The most exciting or most disappointing upboxing ever

[Horton]

https://www.traceup.com/

[RazorRoss3]

It would give us the ability to quantify speed on the ski, that's for sure. Be able to quickly tell if a ski accelerates faster/slower, or the total speed variance, etc. if it could measure angle as well then you'd have a real kicker! Be able to know if the path of one ski is the same as the path of another.

 

I was told there would be no math beyond counting to 6 in this sport.

[Than_Bogan]

Well said. Hope our shared pessimism is proven wrong!

 

(Then again, once I have access to that kind of data I'll have to quit my job to process it all!)

[Horton]

I still have hope. I will ski Friday and the weekend and see what I get. I exchanged messages with the co founder and he side I could get raw GPS at 5Hz

[TylerR]

need numbers..... Shut up and take my money

[wart]

Received my Trace two weeks ago.

URL's provided to learn Trace operation; Web site worked; but now the URL's no longer work.

iPhone app not available yet.

iPhone TestFlight app provided; but the Trace app beta not received yet.

 

So, trending towards disappointing.

 

However, the Trace staff is responsive. Seems they are just having problems getting everything started.

 

[Horton]

@wart have you uploaded gpx data to dropbox? All android stuff seems to work. Just nothing to capture at my desk.

[Than_Bogan]

5 Hz is right on the edge of useful; a lot happens in 200ms in this sport. But I think it WILL be useful IF it's accurate.

 

@RazorRoss3 If you can get position data with sufficient sampling rate, then computing velocity, acceleration, direction, etc. are just a matter of doing the right math. THAT I can do.

[Horton]

@Than_Bogan‌ at this point any data is better than no data (maybe)

[Horton]

ooo I hope it does not turn out like this

http://bricepalmer.com/wp-content/uploads/2012/04/charliebrown-football.gif

 

[RazorRoss3]

@horton, if it does end up like that, please post the video!

[bbirlew]

@horton, I've been emailing with one of the founders as well, and have been working on an android app to process and plot positional/speed/acceleration/pitch,etc data and plot it all with specific focus towards slalom skiing.

 

The hope is that one could 'map' a course simelar to zero-off and my program would process the data and split it into passes. It could plot and replay the data, as well as log it all to a database for comparison.

 

It will store sets by date/time/skier and will store relevant pass info like a ski diary, such as rope length, boat speed, pass direction, z-off setting, ski settings, notes, etc.

 

One day the hope is to correlate that with video data as well so you could select a spot on the pass and see the data for that point as well as the matching video frame (that part of the program isn't done yet!)

 

Unfortunately I have been having a tough time getting any info about the API from the manufacturer.

 

The competing product (Xensr) will be launching shortly too.

 

My program is currently able to parse/plot fake data from google maps (I'll post a screen shot in a couple days when I have my computer)

 

I was hoping to have the sensors integrated with it before sharing this all, but you guys are too quick for me!

Unfortunately my real job gets in the way, but I am hoping to have somthing available for next summer.

 

[ozski]

Just keep in mind we are at the start here guys. Give it time and I'm sure it will get better and I am sure the data will flow. We are starting to work with this type of tech more and more as the commercial potential is being realized. Exciting times ahead.

[Horton]

@bbirlew‌ below is a snip of a conversation between myself and a developer

 

http://developers.traceup.com/Endpoints/Visits.html

dlokshinTrace: and then search for this

dlokshinTrace: POST visits/[visit_id]/export

dlokshinTrace: So that will give you the raw GPS at 5Hz

[Greg Banish]

Stupid questions, but... Waterproof? (Sealed USB?) Shockproof? (Crashes are inevitable) The engineering geek in me is suddenly interested. Next up, combine data from this GPS device with video the same way Harry's Lap Timer works for the road racing (racecar) guys.

[Chef23]

It is probably very cool information but I am still lots of fundamentals to improve. I am not sure I can leverage the information it would provide. For skiers better than me I could see it providing valuable information.

[bbirlew]

@horton,

I don't have a trace yet, and was waiting for a few answers regarding the type/frequency/accuracy/etc of data before ordering one and attempting to support it in my app. They have been a little slow responding, but I'm sure they are pretty busy these days.

Originally I was leaning towards using the xensr once they hit the market, but in a perfect world my app would support both.

Currently I'm continuing to develop and add features with the google maps positional data (with faked speed/accel/etc data) and hope to get my hands on a trace or xensr soon.

I'll keep you guys posted.

 

[Than_Bogan]

@Chef23 I'll take it a step further: I think for a while the net effect of this data would be to ski worse. However, over time (meaning years), we'll be able to develop some analysis tools that may distill out some valuable lessons and/or simple ways to measure if you're doing what you're trying to do.

 

I foresee a long period where this is a Nerds Only exercise that appears utterly useless, until suddenly one day it has turned into gold.

[Horton]

@eficalibrator‌ yes yes and yes but Google it. Trace by Active Replay

[disland]

@Than_Bogan I disagree that analysis will take years to be useful. Being able to know what your "numbers" were from pass to pass and set to set will help any skier optimize his path and aggressiveness to maximize efficiency. Most skiers can crank a turn and try and hold angle. Knowing what the optimum is, is another thing all together. Having an ability to track it would have immediate benefits I think.

[disland]

How about this for a vision. Your watching the US Open webcast. Nate smith is at 43 off and his G-forces, velocity and acceleration are streamed real time on the bottom of the screen like clock is today on an NFL or NBA game. Ski geeks like us will already know what the most optimum numbers are and thus you will see when the skier varies from it, just like you do today when the skier turns too hard for the gates and gets pulled up at the second wake, and you know its going to be a bad 1 ball.

[Horton]

I exchanged emails with a guy at a data logger company a few year ago. Below is what he told me.

 

Your application requires a data logger with a complex set of sensors, called an inertial measurement unit (IMU). An IMU usually contains accelerometers, gyroscopes, GPS, and barometric pressure (altitude). Individually, these sensors can not provide a accurate position assessment due to inherent errors. A Kalman mathematical filter is required to combine the data and reduce the errors such that a usable position is determined. This is how guidance systems work on military weapons. It is not trivial nor is it cheap.

 

[ToddF]

As long as the data is consistent, the degree of accuracy won't matter that much, as long as you have a baseline for your skiing. You should be able to interpret the data to help your skiing.

 

if it is accurate from puck to puck then that will be even more useful.

@Horton was someone playing ping pong in the background?

 

 

[tap]

I read a fair amount of this forum but almost never get around to contributing. However, the timing of this subject was just too perfect and I haven't seen anyone post any actual data yet (if something has already been posted in another thread I missed it). I've been playing around with an accelerometer for a bit and just took some readings this last weekend. A few points of interest on the data: it is a 3-axis accelerometer, the measurements include 1 g from gravity, I recorded at 800 Hz, the time stamp precision was set to 0.1 milliseconds. I recorded several ski sets but did not take notes on which set was which and I tend to jump around a lot with speed and line length, so the data was either 34 or 36 mph and either -22 or -28, but it was a full pass. (I'm a -28@36 skier on a good day, a couple at -32 on a really good day) If I spent the time I could probably figure out the speed based on the time stamp, but right now I'm just playing around with what I can do with the data, not spending too much time on what the data is actually saying... I'll leave that to the forum for now.

 

The chart below is with the accelerometer "rigidly" duct taped to my sternum (center-of-mass'ish). The chart is a combination of the three axis vectors summed into a net "g-force" magnitude: g^2 = x^2 + y^2 + z^2. My first impression... the data is way more dynamic than I was expecting. Peak g's appear to be just shy of 4.5. My plan was to integrate the acceleration into velocity to add to the info, but this is proving not to be as straight forward as I first anticipated. For the moment I'm simply in data overload. And yes, there is a significant amount of error as you integrate acceleration with known error into velocity, and more error stacked on integrating from velocity to position. So, the velocity numbers may be questionable, but I think they may still prove to be interesting. GPS contributed data would be great, but for now all I'm working with is an accelerometer.

 

 

[tap]

I neglected to label my chart axes, Y is total g's, X is the time stamp in seconds.

[Horton]

It is early days but @tap keep at it.

[Than_Bogan]

@‌tap COOL!

 

Crazy noisy, but you can kinda see the 12ish acceleration peaks that occur during the turn (change of direction) and the first part of the pull (increase of speed).

 

I'm not sure you want to count gravity here, although normalizing it out may be hard (more below). Yes, it's technically true that just standing there on the ski there is one g in the z direction acting on you, but we all experience that all the time, and it's not typically what is meant by "g's."

 

It would be nice if you could just ignore the z component to get rid of gravity, but unfortunately your sensor will be rotating, so its z axis will not remain aligned to the earth's gravity. BUT I think we can still do a passable job of removing it. Basically, in SOME coordinate frame the total magnitude came from sqrt(x^2 + y^2 + g^2). Since g^2 is always 1, I believe you can take the squared magnitude, subtract 1 to remove the influence of gravity, and THEN take the sqrt to get the magnitude without gravity. However, at times you will be accelerating some in the direction of gravity (e.g. standing up vs. leaning over), and my idea breaks down in that case. Still, flooring the squared magnitude at 0 to avoid trying to take a sqrt of a negative number might be "good enough" since we're not as interested in the low-g spots anyhow.

 

Worth pointing out: This wouldn't reduce the 4.4 peaks much at all. Would change to 4.3...

[Than_Bogan]

Oh, also: Are you plotting net acceleration or the square thereof? Based on my previous attempts at analysis, 2 g's makes a lot more sense than 4, but of course my analysis was necessarily primitive and could be wholly wrong.

 

IF your chart is the square of the acceleration including gravity, then the results make complete sense to me, with a peak acceleration-beyond-gravity of sqrt(4.4 - 1) ~= 1.85.

 

If not, this data is a little head-scratching to me.

[Horton]

@Than_Bogan‌ don't you have a graduate student you can abuse / assign to this?

[Than_Bogan]

@Horton I'm not in academia, but we do have a similar concept called an Intern! :)

[Rpc29]

Here is another tracking system that looks like it could be very useful for us as skiers, its currently being used by several NFL and major college football teams. They claim to be the only true 10 Hz GPS.

 

 

 

[Than_Bogan]

@tap If you want to continue with a smaller audience, feel free to email me: nathaniel UNDERSCORE bogan AT alum DOT mit DOT edu.

[Horton]

@Than_Bogan‌

I thought you were literally Sheldon Copper but at MIT

[Horton]

@Than_Bogan‌ considering the amount of engineers and smart guys here on ballofspray I don't know why you wouldn't want to keep it open source. if you need some kind of walled garden area within the forum I can also do that

[ntx]

@horton There was a product out about 4 years ago called Shadowbox. I am not sure what happened to it. Cost was right about 500.00 @disland do you remember it? It would have worked on a trick or jump ski also to measure height and distance. It did a over lay of your path onto a Google earth map. The possablities looked great. They did a first generation and before the second rev was released, they disappeared. I wish someone would start it back up. You can find some old info with a google search of "shadowbox tracking device"

[Horton]

@ntx shadowbox was a bust. Went nowhere. I played with some data. Useless.

[Waternut]

@tap that is awesome! I can see each pull through the wakes, turns, and subsequent pulls. It looks like you take a really hard gate. I also think it's neat that it looks like you slammed or got stood up at 3 ball and 4 ball which caused a slightly weaker pull through the wakes afterwards. Meanwhile 5 ball and the pull to 6 looked very calm and composed. Would be really awesome to compare this data with video to see what different mistakes in the course look like.

[tap]

@Than_Bogan‌ , I'll answer the easy ones. I'm having a hard time pulling myself away from the analysis to respond, only so much time in the day. I could argue for or against whether or not to include the 1 g of gravity in the total g's one feels; a counter argument would go something like... if the accelerometer were in free fall it would record 0 g's, if a person were in free fall they have a sensation of 0 g's, therefore an accelerometer reading of 1 g would equate to a sensation of 1 g. Either way, I did not actively leave gravity in the data, I quite simply hadn't yet figured out how to remove it. Removing gravity from a moving coordinate system is a bit of a puzzle. That being said, I like your concept of just removing it from the vector sum with a floor of zero. I don't think it provides much more insight into the progression of the analysis since it does not help remove it from the individual components (x,y,z), however it is interesting.

 

 

To your second question... the charts are in fact g's, not g^2. The sensor records in units of g's, if you set it on a table it would read x = 0, y = 0, z = -1. The first graph (from my previous post) is simply the absolute value of the vector sum, f(x,y,z) = sqrt(x^2+y^2+z^2). The graph in this post is f(x,y,z) = sqrt((x^2+y^2+z^2)-1) with a simple 'if' statement to floor it at zero.

[tap]

@Waternut‌ , I agree, it would be great to compare side by side with video. The sensor I'm using uses a real time clock for the time stamps, so it would be feasible to match it up assuming you calibrate the clocks between the video camera and the accelerometer... and had the time to actually go do it. Unfortunately I don't see myself getting around to that anytime soon.

[tap]

Now for some real fun... I'm hesitant to post these for fear of creating more confusion than clarity, however since there seems to be interest I'll proceed at risk. Please keep in mind as you look through the charts below that the 1 g of gravity is still playing a role in the data. Below are the individual charts for each of the x, y, and z components of acceleration in g's, relative to the accelerometers coordinate system, of the total acceleration graphs previously posted. For reference, I had it taped to my sternum such that positive x was up, positive y was to my right, and positive z was straight ahead.

 

 

 

 

And now for the really cool part... a quick and crude integration of the y-acceleration into y-velocity (again y being left and right). As you can see the data has an artificial drift, I suspect this is the bit of the 1 g of gravity playing a role, or possibly a stacking of error in my crude integration method. But... the information is in there... just need to clean it up.

 

 

[SkiJay]

If I read this the way I would with data from a race car, it looks to me like you pulled a bit hard and a little too early for the gate and got pulled up a tad behind the boat. Everything went great after that until you over-turned at four ball which compromised your cut to five, so you had to dig hard and pull a bit long to make six.

 

This is awesome @tap‌! You're on to something!

[Than_Bogan]

Absolutely fascinating. Significantly more g's measured than I expected. Very cool stuff.

 

I doubt the velocity "drift" (accumulated error) can be fixed without something that explicitly measures position or velocity occasionally. That gets back to the quote Horton posted about military tracking.

[Brady]

I knew skiers were pulling more g's!!!!!! I can't wait to get one and show it to my neighbor. Question, can I put it on my sternum to determine how many g's I am pulling when running from my wife when she found out I was skiing instead of spending time with her?

[Horton]

@Than_Bogan‌ or @trap would anybody like to hazard a guess on what it would cost to build the hardware that we want and the software that would give us something like you would get at a race driving school?. Actual path would nice but I'm much more interested in accelerations, ski angles and actual direction or travel.

[gator1]

About $2k would get you first prototype of a system that would give you handle path position, velocity, and acceleration relative to the course. If you want COM or ski path data that is a tougher deal, but in my opinion handle path is good enough. It'll take us years to figure out what that means, and what it implies for skier form and ideals.

 

The system would be about $200.

[Horton]

I am really not interested in handle path. Maybe I need to think about it more but I don't see the value as compared to ski path and attitude.

 

my thinking is that what the ski does is the result of all skier inputs as well as settings. where the handle is seems like a secondary dataset to me

[Horton]

whatever dataset is required I wonder about a group funded project. I'm not a big fan of one person shelling out the cash, I like the kickstarter model better where a bunch of people chip in two or three hundred dollars

[Horton]

even better would be if @Than_Bogan‌ got an intern to do all the work for us

[Than_Bogan]

I'm actually interviewing a potential intern next Friday...

 

I'm gonna take a little bit of a Goldie Locks position on the handle path: I think it's a lot more interesting than @Horton is suggesting, but that the full path of the ski or c.o.m. would indeed provide some different information.

[Horton]

My locks are golden