Boat speed calibration mysteries

[mglonnro]

With the end goal of creating a boat speed calibration table, I made some initial test runs today.

- GPS = B&G ZG100
- Speed as measured in water = DST800

I'm reading the raw values as they are sent by the sensors onto the N2K network, but I'm sure they internally incorporate some damping/filtering.

I'm scratching my head, though. The attached graph shows SOG in the background and speed in water in the foreground (generally a bit higher).

It's clear that during the top speed stretch, SOG is roughly 6.0 knots, and speed in water about 6.3 knots.

What needs explaining, however, is that during the acceleration they show almost the same speed, but during deceleration, speed in water stays above SOG during the descent.

[tkeithlu]

If you can get two instruments measuring a phenomenon in different ways to agree to that degree, you have a career as a test engineer ahead of you.

Your GPS is comparing sequential locations. It does not actually measure speed. It derives speed by how much time separated readings at two different locations and how far apart they were. There is some damping in how often it makes this calculation and how many readings it includes in the calculation. Your water wheel does read speed, but as time between successive switch closures, or your pitot measures it as differential pressure translated into speed. There are bunches of opportunities for these to differ and to be in actual error by small amounts.

Be very happy with the readings you got.

[mglonnro]

Originally Posted by tkeithlu

Be very happy with the readings you got.

I love your answer

I now I feel like a nerd!

Anyway, in my true wind calculations, if I input SOG = 6.0 and STW = 6.0 I get:

{
  awa: 34.00000000000001,
  aws: 9.999999999999998,
  leeway: 0,
  stw: 6,
  vmg: 2.274151579721473,
  tws: 6.042804919356156,
  twa: 67.72668171233494,
  twd: 81.72668171233494,
  soc: 0,
  doc: 96
}

But if I enter SOG = 6.0 and STW = 6.3, I get 0.3 knots of current! And a 2.7 degrees change in TWA.

Okay, now I really feel like a nerd

{
  awa: 34.00000000000001,
  aws: 9.999999999999998,
  leeway: 0,
  stw: 6.3,
  vmg: 2.1125718671871176,
  tws: 5.935593134478197,
  twa: 70.40746253897214,
  twd: 84.40746253897214,
  soc: 0.2999999999999993,
  doc: 194.00000000000003
}

[Paul L]

What is causing the accel and decel?

Have you calibrated these on a no wind day under power. Splitting the difference between two runs 180* apart?

[Orin]

When your speed dropped to two knots were you turning? Did you set a new course? We’re there any changes in surface currents?
I know it’s a bit pedantic but you would have to take a lot of things into account to keep those nearly exact.
As tkeithlu mentioned because the ways the speed is measured is fundamentally different you cannot expect perfection. A simple example of this might be waves or current. Your impeller, measuring speed directly will measure the surface distance across waves, which will be farther than the map distance.

In order to get a true enough picture of what’s going on you’re going to need to bench calibrate everything and incorporate an accelerometer and compass - or some other instruments to make 3D vectors out of the data.

Then extrapolated calculations regarding current, tws, and other things like that can be a lot more accurate. Not sure you’re going to notice the difference in all but the most extreme conditions.

You’re not trying to get to Hawaii by dead reckoning alone are you?
To me, instruments on a boat serve to give me precious little more than a good general idea of the things I can already feel. A more quantifiable speed etc, so I can figure on if I will arrive with the tide at my current rate.

Even for racing I’ve not needed more than a quantifiable, comparable difference.
Granted I wasn’t trying to get the last 1% for big time racing.

If you need help bench calibrating anything I can offer help but would need more info for the gauges. Feel free to PM.

[mglonnro]

Originally Posted by Paul L

What is causing the accel and decel?

Have you calibrated these on a no wind day under power. Splitting the difference between two runs 180* apart?

- Speed up = throttle up to 2500 RPM (aboutish)
- Speed down = throttle to idle
- Driving on a straight line in no current, little wind conditions.

This is the raw data, no calibrations applied. The B&G calibration is given as a linear 100% over the whole range, and I'm trying to figure out a better model than that (since the difference between sog and stw, for example, can not be explained with just one linear calibration percentage).

[mglonnro]

Originally Posted by Orin

When your speed dropped to two knots were you turning? Did you set a new course? We’re there any changes in surface currents?

Throttle up and down, driving in straight lines without current (see attached picture where I am ).

I know it’s a bit pedantic but you would have to take a lot of things into account to keep those nearly exact.

Yes, I assume they will be off since those are the "raw" values from the different instruments. I'm just trying to figure out what the "off pattern" is. If it would be constantly X% (like the calibration in B&G Triton2 is entered), it would be easy, but it doesn't seem like that.

Might be that the internal damping of the GPS makes the comparison difficult.

As tkeithlu mentioned because the ways the speed is measured is fundamentally different you cannot expect perfection. A simple example of this might be waves or current. Your impeller, measuring speed directly will measure the surface distance across waves, which will be farther than the map distance.

In order to get a true enough picture of what’s going on you’re going to need to bench calibrate everything and incorporate an accelerometer and compass - or some other instruments to make 3D vectors out of the data.

Yes, I can add acceleration, heel, pitch etc to the equation, but for this first run I tried to drive just in very straight lines without wind, current, waves, etc.

If you need help bench calibrating anything I can offer help but would need more info for the gauges. Feel free to PM.

Much appreciated!

And oh yes, the end product of this is a raspberry program that reads aws, awa, stw, sog, cog, heading, variation, pitch, roll from the N2K net and outputs adjusted STW back to the net. (Kind of like some of the B&G 5000 systems.)

[wholybee]

Your STW instrument is *remarkably* close to your GPS SOG. I don't know if it can get any better.

I have tried and tried, and the best I can do is get my STW correct for one specific speed, on one specific tack. Any faster or slower and it get progressively further off. And it reads different on port and starboard tacks.

I have done many calibrations, even swapped the wheel type for an expensive sonic type, with no noticeable change. The problem seems to be, at least with my Raymarine instrument, is that it stores only a single calibration point (which it derives by averaging two runs away and back to remove current). But the response from the sensor is non-linear, and the turbulence over the sensor is different on each tack, as it isn't exactly on the centerline.

The issue you see with STW staying higher then SOG while decelerating, is probably a result of averaging in the STW instrument being different than the GPS does.

[Lodesman]

Originally Posted by mglonnro

Throttle up and down, driving in straight lines without current (see attached picture where I am ).

There's always current, or wind or both. You should do sets of passes over a marked distance in both directions, then average the results. Typically in speed calibrations, you'd run a measured mile, do a Williamson turn then run the same track in the opposite direction.

[Paul L]

Originally Posted by mglonnro

- Speed up = throttle up to 2500 RPM (aboutish)
- Speed down = throttle to idle
- Driving on a straight line in no current, little wind conditions.

This is the raw data, no calibrations applied. The B&G calibration is given as a linear 100% over the whole range, and I'm trying to figure out a better model than that (since the difference between sog and stw, for example, can not be explained with just one linear calibration percentage).

Ok, so you want to create something like a gain and offset calibration. You don't want to take data during accel or decell as there is both a different lag time between the two sensors reporting and a different averaging period.
Take data at constant rpm intervals. Do multiple runs, each run being a pair of 180* opposite direction. Run the rpms so you have data at points like 1,2,4,6,max kts.

[tkeithlu]

Mikael, I doubt very much that any of your instruments read to more than three significant numbers, or one tenth of a knot. The rest of those numbers to the right of the decimal are pure computer-generated garbage.

[Orin]

Warren,
If you are concerned with only the calibration on flat water then you may have to correct for the following error types all at once.

Span error - This has to do with overall scaling. It will be a simple multiplier against all the values. If your values are off proportionally to your speed you have this error.

Zero error - This will be a value will be added or subtracted from the whole curve, if all of your values are off in one direction or the other you have this error.

Linearity Errors - When you are expecting a linear response or signal but don’t get one basically. There are ways to correct for this like taking square roots, approximating curves using multiple point like on a table and interpolating between them, or creating math functions via running regressions on measured points. Linearity error is normally not corrected with simple instrument calibration but via software or a second instrument.

Hysteresis- This error is usually caused by something being bent slightly on each direction of travel or in some way having the forces change because you’re coming up to or going down to a value. This may be complicated by the fact that small ups or small downs may not affect the reading as much as large ups or downs. It will be difficult to correct for this type of error. Fortunately, I think this is not incredibly likely on a water wheel other than throwing out the results from acceleration and deceleration for the first little bit while creating your tables.

Deadband - Sometimes this is created when mechanical linkages create slop in a system. It can also produce a sort of still linear hysteresis as well. It’s an unresponsive area of measurement or an unresponsive amount of change in measurement.

In order to correct for all of these I would get values while not moving (in the case of some instruments you will see values where you shouldn’t) then get out to the test area and move my throttle in measureable increments from minimum to maximum (that you want your reading to be accurate in). A minimum of 5 steps or preferable 9 or 15 steps between 0 and 100% is usually what it takes to account for these errors. Take the same measurements on the way up and down.

Create a table of average values in each direction and share the results.

Honestly a bench calibration fixture might be easier if not slightly more expensive if you don’t have the stuff laying around. (Stopwatch, water containers, pipe, instrument to be calibrated, valve, and a way to measure your results)

[Orin]

Sorry for double posting and such here.

If you choose to share those results I will help to identify the error types and remedies. I personally think being within 5% of span is reasonable. It would be interesting to see how close we can get. There will be limits of course, because I expect that even the salinity of the water will play a little role, density should affect how much friction a water wheel will overcome for instance.

I went to school for exactly this sort of thing.

[mglonnro]

Originally Posted by wholybee

Your STW instrument is *remarkably* close to your GPS SOG. I don't know if it can get any better.

I have tried and tried, and the best I can do is get my STW correct for one specific speed, on one specific tack. Any faster or slower and it get progressively further off. And it reads different on port and starboard tacks.

I have done many calibrations, even swapped the wheel type for an expensive sonic type, with no noticeable change. The problem seems to be, at least with my Raymarine instrument, is that it stores only a single calibration point (which it derives by averaging two runs away and back to remove current). But the response from the sensor is non-linear, and the turbulence over the sensor is different on each tack, as it isn't exactly on the centerline.

The issue you see with STW staying higher then SOG while decelerating, is probably a result of averaging in the STW instrument being different than the GPS does.

Yes! There are lot of interesting factors there! The small "dip" on the downhill, for instance, where STW and SOG momentarily align and then depart again, was when the heading changed a bit, so I guess the issue of lateral speed vector + longitudial speed vector (only thing the wheel measures, I think) affects this.

It's a fascinating subject, for sure!

[mglonnro]

Originally Posted by tkeithlu

Mikael, I doubt very much that any of your instruments read to more than three significant numbers, or one tenth of a knot. The rest of those numbers to the right of the decimal are pure computer-generated garbage.

Yes, absolutely, computer generated floating point "garbage" (10 might become 9.9999999998). I could round them to one or two decimals for sure.