STARPAL DEALER BULLETIN

December 18, 2005
Subject: Thales MobileMapper CE and Sub-meter GPS with WAAS/EGNOS
Version Info – Windows CE v4.2, OSFW 2.59, EBoot 1.17 GPSFW E021 GPSDataServer 1.2.0.3
[Thales has announced a firmware upgrade to fix some GPS problems listed in this report.  We would expect some improvement only in multi-path rejection over the version we tested.]

StarPal has tested the Thales MobileMapper CE and Sub-meter GPS with WAAS/EGNOS (herein called the MMCE) and is publishing the results attached. Please share this bulletin with your StarPal customers who are considering purchase of the Thales MobileMapper CE.

On the Plus side
The unit has a nice mechanical design. It is rugged and waterproof.  It has a nice feel. Mechanically, its ruggedness is comparable to the TDS Recon and Juniper Systems Allegro; the only other Windows CE computers in its class.  It is claimed to be able to drop 1.5 meters (4.9 feet) although we did not try this test. The display is comparable to other color Pocket PC computers. We were excited to test this unit..

Running the MMCE with HGIS
Users should upgrade to at least HGIS version 7.33 when running the MMCE. HGIS version 7.33 and later will work with the internal GPS. HGIS Version 8 and later has addressed more of the issues in the MMCE.  These tests were conducted with HGIS version 8.

Best Case - Long Term Stationary Test
Under perfect conditions, a stationary MMCE can obtain submeter accuracy (repeatability) after about 6-20 minutes.  The plot below shows an 84 hour test with the antenna 20 feet in the air and a clear view of the horizon all around. The unit was in the same location for the entire time. The unit was tested with WAAS (blue dots) and two different RTCM-104 radiobeacons (Station 307 is represented by red/yellow dots and station 310 by green dots)   The radiobeacons were both the same distance from the test location.  You will note that the three differential correction sources gave results that differed from each other by at least one meter.  Always expect a shift whenever you change sources of differential correction as is shown in these plots.  A cold front came through and the sky was overcast which is probably the reason the part of the test using Station 310 didn't track as many satellites or have as tight a pattern as the other two.  With this radiobeacon, the MMCE was only repeatable to 1.26 meters.  A clear sky will improve your accuracy.  The MMCE must be allowed to stabilize for 20 minutes before the test and must not be powered down at any time during the test. The secret to achieving the below patterns is that the MMCE must never fall below tracking of 6 satellites (what happens is discussed below). This is why the antenna needs to be raised above any ground noise clutter. These tests show that the antenna needs to be high enough to track at least 7.0 satellites on average (6 satellites minimum) to have some confidence that you are obtaining submeter results with this unit. A clear sky is probably also necessary to achieve sub-meter results. And obviously, when you return to any site, you must use the same differential correction source as you previously used.




Multi-path Worst Case Test
What results can you expect in the real world in a normal environment with interference, objects blocking the view, GPS antennas only 1.5 meters (4.9 feet) above the ground?  Your results may look more like the plot below.  In the plot below, red dots are fixes that do not meet the MMCE requirements to calculate sub-meter (in most cases the number of satellites is 5).  It is understandable that the red dots are outside the 1 meter radius circle.  Green dots are fixes that do meet the MMCE requirements for sub-meter (6 satellites, PDOP < 4). It is therefore less obvious why the green dots are outside the circle.  But, if we look at the time of each of the green dots, we note the trails of green dots are leading toward the red one meter radius circle. We call the green dots outside the 1 meter circle "Trail-Ins" because the MMCE is trying to reject interference, but does not have enough samples to calculate sub-meter position yet.   These trail-ins are situations where the GPS locked onto a multi-path reflection and it takes a few minutes(up to 20) for the GPS to figure this out. Note that these results are still stationary, the GPS never moved. 




The first thing that reduces multipath errors is the antenna element design. The MMCE uses a very small antenna element that has poor multipath rejection capabilities in itself.  However, the MMCE's multipath rejection capabilites are similar to other antennas of similar size. And the MMCE is much better than any GPS Receivers costing less than US$500.  Raising the antenna above the noise clutter of vehicles and implements reduces multipath signal strength, increases view, and improves accuracy.  Tracking more satellites also helps. A user reading the MMCE screen may block up to 25% of the sky.  This in itself is enough to reduce the accuracy of the MMCE by 2-3 times.

The MMCE has the capability of knowing that it is stationary. If it is tracking 6 or more satellites, it can reduce multi-path errors over time.  The first fix will be in the 3-5 meter range, and then subsequent fixes will improve on that accuracy.  This is called a "Trail-In" because it creates a trail in the plots above as the fixes approach the actual location.  This mechanism will not usually work with a user holding the unit in normal viewing position without an external antenna because it cannot track enough satellites. The unit needs an external antenna that has a clear view of the sky down to 15 degrees(the MMCE default) above the horizon all the way around. If the unit is only receiving five satellites it will not make any improvements.  You should mount the MMCE (or external antenna) as high as possible.  The MMCE restarts this process whenever it is moved; GPS accuracy goes back to the 3-5 meter range and the time starts again.  The accuracy reported by the MMCE can improve reaching the residuals shown in the best case after 15 - 30 minutes.  This capability may be useful for users willing to sit on station for a period of time.    The table below shows the "Trail-In" time to obtain sub-meter accuracy measured in minutes on station:

Time on station	Horiz Error (#satellites = 6)	Horiz Error (#satellites = 7)
1 minute	2.16 meter	2.25 meter
2 minute	2.05 meter	1.93 meter
3 minute	1.93 meter	1.51 meter
4 minute	1.82 meter	1.30 meter
5 minute	1.65 meter	1.07 meter
6 minute	1.51 meter	0.95 metere
7 minute	1.37 meter	0.80 meter
8 minute	1.27 meter	0.81 meter
9 minute	1.19 meter	0.83 meter
10 minute	1.12 meter	0.85 meter
11 minute	1.05 meter	0.88 meter
12 minute	0.94 meter	0.86 meter

Multipath Reception While Moving
While moving, the MMCE accuracy (repeatability) improves somewhat. This is because multi-path errors are more sporadic while moving and the MMCE can reject these errors more easily. If the MMCE is tracking 7 or more satellites, and moving, you can achieve a 1-2 meter repeatability. However, when you stop, the MMCE seems to get confused and goes immediately to a 3-5 meter range accuracy.[There may be some improvement in the newer GPS firmware release in this area.]

Absolute Worst Case
Occasionally, any GPS reports a fix that is just wrong.  Maybe once in one thousand calculations. GPS manufacturers leave these anomalies out of their specifications.  Anomolies that we have seen with the current generation GPS receivers from various manufacturers tend to be short term errors in the range of 25-40 meters once in awhile.  The MMCE generated three fixes (at different times) with errors in the range of 90-280 meters.  Five years ago this would have been acceptable, however this seems large by today's standards. 

Real World Users
We will discuss three different types of users in the real world.
Kinematic means that the GPS is always moving. Yield monitoring would be an example of a Kinematic User. 
A kinematic user with a 100% view of the sky may expect this unit to be within about 2 meters 95% of the time.
A kinematic user viewing the screen (in other words with a 75% view of the sky) may expect this unit to be within about 3-4.5 meters 95% of the time.

Stop-And-Go means that the user stops for a moment to take a reading and then continues on.  Soil sampling would be an example of a Stop-And-Go User.
A Stop-And-Go user drawing points (stops for about 1 minute)  should expect to be within about 2.5 meteres 95% of the time.
The GPS must be stationary with a 100% view of the sky for the entire period stopped; with the sample taken at the end of the period.
The antenna should be mounted on a tripod or vehicle.
The MMCE would only average each point for the period of time actually stopped.

A Stop-And-Wait User might go to a site and have to wait for a machine to dig a hole or do some operation (or simply wait on the GPS)..
A Stop-And-Wait user (that stops for 20 minutes tracking 6 satellites and with a 100% view of the sky) should expect to be within 1 meter 98%-99% of the time.
Note that this measurement is only valid if the user returns to the site with the same differential correction source.
The sub-meter capabilities of the MMCE would be most suitable to a Stop-And-Wait user.

WAAS Receiver Sensitivity
The WAAS receiver in the MMCE has inadequate sensitivity for about one-half of the locations in the United States (in particular it will not work as well in the Northwest United States). This is independent of the above considerations of GPS accuracy.  In most locations, the MMCE will lose track of the WAAS satellite on a regular basis.  The WAAS satellite is directly above the equator and very low on the horizon. At our test location, the WAAS satellite elevation angle is 17 degrees. It can be interfered with by hills, buildings, trees, power lines, fences, cars, 2-way radios, etc. The WAAS satellite is particularly suseptable to multi-path intereference. You cannot place your head, body, arms, or hands between this unit and the WAAS satellite or the MMCE will lose track.  Whenever the MMCE loses WAAS, the GPS accuracy will become significantly worse.  You can expect the accuracy during these times to be within about 7 meters (20 feet) about two-thirds of the time. If you collect data using HGIS, you will see jumps in your data maybe up to 30 feet when the GPS gains or loses track of the WAAS satellite. If you set the HGIS GPS Alert Screen for “Any Differential Fix”, HGIS will alert you whenever the MMCE loses WAAS. Then you should check to see if anything is between the MMCE and the WAAS satellite (such as yourself, another person, or object) and fix the problem if you can. Sometimes the MMCE will lose the WAAS satellite and nothing is in the way(this is usually caused by multi-path intereference); then you wait.  In this case, the MMCE usually regains WAAS within one minute or two.

Differential Radio Beacon (RTCM-104)
Our recommendation is that if you are in an area where a Differential Radio Beacon is available you should use a beacon receiver. Coverage is available over most of the United States as well as near the coastline over much of the world. A beacon receiver can be connected to COM1 to supply corrections to the GPS (instead of WAAS).  There are also Bluetooth beacon receivers available which only need to be near the unit (our experience using Bluetooth around metal vehicles, is that the two units need to be within a couple of meters and in view of each other).  A WAAS receiver works only in line of sight with the satellite; however radiobeacon signals follow the ground (like AM radio signals) so they can be received even though the beacon transmitter may be behind hills.
 
Accuracy versus Repeatability
“Repeatability” means you are at the same location you were before. “Accuracy” means you can get to the same location by completely different measuring means (or differential correction signal).  Surveyors demand accuracy, but most of StarPal customers want repeatability which is usually cheaper.   Most places, in particular more northerly latitudes, beacon receivers work a significantly higher percentage of the time than WAAS receivers like the one in this unit.  We feel that the repeatability of using a single radiobeacon is simpler than trying to use multiple differential sources and trying to take  accuracy of different systems into account.

Serial Ports
Neither serial port meets industry standards. Microsoft ActiveSync is not supported by the serial ports.  The commands SETCOMMBREAK and CLEARCOMMBREAK do not function on COM1.  Therefore, Trimble GPS receivers cannot be fully supported by this unit. Also, opening COM1 twice without powering down the computer tends to freeze the computer. Outputting NMEA on COM1 has on certain occasions after a period of time caused the GPS to stop responding on COM1 and COM2 (did not freeze the computer and GPS still navigated). This happens even when HGIS is not running.  COM2 is the internal GPS interface.  HGIS versions prior to version 7.33 will not work with COM2 (the internal GPS).

RAM Memory
Unlike the HP iPaq and most other Windows CE computers, this unit will erase RAM memory when powered down or reset. Thales recommends that you only store very small files in RAM memory, for example your “My Documents” folder. When the unit is powered down, it tries to backup RAM files into flash memory.  If there are any errors (see flash errors below), the computer discards your files without error message or option to cancel. If you need the faster speed of RAM and thus save your files to RAM, be aware of this and do not “Shut Down” the computer (you can use Suspend).   Connect to ActiveSync and offload your files to a Laptop before shutting down the computer. Sometimes the computer may appear to freeze due to unknown system tasks (possibly some system operation or flash memory failures) but will probably come back within a minute or two. This problem may be worse if there are large files in RAM or flash memory. Other users may have also seen this problem with large numbers of files or folders in the flash memory.  Be warned that if you reset the computer, your RAM files are gone.  None of the Windows Mobile 5 computers we have tested have any of these problems.

Flash Memory Performance
The unit was tested with the supplied Thales 32 MByte flash SD memory and with SanDisk Ultra II flash memory (512 MByte). The Thales 32 MByte SD cards supplied with the unit are error prone when writing files.  HGIS versions before version 8 do not correctly report all the failures which occur using the Thales SD cards or the internal flash memory.  Thales Product Manager reports that SanDisk cards are recommended for the unit.  We had better results with the “SanDisk Ultra II SD” cards and recommend the use of Ultra II cards instead of the supplied Thales SD cards. Flash memory is painfully slow, however there is a performance improvement using the Ultra II cards.

File Write Performance to Flash
Write speeds to flash memory (internal and plug-in SD card) varied from about 1850 Bytes per second up to 11K Bytes per second depending upon conditions and the various cards. The Ultra II cards are designed to write at 9 MBytes per second. The actual write speed is about 900 times slower than the Ultra II SD card’s speed in your digital camera.  

Saving a 11 MByte file to memory takes the following time:
MMCE + RAM = 1.3 minutes (This is the HGIS CPU processing time.)
MMCE + SanDisk Ultra II SD card = 18 minutes.
MMCE + Internal Flash = 29 minutes
MMCE + Thales 32 MByte SD card = 37-42 minutes
(Comparison) HP iPaq 2795 + SanDisk Ultra II SD card =  2.3 minutes
(Comparison) HP iPaq 2795 + SanDisk Ultra II CF card =  1.8 minutes
(Comparison) Juniper Allegro CE + SanDisk Ultra II CF card = 1.6 minutes 
There is a clear problem saving files to flash memory in the MMCE.

SensorTrack Data Acquisition
SensorTrack was tested logging sensors to the Ultra II SD card. Geophysical sensors were connected to COM1 and to the EdgePort USB serial adaptor which provided COM3 and COM4.  Because of the slow response of Flash memory, SensorTrack was only able to log twenty-one records per second (40 byte records) to the Ultra II SD card. Sensor data was randomly delayed up to one-half second while the system was doing other tasks (possibly GPS related system activity as well as SD flash memory activity).

SensorTrack was also tested logging sensor data to RAM memory (see precaution about RAM above). The same configuration was used as in the test with the SD card. The MMCE USB port / System CPU limited the transfer rate to about 4Kbytes per second or about 100 records per second (40 byte records). The maximum usable baud rate for the Serial ports without data loss was 19200 baud.  Sensor data was randomly delayed by 0.3 – 0.5 seconds while the system was doing other tasks (possibly GPS related system activity). The delays were not quite as bad as the earlier test with SD memory, but still unacceptable for sub-second data acquisition.

These random system delays in supplying serial data from the serial ports to SensorTrack when writing to either RAM or flash memory will cause geographic interpolation errors in sub-second sensor data.  Users should not log a single sensor faster than one record per second using this unit. Users should limit their system to one Edgeport module with no more than 8 sensors (each sensor can log one record per second maximum).

USB Port
The unit has Master and Slave USB ports.  Both USB Ports appear to meet only the “Low Speed USB” specification of 1.5Mbit per second. This limits data transfer rates on both ports and was possibly the cause of the baud rate limitations on the EdgePort USB serial adaptor.

Keyboard
The alphanumeric keyboard is a little tricky to enter text, and we prefer the on-screen keyboard which is also available.

Windows CE
The MobileMapper CE usesWindows CE version 4.2(also called PocketPC 2004 or Windows CE.net 4.2). This version of Windows CE was released in 2004.  Windows CE version 5 was released June 2005 and contains numerous improvments. We highly recommend Windows Mobile 5.

Other Reports on the MMCE
Here is a link to another report on Mobile Mapper CE tests.
Trimble White Paper on Mobile Mapper CE


Conclusions
This unit may meet certain user’s needs. If you are creating files containing less than 100 objects per file (points, lines, boundaries, or text), then the flash memory speed should not be a serious problem.  If you are willing to connect an external antenna, and sit on station 6-21 minutes, you can obtain sub-meter accuracy.   If you are using a slow geophysical sensor (such as a penetrometer) that requires you to be on a station for multiple minutes to take a sample, then the GPS accuracy improvements will start to take effect.  This unit is not recommended for high speed data acquisition with our SensorTrack product.

For high speed data acquisition, one option would be Pentium IV Laptop computers with the Edgeport USB Serial module(2, 4, or 8 port).  These systems can collect up to 500 records per second spread across up to 8 serial ports with random data delays(jitter) on the order of 0.1 seconds.

The TDS Recon, Juniper Systems Allegro, or other Handheld computers with Windows Mobile 5 may also be used with the 232BSS4 4-port serial Multiplexer to log up to 3 Geophysical Sensors (or more if they have spare Serial COM ports).For Data Acquisition of single geophysical sensors (dual sensors if used with a Bluetooth GPS), you may consider the Juniper Systems Allegro computer series which has two high speed serial ports built-in. The TDS Recon with built-in GPS can also be used for data acquisition through its external serial port.  HGIS will run with Windows CE computers Version 3.0 and later (the most common versions of Windows CE were 3.0, 4.2, and 5.0/5.1) called by various names: PocketPC 2003, Windows CE.net, and Windows Mobile.

I hope this information aids in your decisions regarding the Thales MobileMapper CE.

Richard Herrington
http://www.starpal.com