Blog

Read more on phases of rowing. Created diagram based on what I've read & web sources: 
Figure 1: Diagrams to parts of the rowing stroke. 4 main parts: Catch, drive, finish, recovery.

Got in contact with Gregg about using Vicon system. A guy named Michael from Prof Jane Wang's lab opened the door for us. We came in on Saturday and Gregg taught us how to calibrate and setup the cameras. We tested with various camera positions, but due to time constraints, did not get to actually collect data and do a trial run. 

After the cancelled skype meeting with Prof Ruina on Sunday, we went to Dollar Tree and Target to buy some supplies like black table cloth, posterboard, and trifold (cardboard) to prep the erg for use of the Vicon system. (The Vicon system uses infrared to track the points... any reflective surfaces such as shiny metal plates have to be covered.) 

We went back to the Vicon system on Monday to try to figure out how to collect data. We were able to create a "subject" and specify segments and joints on the system, but have no idea how to save data or where numerical values are stored. The interface for Vicon has very limited options and the only file types we have been able to find are .xcp and .cp, both of which we also do not know which other programs can use/open, or what specifically those files save. Gregg had said on Saturday that he was able to get position graphs from the system, while Song Chang (grad student of Prof Jane's) said that they had used another program to extract data from Vicon. Gregg seems to know the setup better, though, so we shot him an email again. 
Figure 2: The only graph we have been able to generate so far is the Trajectory Count graph... but we're not sure what this entails.

Saturday:

Gregg showed us the basics of the Vicon system.

• set up and calibrated cameras
• did not bring erg

The Vicon system emits infrared radiation and detects the radiation reflected back from these reflective spherical markers. So, in order to remove interference, we have to cover up any shiny parts on the erg that could also reflect infrared. Michael 
Sunday: 

We planned to Skype Andy at 2pm, but we had to push back the meeting to Tuesday. Instead, we went over to the Ithaca Mall to grab masking tape and cardboard in order to cover up the shiny metal parts on the erg.
Above: Masking tape masked the reflective surfaces (seat bar, nuts, chains, etc.) on the erg.

*There is no universal ideal force curve: *Kleshnev writes that a rower's force curve depends on his position in the boat. Volker Nolte also notes that in a double scull, the stern and the bow rowers exert peak force at different times to be a more effective pair. Lightweight rowers tend to have a more rectangular force curve, and heavyweight rowers a more triangular force curve.

http://highperformancerowing.net/journal/tag/biomechanics

Spent the day at Olin Library and finished Rowing A Scientific Approach. See book notes. 
The book gives a lot of good pointers on how best to row and the postures involved, but did not provide enough detail on the accelerations. 

Things to consider / learn:
- Angle of foot rest should be made so that at starting position, knee / thigh to calf are 90 degrees from each other. The book claims that angles less than this will cause a moment that will tend to lift the rower out of the seat. 
- For drive, the lower back / hip is what should begin the drive, not the legs
     > This is to prevent "jack-knifing" the back and snapping the spine. When the lumbar (lower back) spine is in lordosis, the extensor Erector muscles can act as a tie beam to support the load.
     > Keeping the head up helps do this. 
- At the end of the drive (propulsive) phase, the rower's back should be mostly vertical (100-120 degrees tilted)... angles greater than this will result in poor form that compensates physical safety of the rower's spine and muscles. 
     > I noticed this difference between Ellen's rowing and that of other rowers. Ellen always kept her back straight (which is good) and moved with her seat first, whereas other rowers tend to lean too far back and swung their bodies back first. I guess this is the difference between an Olympic rower and an amateur one. 
- At the end of the propulsive phase, the hand pulls bowwards and downwards, to lift the paddle out of the water in preparation for recovery and the next stroke.
     > I always thought that the handle was supposed to be kept level or even pulled up slightly... 

Contacted Gregg to see when he could come in this week. 

Dragon boating update: we won 3rd place in our division (B Major)... beat Upenn but not Harvard. More info

Dinner at Petr's update: see folder for photos and videos of us playing Settlers of Catan

Almost finished with Rowing A Scientific Approach. I will post notes here later. 

Got a stellar read from the library titled "Journal of Sports Engineering and Technology: Special Issue on Technology in Rowing" from the Institution of Mechanical Engineers. It can't get any better than this.
The publication has 11 papers on dynamic movement of the oars, boat, and mechanics of the body. I had gotten it solely for Kleshnev's 2010 paper (which we surprisingly don't have amongst his large collection) titled "Boat Acceleration, Temporal Structure of the Stroke Cycle, and Effectiveness in Rowing," to get more detailed graphs of acceleration curves. 

Weekend plans
Andrea - Back packing in Adirondacks 
Lily - Dragon Boat Festival! Cornell (Dragonboat Z) is racing against Harvard and UPenn at 8:40am. (See schedule). It'll be a long day, so maybe I'll get started on the publication. Surprisingly, some of what I've read so far can be applied to dragon boating. (In dragon boating, the crew faces the direction of movement... and you're rowing with 19 other people).

Wednesday
- Read half of Rowing A Scientific Approach by J.G.P. Williams and A. C. Scott
- Gregg Steisberg was supposed to come in today to teach us to use the Vicon, but rescheduled for Friday
- Asked my ORIE friends how to compare curves for similarities. (I haven't taken stats yet). They said to take the difference at each point, square it or abs value it (to get rid of negative sign cancelling), then sum it up to get a number. We can try this with the curves of rowing on a boat and rowing on the various machines to quantify the similarity of the curves against real rowing. The issue now is how to normalize the curves so that each data point corresponds to the same action / part of the stroke... 

Thursday
- Add and Connect links (aka roller chain offset link) came in. (#25) We bought it to better connect the flywheel chain and future adjustment to chain lengths. Because chains have an inner and outer link, adjustments to chains have to be done in twos, with the removal or addition of an inner AND outer link. Offset links are like inner and outer link converters, allowing just one link (the offset link) to be added or removed. 
- Increase pretension in soft bungee to 11 inches
- Decreased pretension in stiff bungee to 3 inches
- The machine still makes strange noises, but it's hard to tell where they're coming from and what exactly is the problem.
- Handle still slips
- Seemed to work better when super soft bungee in parallel with axle chain is taken off 

Monday
- The erg seems more robust now. It can survive many minutes of rowing without fail. Infact, no chains, bungees, or other contraptions broke while testing today. :D
- However, the handle chain still retracts too slowly during recovery. Andrea thinks the soft bungee needs more pretension. I suspect that there is more to it and we will need to add an additional part between the differential sprocket and the soft bungee to quickly take up the slack.

Tuesday
- Visited Olin and Uris library stacks to read up more on the biomechanics of rowing.
- So far, most of the position, velocity, and acceleration curves for rowing that we have are all from Kleshnev and his newsletters. I want to see what others have to say on the subject and do comparisons.
- Cornell surprisingly doesn't have that many hardcopies (books) on the subject matter, but we did manage to find some.
- The Journal of Sports Medicine & Physical Fitness yielded an article on the muscle groups involved at each particular stage of rowing.
- A German book titled Leistungsanalyse Bei Rudermannschaften by Erich Schneider looks promising.... if we know how to read German. It does, though, have a great bibliography page that I've been looking through (for English articles) for more sources.
- Rowing A Scientific Approach by J.G.P. Williams et. al. also looks promising.
- More sources. Most of these I will need to scour through Cornell's databases because databases require $$ to view unless I click it through Cornell first. But Get It! Cornell works inconsistently. :\
- A nice librarian at the cirulation desk (Sarah E. How) helped me out in searching for the books. Luckily, she's also a rowing enthusist and told me that Ithaca its own share of Olympic rowers! Andrew Byrnes and Caryn Davies (men's and women's 8 ... gold in 2008 Beijing).  They both rowed for Cascadilla Boat Club.

"The more you fail, the more you will succeed." - Andy Ruina

Ordered five add-and-connect links from McMaster, $2 each.

Proper way to install spring clip:

This info could help us in preventing chain breaks, because those are the most common errors we've encountered. See also the very complete The Complete Guide to Chain.

We went to Olin library to do some research. I found a paper called "Electromyographic Analysis of Rowing Stroke Biomechanics (1987)," in which the authors, Rodriguez et al, found which muscles are used when and to what intensity during the rowing stroke.

I checked out Rowing Faster, by Volker Nolte, and Rowing Fundamentals, edited by John Ferriss. Nolte covers every thing from training programs to the physics of an oar.

We fixed the handle bungee system housed inside the bar.

We changed pre-stretch to 8" in order to increase the back tension on the handle sprocket to reduce chain slippage.

One end of the handle system:

The picture below has been modified to outline the components of the handle system.

• The red line follows the metal bar that provides the structural basis for the system.
• The green line follows one end of the handle bungee, and this wraps around a cinch point that anchors down a part of the bungee cord, so the free end won't slip through. The free end can be lengthened to increase pre-stretch.
• The blue line follows the handle chain, which wraps around a freely-rotating sprocket located in the black plastic trapezoidal housing.
• The yellow line follows the other end of the handle bungee that is the attachment point between the bungee and the chain. The bungee wraps around a pulley (hidden in the back) and is then fixed to the black plastic housing, the intermediate connection between the bungee and the chain. The bungee is fixed to the plastic housing, while the chain can run along a freely-rotating sprocket in the housing, so this is essentially a floating pulley. The free end of the bungee can also be lengthened in order to increase pre-stretch.

The handle system in its entirety:

Because the chain wraps around a sprocket-pulley, it's as if there are two chains running right next to each other, in opposite directions. We made sure the chains didn't criss-cross in order to make sure the chain runs smoothly, noiselessly, and without wear-and-tear.

This process took a lot of concentration and time in order to get right. Some things to keep in mind for next time:

• Each end of the metal bar has two opposing metal axles for the pulleys and black plastic housing. One of these axles is longer than the others, and the black triangular shaped plastic chain feeder is fixed there. Don't use a retaining ring on the triangle piece yet, but just note where it goes.
• The white pulleys have to go on opposite sides and opposite ends of the metal bar. Place them on the axles and fix them using retaining rings.
• When setting up the bungee system, eliminate all pre-tension. When the chain and bungee and plastic bits are all on the metal bar, slide the system into the erg, and then pre-tension is easily adjustable.
• Make sure the chains do not criss-cross.
• The black triangle piece must point upwards in order to feed the chain upwards.

Physical Erg
To address the issue of excess slack in the handle chain, we took a couple of measures:
- Andrea had some bike chain lubricant (Teflon Bike Lube) that we used to loosen stiff links
- We put in more pretension in the soft bungee (8 inches)
- We plan on adding another pulley/sprocket for the handlebar chain, as an intermediate between the bar encasing the soft bungee and the differential. We need to figure out a way to hold up such an assembly, but the purpose is to add more "back tension" -- tension on the chain, in the direction opposite the pull of the handlebar, so that the chain has a tighter grip on the sprocket.
Figure 1: My depiction on MS Paint of adding an extra sprocket/pulley for the chain to ride on to create back tension on the handlebar chain. Not the best illustration, but gets the point across. We're getting really good at churning out diagrams of ergs. :D

Software
- As mentioned last time, we got the data curves for the 4 types of ergs (stationary, on slides, Dynamic, and Rowgometer).
- After comparing to online videos and pictures, we think that what we thought was the Dynamic might actually not be the Dynamic erg. :O Andrea thought it looked like a DYI transformation from a stationary erg to a dynamic erg. Concept II's Dynamic erg has a stationary display which is not attached to the sliding feet rests, whereas the erg we tried connects the display with the footrests. The Row Perfect's dynamic erg has the same setup as the erg we used, so the two should be equivalent.

Left to Right: Figure 2 - Concept II's Dynamic Erg; Figure 3 - Teagle Rowing Room's dynamic erg; Figure 4 - Row Perfect's dynamic erg

- Prof Ruina urges us to use the Vicon Motion Capture system from Prof Jane Wang's lab. We contacted her and she is hesitant to let us use the equipment, citing its complexity to learn to use and that her grad students are all busy and would not be able to train us until the fall. We visited the setup, which consists of an area in the range of 3 cameras connected to a computer system. It is quite is limited in space and I am doubtful as to whether the cameras' range would accommodate all of our erg's movements. Additionally, there is the problem of how to capture the motions of the other ergs, seeing as we cannot take them from the rowing room, and the Vicon system is fixed.
- If we can get around the issue of space, the Vicon system looks quite promising. Prof Wang sent us the file of an abbreviated user manual: 2010-Vicon-UserManual.docx The entire user manual. This guide from the web might be more detailed.

Video: Short video using Vicon to track motion on treadmill.

- Prof Wang also mentioned ImageJ as another tracking software. I briefly used this for my ENGRI course to count and track air bubbles, but after looking more into the software, I think that it is more suited for counting small particles. It also doesn't seem to have the autotracking feature found in Tracker, which means manually clicking through thousands of frames :(

- Gregg (sp?) used Vicon in the Ruina lab before... we're contacting him to learn more and see if he can teach us to use it.

Finally mastered the autotracking feature of Tracker! (Tutorial page to come.) The basic idea is... 
- SHIFT + CTRL + click to set the "template" for what they will search for the match the point of interest in each frame 
- Make sure that there is both a "template" image AND a "match" image after your click. Autotracker will not work unless there is a "match" image. I have no idea why... this is what kept tripping me up
- Sometimes, tracker will incorrectly match a frame. When this happens, you have 2 options. SHIFT + click to manual define a point, SHIFT + CTRL + click to redefine the "template". This is useful to keep the autotracking going for the next couple of frames, if the image has changed drastically from its original look as defined in frame 0.  
- Make the dotted square (area that the program will search in to find the image) around it large enough to include all positions the point of interest will take on during the course of the video. This way, you don't have to keep stopping and expanding it. 

Tracker only seems to save screenshots ("snapshots") of the graphs in 4 file formats: .EPS, .JPG, .GIF, .PNG
I would have preferred if we could save the graphs in vector file formats, so we can zoom in and out and crop parts of the graphs without losing quality. To compensate for this, I used the dual monitors at Mann to expand the graphs as much as possible before "snapshotting" the image and saving it. .EPS is a high quality format that only drawing programs such as Adobe Illustrator can open. .PNG retains the highest quality out of the other 3. Thus, I decided to save all graphs in both file formats. 

I did video analysis of Rowgometer. 

Figure 1: Position (red), velocity (blue), acceleration (green) curves for the erg. 

Figure 2: Position (red), velocity (blue), acceleration (green) curves for the handle.

Figure 3: Position (red), velocity (blue), acceleration (green) curves for the seat.

With so much data, it's difficult to see the relevant information. I noticed that every 3 frames would give the same position data, creating piece-wise functions for each curve, and making things look jagged in the comprehensive view, especially for the acceleration. If I adjusted it so that Tracker only graphed the data for every 3 frames, the curves are a lot smoother. 

Figure 4: Acceleration vs Time, Marked at Every Frame. Curve is very jagged and it's hard to see what is going on. 
Figure 5: Same graph, but with data only marked at every 3 frames. This looks a lot cleaner and it's easier to see what is going on. 

Monday
- We Skyped Prof Ruina again (I might have my dates mixed up) and he challenged our statement about the seat of the Concept II Dynamic Erg being essentially steady during rowing. While the seat is able to slide, during the physical act of rowing / working out on the machine, we noticed that in actually, the seat and person hardly moves. However, Prof Ruina said that this is not sufficient to conclude that the seat essentially does not move and we will need to do more research on this. 
- Andrea was not yet back from a hiking trip, so Steven and I split tasks. He read up on some papers while I used Tracker (the tool mentioned in Fri 2013.06.07 LL) to track the seat position on a Concept II video of rowing on the Dynamic Erg that I found on Youtube. Unfortunately, the video does not have marked points to make video tracking easy. For example, the seat is all 1 color, making it difficult to distinguish 1 point on the seat from another. Thus, I had to manually click through 1000+ some frames of video to get the position graph. This seems ridiculous for 21st century technology... there has to be a better way to do this. 

Figure 1: Seat position VS time for Concept II Dynamic Erg. Not sure if it's due to human error in manually clicking each point, but there doesn't seem to be much of a pattern in movement. However, the average movement is within 2-3 inches, which is small enough to not move the COM and support our above point. 

Tuesday
- I found that I had been using the program incorrectly. =( Velocity and acceleration curves are supposed to be automatic and calculated in conjunction with the position data. However, they were empty for my set. Did some digging online and found this manual. 
- We were curious to see how the Rowgometer compared to the competition. After fixing the chains and other loose parts, we took the machine downstairs to film in the hallway of Thurston. Before, all of our videos were at an angle, which makes it difficult to analyze the distances in a video for Tracker. Thus, we filmed the erg straight on while Andrea rowed. Steven volunteered his camera. Video footage: https://cornell.box.com/s/uoc5qbe0k5p9lzn3bd9p 
- There was some glare coming from the windows that we hope won't impeded the video.

Wednesday 
- Analysis of the video showed that the quality was not as high as we had hoped. Thus, we borrowed the fancy Canon VIXIA HF M52 HD camcord at Uris to do the filming. Since we had the camera, we figured that we might as well as officially film the other machines (stationary, erg on slides, dynamic) from the Rowing Room in Teagle. Having the same person row for all 4 videos will help keep the stroke movements somewhat consistent. 
- We cut circles of neon green duct tape to use as tracking dots for our points of interest. For the erg on slides, Dynamic, and Rowgometer, we placed points on the handle, seat, and erg. For the stationary erg, since the erg does not move, we only had 2 points of interest. We used such bright dots because the Tracker program needs points that are clearly defined from the pixels around it, and uses oval shapes to define the point area. 
- Everything was fine and dandy until after we downloaded the video files from the files, we saw that due to some camera setting, all footage was filmed in .MTS file format. We will need to convert this to a more common file format such as .MOV or .MP4 for Tracker to analyze. See video files in the 2013.06.26 Rowing Filming folder on BOX. 

Figure 2: The Rowgometer got some sun in our outdoor filming session. Bright neon green dots indicate the points of interest: the Erg, the handlebars, and the seat. 
Thursday & Friday
- We hit up Mann Library because they have a lot of video editing software on their Macs and PCs. Plus, school computers have more processing power than laptops, generally, and it's more secure to try and download programs on them, since the changes are deleted when the computer restarts.
- To convert .MTS to .MP4, we began with a free software called Brosoft, which is supposed to be one of the best programs out there. However, with the free trial version, all videos come with their company's logo watermarked in the middle. 
- Although the watermark would not have impeded the analysis, we wanted our video to look more professional and thus looked into some other options. The 2nd conversion was with Arcsoft, whose video quality seemed a little lesser than Brosoft, with a more dodged (whitened out) look. We decided to go with this one, since the quality was still pretty good. 
- Next, we used yet another software called Free Video Dub to shorten the video to include only the frames with good rowing. Since we had already altered the video via file type, we wanted to minimize additional quality loss from compression. FVD does not re-encode the video, so no quality is lost. 
- After the videos were finally converted, we broke up the work with 1 video per person to analyze on Tracker, and whoever finishes first working on the last video. Tracker supposedly has an autotracking feature but it was extremely tricky and glitchy to use. Tracker, in general, crashed a few times until I figured out that the .JAR file or older versions of the .EXE type works better and does not require installation on the computer (which is great). Just before the library closed for the day, I found this helpful video of a guy running Tracker 3.10 (current version is 4.8) and using the autotracking feature. 

We haven't dirtied our hands in a while...

Contacted Andy for guidance. He gave us a good pep talk and some advice.

Professor Jane Wang gave us a tour of the Vicon motion tracking system on the 2nd floor of Kimball. If we intended to use it, we would have to:

• go through an initial training period
• calibrate the video system for each run
• find a way to transport a stationary erg, erg on slides, and dynamic erg from Teagle to Kimball, if we want to compare the acceleration graphs among the different ergs.

I'm not sure that using this system is worth our time, but I've emailed Gregg, one of Andy's students who has used Vicon, asking for his advice.
Andy's suggestions to reduce slippage of handle chain on sprocket:

• increase number of teeth the chain engages - currently we can't find a way to implement this
• increase back tension on the chain - this can be solved by increasing pre-stretch in the handle bungee system in the erg's bar.

Happy Fourth!

As mentioned last week, there were lots of strange clinking sounds coming from the erg and the pull during the drive didn't feel quite right and the chain was always slack. For this week, we worked to fix these problems. 

Noise & Alignment
- We tightened all nuts and bolts to eliminate rattling of loose parts as the cause of the problem
- The drive sprocket and casing which contains the differential are supposed to be parallel and attached with screws and bolts. We worked to align these in a more parallel fashion, but it was difficult to completely tighten the bolts, due to the tight spacing of the differential.
- We decided that we had previously loosened the flywheel chains too much, and thus tightened the chain again. 

Chain & Parts
- Parts of different chains keep falling off. One of links of the chain attached to the stiff bungee split off and we had to buy a master chain link from the Bike Shop in Collegetown to replace it. 
- We need more u-clips to hold master links together... they get bent easily and break.
- We noticed that not having long enough threaded rods or threaded rod of a certain diameter was a reoccurring issue, so Andrea ventured to get threaded rod from Lowe's. She bought 3 feet each of the 3 smallest diameter rods: 8-32, 10-24, 1/4"-20. However, she did not get nuts for them, in case we have the correct sizes already at the lab. We tried the existing nuts (which did not fit) and purchased 10 of each type from the stock room in Clark Hall's Shop.

Bungee & Frame
- In an effort to make the stiff bungee contraption more accessible and fixing parts easier, we took the stiff bungee out of the top bar and fixed the ends along the vertical bar holding up the smaller wheels. We then transferred the soft spring into the top bar, since we usually do not need to access it and the way that it is assembled (on a flexible rod) makes good use of the enclosed bar to support itself. Now the erg feels a lot lighter in mass!
- There was still too much slack in the handle chain, so we while we tried out the erg, we temporarily wrapped some of the chain around the handlebar and pulled. This turned out to be a bad idea, as now the chain was too tight, resulting in the soft bungee coming unraveled in the rod attachment (which is now in the top bar). Getting the bungee+rod contraption out proved to be quite a challenge, as bolts holding down the bar kept getting in the way, and the pulleys of the rod were also tightly fitted. This made us contemplate building a see-through frame for both stiff and soft bungees so that they could both be out in the open. I worked on the CAD for that. 
Figure 1: Collage of CAD. The see-through frame would be bolted on the top bar(only existing bar on erg) and vertical bar (attached to smaller wheels in the back). Material is probably acrylic or some sort of thin, see-through plastic. L-shaped metal hinges (fixed and movable) would keep the shelves attached together. There are a total of 3 shelf compartments (1 for the stiff bungee, 1 for the chain attached to the stiff bungee as it loops around the axle sprocket, 1 for the soft bungee). One side can have hinging "doors" to allow access to the compartments during fixing, and provide walled support otherwise. Attachment to the footrest (slanted bar in front) is tricky and will need to be better thought out. See 2013.06.19 Bungee_Spring_shelf idea folder on BOX for official files.
Presentation
- We Skyped Prof Ruina to further hone the presentation for Concept II. For data, he said that the current graphs (position, velocity) that we have are not useful, and that acceleration graphs are what is more important. We will need to contact Ellen to obtain those, and to get higher quality graphs. 
- We also whittled down the slides to reflect the main ideas of our design. I would like to have more numbers and data to prove our point that Rowgometer is better than the other ergs, however. 

July commences with an onslaught of cloud blood.

Today we used Cabrillo's Tracker program to finish gathering the position vs. time data from each of the four videos (stationary, slides, dynamic, Rowgometer). The only problem seems to be that every three frames are the same; Tracker seems to calculate velocity - 1st derivative of position - merely as the slope between two successive position data points, so there are many near-zero values in the velocity data, and the acceleration is pretty wonky.

Ex. three successive position data points:

...and the corresponding velocity data points - note the zero value in the middle:
These jumps in velocity values result in acceleration values that oscillate from negative to positive to negative and so on. We need a smoother graph of acceleration.

EDIT: This redundancy problem is easily solved by changing the frame step size to 2 or 3 (located to the right of the frame slider, underneath the video display).