Thread: About 37xx converted to 34xx satellites reception accuracy

Originally Posted by Kanopus

Both tracks, for 3790 and 3490, look very good and show no trace of problems

37xx(red) is number one , but converted to 34xx(yellow) isnt so bad

Originally Posted by catymag

37xx(red) is number one , but converted to 34xx(yellow) isnt so bad

What do you convert a 3790 for green in?

@Caty
I'm guessing that the grey line is 3590 and the green is 3490 with kunix's flash then. Interesting as a side-observation is that the 3790 [red] seems to be more consistent than the 3590 [grey] if that's that case.
@all
It would be worthwhile to test 3790 against 3490 in adverse 'city canyon' conditions so i'll do a side-by with those 2 in the big smoke when i can. The area used yesterday has quite big trees in places but there's nothing quite like the satellite signal blocking effect of tall buildings, is there? I really don't want to venture into the city with 4 devices strapped to the windscreen though. I attracted some unwanted hard looks from Sunday afternoon power walkers and joggers while driving aimlessly around in their leafy suburb and i'm sure more than one of them noted down my plate number 'just in case' after i had passed them several times. I didn't quite look like this, but it felt like it:



Maybe i'll reconvert the 3490_kunix later to conventional fw 3490 and add it in as well.

@Giomen
Huh???? Green line in Caty's google maps, i'm sure, is the 37>34 flashed with kunix's modded fw [34x0 with 37x0 gps fw].

@Neil and Giomen
I just downloaded this:

Originally Posted by dasilvarsa

All Tracklogs Combined in GDB Format.

[Only registered and activated users can see links. ]

and dropped it in GPSMapedit.
Colors have been choosen by dasilvarsa and you can check them in MS tracks tab
[Only registered and activated users can see links. ]

3790>3490= yellow
3790>3490_kunix= green
Real3590= gray
Real3790= red

The great material produced by Neil can be analyzed from different perspectives, or more objectively, under various zoom levels, which can result conclusions somewhat different from those shown above. Our comrade Kanupus seems analyzed these data with a zoom farthest and concluded that "Both tracks, for 3790 and 3490, look very good and show the trace of problems. I can not see any reason for concern about accuracy .. ". I, on the other hand, used a zoom closest of the tracks taking the cue Caty's, working over Google Earth images which allows a realistic view of the terrain. Based solely on these observations I have come to some conclusions:

The first and unquestionable conclusion is about the 3790 real, which confirmed being in fact very accurate, despite the "evil tongues" lol, because your tracklog remained almost 100% of the path correctly positioned on the roads, an excellent device! As I currently have a 3490 new, soon I will make a comparison with the 3790 real, but I believe it must be equivalent in terms of accuracy.

Neil’s test also made it clear that the converted 3490 is far less precise than the 3790 real, as shown in the screens below attached of Google Earth. So if it can be concluded that without doubt the accuracy in gps screen is linked to the tracklog recorded and is not mere "impression" or mistake of uninformed users. The lowest precision shown on the screen of 3790 converted implies greater quantities of positioning errors on the ground, as shown by the tracklogs. The real 3790 was really accurate, the real 3590 not so much, the 3790 converted have substantial failures during the path, as shown in the attached images, and the 3790 Kunix 's, a case that I'll comment the part. I was careful to follow the tracklogs with extended zoom from the beginning to the end of the paths and captured many images of inaccuracy of 3790 converted, but for practicality I am publishing only a few. To make it easier I have attached the kml files for viewing in Google Earth for those interested. It is good to remember that accuracy in screen is not an absolute datum, but an order of magnitude, and not necessarily instant information. It may take some time for the system to change the accuracy shown on the screen, for example if you have a 3 meters brand in the sky opened and go into place without or with low signal It may take some time to appear on the screen the new "precision".

In relation to "notable" inaccuracy of 3790 converted with Kunix firmware, what seems to be a huge disappointment, I think we can see optimistically. That's because somehow the Kunix changed something in a region that has a direct relation with the accuracy of gps, it now remains to find out exactly which. Such inaccuracy also proved more than evident that the firmware, or part of it, is directly related to the positioning accuracy.

All these tests and analysis was an apprenticeship: really the best way to compare the accuracy is analyzing tracklogs of several gps, preferably on a reliable bases, in many cases of urban perimeters, with satellites photos as Google Earth (maps may be more susceptible to errors), since equality criteria are followed in the measurements, as previously mentioned. Also important to note that many of the errors presented are corrected, or rather, masked, by the Lock on Road of gps, so that most users do not give much importance to "accuracy" of 3790 converted. Anyway, it would be interesting to also do tests with lock on Road enable, because the inaccuracy can fix the vehicle in other streets and other aberrations, as I've seen it happen several times. The resulting tracklog would be funny..

Finally, a friendly advice to Neil: dispose of your 3590 real quick, lol...



ps.: yelow-> 3790 converted; red-> 3790 real; green-> 3790 Kunix fw; blue ligth->3590 real

* more images and the tracks in kmz format (GEarth), in the link:

[Only registered and activated users can see links. ]

** click on image to zoom
[Only registered and activated users can see links. ][Only registered and activated users can see links. ][Only registered and activated users can see links. ][Only registered and activated users can see links. ][Only registered and activated users can see links. ]

[Only registered and activated users can see links. ]
Here is real 3490 walking log for Google Earth and Mapsource.
It seems to match the 3m accuracy of the 3490, A bit shaky for my liking, Ignore the first few point as the unit was still acquiring sats.(finding it's feet)
This was my lunch time walk today shared with a Large Ice Cold Coke.

Ok, i'm maybe gonna chuck '[Only registered and activated users can see links. ]'. My habit of making outrageous statements might be irritating to some ppl but i believe we need to be openly polarized about this 'problem', i.e. I think there are 3 main camps:

(1) those who think there is no real problem and therefore nothing needs to be done about it;
(2) at the opposite end are those who not only believe that the accuracy problem is real but who passionately seek a solution;
(3) in the middle are those who find this investigation interesting and are participating mostly out of interest and curiosity.

I started in Camp 1 as evidenced by my statement in post #6 linked above. I've now moved to Camp 3. I may be one of the few users here who have used GPS (i) on road in cars, heavy trucks & on bikes; (ii) outdoors both day and night [in military situations on foot navigating in difficult terrain]; and (iii) in aviation. I can state categorically right here and now that automotive use on road does not require the same degree of horizontal or vertical accuracy as the other two. Normal auto use particularly does not need highly accurate vertical readings, as evidenced by the fact that nobody except me has commented on the wildly varying altitude readings between the converted and unconverted units. Have a look at the readings. They generally don't agree. Maybe none are correct, i haven't checked but that's not my point. It's that the divergence is much wider than their horizontal 2 D divergence and i believe that is perhaps to do with hotfix probably not geared to work on altitude, only on coordinates. None of these unit rely on SBAS (WAAS) or LAAS or any other physical (real) means of positional error correcting. What they have is software help with that, e.g. lock on road, hotfix.

To some degree i think this is a 'storm in a teacup'. Nobody is going to die in a fiery plane crash because he used his 37>34 convert to make a rainy/foggy/night landing approach only to find out too late that he was 50' to the right/left of the centreline and/or 100' lower than he thought. Nor die because he accidentally stumbled onto an enemy firepit during a nighttime recce intending to set up an ambush or sniper post. Where such close accuracy is important there are specialized precision gps units designed for those jobs. I'm not being dramatic here, just trying to put this in context and perspective. So i believe this absolutely: Vertical (3D) precision is of little importance for auto devices and 15 metre/50 feet 2D accuracy is good enough. No doubt others hold a difference view and i'm not saying they are being precious about wanting 3 metre/10 feet 2D accuracy in an auto device. But if you do want it, and kunix cannot sort this out with a fw mod, then convert back to 3790.

Regardless, i'll do a side-by-side of virgin and convert in the city when i can. If kunix has come up with some new tweak i'll test that too. If not, i'll revert the 3490_kunix to regular 37>34 convert and include that just in case it's different gps hw to the others [as kunix said, maybe his mod didn't actually flash anything]. I only wish i knew someone locally with an original 3490 too. God know's dasivarsa won't send his over to me.

PS: Angelo, do you wanna buy my 3590? Make me an offer. I must say i'm a little surprised that the smaller-cased 3790 is more accurate than it.