Originally Posted by valkyriesRide

It would be nice but I don't have compatible receivers and my drone can't carry the weight of a windows PC

If I'm reading this correctly, these guys implemented it on an intel edison.

http://www.emlid.com/reach/

Originally Posted by valkyriesRide

It would be nice but I don't have compatible receivers and my drone can't carry the weight of a windows PC

rtklib runs on a raspberry pi

all ublox 5/6/7/8 modules internally read code and carrier, much of the information is available through debug messages @ 1Hz . M8N's latest firmware 3.01 breaks this, just downgrade to 2.01.

http://wiki.openstreetmap.org/wiki/UbloxRAW
http://gpspp.sakura.ne.jp/paper2005/...2014_ttaka.pdf

navispark raw is $80 without antenna, which is only slightly more expensive than a m8n and much cheaper than a ublox M8T. The Skytraq Venus 8 module cranks out at 20Hz, which beats the 5Hz you get from ublox modules with unlocked (P/T) firmware.

Hi, thanks for those helpful comments. I'm starting see why the two GPS didn't agree. So now I think it will be interesting to see what results can be obtained using RTKLIB with these (possibly fake) receivers and raspberry pi zero.

I've enabled the raw data messages on one of the receivers but the other receiver won't take the message, I'm ordering a new one so I'll post my findings when that arrives.

Thanks for giving it a try! I followed the guide here https://www.dropbox.com/s/chki3u19u9...o-M8N.pdf?dl=1 which was linked in the thread http://gpsforum.geospector.de/phpbb/...=3503&start=30

edit: ntrip data sources http://gpsworld.com/finally-a-list-o...ns-in-the-u-s/
static PPP solution on the web http://www.ngs.noaa.gov/OPUS/

Originally Posted by sesat

I followed the guide here https://www.dropbox.com/s/chki3u19u9...o-M8N.pdf?dl=1

This guide has been really useful. On the Raspberry Pi the CUI programs have to be compiled from source in GCC, then everything is run and configure using command line, the windows GUI on the other-hand runs with a download and a double click, so this is turning out to be harder than I thought.

I was able to compile all the tools on Raspbian Jessie by just running the makeall.sh script in the /RTKLIB/app/ directory. A working config file from the GUI can be saved and edited for use on the pi (renaming serial devices, etc).

Originally Posted by sesat

I was able to compile all the tools on Raspbian Jessie by just running the makeall.sh script in the /RTKLIB/app/ directory. A working config file from the GUI can be saved and edited for use on the pi (renaming serial devices, etc).

Good to hear it went so smoothly, I was caught-out several times setting this up.

I'm using the Pi's UART on the GPIO to connect to the GPS receiver so that meant disabling the serial console login and stopping and disabling the tty service. My Pi is headless so I needed another way to login, which meant setting up wi-fi and SSH.

The dev version of RTKLIB on github (2.4.3b8) currently wont make on the Pi unless some un-committed changes are patched in, when that's done. Its just a case of bashing install.sh in the app folder.

Then a couple hours flew by while I tried to work out why there was no stream coming in from the GPS, convinced it hadn't installed correctly, it turned out to be the wrong baud rate (pi default is 115200) and the serial device not being specified correctly in the .conf file (the .conf file expects ttyAMAO not /dev/ttyAMA0 like normal).

One last gotcha was the startup bash scripts in rtkrcv/gcc need to be given executable permission or the rtkrcv server wont start.

And when I eventually found my way out that little rabbit hole - it works like a charm.

RTKLIB is definitely improving the accuracy from these NEO-M8Ns but I'm not blown away by the results so far.

The solution is within 2-3m using a pi model B as the base station (point 2 in the plot) and a pi zero as the rover (point 1). This is with the receivers indoors by the window on an overcast and rainy day.

What else can I do to improve the results? I read about replacing default kernel with PREEMPT_RT kernel 3.18.9-rt5-v7+ on the EMLID website but would that affect the solution output by rtkrcv at all?

Neo-8M's can easily give ~5 mm horizontal accuracy on baselines of several kilometers. This is with tens of minutes of data collection under static conditions. Kinematic accuracies can be on the order of 2 cm horizontal after the integers have been fixed. This may take up to several minutes even under ideal GPS receiving conditions. Less than ideal conditions mean more time is needed. Once the integers are fixed and no cycle slips happen ~2 cm accuracy will be maintained.

This isn't that difficult under good conditions. One thing to note about RTKLIB though is there is a bug in the software that makes RTK more fragile than normal. It doesn't handle the introduction of new satellites correctly. Many newcomers to RTK have only tried RTKLIB and believe it is the standard and think RTK isn't reliable. Well typically RTK isn't really robust, but it is better than the current RTKLIB suggests.

Most people immediately want to slap a GPS receiver to a multicopter and achieve cm level accuracy for the vehicle while it flies around. That is possible, but you do have to know what you are doing.

It is best to learn to walk before you try to run. Along those lines if you want to use GPS to measure baselines to cm level accuracy this is relatively easy. I suggest you put two receiver/antennas out in an open sky area and measure the distance between the antennas with a tape measure. Collect ~twenty minutes of raw data (RINEX) and then process it with RTKLIB. The baseline should easily process to within a couple cm of the true distance.

This isn't that hard. I'll gladly walk you through it. Weather does NOT matter. Fog so thick you can't see the end of your outstretched arm does NOT matter. Anything that blocks/reflects a 1.6 GHz signal does matter. Metal siding/roof = bad, open water like a pond = not good, most buildings have something in them that is bad, etc.

Take your receivers to an open area, measure the distance between the antenna with a tape measure (to mm level), collect just ten minutes of data and with RTKLIB you should be able to process the data to a ~ 1 cm accuracy baseline.

That's exactly what I had planned to do next Thanks very much for your offer to walk it through, I will collect some data, have a go at post-processing it and report back.

Regarding the bug with new satellite introduction, this sounds similar to the ar lock count and elevation mask issues which I read about here and was planning to see if I could reproduce the error (and resolution) with my setup.

The RTKLIBExplorer site should be perfect for your purpose. I haven't gone through all his stuff, but it looks pretty good.

One thing I found with my M8N is that it had a lot more cycle slips than a LEA4T or an M8T - when using the same quality antenna. This makes RTK much harder to do reliably. It doesn't bother static positioning as much. If my M8N receiver is representative they just aren't going to be good for any type of kinematic work. Not without some very good cycle slip repair software.

Maybe the RTKLIBExplorer guy will come up with something to make $30 modules work, but for cheap RTK I would definitely spend $80 for a M8T receiver for the better cycle slip performance alone.

I introduce you the new RTK M8P GPS on the Indiegogo campaign https://www.indiegogo.com/projects/t...recision-gps#/

Ready to ship
http://www.csgshop.com/product.php?id_product=232

Hello,

We are prood to introduce you the Tiny RTK for drone application.
We have integrated the antenna, the compass, the Bluetooth and charger for lipo.
https://drotek.com/shop/en/home/792-...e-edition.html

Sky-Drones introduces SmartAP RTK - the kit based on UBlox M8P which provides centimeter level accuracy. The module is fully compatible with all SmartAP autopilots and integrated in SmartAP GCS, standard telemetry is used to send RTCM corrections. Here is a short demo:

SmartAP RTK (2 min 8 sec)