u-blox TIM-LP (and TIM) OEM Receivers Tested

In the period from April to August 2004 the Swiss firm u-blox loaned me one and donated me one (!!) of their TIM-LP OEM receivers for evaluation. This receiver is based on the ANTARIS chipset, developed by u-blox and Atmel.It is a 16 channel receiver which delivers raw pseudoranges and integrated carrier phases at up to 10 Hz (!). I ran my standard tests on the receivers, mainly to find out how good the integrated carrier phase is, and if it allows double difference ambiguity estimation and cm level carrier phase processing. In order to assess their relative performance, some tests were run together with a u-blox TIM (SiRF II chipset) OEM receiver (also donated by u-blox and tested earlier),

The following tests have been executed:

  1. In a so-called zero baseline setup the multipath- and noise of the raw C/A code and carrier were determined (see my multipath and noise pages for an explanation of the method).
  2. Two short baseline tests, one in a multipath benign environment and one in a multipath heavy environment, allowed me to gain more insight into the receiver behaviour under favourable conditions and under realistic conditions.
  3. Finally, the receiver tracking performance results in an urban area and in the woods are reported.

Test 1 was done with two TIM-LP's, in test 2 and 3  I used one TIM-LP  and one TIM.


Experiment 1.

In the first experiment, I used Tri-M's Mighty Mouse antenna on a small groundplane (25 cm dia) on top of the roof of my house, with a nearly unobstructed view over the horizon. The antenna was connected to a two-way signal splitter, the splitter RF outputs were connected to the receivers antenna input (the two receivers in a 'zero baseline' or ZBL setup). The receivers generated raw measurement data at 1 Hz in the TIM-LP binary format (RXM-RAW). The data was downloaded to the serial ports of my PC. About three hours of data were recorded on 13 May 2004 and processed.


1.1 C/A code Multipath

The graph above shows the C/A code multipath magnitude of both receivers. The relation between the multipath (1 sigma) value and the satellite elevation is not very strong. The maximum value 3 m, the average value around 1 m. The individual data points of both receivers are very much alike, which shows that the multipath performance  of both receivers is very much alike.
The high multipath values at high elevation show that my antenna location is

  1. not free from multipath effects, and
  2. multipath reflections occur at high elevation angles.


1.2 C/A code noise

There is a clear correlation of noise with elevation. At elevations higher than 20 deg the noise is about 0.1 m, at lower elevations the noise increases to less than 0.4 m. This graph (where multipath is cancelled), and the graph in section 1.1 (with multipath present) demonstrate the importance of multipath free observations.


1.3 C/A carrier noise

In a zero baseline setup the double differenced carrier observations (in cycles) ideally scatter slightly around integer values.For the TIM-LP this is the case, and the scatter (noise) is very low indeed: a value of 0.003 cy (0.6 mm !) at elevations higher than 25 deg, and less than 0.006 cycles at low elevation. This test suggests that the noise level is low enough to allow double difference carrier phase position calculation at the cm-level. The proof is given in section TBS below.


1.4 Positioning results

Raw data has been collected from the TIM-LP's during the zero baseline test from GPS seconds of week (sow) 407275 to 4214588 at a rate of once per second: 14328 epochs. For each epoch the raw data have been processed in a number of modes:

  1. Stand-alone position calculation based on pseudoranges, navigation message ephemeris and ionospheric parameters, and standard atmosphere values for calculation of tropospheric delays. Average deviations (bias) from the true position (known at cm-level) and standard deviations (sigma) have been calculated. The first line of the table below gives the results for one receiver, results of the second receiver are identical.
  2. Single difference position calculation based on pseudoranges. Average deviations (should be zero in a ZBL test) and sigma's have been listed in the second line of the table.
  3. Double difference position calculation based on integrated carrier phases. Biases (should be zero) and sigma's have been listed in the last line.

rcvr mode Lat bias
[m]
Lat sigma
[m]
Lat 95%
[m]
Lon bias
[m]
Lon sigma
[m]
Lon 95%
[m]
Heigh bias
[m]
Height sigma
[m]
Height 95%
[m]
TIM-LP stand-alone 1.0 1.3 3.6 -0.6 1.3 3.2 1.6 3.0 7.6
TIM-LP single difference -0.03 0.19 0.41 0.01 0.12 0.25 -0.03 0.31 0.65
TIM-LP double difference 0.000 0.001 0.002 0.000 0.001 0.002 0.000 0.002 0.004

95% errors in the stand alone mode are 3 to 4 m for the horizontal position components and about twice as much for the vertical component.

Single difference errors are very small compared to the stand alone errors, this is mainly due to the nearly perfect cancellation of multipath in a ZBL test using identical receivers. And also the low pseudorange noise value (section 1.1) plays an important role.

Finally double difference errors are at the level of a few millimeters only, again due to multipath cancellation and low carrier noise.

Click the thumbnails below to inspect the details of the scatterplots. Use your browser 'back' button to return to this page.

It must be emphasized that since multipath errors cancel so well in a ZBL test, this test is is not very realistic. Hence a number of short baseline tests have been executed, see section 2 below. But a ZBL gives a good indication of error sources such as pseudorange noise and carrier noise.


Experiment 2.

A Tri-M Big Brother antenna feeded 4 receivers via a 4-way signal splitter: one u-blox TIM, one u-blox TIM-LP, one Furuno GN-77M, and one Furuno GS-80. The antenna was placed on the rooftop of my car. Data was collected on 4 August 2004 for about 30 minutes at a location without any obstructions, for about one hour at a high multipath location, driving through an urbanized area (certainly no 'urban canyon'), and driving under heavy foliage. Raw data was recorded at a rate of one Hz on two PC's in my car via their serial ports .

A fifth receiver was running at home (location: see section 1 above) and served as reference for single difference and double difference position calculations.

Test results of the u-blox receivers are reported below. The Furuno test results are reported in another page.


2.1  Short baseline, no obstructions.

A location about 4 km away from home without any obstructions was used as 'very low multipath' site. Data was recorded from GPS sow 323161 to 324900: 1740 epochs. The true position was found by double-differencing the TIM-LP carrier phase observation with those of the reference receiver at home. About 15 minutes of data were required to find the correct integer ambiguities. Next, accurate positions (accurate relative to home) were calculated using again the double differenced carrier phase observations.Finally positions were calculated using stand alone and single differenced pseudoranges, these positions were compared with the true position. Results are given in the table below. Since the double difference position served as truth, biases could by definition not be determined.

Single differencing the TIM-LP pseudoranges with the reference receiver data removes the bias components largely, but increases the noise components. This increase is due to the noise of the reference receiver in the single differenced observations being of identical magnitude to the TIM-LP noise. The TIM noise is larger than the TIM-LP noise, single differencing obviously removes noise components which are common to the TIM and the reference at home.
rcvr mode Lat bias
[m]
Lat sigma
[m]
Lat 95%
[m]
Lon bias
[m]
Lon sigma
[m]
Lon 95%
[m]
Heigh bias
[m]
Height sigma
[m]
Height 95%
[m]
TIM-LP standalone -1.4 0.8 3.0 1.0 0.7 2.4 -4.7 1.6 7.9
TIM-LP single difference 0.3 1.7 3.3 -0.5 0.9 2.3 0.2 2.2 4.6
TIM-LP double difference   0.007     0.006     0.023  
TIM standalone 0.6 3.1 6.8 -0.2 0.8 1.8 -2.7 2.5 7.7
TIM single difference -0.1 2.2 4.6 -0.4 1.0 2.4 -1.5 2.2 5.9

Cllicking on the scatterplots below allows a more detailed inspection. Striking is the fact that the horizontal scatter plots show two 'clouds' of points, which is most prominent in the upper left plot. Also a jump in position can be seen in the height time histories (the lower left and lower middle plot at about 324100). The jumps concur with prn 4 entering the calculation, but I can not explain why this jump happens. Anybody ??


2.2 Heavy multipath

I parked my car about 5 m in front of a metal warehouse with a height of about 15 m. The warehouse acted as a nearly perfect 'mirror' for the satellite signals, maximizing the detrimental effects of multipath reflections. I logged raw data from both receivers from GPS tow 325140 to 328800 (3660 epochs) at a rate of once per second.  Unfortunately I have not been able to calculate the true (double difference carrier) position, so the calculations are compared to a best guess truth, beleived to be accurate to 2 m horizontal and 1 m vertical. But the test results below compared to the results reported in section 2.1 show clearly that position accuracy decreases dramatically, by a factor of about 3, at a high multipath location. This is true for both the TIM-LP and the TIM.

rcvr mode Lat bias
[m]
Lat sigma
[m]
Lat 95%
[m]
Lon bias
[m]
Lon sigma
[m]
Lon 95%
[m]
Heigh bias
[m]
Height sigma
[m]
Height 95%
[m]
TIM-LP standalone -2.0 5.4 12.8 -0.9 3.0 6.9 -4.6 6.9 18.6
TIM standalone -2.7 8.0 18.7 -2.0 4.6 11.2 -3.5 13.1 29.7

The scatterplots also show large position excursions due to multipath reflections. The rightmost plot shows multipath errors on prn 30 during 1000 seconds of both receivers. The curves are deliberately offset to show the differences, hence the absolute value of the multipath errors is meaningless. But peak-peak values are up to 16 m for the TIM-LP and close to 20 m for the TIM. It is also noteworthy that the shape of the error  plots is very similar, and that the period of the multipath is around 110 seconds (a little less than 2 minutes), which is a very typical value for multipathe errors.

The effect of pseudorange multipath errors on position accuracy is well visible in both height time histories (left middle and right middle plot).


3 Tracking performance in urban area and under heavy foliage

These tests were also performed on 4 August 2004, and preceeded the tests reprted in section 2. Hence the same instrumentation setup applies.

The main purpose of these test is to compare the satellite tracking capabilities of the TIM-LP with the TIM. I tried to find an environment where the receiver tracking loops are challenged: in an area with much high buildings, and under heavy foliage. It is ephasized that for this analysis only the raw data (pseudoranges, integrated carirrier phases) was used, and not the position solution provided by the receivers (the receiver calculated position was also not used in the above described tests).


3.1 Urban area

Five circuits were driven in an urban area (certainly no urban canyon, closely spaced sky scrapers are not present in the area where I live) from GPS tow 321230 to 321720 (491 epochs at a rate of one per second). The leftmost figure shows the number of epochs with 0, 1, 2, 3, 4, ... , 12 satellites in track, for the TIM-LP and the TIM. It is clear that both receivers always tracked 4 or more satellites, so a standalone position calculation could always be performed. The TIM-LP seems to perform slightly better than the TIM. This might be due to the fact that the TIM-LP is a 16 channel receiver with sufficient spare channels to track EGNOS/ WAAS satellites, while the TIM has to share its 12 channels to both GPS and EGNOS/ WAAS satellites.

Standalone positions based on raw pseudoranges were overlaid on a map (map to still be added) of the area, for both TIM-LP and TIM. Note the smaller deviations from the road of the TIM-LP w.r.t. the TIM.


3.2 Heavy foliage

Following the urban area test five circuits were driven under rather heavy foliage from 321900 to 322440 (540 epochs at one update per second). Tracking statistice are given ion the table.. Again 4 or more sats were tracked all the time. And again the TIM-LP performed slightly better than the TIM.

Scatter plots show identical behaviour as in section 2.3 (map to still be added) .


3. Conclusion

Tests were carried out

It is concluded that:

  1. Pseudorange (code) and carrier noise of the TIM-LP are low and only weakly depend on satellite elevation. Double differenced carrier phase observations converge well to integer values in a zero baseline test.
  2. Positions calculated with raw pseudoranges are accurate to about 3 m horizontal and 6 m vertical (95%) at low multipath locations, and are about three times as high at a high multipath location.
  3. High accuracy relative positioning using double differnced carrier phase observations is possible at low multipath locations, about 15 minutes of data was required to resolve the integer ambiguities. horizontal accuracy is at the 1 cm level, vertical accuracy about 2 cm (95%). High accuracy relative positioning for the TIM is not feasible, see the TIM testpage.
  4. Positioning results using raw pseudoranges of the TIM-LP outperforms results of the TIM by a factor of 1 to 2. Positon scatter during the urban area test and the heavy foliage test is signifivcantly less for the TIM-LP compared to the TIM.
  5. Satellite tracking performance of the TIM-LP is slightly better that the TIM, possibly due to the increased number of channels.

Again my thanks go to the people at u-blox, who enabled me to play with their exciting products.

Back to index