New uBlox PPP module to consider

[Chris Anderson]

Guys, 

I had a good meeting with the uBlox team today at CES and they are recommending that we also offer a higher-end uBlox (the 3DR module would cost about $99, compared to the current one's $75) that uses the Neo 6P chip set that uses Precise Point Positioning, which they say should give us sub-meter accuracy. 

They say it's as good as using a DGPS beacon, without needing another piece of hardware. 

It's like a regular uBlox, but when you're loitering in ca opter, the PPP kicks in and can really tighten the zone, they say:

u-blox’ industry proven PPP algorithm provides extremely high levels of position accuracy in static and slow moving applications, and makes the NEO-6P an ideal solution for a variety of high precision applications such as mapping, marine or agriculture. 

 

Ionospheric corrections such as those received from local SBAS geostationary satellites (WAAS, EGNOS, MSAS) or from GPS enable the highest positioning accuracy with the PPP algorithm. The maximum improvement of positioning accuracy is reached with PPP+SBAS and can only be expected in an environment with unobstructed sky view during a period in the order of minutes. 

Unlike the previous uBlox PPP chipset, this the 6P operates at 5Hz. There is 7P coming, too, but it's just lower power, not higher performance. 

What do you all think?

Chris

--
Chris Anderson

CEO, 3D Robotics

ch...@3drobotics.com

[Craig Elder]

Sounds interesting.

Did they say anything about footprint and binary message compatibly? 

--
 
 

[Chris Gough]

I'm pretty sure the footprint is same as other "Neo" modules, and it it will be binary message compatible.

Will it really help though? I looked at that one and thought it would just mean that if you hover in the one spot for a couple of minutes, then your position accuracy will start to improve. I suspect that as soon as you move hither or tither, it will drop carrier phase lock and go back to being a normal GPS. Worth a try though if you have one to hand (2 experiments required, with and without SBAS).

Chris Gough

--
 
 

--
.

[Andrew Tridgell]

Hi Chris,

I have a 6P module on my roof that I've been using for doing DGPS
testing. It is this one:

http://www.csgshop.com/product.php?id_product=103

I can say that the PPP feature works well when not moving, though it
does take a few minutes to kick in. I have a current 3DR uBlox mounted
next to it, and made a video of the position estimate from the two
modules as it evolves over time:

http://uav.tridgell.net/GPS/gps-ppp-compare.ogv

(sorry for the crappy screen capture video).

The XY scales are in meters. As you can see, the PPP module does produce
a tighter position estimate.

What I don't know is if it will do well in a plane/copter. I haven't
tried flying it.

There is also one big disadvantage of the PPP module. It doesn't accept
DGPS corrections as inputs. So if we do get DGPS working on a ground
station then it wouldn't be able to be used in a plane.

It does produce raw data however, so it could be used as a DGPS ground
module, and that is why I bought it. It was cheaper than getting a 6T.

Cheers, Tridge

[Andrew Tridgell]

> Did they say anything about footprint and binary message compatibly?

They use the same protocol, except that it doesn't accept the RTCM
inputs for DGPS. It does generate the RAW messages you need to generate
RTCM though! It is ideal for a DGPS ground station.

Cheers, Tridge

[Jesse Brain]

Let me know if you get any of them Chris, I'll buy one, I want to see how good they are in Australia... I really think that GPS performance just lacks down here... 

[Chris Gough]

  http://uav.tridgell.net/GPS/gps-ppp-compare.ogv

(sorry for the crappy screen capture video).

The XY scales are in meters. As you can see, the PPP module does produce
a tighter position estimate.

What I don't know is if it will do well in a plane/copter. I haven't
tried flying it.

My theory is that the P type modules get their precision by finding/holding a carrier phase lock, and using it to measure a phase angle (==small distance adjustment). If so, and because GPS L band is ~19cm wavelength, then sudden antenna movements >10 cm (i.e. >180 degree phase shift) could present it with difficulty. When such motion introduces "carrier phase ambiguity" faster than the uBlox bag of tricks can resolve it, the P module will drop out of of PPP mode (or fail to acquire it). Obviously trigonometry applies here: the ~10cm movements need to be towards/away from the satellites. Movements orthogonal to that line will have an infinitesimal effect on carrier phase.

Of course, just chucking a pair of GPS' in a few types of vehicle would be definitive if the P module worked significantly better in any of them, but if not it wouldn't prove that the P module is no use to us in the air. 

For a few different directions of motion, if we could determined the the maximum velocity that it could hold PPP, and the minimum velocity it could acquire PPP, we might be able to characterise it with an "effective per-satellite phase sample rate" for loosing and gaining PPP. Combining these characteristics with the position of the satellites, we could estimate the range of motion that would cause it to loose/gain PPP mode (for a given constellation). I think these characteristic will be very different with/without SBAS coverage, because a lot of the tricks in their bag will utilise SBAS. In other words, working SBAS will make it seem like the phase is sampled more frequently, so we could move faster without loosing phase lock.

It does produce raw data however, so it could be used as a DGPS ground
module, and that is why I bought it. It was cheaper than getting a 6T.

@Tridge, I still have a LEA-?T that I could swap out on a module for you. I didn't get round to doing it over Christmas. If you want attack DGPS soon then let me know and I'll push it to the front of my queue.

Chris Gough

 

Cheers, Tridge

--

--
.

[Robert Lefebvre]

I agree with the others, that if if the PPP really only works after being very stationary for something on the order of minutes, then this will have limited benefit on a copter.  Maybe it would be useful as a DGPS for a ground station.

I think inertial XY is what we need for copters.

--
 
 

[Doug]

Chris - I am skeptical of your thoughts on how the 6P modules are getting the precision.  I have a 6T module I have collected carrier phase data with it - the 6T and 6P modules will both output raw carrier phase measurements.  The consistent availability of carrier phase measurements is a significant issue.  IMO there is about zero chance that the module is actually solving the carrier phase ambiguities (especially considering the frequency with which carrier phase lock was lost in my testing).  Single stand alone survey grade modules usually cannot solve the CP ambiguities very fast (eg takes more than several minutes) as you have to wait for the satellite geometry to change sufficiently and this module is not of that quality.  Also in this case there is no expectation that the module is truly not in motion which can be used - IE if the pseudorange solution appears stationary that is really a poor indication that the module is not in motion at the centimeter level and a single repositioning at the centimeter level would screw over the ambiguity resolution.  

I suspect that their PPP solution is just a good implementation of SBAS with carrier phase smoothing applied to the solution.  Carrier phase smoothing is MUCH more robust than carrier phase ambiguity resolution and can give a great improvement to the solution characteristics, but cannot guarantee an improvement in absolute accuracy and still leaves an order or two of magnitude between the solution quality available from a real carrier phase solution.

[Chris Gough]

Doug, thanks for the analysis, except for the SBAS bit I think you are spot on. Here in AU there is no SBAS, yet Tridge is still getting PPP.

This page discusses HOFMANN-WELLENHOF Carrier Phase smoother algorithm: 

http://www.gmat.unsw.edu.au/snap/gps/gps_survey/chap6/6411.htm

And says the CP smoother is reset if a cycle slip is detected (but the cycle slip is _not_ corrected, i.e. the phase ambiguity is not resolved, only detected).  I think we are talking about the same thing, although you were using more appropriate words :). Had I googled "CP smoother" first, I would have written "create a CP cycle slip" rather than "introducing CP ambiguity". I hadn't thought of using dopler with phase to detect cycle slips, that's neat.

Chris Gough

On Sat, Jan 12, 2013 at 4:45 AM, Doug <dewe...@gmail.com> wrote:

Chris - I am skeptical of your thoughts on how the 6P modules are getting the precision.  I have a 6T module I have collected carrier phase data with it - the 6T and 6P modules will both output raw carrier phase measurements.  The consistent availability of carrier phase measurements is a significant issue.  IMO there is about zero chance that the module is actually solving the carrier phase ambiguities (especially considering the frequency with which carrier phase lock was lost in my testing).  Single stand alone survey grade modules usually cannot solve the CP ambiguities very fast (eg takes more than several minutes) as you have to wait for the satellite geometry to change sufficiently and this module is not of that quality.  Also in this case there is no expectation that the module is truly not in motion which can be used - IE if the pseudorange solution appears stationary that is really a poor indication that the module is not in motion at the centimeter level and a single repositioning at the centimeter level would screw over the ambiguity resolution.  

I suspect that their PPP solution is just a good implementation of SBAS with carrier phase smoothing applied to the solution.  Carrier phase smoothing is MUCH more robust than carrier phase ambiguity resolution and can give a great improvement to the solution characteristics, but cannot guarantee an improvement in absolute accuracy and still leaves an order or two of magnitude between the solution quality available from a real carrier phase solution.

--
 
 

--
.

[cont...@nerim.net]

What do you mean by inertial XY ?

An inertial GPS like those ones ?

http://www.oxts.com/default.asp?pageRef=1


Those inertial GPS are big and expensive, some of them can even have dual antennas inputs for precise heading output. They can give 250 Hz output with 10 cm precision.

Would be nice to have a home made inertial GPS on a multicopter but i'm afraid by the code complexity to merge inertial measurements with GPS pseudo range and carrier phase data.


Olivier

[Jesse Brain]

@Tridge, have you done much DGPS work?  have you got some pointers on where I could start if I wanted to integrate DGPS into a ground station to improve the fix on my copter?  I've been working with GPS for years, but never really touched on DGPS as I never had a suitable source... 

[Craig Elder]

You need to deliver real time kinematic (RTK) messages in a format called RTCM to the onboard GPS via the second com port on the GPS..  

At the moment these could be provided via a 3DR telemetry radio or potentially some structure via MAVlink but as yet we don't have a way to access the second com port in the GPS.

--
 
 

[Jesse Brain]

well that's a bit annoying, I see what you mean, unless pins 21, 22 & 23 are reserved for uart2 in the future than the package doesn't actually have a second uart for feeding DGPS data into it... 

tell me, how much better are the ublox GPS modules compared to the mediaTek ones?   is it worth me buying a ublox unit to get a better loiter?

[Hagen Felix Piotraschke]

This explanation from Doug describes 100% exactly what the NEO-6P is doing!

Even the discussion itself is already some somths ago, it becomes much more interesting again due to the new successor now (NEO-7P).

I had assembled and sold some hundreds of different devices based on the NEO-6P, that's why I have quite a bit of experience with this really nice positioning module. :-)

But there's one more detail which cooperates with the other two (the really good SBAS implementation and the carrier smoothing)!

It's the implemented dynamic profiling which is even not completely new (u-blox modules are eqipped with this feature already for several years) but in this partnership still an important part of the whole solution.

So if you have a reliable board (or complete device) with the NEO-6P and a little better antenna, this module should be still as a primary choice.

The "truth" of tracks (not only the absolute precision) is still ages better, compared with any other receiver module at the moment, worldwide.

This benefit is also availabe in moving applications, even the best results are done in stationary mode.

Currently I'm testing first samples of then new module, especially of course concerning the new features RTCM-Input and 10 Hz update,

If I see the copter topic now, I would agree that a rock-stable PPP is always netter than

Am Freitag, 11. Januar 2013 18:45:18 UTC+1 schrieb Doug:

...

[Ben Nizette]

Hi

On 25/10/2013, at 10:10 AM, Hagen Felix Piotraschke wrote:

> The "truth" of tracks (not only the absolute precision) is still ages better, compared with any other receiver module at the moment, worldwide.
> This benefit is also availabe in moving applications, even the best results are done in stationary mode.

Do you have any data sets you can share here, especially without SBAS? I've never had anyone from uBlox try and tell me that the PPP mode is worthwhile when moving, and given that even their marketing guys didn't believe it then I didn't bother to test it!

> Currently I'm testing first samples of then new module, especially of course concerning the new features RTCM-Input and 10 Hz update,

I'm looking forward to the 10Hz rate being specified, will be interested to see if/how accuracy changes. I know some of the 5/6 families, especially running from Flash, would have their accuracy degrade as you pushed them to 4-5Hz (especially with SBAS enabled).

The RTCM input was introduced halfway through the 6- series and I'm doing some experiments with it now, alongside Tridge. In my experiments, a good source of RTCM corrections does show significant improvement, but perhaps not as significant as one might expect. It appears that when supplying these corrections, uBlox disables much of its internal wizardry that really shouldn't have to (such as multi path detection) which goes some way to negating the DGPS advantage.

I hear tell the RTCM implementation in 7 will be somewhat better while they're expecting it to be reasonably complete by 8 (including RTCM 3 input, not just 2.3).

Anyway you can check out https://github.com/benizl/pyUblox if you're interested in code, I'll push up some actual data/results shortly.

Cheers,
--Ben.

> If I see the copter topic now, I would agree that a rock-stable PPP is always netter than
>
> Am Freitag, 11. Januar 2013 18:45:18 UTC+1 schrieb Doug:
> ...
> I suspect that their PPP solution is just a good implementation of SBAS with carrier phase smoothing applied to the solution. Carrier phase smoothing is MUCH more robust than carrier phase ambiguity resolution and can give a great improvement to the solution characteristics, but cannot guarantee an improvement in absolute accuracy and still leaves an order or two of magnitude between the solution quality available from a real carrier phase solution.
>
>

> --
> You received this message because you are subscribed to the Google Groups "drones-discuss" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to drones-discus...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.