HighAngles wrote:Where did you get the device?
There are a variety of IMUs at all sorts of cost points, but the long pole in the tent is to log the data. So either you need to carry a computer or some dedicated datalogger, and there are lots available.
HighAngles wrote:Did you modify it for application in a ski boot?
I just took the IMU out of the housing and "dremel tooled" a pocket into the sole plate and added some titanium to carry the foot loads around the IMU. Then added a new LEMO connector out the side of the boot (photo sometime in the future).
HighAngles wrote:Is it expensive?
(ha ha) Nothing is cheap...
polecat wrote:From the graph, It looks like output from a data acquisition system with an inertial measurement unit. (They're often used for dynamic analysis of motor vehicles by car manufacturers, racing teams, safety testers, etc).
Yes exactly. I am using a motorsport datalogger and an IMU at 100-Hz, also GPS at 10-Hz.
One advantage is that the graphing and support is good.
Also .dll can be used to add in more complex math.
Similar issues with orientation and performance are being done at the top level of MotoGP, but that seems pretty closely held info.
polecat wrote:How do you set the reference datum? Is it fixed at the beginning of a run? Is the roll angle corrected for the hill's slope (to give tipping angle) or is it relative to inertial space?
All the angles are ECEF, so it is not corrected to the slope.
Ideally one would use a KALMAN filter to correct everything to the periodic "truth" provided by the GPS.
Recently mentioned by a knowledgeable ski-science guy to use quaternions. And other knowledgeable people in similar fields mentioned similar things. I have a mathematician working on the KALMAN filter this summer (Northern hemisphere summer). I also have a friend who does a lot of small helicopter work looking at conditioning data or putting the KALMAN filter onto a blackfin(R) board and that will dovetail in with his work.
The orientation to the "slope" is a difficult problem and the "slope" is riddled with moguls etc, so at some level you can use pitch and GPS-vertical drop (plus distance) to work out slope and gradient (fall line), but it is not quite the same as a know, fixed, and surveyed surface.
In the end I think one may want to know ski roll angle (edging) relative to the ski's track. i.e. is the ski over-edged (and by how much), or under edged, etc.
polecat wrote:Do you have more inputs available, perhaps for strain gauges? It would be really cool to track ski weighting, for/aft balance, etc.
Yes I have more inputs for foot pressure sensors, but I have not gotten those in as of yet.
Strain gages are also possible, and there are pull-string potentiometers that could be used - but although I have a few I am not sure how they would work in the system. There are 1000 channels possible in the datalogger, so there would likely be enough to go around. A/D converters give a couple dozen channels each.
I think that force transducers between the binding and the ski would also work, but neither shows body position...
They only show effect not cause.
So if one wants to correlate cause with effect then video, body position knowledge, foot-pressure knowledge, etc, are needed to determine what produces the best effect.
Somehow one needs to correlate the effect and cause if the goal is to have causes that produce better effects.