All things GPS

[Mr. GPS]

First off, a heaping helping of thanks to Mr. GPS for a little light in the darkness here.

You're obviously more expert (by a lot) on these matters than I am, but I do still have one reason for being concerned that there may be at least some hardware issue at play here in addition to the software problems you describe.

I was at lunch on Friday with my boss, who has an HTC Evo. We were in a restaurant, which was a standalone brick-faced, steel frame commercial building. We were in the interior, probably 40-50 feet from any window or door. We put our phones on the table (his Evo, my Captivate) and both launched the GPS Test app. His saw 6 satellites with 4 fixed almost immediately-- my Captivate saw nothing at all for several minutes, then eventually saw 1 sat, but couldn't lock it. After lunch we repeated the test out in the parking lot. By the time the GPS Test splash screen had cleared, his already had 6/8 and went to 8/10 within a couple more seconds. Captivate again started with nothing, gradually adding one at a time. Eventually Captivate did see 8-10, but had a lot of trouble holding a fix. His was steady at 8/10, mine kept going 3/8, 2/8, 6/8, 2/8, 0/8 etc, with the accuracy fluctuating wildly as the sats were acquired and lost. Also, the SNR on the Evo was 30-45 on most of the sats, while on the Captivate it was in the 20-35 range, and kept bouncing up & down, while the Evo's were pretty steady (not completely steady, but a lot more so than the Captivate).

Since GPS Test appears to be displaying the raw sat data including which sats it sees and the SNR on each one, the fact that the Captivate had weaker signals even once it saw the sats said to me that it had inferior hardware, like maybe an insufficient or poorly placed antenna. Am I way off base here? Do the symptoms that I saw sound like they can be fixed with a firmware or driver update? Maybe with better software the data that it got would be sufficient to lock more sats and derive a good location, but the fact that the SNRs were always worse just seems to say to me that no matter what you do in software, you're always going to have less to work with. Ie., a hardware problem of some sort, antenna being my best guess.

Fisrt we need to clarify
1. GPS Test reflects the computed values from Samsung GPS driver since
it is Samsung LabtestMode program.

2 Antenna and GPS chips are the factors for receiving raw signals.

So it could be the case, the GPS chips/anntenna contains X signals
and a flaky GPS driver reflects (computes) X-Y signals in weak noise satelltte

The issue is how to compare the GPS driver capabilty and GPS chips/antena separately at the same location.

My AT&T store testing shows that we can get sufficient and fair signals from BCM4571 compared with iPhone 4 in the same location.

 

[samtr87]

Do you mind going back to your first post in this thread and explain how you "train" the gps in this phone? Reading through this thread I'm a bit overwhelmed by the info.

 

[SamsungVibrant]

Once a dumb GPS receiver is given with sufficient almanac, for instance, too many testings by using GPS Status, or LabTestMode in an area, it can

 

[jfenton]

Just tried out GPS test on my EPIC.

GPS has been working fine for navigation in the car but I had no clue what it was actually doing.

Tried in 3 different areas on the way home.

In all cases, 9-11 IN VIEW, 3 IN USE and no matter where I am, a 98.4 foot accuracy...so my guess is that this # is burned into the phone by Samsung

 

[SamsungVibrant]

Just tried out GPS test on my EPIC.

GPS has been working fine for navigation in the car but I had no clue what it was actually doing.

Tried in 3 different areas on the way home.

In all cases, 9-11 IN VIEW, 3 IN USE and no matter where I am, a 98.4 foot accuracy...so my guess is that this # is burned into the phone by Samsung

do you have use wireless networks turned on or turned off. please turn it off and go for a long drive. then throw loops to the gps, when it tells you go right, make a left, take alternative routes, lets see if the gps can reroute accurately.

let us know

 

[mullingar]

Mr. GPS, sorry I was at work ...

Actually, I returned my samsung vibrant (had a totally unrelated problem, partly my own fault). I'll get a new one in few days.

As for the driver, they have apparently have a GPS daemon. I'll dig in once I get my new phone.

BTW, thanks for the rigorous analysis. It clearly sheds light on the issue.

 

[Mr. GPS]

Just tried out GPS test on my EPIC.

GPS has been working fine for navigation in the car but I had no clue what it was actually doing.

Tried in 3 different areas on the way home.

In all cases, 9-11 IN VIEW, 3 IN USE and no matter where I am, a 98.4 foot accuracy...so my guess is that this # is burned into the phone by Samsung

Your case is like a "no-problem" EPIC case as we have "no-problem" Captivate/Vibrant cases. But when you check this forum: EPIC 4G, There are many EPIC cases having GPS issues.


As my posts, I can interprete both "No-Problem" and "Problem" Captivate/Vibrant cases from the same Samsung GPS driver based on test deisgns and reports

EPIC GPS driver may be dfferent from Captivate/Vibrant GPS driver.
Or Captivate/Vibrant phones can have same results under the same condition.

My concern is that Samsung could have changed the defect driver codes in an ad-hoc way for EPIC. It could improve something, and break something else. Otherwise, it should not have many GPS problem cases reports today.

The right approach to demystify EPIC GPS is to have proper test designs, which can lead to new discovery for EPIC.

 

[baddress]

Also, the SNR on the Evo was 30-45 on most of the sats, while on the Captivate it was in the 20-35 range, and kept bouncing up & down, while the Evo's were pretty steady (not completely steady, but a lot more so than the Captivate).

I've said it before but the Captivate has sh!t antennas or antenna coding or whatever - compare them to any phone in any cell phone store side by side, and the Captivate will get worse SNR and bounce around all over the place, wifi gps cell and everything, it bounces like crazy. Almost makes you wonder if something is shorting out the antenna like the iPhone 4 problem.


edit: Also to the guy who enabled use wireless networks, like was explained before this lets the phone use the cell tower to get a rough position and then uses the phone's data connection to preload the gps almanac for that location. So then the almanac is stored for a while and that is why sometimes it will work even if you turn off the use wireless networks again.

 

[amdrider]

Why are folks saying a lower SNR is better? On the cell singnal, isnt' lower dbn better? why would SNR be opposite on GPS?

Also, how could the GPS Test app predict accuracy if it doesn't know where you are in the first place? I mean if it doens't know where I am, how does it know that it's 32.8 feet accuracy?

 

[askpcguy]

First off please explain Cold Start and Hot Start.

Basically what this means is if the GPS driver or app had a 3D fix, and you do not delete the GPS data, the GPS driver already has a pretty good idea of your position on earth. Then it's a matter of some math calculations and you have a new 3D fix. Examples are navigating someplace, and being indoors WITH USE WIRELESS NETWORKS DISABLED 30-45 minutes passes. If you walk outdoors and try the Google maps or GPS status or even LbsTestMode, it will obtain a lock usually pretty quickly.

Cold Start is different, each time the GPS driver is initialized, it has to run the routine to "zero" in to your position on earth. The following link is not directly related to GPS being implemented in mobile phones; however the technology behind it is very similar http://www.ikegps.com/downloads/TTFFstartup.pdf

Now again I ask, when everyone complains about getting poor GPS performance, is the USE WIRELESS NETWORKS enabled or disabled?

My theory behind the Galaxy S GPS is that the start up mode is dual phased. First GPS signal is obtained from Cellular Towers to give a coarse location. Then that can be refined if you will using the GPS satellites to give a precise location (street level). I will continue to state I do not want to enable Use Wireless Networks.

My prior Bluetooth GPS receiver did not use that (to my knowledge) but generally obtained a fix in a minute or two. Yes, even while driving. It never jumped streets while driving on freeways.

I suspect the Captivate GPS problem to be a mixture of hardware in the antenna placement might not be 100% ideal and the GPS driver being software.

 

[Mr. GPS]

Also, how could the GPS Test app predict accuracy if it doesn't know where you are in the first place? I mean if it doens't know where I am, how does it know that it's 32.8 feet accuracy?

Position, time, and error are variables. We only know the constraints on the variables together. Math calculation is to estimate the variable values under the constraints.

 

[amdrider]

Thanks Mr GPS.

Also, to answer my own question about SNR. I had forgotten than on our cell strength meter, "lower negative" is better than "higher negative". Ie: -51 dBm is a better signal than -100 dBM.

Just to provide another data point. My GPS trackes me down to which lane I am in on a 3 lane highway. I have only used my phone in one city.

 

[mratlndmrk]

am i crazy or does my GPS work totally fine? ive used it 10 or so times in different locations and never really had a problem with it

I also have no GPS problems. Locks on within 10 seconds if I'm outside and it's dead on. I haven't played with it inside to see if the GPS assist works.

 

[Mr. GPS]

I also have no GPS problems. Locks on within 10 seconds if I'm outside and it's dead on. I haven't played with it inside to see if the GPS assist works.

Thanks for the report.
A GPS driver needs to check different conditions(signals, error, almanac, ..., ) and use proper procedures to compute the position. For each condition, the procedure/strategy could be diffeernt.

The conditions may be defined differently in each GPS/AGPS system.
AskPCGuy posted a pdf reference for a GPS in this thread yesterday. For Galaxy S phones, you can use LabTestMode to find out possible conditions (settings) Samsung GPS driver uses, but no detailed parameter values of this condition and check sequence information.

Apparantly, Samsung GPS driver failed very badly in certain conditions.
We need to figure out which conditions, why, and any workaround before real driver overhaul.

It seems that this Samsung driver often executes an unnecessary "COLD-COLD-START" (my Task 1 condition) for the first fix. The implemention of this procedure is slow and also not work at all under low signal environment.

 

[Mr. GPS]

Thanks Mr GPS.

Also, to answer my own question about SNR. I had forgotten than on our cell strength meter, "lower negative" is better than "higher negative". Ie: -51 dBm is a better signal than -100 dBM.

Just to provide another data point. My GPS trackes me down to which lane I am in on a 3 lane highway. I have only used my phone in one city.

It will make it easier for GPS driver implementation if GPS hardware can provide higher SNR value. Some post mentioned that Samsung GPS driver has difficulties in the condition of SNR < 20

 

[ulbonado]

At the risk of obfuscating things further, I ran across an app I hadn't seen before yesterday, called TrackerBooster: Tracker Booster - DroidProjects

It runs in the background and constantly triggers GPS requests, which it claims will keep the location data more up to date. The link above shows two tracks, one made with it running and one without, and the one with it definitely looks cleaner and more accurate. He also even mentions using it on a Samsung Galaxy device, although it's not clear if the posted tracks are from a Galaxy.

But anyway, I installed it and ran it yesterday evening on my way home from work, with the My Tracks app recording my track. Results were a bit odd-- TrackerBooster caused the GPS icon in the notification bar to light up, and it actually locked on steadily instead of blinking back and forth like it usually does, which seemed to be an improvement. But the track that My Tracks recorded was no better than normal, wandering off the road I was on and into people's yards and houses pretty frequently. So I don't think this really helped me, but I'm curious why the icon showed it apparently locked, which I normally can't keep for long.

 

[pkoutoul]

I learned more about the workings of GPS and about this particular issue from this thread than anything else I have seen recently. Thanks.

 

[Mr. GPS]

At the risk of obfuscating things further, I ran across an app I hadn't seen before yesterday, called TrackerBooster: Tracker Booster - DroidProjects

It runs in the background and constantly triggers GPS requests, which it claims will keep the location data more up to date. The link above shows two tracks, one made with it running and one without, and the one with it definitely looks cleaner and more accurate. He also even mentions using it on a Samsung Galaxy device, although it's not clear if the posted tracks are from a Galaxy.

But anyway, I installed it and ran it yesterday evening on my way home from work, with the My Tracks app recording my track. Results were a bit odd-- TrackerBooster caused the GPS icon in the notification bar to light up, and it actually locked on steadily instead of blinking back and forth like it usually does, which seemed to be an improvement. But the track that My Tracks recorded was no better than normal, wandering off the road I was on and into people's yards and houses pretty frequently. So I don't think this really helped me, but I'm curious why the icon showed it apparently locked, which I normally can't keep for long.

I plan to post a summary with more complete theory behind by this weekend. So stay tune. For now, my suggestion is
to use better GPS applications to feed the almanac, and it needs multiple times to fill up the missing GPS "nutrition" locally. But thanks for your report, it will help quantify the issues.

There is a quality issue for local almanac. This is related to how to compute the current position by using which satellites information from the local almanac.

 

[NightAngel79]

Gonna move this thread to the Support/troubleshooting section. Any questions just shoot a pm to a mod/guide.
Thanks

 

[askpcguy]

I am suprised that nobody has suggested debugging the gps driver see link Using the Dalvik Debug Monitor | Android Developers

If I get some time I'll give that a go this weekend and will reply back with my findings.

Edit1: If you are going to give the andriod development kit a go and want to try the Dalvik Debug Monitor, ensure you place the ADK into a folder which contains NO SPACES. Also verify your java.exe is on the path in your environment variables. I believe you could copy the java.exe into the \tools\lib folder but I might be wrong.

I fired it up just now and it looks good. Hopefully the logs of the LbsTestMode app trying for a GPS lock won't be millions of lines long.

 

[Mr. GPS]

Thank you for this excellent thread.

So I learned a new trick from this and other GPS threads.

1. Start AT&T Navigator.
2. Let it sit on the Terms of Service page for a few seconds.
3. Click Exit. (Don't agree, don't sign up, don't pay anything.)
4. Use GPS Test, Maps or other apps that need GPS

I get an excellent lock on all of a sudden. It's accurate and stays accurate. Typically 15-30 meters accuracy indoors, sometimes 5-10 meters outdoors or in car.

So that AT&T bloatware is good for something. More proof it's a software fix.

You deserve an award from AT&T for two things:

1. Best apology from AT&T to the Captivate customers and let custmers freely utilize the GPS iinitialization function from AT&T Navigator. Furthermore, AT&T should ask Samasung make steps automatically by adding an option in system level GPS setting: GPS check with automatic AT&T GPS self-initialization in one-click

2 Open a new GPS business model for AT&T. At&T will charge Samsung and only Samsung for each click for turning-on GPS with automatic AT&T GPS self-initialization from all Galaxy S customers. This can push Samsung harder and make custmers less suffer of using current defect GPS and recover the name from Samsung Abused GPS to New AT&T Assisted GPS

 

[bjpjjs]

Mr. GPS,

Can post or sticky a step by step for ytaining the captivate.

thx

 

[Andyman10]

It will make it easier for GPS driver implementation if GPS hardware can provide higher SNR value. Some post mentioned that Samsung GPS driver has difficulties in the condition of SNR < 20

I too have problems with the GPS on my SGS and having worked in communications product development for many years (DSL, 802.11, HSPA, GPS etc) I can see the logic of what Mr Gps is saying, as if the GPS receiver cannot download accurate data from the satellites and process it, then it cannot possibly function correctly..but I

 

[richb500]

Thank you for this excellent thread.

So I learned a new trick from this and other GPS threads.

1. Start AT&T Navigator.
2. Let it sit on the Terms of Service page for a few seconds.
3. Click Exit. (Don't agree, don't sign up, don't pay anything.)
4. Use GPS Test, Maps or other apps that need GPS

I get an excellent lock on all of a sudden. It's accurate and stays accurate. Typically 15-30 meters accuracy indoors, sometimes 5-10 meters outdoors or in car.

So that AT&T bloatware is good for something. More proof it's a software fix.

Now I wish I hadn't removed AT&T Navigator. Can someone post it for me (ATTNav-Samsung-Vega-1488.apk)?

Thanks,
Rich

 

[Mr. GPS]

I