There are also "official" Windows compiles of NTPD, on the ntp.org web site, I think. Micon-GHS clock Also check out Micon-GHS clock which uses a rockwell Jupiter module.

Its important to set the GPS to only emit one specific sentence (which one!? though? still learning about this..)
in the GPS mode to prevent it from inserting two NMEA sentences where you only want one, perhaps causing some wobble.. Actually, my logging host use case is a bit different than most in that I have it arranged so that it writes logs to two places, short, terse logs get written to the flash every five minutes, and long, verbose logs get written in realtime to a hard drive.. for now, I think I may set up an old Cobalt Qube 2 (remember them? They still look modern.) to do some of the logging as well. I don't think its asking too much to put asterisk/freepbx,postgres - with an extremely light load, and syslog and soon, hopefully, owfs+one wire dongle+hvac control on the same box. For Asterisk the box has no problem with two phone conversations at the same time (there only are two of us here) and doing everything else. The solid state drive is actually faster than a regular hard drive, and completely silent.

The only PITA is that currently I am using RPMs for everything and have to do the building of them on my Macbook using VirtualBox since the main filesystem of the thin client SSD shouldn't be written to any more than necessary and even if I didn't care the space is limited.. (its actually a HP t5720, an extremely good deal for the money -$100 used. They also make excellent carputers for road warriors.) The total power draw *with* the extra hard drive is still under 10 watts! Very little heat relative to a "regular" system. And the only sound it makes is almost inaudible.

I found the datasheet yesterday showing how to expose what they claim is the P1PPS out of the Venus chip- at a 1.3 - 3volt TTL level.. (I'd have to solder one wire plus a ground to the GPS chip.) Ive had some experience with fine soldering, so I don't think that would be too difficult. But the thing that is daunting is testing that signal to verify that indeed it is adequate to the task. I don't even have a real oscilloscope, although I could probably rig something up with my computer and some added parts.. there are audio frequency oscilloscope apps..say, I could use the signal to back-bias a diode and interrupt an audio tone. Or whatever.. cross that bridge when I get to it. Or, I guess I could just buy the Sure board and be done with it. But my wife would probably prefer me to use what Ive already got.. which sounds like a fun challenge, getting a super accurate clock for our home net for $18, $2 in parts and a bit of soldering. (and a bunch of time.. shhhh!!! BTW, yesterday I stumbled across source for the Jupiter modules used.. Under $10. But I get the impression - it being old technology, that the Venus is far more sensitive.
laptopgpsworld.com About