| |
| ProfessionGPS |
|
|
| POIenFrance |
|
|
| Visiteurs |
Actuellement en ligne: 143
Dont membres: 0
+ d'infos...
|
|
|
| GPS Chipsets and Receivers Compared |
3. On the Road
 So far we've looked at the specs and I've shared my impressions on the overall behaviour of these various GPS receivers based on my daily testing routines and general use with road and moving map software. That's a good start but now let's dig a bit deeper and see if these impressions match the facts. To do that I set up a testing method that relies on side by side testing with simultaneous logging of raw GPS data. This requires a lot of time and equipment (old PocketPCs come in handy !) but I feel it's really the only way to level the playing field and make meaningful comparisons. Keep in mind that road navigation software "snaps" the position to the closest road so a lot of the raw GPS data will be smoothed out, still the better the underlying data, the less "snapping" guesswork will have to be done and the more accurate and faster the guidance will be.
Starting off with the "road driving" environment, which is how most GPS receivers will be used. The good news is that on the open road, I've found that all the GPS receivers using various chipsets performed equally well, accurate tracking and no visible lag. Where things get more interesting though is when the "going gets tough" for GPS signals in dense urban environments where you actually find yourself needing to rely most on GPS guidance to find your way. This is what we will focus on.
Downtown Paris - ST/XT2/SS3/NX/uNAV/Garmin
The first series are tests done in the dense urban environment of Paris where there's a lot of interference on hand for GPS signals although not the urban canyons one would find in Manhattan where I will be doing some testing shortly too. Here are pictures of the results and comments are below:
   Fig.d1 - Fig.d2 - Fig.d3    Fig.d4 - Fig.d5 - Fig.d6
This first series of pictures gives a few clues as to how these various chipsest deal with these adverse reception conditions and various assortment of underpasses and narrow streets. Let's start with the problems and it appears that the NemeriX chipset as tested with the iTrek is having a hard time producing a clean GPS position even though it kept the fix with a good number of satellites throughout (Updated 04/2006 - NemeriX have now released a v4 firmware with very impressive performance - and multipath being no longer present - Updated 07/2005 - NemeriX has released a v3.1 firmware that fixes these issues for the most part - no upgrade is possible at this time, but new models such as the Haicom 305N are shipping with that version). it's all over the place really and it's going to be difficult for the navigation software to give meaningful guidance at times and you will likely see some confusing map rotation. My guess is that they are not filtering enough of the multi-path (GPS signals bouncing around) although this gives the impression of high sensitivity with many satellites in the fix. In fact no chipset is immune to that problem and SS3 shows it for a few seconds too at the beginning of the test at low speed (Fig.2), more on that in the pedestrian testing. The uNav and Garmin chipsets also struggle when reception gets difficult (as seen in Fig.6 with a passage between appartment blocks) with significant jumps in the position. On the other hand the SiRF chipsets (ST, XT2, SS3) provide much smoother tracking although as seen in the same picture, the track veers off course slightly after the first turn on the right, still, this is something that "snap to road" will deal with easily.
Downtown Paris - ST/XT1/SS3/Garmin
More testing in downtown Paris with Xtrac v1 added for comparison purposes:
   Fig.d13 - Fig.d14 - Fig.d15
As expected XT1 shows some "rounding" that was typical of the first version of Xtrac in addition to the lag not visible here. Less expected is Garmin's "blind tracking" feature that produces results likely to confuse navigation software. Apparently when the fix is lost (in d15 the turn takes place in a tunnel), the GPS continues to output data based on previous fixes, just in case you kept driving in the same direction...While this type of approach may have made sense a few years ago, especially with the 100 meter accuracy induced by SA, it doesn't seem very appropriate in comparison with what the competition achieves these days.
Downtown Nogent/Marne - ST/SS3/NX/Sony
The next set shows some comparisons in the small streets of Nogent sur Marne in the East of Paris:
  Fig. d7 - Fig. d8
Again, as seen in fig. d7, the NemeriX chipset does not perform very well and the Sony chipset is also struggling to keep up. Fig. d8 focuses on ST vs SS3 and this is where we start to see the performance boost brought by SS3, just as accurate as ST in "normal" conditions and more dynamic and accurate when they get more difficult, likely due to computation power brought by the correlation boost.
RER in Vincennes - ST/SS3
Not technically driving, but with similar dynamic conditions to driving through a succession of tunnels, this test was done on the RER commuter train just outside Paris in Vincennes when it alternates between underground and "dug-in" portions with a (limited) view of the sky. Train was travelling East to West.
  Fig. d9- Fig. d10
More of what sets SS3 apart from ST, with both receivers tracking until the underground portion for the Vincennes station and then only SS3 being able to start tracking again in a meaningful way in the "dug-in" portion. Granted it's slightly off mark, but not enough to be a problem for navigation software.
Urban Canyons in Toronto - ST/SS3
Marvin did a test run (West to East) in Toronto on King Street in the middle of tall buildings where GPS signals can do a lot of "bouncing". He was using a Globalsat BT-338 and a Copilot mouse (Rikaline 6010 with ST FW220), this is what it looked like:
  Fig. d11- Fig. d12
The first half of the drive is easier for GPS without many tall buildings, but after that SS3 comes into its own, tracking close to the route even if some drifting can be observed. Not seen here, but Marvin pointed out that in "heading up" mode, the map was in constant rotation with the ST GPS, almost enough to make him dizzy! Also of note are the lost fixes with ST with 367 recorded points versus 408 for SS3.
TTFF (Time to First Fix) - ST/SS3
Now that we've seen how SS3 could improve on ST when the environment gets tough, another aspect of SS3 that enhances the user experience is the faster speed of TTFF, when the GPS receiver starts from scratch to acquire a position. There are many users who complain that leaving an underground parking garage in town will result in 5+ minutes of driving without a fix. My answer has been to suggest waiting for 2 minutes upon exiting the garage, but that's not always possible and hardly user friendly. From my testing so far, it seems that SS3's massive correlation power allows for consistent TTFF, about 45 seconds, whether you're moving or not:
 Fig. d16
You can see above that the ST based Royaltek RBT-3000 starts tracking quite a bit later than the two SiRFStarIII receivers (Globalsat BT-338 and Leadtek 9553) in an easy open environment and still needs some time after that to produce a "good" track.
If you have questions or comments, you can use this discussion thread in the forums
|
|
|
|
|
|
