Skunk works

[banaari 27]

(...) difference in accuracy of the XTE NMEA sentences.

Damn straight - the NMEA stuff all smells as if it were written around Loran accuracy levels, grrgh.

[banaari 27]

Adding injury to the NMEA insult it turns out that due to the fixed 4800 BPS limit and the number of sentences needing to be transmitted every cycle, the eTrex is only emitting NMEA updates once every 2 seconds. Good news is that Garmin have actually documented (learn, Raymarine...) their serial protocol which is binary, operates at twice the speed, and allows the host to ask for more or less anything they want from the GPS. Wrote a little hexadecimal terminal program to exercise it... bit of a bitch manually computing the checksums in the data packets, but it does actually behave as advertised.

Only quirk is there's no heading information as such; but velocity is supplied as separate North and East components...

[Atom Ant 0]

I've just read all 3 pages of this thread and now my head hurts. But I take my hat off to you even though I don't understand a thing you've said. But I gather its about not running aground while going up a narrow winding waterway and relying on a gps, a computer and a tiller pilot to get you there automatically.

 

I'll keep reading this with interest. Can't wait for the next instalment.

[banaari 27]

Let's just say the nights are long in Northland...

[banaari 27]

Code written to take course-over-ground data from the GPS in Garmin format (we're now on our third protocol...) at one update per second, and feed it to the autopilot as heading data from an external fluxgate.

Out and about early this morning to catch the tide; testing results "not wonderful". Yes, the pilot will now hold a specified course-over-ground, to about the same accuracy as it will hold a magnetic heading from its own internal compass. The COG from the GPS has considerable jitter, lag and stiction though, worse obviously at low speeds. Not sure performance is adequate for the track-keeping I have in mind. Might try averaging the last X fixes, but that would add lag into the system.

 

An interesting side effect of mimicing an external fluxgate: The pilot seems to have recalibrated its own internal compass in light of the external data; is now operating in a close approximation to degrees True.

 

Next thing to investigate is the precise nature of the adverse effects of wind on the track keeping. Are we being blown sideways off track while maintaining the requested heading, or is the nose being blown off downwind, or some combination thereof? If the nose is being blown downwind that might be amenable to more aggressive steering commands and maybe COG isn't so necessary.

[Island Time 1,235]

Yep, sometimes you can find that the manufacturers are not actually too bad at what they do with the budget equipment they use. A Rate gyro would help considerably, but complicates the maths - putting rate of turn into the equation for rudder response. The Simrad RC42 rate gyro is pretty good.....

No, I don't sell those (Yet, working on it)

 

Matt

[banaari 27]

A rate gyro would help considerably

Glad to hear to you say that, because I've actually got one available... ex robotics kit... might take a few days to work up the energy to build yet-another-interface

[rigger 47]

You mentioned early this morning.

When I was working DP often we had weaker GPS signals around sunset and sunrise, a well known issue in the industry.

 

Would it be worth checking the positions that the GPS is passing on. Do they follow a smooth trend or jitter?

Perhaps the GPS unit is not getting the best signal, -

- How many sats is the GPS picking up?

- Is there the possibility of multipath error?

- Check antenna position - is it clear? If at the mast head and vessel is rolling you may expect the heading to go all over the place unless you have a MRU or something similar

 

You could try 2 GPS units where you average the positions.

 

Might be worth considering some of the gear that a DP1 vessel has:

- 2 Gyros,

- min of 2 GPS units

- min of 2 MRUs (motion response unit)- - when these fail position keeping turns to $#!^ if there is any sea running.

- min 2 anemometers

- alternate position fixing systems.

- computer systems (you are building this part)

 

 

Awesome progress in a short time

[banaari 27]

Hehehe, the installation I've got doesn't even aspire to the quality of problem you're talking about

No, what I'm getting is pretty much what I'd expect from the velocity components emitted from a consumer-grade handheld; combined with the underlying physics... i.e. _Position_ can be accurate to a very high level indeed; but since velocity (and thus COG) is derived from the delta between consecutive positions it only takes insignificant errors in position to produce absolutely huge errors in velocity, unless averaged... and the higher the update rate, the greater the error.

 

Having said all that, you're right in that I really ought to park the handheld in a bracket of some kind, preferably on the roof, as opposed to sitting on the cockpit seat!

 

In a previous lifetime I had the luxury of working with Trimble RTK kit... 1 centimetre accuracy... ...

Someone did point me at this http://www.vectornav.com/products/vn200-rug which would be tempting if I wasn't trying to do it entirely with existing hardware.

[banaari 27]

ps. What's "DP" ?

[Grinna 2]

differential positioning?

[rigger 47]

ps. What's "DP" ?

 

Dynamic positioning.

Computer controlled position of a vessel, can include track control.

There are different levels:

 

DP-1 - Loss of position may occur in the event of a single fault.

DP-2 - Loss of position should not occur from a single fault of an active component or system such as generators, thruster, switchboards, remote controlled valves etc. But may occur after failure of a static component such as cables, pipes, manual valves etc.

DP-3 - Loss of position should not occur from any single failure including a completely burnt fire sub division or flooded watertight compartment.

 

Vessels are built to a rating / class notation.

 

DP vessels cost more to build:

As an example in 2005:

- non DP vessel = US$10million

- DP1 vessel = US$14million

Identical hull design, accommodation, winches / deck hardware.

Different engines, generators and of course the DP system.

 

A DP3 vessel is very expensive to build compared to a DP1 vessel.

[banaari 27]

Disregarding wind issues for the moment, I decided to do something about the abrupt changes of course at each waypoint, so code written to import the route from the GPS, and smooth it according to a minimum turning radius.

 

This turned into an evil chain of "but first" problems. Amongst other things I didn't fancy the notion of testing geometry-intensive code without a graphical interface, so the software is now a fully-fledged Windows application, with all the attendant development overhead, coordinate system/projection issues, etc.

 

The final smoothed route comprises a lot of very tiny segments; this together with the need to use Garmin format to get data out the GPS in a timely manner, means the GPS is reduced to supplying the initial route when asked, and basic position information... it will be up to the software to determine on each position update which route segment is closest to the current position and derive heading, cross-track etc, from there.

[banaari 27]

Went out and tested the latest iteration of the software. Significant changes are

 

(a) 1 Hz update using Garmin's protocol, not the 0.5 Hz over NMEA

( The software's now doing all the grunt navigation work

© The route is smoothed around corners

(d) We have a nice visual interface to see what's going on

 

And... we're almost there with this. The autopilot (until I remembered to factory-reset it) was still operating in degrees True; while being fed steering commands in Magnetic. Even with that thrown into the mix, things have improved dramatically. Apart from a couple of hair-raising moments in the narrower part of the channel, I was able to sit in the companionway and just monitor things as the rig obediently turned corners on cue. Very very cool being steered hands-free around a curve between two channel markers.

 

I need to do some work on the waypoints; some of them are just not in the right place; combined with the curve-smoothing, a couple of corners did require user intervention (read: the software was relieved of command...) Also, while the smoothing has improved things, because of the way the cross-track correction works, the boat will always be several seconds behind the 8-ball and thus track a few metres to the outside of a bend. Think the fix here is to take the desired "course" from a section of track M metres in front. i.e. if we're off course to the left and would normally correct right, BUT we're on a left-hand bend, the two should be allowed to cancel each other out.