Tuesday, December 18, 2012

Fruit-Bat Tracking with Robin GPS Logger

I have recently received the results of an interesting experiment done with Robin, Cellguide's snapshot GPS tracker so I thought I should share it with the Community.

Figure 1: Yes, a fruit bat
Becuase of its size and weight Robin enables new applications of GPS technology, including wildlife tracking of course. Particularly of species on which was very difficult in the past.

Test Setup 
  • CellGuide Robin GPS Logger with 100mAh battery (2.2 grams): total weight 4.75 grams
  • Programmed to operate every day between 5pm to 5am
  • Total length of the experiment: 72 hours
  • Operating regime:
    • Fix every 3 seconds
    • Medium sensitivity (GPS snapshot length of 128ms)
The test was conducted by researches from Tel Aviv University. They packed Robin in a light weight plastic wrap along with a UHF beacon device, and attached everything to the bat using a special medical glue that only sticks for a few days.

Figure 2: A picture of the fruit bat with Robin on the back.
Figure 3: The test area.
During the first night, the bat visited three other trees along its path. On the second night, the bat took a bit of a different path but still quite similar to the one from the previous night and flew to the same tree. Then, pretty much as the two previous nights, the bat flew to his favorite tree again.

 Figure 4: From left to right: first, second, and third night

The Robin logger was found two weeks later inside the cave, using a UHF beacon device. A summary of the experiment is given below.

Figure 5: All nights on the same map.
Despite this experiment being simply very cool for a GPS passionate as I am, I shall probably mention here that I don't work for Cell-guide so for further information on the Robin GPS Logger please do contact the experts!


Friday, December 14, 2012

Full Galileo IOV

After a long wait, it was a pleasure this morning to record the full IOV constellation.

Figure 1: Orbitron prediction of Galileo visibility this morning.
The NV08C-CSM receiver with a R&D firmware could track all 4 IOV Galileo satellites, as well as another 21 SVs between GPS, Glonass, and EGNOS.

Figure 2: NVS Storegis recording NMEA with all 4 IOV satellites.
An excerpt of the log file:


I also turned on my SdrNav00 modified with a 26MHz oscillator and obtained the following:
Figure 3: SdrNav00 signal properties.
Figure 4: SdrNav00 signal power spectrum

Figure 5: Galileo acquisiton on E1b signal.

Galileo pseudoranges were logged by NV08C-CSM and by my SdrNav00, so the next step is to try PVT with the TUM orbits.

Tuesday, August 21, 2012

The GPS tracker to beat (smallest, lightest, and most durable)

Figure 0: Please read below.
From the very early days of GPS snapshot technology has been proposed as a way of reducing power consumption and footprint of a radio positioning device. Some papers about NASA microGPS date back to late ‘90ies:

Later on many Companies tried to patent this technology and bring it to the market. Most of them have been acquired or have merged - see the latest acquisition of CSR “mobile business” (whatever that means!) from Samsung.
To our knowledge until today all development efforts led to products with disputable market readiness, if otherwise mostly devoted to picture geo-tagging.
Instead, the assumption that a snapshot approach could beat (from the cost and power consumption point of view) a hardware baseband has always been contradicted by Moore’s law and market demand. Lead manufacturers now release complete GPS engines on a chip smaller than what was before the RF Front-End part only. Compare e.g. the CSR GSD4e (3.5 x 3.2 x 0.6mm 42-ball WLCSP) and the Skyworks SE4150L (4 x 4 x 0.9 mm 24-pin QFN). That is probably why the complexity associated to a snapshot positioning approach has never won over a slightly more expensive and power consuming standard GPS receiver.
Probably the real value of a snapshot approach has not been identified yet and in our opinion is to be searched in other domains such as geoencryption and GNSS authentication:

So we longed for a Company that would own the technological skills to provide a new standard in terms of snapshot positioning. 
Cellguide has been in the business several years now but only recently released a real hardware module – simple to use and available for sale – that works with a snapshot positioning approach and has specifications that could dramatically change the tracking business: the Robin.

We bought an evaluation kit (expensive, but breaks the ice) and we are testing it together with other 
enthusiastic engineers. 

A Robin device on its cradle (used for charging, programming, and downloading the snapshots on the PC) looks like this:

Figure 1: Robin device on its cradle
 A couple of Robin devices next to each other:

Figure 2: Two robin devices next to each other

The Robin brochure says already a lot, but I wanted to show the results of a real test (on the beach – it is summer after all).
The reference path was roughly this one:

Figure 3: Approximate reference path
The little wire antenna in the pictures above was used (which is terrible compared to a patch antenna of course). 
Please note that each snapshot is - by design - completely independent from the others.

64ms snapshots (at 1Hz) lead to the following positioning results:

Figure 4a: 64ms snapshots, raw
Figure 4b: 64ms snapshots at 1Hz, Kalman filtered
Figure 4c: 64ms snapshots at 0.2Hz, Kalman filtered
The same was test was also reproduced by taking 512ms snapshots (higher sensitivity) at 1Hz.

Figure 5a: 512ms snapshots at 1Hz, raw
Figure 5b: 512ms snapshots at 1Hz, Kalman filtered
Figure 5c: 512ms snapshots at 0.2Hz, Kalman filtered
 Results can be summarised in the following:

Figure 6a: 64ms snapshots, raw (red) and filtered (green)
Figure 6b: 512ms snapshots, raw (red) and filtered (green)
Figure 6c: filtered output for 64ms (diamonds) and 512 (circles)

Robin was born at Cellguide for wildlife tracking but as we came across it we could not help thinking it shall be employed in lots of other cases as well. The Aclys chip (earth of Robin) has a low data rate, low pin count SPI bus compatible even with simple 16bit micro-controllers. It can therefore be embedded in very small and cost effective devices to enable tracking capability where it was not possible before.
We see potential in this technique and we hope you enjoyed reading as as much as we did running the tests!


P.S.: For further information please contact directly the Cell-guide guys at info@cell-guide.com

Tuesday, April 10, 2012

Spring news in the GNSS and SDR domain

I have been following in the last few days interesting developments in the GNSS and SDR domain.

1. NV08C-CSM and dual constellation RAW measurements

It's recent news that NVS has lifted the constraints on the firmware with RAW data for GPS and Glonass on L1. In my opinion this is one of those rare times that the rules of the game are changed by one of its players.
Essentially the NV08C-CSM provides real-time kinematic capability at 10Hz, with a price point as low as 35EUR/piece in small quantities. It is already possible to download an unofficial version of RTKNavi.exe and RTKConv.exe here, but RTKLIB will officially support the receiver from next version anyway.
Below are some pics of what Denga10 and the navXperience 3G+C could do with the good old 2.4.1 "Static Precise Point Positioning" over three hours.. bringing down the error to less than 20cm in complete standalone mode (by using only broadcast products).

Rtknavi (NVS mod) doing static PPP with 14 sats

Rtknavi (NVS mod) available satellite close-up

Glonass broadcast navigation data

GPS broadcast navigation data

GPS/Glonass observations.. 21!

Satellites used in the fix

From cold start, first three hours, ground track
From cold start, first three hours, position

From cold start, third hour, ground track

From cold start, third hour, position

The likes of Novatel, Trimble, Hemisphere, etc are not going to like it I suppose. On the other hand, users have perhaps a cheaper option to enter the high-precision domain.


For once, the group of innovators is European.
These guys of osmocom are smart and -what's even better- their work with the open source team. RTL-SDR is, I believe, one of the freshest finds in the Software Defined Radio domain. The Fun Cube Dongle, developed in UK, is already a very clever tool. But finding the super-tuner Elonics E4000 (note, again a UK Company) in 25$ USB DVB-T dongles and being able to grab data in a 3MHz bandwidth between 64MHz and 1.7GHz... that is really cool.

I had to buy one! ...actually two before I found the E4000..
A Terratec Cinergy T Stick Black rev.1, with the FC0012 which is not suitable for GPS
A Newsky TV28T, with the E4000.

The Realtek RTL2832U demodulator uses a 28.8MHz crystal by default.. which has too poor accuracy for a GPS receiver.. but a SDR acquisition algorithm can easily handle 100+kHz of apparent Doppler shift in the frequency search :)
So for a nominal sky plot as the one below:

And a signal looking like as follows:

Power spectrum around GPS L1, FS=2.728MHz
The ADC is saturated.. the gain I set was too high.

I could acquire the following birds:

SV 9: Doppler +112000.0 CodeShift:     99 xcorr: 553566.6
SV12: Doppler +113500.0 CodeShift:    502 xcorr: 301126.3
SV15: Doppler +110000.0 CodeShift:   1847 xcorr: 871305.9
SV17: Doppler +110500.0 CodeShift:   2225 xcorr: 401100.6
SV18: Doppler +110000.0 CodeShift:     38 xcorr: 368766.1
SV26: Doppler +107000.0 CodeShift:   1779 xcorr: 511844.4
SV27: Doppler +110500.0 CodeShift:    264 xcorr: 650033.9
SV28: Doppler +107500.0 CodeShift:    203 xcorr: 320405.6

The next step is to have my new version of GPS-SDR processing the data in real-time.. won't be long!
I managed to upload the binary file which is long enough for everyone with a SDR receiver to calculate the position. I used FS=2.048MHz and a nominal IF of 0.0Hz (but because of the crystal inaccuracy the IF actually falls at about 110kHz). The data type is I&Q interleaved int8_t.


3. Open source FPGA receivers

I waited too long for the code from the University of Tampere: they were promising to licence the TUTGNSS and never happened.
Finally the Open Source community is bringing to the reality an Open Source FPGA implementation of a GPS receiver based on the Namuru / Zarlink GP2021 correlator structure and soft CPU cores (the LatticeMico32 Milkymist and the Altera NiosIIe). As also scientists at the University of Tokio (remember RTKLIB?) are involved, this time I know it is going to happen.
Some useful links:

Stay tuned, GNSS is evolving rapidly!

Monday, February 20, 2012

Low cost RTK performance round-up

Having done quite a bit work since the first Yuan10 was built, I think it's time to write a small summary.

I recently put up a navXperience 3G+C antenna that will serve as reference for all future RTK work in this area. With the support of the reference station University of Pisa and their reference observations at 0.2Hz I run RTKLIB (RTKNAVI and RTKPOST) for several days using several low-cost receivers that I assembled.
This should be somehow considered the best possible scenario, having a short 1.5Km baseline.. nevertheless the results are quite exciting I believe.

Figure 1: Yuan10: Skytraq S1315F-RAW

The plots above show solving of the integer ambiguities in about 5 minutes and reaching an accuracy of about 4cm square.

Figure 2: Rappen10 mMCX version: uBlox NEO-6P (RTK mode)

Again, the NEO-6P brings the position down to 4cm square accuracy in less than 10 minutes.

Figure 3: Denga10: NVS NV08C-CSM

The NV08C-CSM solves the carrier phase integer ambiguities and also brings down the accuracy to about 4cm square.

Each of these receivers has his specialty:
  • the S1315F-RAW has 20Hz raw measurements update rate, which is still unmatched performance by other low-cost receivers
  • the uBlox NEO-6P has integrated PPP for high accuracy standalone mode, no other low-cost receivers claims that
  • the NVS NV08C-CSM has Glonass!

P.S.: I attached a trimmed version of the data used to obtain these results here. Please note how the NVS observations are covered by NDA and therefore not included.

P.P.S.: The latest FW from NVS (dated 27 Feb 2012) improves dramatically the quality of the NV08C-CSM raw measurements: results above are not representative anymore.