Sunday, April 10, 2011

The benefit of Glonass?

Another study I am carrying out is how much benefit the availability of Glonass satellites really brings. Low-cost dual constellation GPS-Glonass standalone receiver are starting to show up in the market: Skytraq released the S4554GNS-LP and the S2525G2F, and NVS is soon coming out with their NV08C.

Well, I assembled a few S455GNS-LP on a carrier board:

Figure 1: Kopeke10 USB GPS-GLonass receiver
..and drove in the countryside doing some tests.

Figure 2: Test drive with two dual constellation antennas: a non-branded one (20EUR, left) and a Tallysman TW2400 (80EUR, right).
On the left antenna I had:
- a S1315F-RAW
- a S4554GNS-LP

on the right antenna I had:
- a uBlox LEA-5H evaluation kit
- a S1315F-RAW
- a S4554GNS-LP

So that I would make sure to test my receivers with both antennas and compare them against a good reference. Some preliminary results are shown below.

Figure 3. The roundabout is always good to test how much filtering the PVT solution has. Considering that Google Earth might not necessarily be super-accurate, all the receivers perform pretty well.
Figure 4. Driving under a big junction does not affect the receivers very much. Well done!

Figure 5. Uhm? What the heck is that? The Skytraq receiver obviously picked up quite a bit of multipath from the South-East building when I stopped. The good LEA-5H -on the same antenna- did not make a move, and neither the other S4554GNS-LP.

Figure 6. Approching a parking slot just after the road bends, running along a bunch of trees. Multipath has again an interesting different effect on the receivers. But I am more likely to have parked where the uBlox said :)

I will soon add some results in urban environment, where having twice as many satellites should make a real difference.

EDIT 26/08/2011:
Recently, I went out again to compare data from the Sytraq S455GNS-LP and a NVS NV08C-CSM which I soldered onto a carrier board called Denga10 (shown below). The antenna was a low-cost GPS/Glonass magnet-mount patch.

Figure 7. Denga10 carrier board for NV08C-CSM GPS/Glonass receiver.

NVS provides easy software to manage their receivers and translate what is a unconventional NMEA stream:


Figure 8. Navis converter to generate a KML from the GPS/Glonass NMEA output of the NV08C-CSM
Also Skytraq is working hard and recently added a GNS Viewer to their software in order to allow customers to quickly evaluate what is again a non conventional NMEA stream:


Figure 9. Skytraq GNS Viewer.
Some results are shown below. Seems that the NVS tends to filter a bit too much. Luckily it's possible to upgrade the firmware of these receivers and hopefully NVS has already worked on it to improve accuracy.

Figures 10. The NVS receiver tends to filter too much compared to the Skytraq.

EDIT 07/12/2011: NVS released recently a new firmware for their NVS08C-CSM. I will update Denga10 and have another test session very soon!


To be continued (again),

Skytraq s1315f-raw (under the magnifying lens)


Lately I have been working with one of the few mass market receivers which outputs raw observables: the Skytraq S1315F-RAW. The only other "low cost" alternative to it is the uBlox LEA-4T, LEA-5T, and LEA-6T.

My purpose was to assess its capabilities, to understand if there are applications in which this device can really be useful. I don't know the answer yet, but I can publish some preliminary results obtained with RTKLIB.

So, I assembled a few of these on a board together with a FT232RQ UART-to-USB transceiver and run Cutecom on my Linux machine.

Figure 1: Yuan10 20-Hz raw measurements GPS receiver.

To my surprise, this does not work like the uBlox chips, which can output both NMEA and binary. The S1315F-RAW only outputs binary data. The news you don't read anywhere else is that the observations are taken at 20Hz, and do not include a PVT solution. Also, the clock bias is let running free and therefore the measures can be quite far off (tens of milliseconds) the real GPS time.

Ultimately I wanted to know what is the quality of the carrier phase, since this will tell if the receiver is fit for low-cost RTK applications.

Scenario 1: Zero Baseline, Open Sky

Figure 2: Zero baseline track results in 1 hour.

Figure 3: Zero baseline ENU results in 1 hour.

Scenario 2: Short (known) Baseline, Open Sky
To be done still :)

Comparison 1: Orcam GPS-26 OEM (SiRF Star II)
Using the board below I collected RAW observables (pseudorange, carrier phase, Doppler, and C/N0) from two SiRF Star II receivers.

Figure 4: Dual SiRF Star II receiver board assembly
I wrote a simple SiRF binary to Rinex3 parser and the observations look like:

> 2011 04 08 17 23 47.7684297  0  8                                            
G26  50696315.376     1063480.948       91453.281          46.500              
G17  51072993.018     1467528.376       86392.422          43.500              
G27  48803613.805     2145210.883       87538.367          48.000              
G09  48571516.433      925322.182       85910.797          50.000              
G22  51765421.846     1955361.390       86170.438          40.800              
G12  51568741.123      921796.354       84204.570          42.900              
G28  51649721.912     1347484.069       90745.469          42.900              
G15  48654575.927     1362007.686       89023.945          53.800              
> 2011 04 08 17 23 48.7684297  0  8                                            
G26  50713719.228     1154933.716       91452.766          46.000              
G17  51089432.729     1553920.546       86392.164          43.000              
G27  48820272.359     2232748.763       87537.875          47.900              
G09  48587864.077     1011232.602       85910.422          50.400              
G22  51781819.554     2041531.555       86170.164          41.000              
G12  51584765.483     1006000.356       84204.008          42.900              
G28  51666990.473     1438229.023       90744.961          42.900              
G15  48671516.093     1451031.430       89023.742          53.300              
> 2011 04 08 17 23 49.7684297  0  8                                            
G26  50731122.086     1246383.916       91450.203          46.000              
G17  51105869.630     1640310.413       86389.867          43.000              
G27  48836930.232     2320284.139       87535.375          48.000              
G09  48604211.074     1097140.644       85908.039          50.000              
G22  51798218.833     2127699.431       86167.875          40.900              
G12  51600789.059     1090201.716       84201.359          42.700              
G28  51684257.464     1528971.343       90742.320          42.600              
G15  48688455.438     1540052.871       89021.438          53.000              
> 2011 04 08 17 23 50.7684297  0  8                                            
G26  50748523.408     1337830.931       91447.016          46.300              
G17  51122307.805     1726697.388       86386.977          43.100              
G27  48853586.988     2407816.398       87532.266          48.100              
G09  48620558.438     1183045.689       85905.047          50.000              
G22  51814616.209     2213864.420       86164.992          40.800              
G12  51616810.906     1174399.828       84198.109          42.300              
G28  51701524.646     1619710.464       90739.117          43.000              
G15  48705394.801     1629071.432       89018.555          53.000                

but unfortunately RTKLIB refuses to process the raw measurements and I am still trying to work out why!

Anyway, the first results with the S1315F-RAW seem promising enough, showing 2-3 cm errors in Kinematic mode. More will come on this post soon.