Getting the Most Out of Your Video

Discussion in 'Tennis Tips/Instruction' started by diredesire, Oct 2, 2006.

  1. diredesire

    diredesire Super Moderator

    Joined:
    Mar 16, 2004
    Messages:
    6,712
    Lots of threads have been popping up recently about videos, calculating speeds, and other related topics. I'm going to (for now) sticky a thread where you can post links and other resources for how to get the most out of your video...

    Thanks to user Mike Cottrill for suggesting this thread, and the first link of the sticky:

    http://tt.tennis-warehouse.com/showthread.php?t=99228
    "Calculating Serve Speed from a digital Video"

    Please help out by adding more links!
     
    #1
  2. MasturB

    MasturB Hall of Fame

    Joined:
    Sep 12, 2006
    Messages:
    1,727
    If you want to check your speed, use a free program like Windows Movie Maker, pause it at the point of contact and fastforward it by frame.

    It's automatically set to 30 FPS (I think)
     
    #2
  3. Mike Cottrill

    Mike Cottrill Hall of Fame

    Joined:
    Sep 3, 2004
    Messages:
    1,626
    Looking for consumer Camcorder for sports? Sony DCR-DVD505

    Looking for consumer Camcorder for sports? Sony DCR-DVD505 Smooth Slow Recording 240fps

    Looking for a consumer Camcorder for sports? As of 10-3-2006, Sony has a selection of camcorders that can record short clips at 240fps. Great for analyzing strokes and for calculating serve speeds.
    Sony:
    “Smooth Slow Recording
    By increasing the record rate from 60 fields per second (fps) to 240 fps for 3 seconds allowing you to capture 3 seconds of fast motion and play the video back in 12 seconds. This is great for analyzing a golf swing or a viewing a bird in flight”
    At times in normal mode, it appears that the camera playbacks 30 or 60fps. Most likely just an interlacing adjustment the camera does with on board processing.

    I personally have this one:
    [​IMG]

    3 seconds if enough time to capture the serve motion and the ball reaching the fence if serve is at least in the upper 80’s. Resolution in smooth slow recording mode is far less than in normal mode, but still very useful for analyzing sport strokes. Nice camera for every day recording as well.

    Here is a link to the selection of camcorders:
    http://www.sonystyle.com/is-bin/INT...mcordershop&ref=http://www.sony.com/index.php


    Examples and editing to come soon.
     
    Last edited: Jan 11, 2007
    #3
  4. onehandbh

    onehandbh Hall of Fame

    Joined:
    Dec 19, 2005
    Messages:
    3,071
    Unfortunately I "upgraded" to windows moviemaker2. It added editing
    features but reduced fps down to about 14 or 15. Anybody know how
    to get back this feature?
     
    #4
  5. maverick1

    maverick1 Semi-Pro

    Joined:
    Jul 23, 2006
    Messages:
    573
    I am pretty sure it depends on what format you choose when capturing.
    The default format is a highly compressed one. If you choose DV-AVI, it will take up a lot of space but it will preserve all the information that is on the tape.

    There are intermediate formats that compress and still keep 30 fps.
     
    #5
  6. maverick1

    maverick1 Semi-Pro

    Joined:
    Jul 23, 2006
    Messages:
    573
    Onehandbh,
    You seem to be right that Windows movie maker displays about 15 fps in its preview window, even if you capture DV-AVI.

    But the captured file does contain 30 fps. You can use a 3rd party player that shows frame-by-frame.

    For example, Elecard. A 21-day trial version is free(http://www.elecard.com/download/index.php)
    To see frame by frame, you have to pause and hit the arrow keys. BUT there is a setting that controls how much each arrow key advances to film. Make sure you set it to 1 frame.
     
    #6
  7. onehandbh

    onehandbh Hall of Fame

    Joined:
    Dec 19, 2005
    Messages:
    3,071
    hey maverick1,

    Thanks for the tip on elecard. It's a nice, simple video player.
    Not sure if I totally understand the whole frame counting
    thing. Here's a video of my old service motion.
    how many frames would this be? (and speed?)


    http://media.putfile.com/old-service-motion
     
    #7
  8. maverick1

    maverick1 Semi-Pro

    Joined:
    Jul 23, 2006
    Messages:
    573
    Onehandbh,

    Putfile doesn't allow me to step through the video.

    The serve seems to have nice form and speed, but I am not experienced enough to judge speed from watching.

    If you are computer-savvy enough to download Elecard and take a look, I am sure you can do the rest.
    Open up the captured file(should be a large one, 1 Gigabyte for every 2 minutes) in Elecard, pause it before you are about to hit the serve.
    (I don't have this software at work, so this is from memory)
    There should be something at the bottom of the Elecard controls window to click, that expands the window to display more controls. One of the controls is for how much the video advances every time you press the left/right arrow keys. Set it to the minimum(1 frame or 1/30th of a second).

    Then press Right Arrow until you reach a frame where you have just hit the ball. (It is likely that the you won't catch the exact instant of contact, and ball is already on its way. In this case, just use your judgment to estimate a half or 1/3rd frame or whatever)
    Then keep pressing the Right arrow key until the ball lands on the court, and of course count the number of times you had to press the key, and let me know.
    Please try it and if you still have trouble I will do my best. It would be a pity to give up now after you have done 95% of the work.
    You can send me an email by clicking on my alias on the left.
     
    #8
  9. onehandbh

    onehandbh Hall of Fame

    Joined:
    Dec 19, 2005
    Messages:
    3,071
    Okay.

    I get 13 frames: if I don't count the frame at the moment of contact w/racket
    and click until it hits the court.(and including that frame where the ball
    hits the court) Is that right? According to the calculator (for 30 fps)
    the serve would be 111 mph.

    If count the frame at the moment of contact then 14 frames would come
    out to 103mph.
     
    #9
  10. maverick1

    maverick1 Semi-Pro

    Joined:
    Jul 23, 2006
    Messages:
    573
    It sounds right that you should NOT count the frame at contact.
    Congratulations, that is very fast.
    Make sure you account for the distance correctly. It is 60 feet you strike the ball directly above the baseline and it lands on the service line, down the middle.
     
    #10
  11. onehandbh

    onehandbh Hall of Fame

    Joined:
    Dec 19, 2005
    Messages:
    3,071
    maverick1,

    Do you happen to know which web/video hosting sites allow you
    to step through video frame by frame (and preserve the original 30fps rate)?
     
    #11
  12. maverick1

    maverick1 Semi-Pro

    Joined:
    Jul 23, 2006
    Messages:
    573
    I haven't found any. Not youtube nor google.

    You can try the filesharing sites like putfile.com. You can edit in Windows movie maker and produce a 0.5 second to 1 second video with 15 to 30 frames. Even at high resolution, it should be a pretty small file. You can upload to one of these sites. Users will still need something like Elecard to step through it.
     
    #12
  13. onehandbh

    onehandbh Hall of Fame

    Joined:
    Dec 19, 2005
    Messages:
    3,071
    how to estimate serve speed (using maverick1's method)

    Here's an example on how to use maverick1's serve
    speed calculator (using my one of my serves). (with lots of
    help from maverick1!) Hope it helps others how to use his
    calculator.

    Video was recorded at 30 frames / second (29.97). I used
    the elecard video player on my computer that maverick1
    suggested. Here is a frame by frame jpeg captured from
    the video using elecard:

    [​IMG]

    1) Calculate the distance the serve travels (and adjustments):
    60.0 feet (distance to T)
    + 0.5 feet (for additional distance from striking ball 8+
    feet above the court. I am 5' 11 3/4")
    + 0.5 feet (my serve was in the middle of the box)
    - 1.0 feet (distance inside the service line the serve landed)
    - 1.5 feet (distance from baseline at contact point.
    I land inside the court)
    = 58.5 feet traveled by ball.

    2) Calculate total number of frames (time) the ball is in the
    air.

    a) Frame 1 appears closest to the point of contact.
    According to maverick1, if you look carefully you'll
    realize that it is actually just before contact. The main
    clue is that arm is not vertical yet - it is leaning back.
    Also, the ball is still in the same focus and is not
    blurry or squashed from the contact yet. (looks the
    same as frame 0)

    b) In frame 14, it appears that the ball has struck the
    court and in frame 15 it is clearly bouncing back up.

    c) 14 - 1 = 13 frames. Then subract 0.5 frames since
    frame 1 is just before contact.
    So we have 12.5 frames.

    3) Plugging 58.5 feet and 12.5 frames into maverick1's
    calculator (http://mavericks.cc/tennis/serve_speed.html)
    I get:
    112.32 mph.

    It seems like a bunch of steps but once you actually get down
    to doing it it isn't hard and is actually kind of fun. Another
    thing that maverick1 pointed out was that the racquet
    travels about 90 degrees from from 0 to frame 1 and that
    shows the racquet head speed.
     
    #13
  14. drakulie

    drakulie Talk Tennis Guru

    Joined:
    May 3, 2004
    Messages:
    24,465
    Location:
    FT. Lauderdale, Florida
    One hand, are you sure the way you are calculating is correct? I ask because this photo, along with the video were provided to mav and mike when they were working on their formulas. They used it to see the accuracy of their formula compared to the radar gun I used in the video. The ball landed just inside the baseline up the T-it almost landed on the line. (60 feet). The radar in the video shows 101 mph, and using the calculator I got 103+. However, doing it the way you are I would get a speed of 115.7 MPH. That just doesn't sound accurate to me. That is a difference of 14 mph between the formula and the radar. It would mean the original videos I posted were between 118-122 MPH, which I know can not be possible.

    http://www.youtube.com/watch?v=5EJe64Ky7rE

    [​IMG]
     
    #14
  15. maverick1

    maverick1 Semi-Pro

    Joined:
    Jul 23, 2006
    Messages:
    573
    Drakulie, What do you think he is doing differently from you?
    The distance estimates seem reasonable. As far as frames go, I thought 12.5 was a good estimate and I suggested it to him, but now I think it may be closer to 13.

    Onehandbh has numbered his frames from 0, but since the 0th frame is well before contact, I think your sequence and his sequence can be compared very nicely.

    I think in frame 1, his wrist is still cocked back so it is some distance from contact. So I thought it was reasonable to assume contact happened at frame 1.5. But considering how much the racket moved between frame 0 and frame 1, contact probably happened at frame 1.1 or 1.2. Since the ball seems to be touching the floor in frame 14, the net number of frames is 12.8 or 12.9

    In your frame sequence, your first frame is closer to contact than onehandbh's "frame 1" because your racket has passed vertical.

    The ball is touching the court at frame 14 for both you. So I would say your serve took about 0.1 frames longer but it landed closer to the service line. I don't want to speculate on which serve was faster.
     
    #15
  16. drakulie

    drakulie Talk Tennis Guru

    Joined:
    May 3, 2004
    Messages:
    24,465
    Location:
    FT. Lauderdale, Florida
    Mav, the reason I asked was because the following confused me:

    "c) 14 - 1 = 13 frames. Then subract 0.5 frames since
    frame 1 is just before contact.
    So we have 12.5 frames."

    14 - 1? then another .5 was subtracted. I still don't get it.

    Wouldn't it be: 14 - .5 frames = 13.5? Where did he loose another whole frame?

    On mine, there are 14 exact frames between the very first frame and when the ball strikes (forget about the last two frames). On mine, I have not yet made contact with the ball (almost there), on frame 1, although it is extremely close. On his # "1" frame, contact has been made as you can see the ball blurred.

    So I was curious how he lost 1.5 frames, even if he had not made contact on frame 1, and both of our serves, as you said landed on frame 14.
     
    #16
  17. maverick1

    maverick1 Semi-Pro

    Joined:
    Jul 23, 2006
    Messages:
    573
    May be it will easier to see if we count in miliseconds.

    The interval between frames is a little greater than 33 milliSeconds.
    Let us say the 1st frame is taken at time 0, the second would be at time 1x33 milliseconds, the 3rd would be at 2x33 milliseconds...the 14th frame at 13x33 = 429 = milliseconds.

    If contact between racket and ball happened half way between the first and second frame, that means contact happened at time 16.5 millisecs. If the ball hit the court in the 14th frame, that corresponds to a time of 430 millisec. So the time of travel is 429 - 16.5 = 412.5 milliseconds
    412.5 / 33 = 12.5 frames.

    The 0.5 frame may be too much of a deduction and should be .1 or .2, but that is a different issue.

    By the exact same process as above, your serve comes out to 13 frames.

    It may be confusing because he has included an extra frame at the beginning that you don't have. If you count from his second frame(labeled "1") and your first frame, it is exactly the same number of frames to bounce.
     
    #17
  18. onehandbh

    onehandbh Hall of Fame

    Joined:
    Dec 19, 2005
    Messages:
    3,071
    drakulie,

    Basically, just count then number of frames (from but excluding
    frame at point of contact w/racquet) up to and including
    the frame where the ball strikes the court.
    On both of our videos that comes out to 13 frames.
    But frame 1 of my video is slightly prior to contact verses
    your frame 1. Hence maverick1 suggested I subtract 0.5.
    (but corrected to 0.1 or 0.2 now).

    There are other variables involved like height of contact point,
    forward lean of body, depth of actual serve, and location in
    the service box so it's hard to say which serve traveled faster
    despite mine taking less time to reach the court.
     
    #18
  19. drakulie

    drakulie Talk Tennis Guru

    Joined:
    May 3, 2004
    Messages:
    24,465
    Location:
    FT. Lauderdale, Florida
    How?

    In my first frame I haven't made contact and his first frame labeled "1" he hasn't made contact; and both of our serves make contact with the ground on frame 14.

    How do you calculate "13" for mine, and 12.5 for his?

    This does not make sense.

    In addition, 13 frames for mine would mean my serve was actually 111 mph. A difference of 10 mph from the radar, which is a lot.

    I'm aware of that, and did not take that frame into consideration.

    Exactly. So how do you conclude 12.5 frames for him and 13 for mine. Even if I am little closer to contact on frame 1.

    Sorry for the confusion mav, but this makes no sense.
     
    #19
  20. drakulie

    drakulie Talk Tennis Guru

    Joined:
    May 3, 2004
    Messages:
    24,465
    Location:
    FT. Lauderdale, Florida
    OK, so far we are on the same page.


    Actually, I haven't made contact on my frame 1 either. Here is where the confusion begins.

    Going by how you are doing it, that would mean I should put 12.5 in the calculator as well. Which would come out to 115 mph serve. That is a difference of 14 mph between the radar and the formula. This is a huge difference.

    Even if I go to 13 frames (saying I am closer to contact than you on frame 1) I get 111 mph. A difference of 9 mph between formula and radar which is still a huge difference.

    By the way, I'm not concerned about who's serve is faster, rather more concerned about the accuracy of the formula.

    PS: forgot to mention in my first post, you have a real nice service motion:)
     
    #20
  21. onehandbh

    onehandbh Hall of Fame

    Joined:
    Dec 19, 2005
    Messages:
    3,071
    drakulie,
    Actually I like your motion better. It's more efficient. I'm trying to incorporate
    more shoulder turn, coil, a higher toss, smoother uncoiling -- a lot of the things
    your serve already has. btw, great recommendation on TT (for the "sonic serve" video).
    Just got it today. It really breaks it down into simple steps.

    My eventual goal is to be able consistently hit about 120 w/good placement and disguise.
    Not sure if it's possible or even realistic since I don't practice serves a lot and I always
    have some topspin on my serves. but if I can generate more racquet head speed then
    my kick & slice will have a lot more action as well.

    btw, did you notice that adding lead increased your
    serve power? I've never added lead to my PS 85's b/c I was
    afraid it would be harder to hit reaction volleys with in
    doubles.
     
    #21
  22. maverick1

    maverick1 Semi-Pro

    Joined:
    Jul 23, 2006
    Messages:
    573
    Drakulie,

    I think we are very close to agreeing.

    As I mentioned in both my posts, the deduction of 0.5 frames was too high and it should probably be 0.1 or 0.2 frames.

    As you say yourself, you are a bit closer to contact(racket looks past vertical) than he is(looks not vertical yet). The deduction for you should probably be 0.1 frames or less.

    Looks like your confusion would be resolved if we say your framecount was 12.9 and 1hbh's 12.8. At this minute level, obviously I don't really know. I am pulling numbers out of my *** :) If someone else thinks they should be 12.75 and 12.7, I can't argue with that.

    Sorry it took so long to clear up. For a while, I wasn't sure whether your issue was deduction of the extra frame at the beginning of the sequence or the 0.5 frames for not yet having made contact in the next frame.
     
    #22
  23. drakulie

    drakulie Talk Tennis Guru

    Joined:
    May 3, 2004
    Messages:
    24,465
    Location:
    FT. Lauderdale, Florida
    Hey Mav, hope all is well. My confusion is more in part with the accuracy of the formula. Even If I was to plug in "12.9" frames to my serve, the speed would be 109.924 MPH, which is still very high compared to the radar. (+8.9 mph difference).

    The way we had originally calculated my serve speed, there was only a difference of 2-3 mph between the calculator and the radar.

    This is a big spike in serve speed.

    Also, I know my radar is very accurate and calibrated. I have taken it and compared it to those professional baseball radar guns, and the difference is
    minus -1 to 2 mph. A lot of times it displays the same speed.
     
    #23
  24. Mike Cottrill

    Mike Cottrill Hall of Fame

    Joined:
    Sep 3, 2004
    Messages:
    1,626
    The formula is good. It is the “d” and the fps that must be input that draw differences. The small differences in frame count make the difference. The frame count becomes more clear with the 240 fps thus the more accurate a calculation.
     
    #24
  25. maverick1

    maverick1 Semi-Pro

    Joined:
    Jul 23, 2006
    Messages:
    573
    Even if the radar is perfect, it can only measuere the speed when it "sees" the ball. The ball slows down by at least 3 mph in the first 5 feet.
    I once suggested to you a simple experiment to test the range of your radar - place it on the other side of a fence and hit some simple forehands at the fence. Vary the distance between the fence and the radar until you find the maximum distance at which it shows the ball speed. That is the range of your radar.
    Once you know the range, we will be in a much better position to discuss the accuracy of the formula.

    The other thing is you may be overestimating the distance you plug into the formula. You could have struck the ball deeper inside the court than you thought, or the ball could have landed more than a few inches short of the line. People underestimate distance from the line when looking at the bounce from the other side of the court. When I feel like I just missed a shot long, I sometimes ask by how much it missed, and I am invariably shocked when they 1 foot or even 2 feet. Other times, I think the ball was on the line but they will tell me it was well in.
     
    #25
  26. drakulie

    drakulie Talk Tennis Guru

    Joined:
    May 3, 2004
    Messages:
    24,465
    Location:
    FT. Lauderdale, Florida
    yeah, I haven't done that yet, but will give it a go next time I get the chance, and follow up with you.

    The spot where the ball landed left a mark so the distance is accurate. In fact, the day I took the video, I kept serving over and over until I got a serve as close to 100 mph, and right on the cross of the T. That is why I used this particular serve. It was 101, and right on the T.

    Again, when we first did the calculation, the formula and the way we were using it calculated my serve speed at around 103+ MPH (+2 to 3 mph), from the radar speed of 101.

    Now it is calculating it at 109-111 MPH (+8-10 mph faster) than the radar.

    No way this is accurate. Maybe it is not the formula that is innacurate, but the way we are using the formula.

    If I stay at 13.5 frames I am much closer to the radar's 101, than if I go to 12.5 frames, 12.9 frames, or 13 frames.
     
    #26
  27. maverick1

    maverick1 Semi-Pro

    Joined:
    Jul 23, 2006
    Messages:
    573
    The one thing I am sure about is the number frames. It is 12.9 or 12.95 or something like that.

    If the formula is wrong, it is wrong. So far I have no reason to believe that. But it is not for me to pronounce it correct. I have posted all the details of the formula on the website along with my email address, but I have not had any comments on its vailidity.

    We will have to wait for you to do the range test now. If you come back and say the range is 25 feet, that would explain an 8 mph difference. You have seen yourself that the speed registered drops rapidly as you move the radar further away from you.

    You may think that you have it placed it just the perfect distance because moving it any closer doesn't increase the speed resgitered. However, as you move it closer, a different reason for understating speeds kicks in - the angle beween the line of the ball's movement and the radar's line of sight to the ball becomes larger and larger. For example, if you place the radar 9 feet from the baseline, and make contact at a height of 9 feet, the radar will show a speed that is 70% of the actual speed.

    Having a technical background, I am all too aware of the two above mentioned deficiencies of low end radars. This is why I haven't yet got it even though I am curious about speeds and the price is less than that of a racket.
     
    #27
  28. maverick1

    maverick1 Semi-Pro

    Joined:
    Jul 23, 2006
    Messages:
    573
    http://fcchandler.home.comcast.net/stable/index.html

    This is a free video editing software that seems to be able to play most formats of video, though it doesn't have the frills of commercial software.
    Most importantly, you can use arrow keys to step a frame at a time.

    It doesn't have an installer. You have to unzip the files into folder.
    It is convenient to create a shortcut on your desktop for the main program, VirtualDub.exe
     
    #28
  29. Mike Cottrill

    Mike Cottrill Hall of Fame

    Joined:
    Sep 3, 2004
    Messages:
    1,626
    Mav,
    Nice little program for free. Better than wind DVD from what I have seen so far. It appears to show 240 fps as 120 fps and video is clearer. It must be the interlacing thing. Cyber Link still shows frames a little better IMO but video is less sharp. Virtual Dub program counts 53 frames and Cyber link shows 106. Another video in normal mode shows 30 frames with cyber link and ~15 or 16 with Virtual Dub. Odd thing with Wind DVD is appears the ball lands out but it is clearly in with the other two programs. Counting the frames accurately is downfall with 30 fps video.
    Odd thing, as I remember I did get a little more on the slow mo serve, but it did not seem like that much more. The serves are just about identical.
    Nice program.
     
    #29
  30. onehandbh

    onehandbh Hall of Fame

    Joined:
    Dec 19, 2005
    Messages:
    3,071
    maverick1,

    I believe you made an error in calculating the number of
    frames in drakulie's video. Go back and replay the
    serve video and you will find that there is actually one
    more entire frame that is before the first frame in the
    jpeg sequence of drakulie's serve that will show the
    actual contact point. The ball has already traveled about
    a foot at the point of the first frame in the montage of
    serve sequences posted earlier that was used to count
    the number of frames in his serve video. The
    ball just happens to be in front of the racquet from the
    angle of the camera and that's why it *appeared* to be the
    contact point but is not.

    If we assume that the total distance traveled by the ball
    is: 60 feet - 6 inches (distance inside service line) and
    drakulie contacted with the ball at 1 foot inside the service
    line (since drakulie said he landed 1 -2 feet inside the line)
    then we get 58.5 feet as the distance.

    Then plugging 13.9 frames into the calculator, we get...

    ** 101.008780 mph **

    ... which is stunningly close to what the radar recorded.
    Very well done creating the formula!



     
    #30
  31. drakulie

    drakulie Talk Tennis Guru

    Joined:
    May 3, 2004
    Messages:
    24,465
    Location:
    FT. Lauderdale, Florida
    This is completley incorrect.

    The "16 frame action shot" is a frame by frame (one after the other) captured from the video to the software that comes with the camera. Contrary to what you state, in the 1st frame in the "16 frame action shot", the ball has still not been struck.

    The software inside the actual camera allows you to view frame-by-frame, while the entire video is still saved in the camera. The frame before the first frame I posted, in the "16 frame action shot", clearly shows the racquet head is not close to contacting the ball, and has not yet reached a completely vertical position, therefore I did not bother including it. So there is no way the ball has been struck on the frame before.

    In addition, 3 things do not support what you are stating:

    1. The camera was not placed directly behind me, rather behind me and to my right on the fence. I am serving from the left side of the center hash mark, and the direction of the serve is going to the right (AD Side). In other words it is traveling into the "sites" of the camera, not away. So if it has already travled a foot,,,,,,, because of the position of the camera and direction the ball is traveling you would clearly see the ball without any obstruction from the strings.

    Your scenario would only apply if the camera were directly behind me and/or more to the left of me for this to be possible.

    2. If I already struck the ball, and it traveled a foot, my hand, arm, and racquet face would already be pronating downwards, with the racquet face side that hit the ball going to the right (like in your frame numbered "2"), which they are clearly not in the first frame. However they are in the second frame.

    3. This radar does not pick up the ball immediately at contact, rather as the ball approaches an area in front of the radar. So we know for a fact the speed at contact is faster than the speed shown on the radar. In other words, for the video I recorded at 101 mph, the actual speed is 2-3 mph faster at contact. 103-104 mph, at minimum.

    For your calculations to apply, it would mean the radar speed would have shown 97-99 mph, and the calculator would show 101.
     
    #31
  32. drakulie

    drakulie Talk Tennis Guru

    Joined:
    May 3, 2004
    Messages:
    24,465
    Location:
    FT. Lauderdale, Florida
    Hey Mav. I did the "fence test", LOL.

    The radar begins to have problems picking up speeds outside of the 18' mark. So, when it gets with in the 18' foot range it starts reading the mph.
     
    #32
  33. onehandbh

    onehandbh Hall of Fame

    Joined:
    Dec 19, 2005
    Messages:
    3,071
    [​IMG]

    drakulie,
    First, Here is a description of the frames in the picture:
    FRAMES p1 to p4 are 4 consecutive framesof the toss as it is going up to
    show the path of the ball.
    FRAMES -2, -1, 0, 1, 2, 3, 4 are consecutive frames leading up to the
    moment of impact and after it. Frame 0 is closest to where contact
    occurs.

    I'll provide a short gist of the analysis followed by a more detailed one
    below to fully explain my analysis. I think what happened is that your
    camera gave you 16 frames but what happened is that from the
    small LCD display it looked like the contact point but actually it
    wasn't. I would have thought it was too, until I was able to see the
    previous frames in detail using the elecard software, which can
    capture all video frames individually in sequence.

    I apologize if it comes across abrasive or anything. I was just trying to
    find out why there was just a discrepancy between the radar and formula.

    Here's is the short explanation of how I know there is one more frame.
    Looking at the sequence of frames:

    - The toss moves the ball from right of the picture to left, as is consistent with pros' tosses. Moving in a parabolic motion from right to left.
    - In this case the ball has clearly moved to the right from one frame to the next, so the ball must have been struck to cause it to move back toward
    the right.

    ---- warning: long detailed analysis below ----------------

    Frame:
    (p1 - p4):
    We see that the ball is traveling upward on the right side of the little
    tree. It appears to go up in a more or less straight path.

    (p4):
    Here we see taht the ball exited on the ride side ofthe little tree.

    (-2):
    The has not droppe back into view of the camera yet. It will come in at
    about the peak of the little tree.

    (-1):
    If you look carefully, you will see that the ball has just begun to enter
    the frame where I have drawn a red arrow.

    (0):
    The has continued to drop into frame and the racquet has either made
    contact or is about to make contactd. Just like the ascent, it is traveling
    in a more or less straight path down. It is, however, slight to the
    left of where it exited the frame on its way up. This tells me the toss
    traveled it a parabola going right to left.

    (1):
    The ball has already been struck. How do I know? Because it is now A
    LITTLE TO THE RIGHT of it's previous location in frame (0). This cannot
    be due to the toss because the toss has a slight RIGHT to LEFT
    path. Therefore, the ball must have been struck. The serve is going
    toward the ad court so this makes sense. Why does the racquet appear
    to be in front of the ball? Because the camera is slightly to the right
    of the center line and drakulie has begun his pronation going left to
    right. Just by coincidence of the location of the serve and camera, for a
    brief moment the racquet is still in front of the ball. This is probably why
    you thought that was the contact point.

    Also, the racquet angle with respect to the court of the frame in which I strike
    the ball is very similar to the frame 0 in the picture above. If your serve has
    a lot of topspin then it is unlikely that the angle would facing so much more
    down toward the ground than my serve, which has topspin but is not my
    heavy kick serve.
     
    #33
  34. maverick1

    maverick1 Semi-Pro

    Joined:
    Jul 23, 2006
    Messages:
    573
    FWIW,
    I think onehandbh made a good observation wathing the ball relative to the tree in the background. It proves that contact has already happened in the frame that we previously thought was just prior to contact. So we are talking total frame count over 13 rather than below 13.
     
    #34
  35. drakulie

    drakulie Talk Tennis Guru

    Joined:
    May 3, 2004
    Messages:
    24,465
    Location:
    FT. Lauderdale, Florida
    Onehand, don't worry as I did not take your post as abrasive in any way. As I said, I am as anxious as you to ensure the formula is accurate.

    Your above post is a very good one, very good sound observation, and as you said, I would be unable to see this detail on the small LCD screen. My apologies! :)

    Unfortunately, taking your observation-- then it is clear proof the formula is innacurate.

    As I pointed out in my previous post, "point #3" ---With your observation and calculations you got ** 101.008780 mph ** using the formula. The radar showed 101 mph. There is no way, for the radar and the formula to coincide so decisively. If my serve is 101.008780 using the formula the radar would have to show a much lower recorded speed.

    Remember, the ball has already slowed down significantly when it has entered
    the radars path in which it picks up the ball speed. Therefore, if the formula were accurate, then the radar would have displayed a speed much less than 101.

    The radar was placed for this serve approximately 30 feet from the baseline. Based on the test I did today, the radar did not pick up any serve speeds 18 feet away. It began consitently picking up speeds inside and including the the 15 ft range, although it did pick up a couple from 16, and 17 feet, but not as consistently as 15 feet. So, in my serve the ball travled at least 15 feet (almost to the service line) before the radar calculated the speed (see frame # 7 of my original action shot when the radar light shows a display for the first time).

    Maverick has previously pointed out, the ball loses about 3 mph in the first 5 feet. So the radar would have shown at minimum a speed 98 mph not 101. In my case, the ball traveled an additional 10 feet, in which there would be even more decreas in speed than 98 mph.

    So one of two things is possible:

    1. The formula is innacurate, and actually calculates a slower speed than actual speed at contact---or;

    2. Pilor error-- The number of frames that are suppose to be counted and entered into the formula is innacurate.
     
    #35
  36. maverick1

    maverick1 Semi-Pro

    Joined:
    Jul 23, 2006
    Messages:
    573
    1hbh used 13.9 frames in his calculation.

    I don't agree that 0.9 frames have passed after contact. I just calculated that, in 0.9 frames, a 100mph ball should travel 4.4 feet. It clearly hasn't traveled that far. Because of the angle, It is very difficult for me to say if the ball has traveled 1 foot or 2 feet in this frame. If it is 1 foot, you should be using 13.2 frames, which gives a speed of 106.37 mph with 58.5 feet of distance.

    That gives you a reasonable 5.37 mph for the speed by which the ball slowed down beore the Radar picked it up.
     
    #36
  37. drakulie

    drakulie Talk Tennis Guru

    Joined:
    May 3, 2004
    Messages:
    24,465
    Location:
    FT. Lauderdale, Florida
    I think we could all agree on this.

    Actually, in this scenario, the ball traveled an additional minimum of 11-13 feet before the radar picked up the speed of the ball, in which there is more loss of speed, that are still unnacounted for.

    Mav, hope you haven't lost your mind yet. LOL :)
     
    #37
  38. onehandbh

    onehandbh Hall of Fame

    Joined:
    Dec 19, 2005
    Messages:
    3,071
    LOL. hey maverick1, just to make things even more complicated for you,
    some more things to take into consideration:

    - condition of the balls and type of balls. (how fuzzy and old are they?)
    - Which brand? (e.g. new dunlop balls seem to be very fast relative to some
    of the other brands).
    - Also humidity and temperature of the outside air
    and balls.
    - elevation? (e.g. serves at mile high would be faster)

    drakulie, can your radar work if it points at the serve from the side?
    so that you can get the moment right after contact w/the racquet?
    Is that how the pro serves are measured?

    The angles of the serves in our videos are best for viewing but unfortunately
    it's hard when to see when the moment of impact is. It may be that the formula
    will always have a limiting factor due to the quality of the footage and our
    ability to figure accurately approximate the moments of impacts.

    P.S. drakulie, you should show those two frames to ball control tennis &
    tell him/them/it that you mastered the prolonged contact serve where
    you can keep the ball on the racquet for 2 feet, so there 4 ms
    contact is nothing.

     
    #38
  39. maverick1

    maverick1 Semi-Pro

    Joined:
    Jul 23, 2006
    Messages:
    573
    It is definitely possible to tweak the estimated numbers to get exact agreement.

    First, if the Radar detected the ball after about 11 feet, I would expect a slow down of about 6 mph. (If the slowdown is 3 mph in the first 5 feet, the next 5 feet would see a smaller decrease).

    The distance traveled in that frame could be 6 inches rather than 1 foot, in which case the frame count would be 13.1 and not 13.2
    The total distance might have been 58.9 instead of 58.5. That gives 108 mph
    The Radar may well have rounded 100.5 to 101. That gives a difference between radar and formula of 7.5 mph.

    So we now have *too much* of difference - we expected 6 mph difference but have 7.5 mph.

    I hope you are now getting the drift. The entire error can be explained by an error in estimated inputs, without needing to question the formula itself. Of course, this doesn't PROVE that the formula is right. All I am saying is there is no reason to doubt the formula; there has never been in my mind throughout this thread.

    Of course, you can question the usefulness of the formula if it is possible to get a 5mph difference by estimating the frames/distance slightly differently. But the Radar is also liable to give very different readings depending on how far away and at what angle it is placed.

    To answer 1hbh's points about the effects type of ball, humidity etc., No model of a real world phenomenon can take into account every variable that might have an effect on the phenomenon. The best you can hope for is that you have considered the most important variables and the effect of the rest can be ignored.
    It is true of every gadget you take for granted. The odometer in your car is counting the number of revolutions of a wheel and muliplying that by some constant to get distance traveled. The actual distance would be affected by wear in the tire, air pressure in the tire and other variables.
     
    #39
  40. drakulie

    drakulie Talk Tennis Guru

    Joined:
    May 3, 2004
    Messages:
    24,465
    Location:
    FT. Lauderdale, Florida
    Yeah Mav. That is why I say, it could very well possibly be pilot error when inputing the # of frames and distance into the calculator.

    You have been such a good sport, and very helpful. :) Do you work for Nasa?

    Anyways, this ones for you,;:

    For he's a jolly good fellow, For he's a jolly good fellow,For he's a jolly good felloooooow.....

    Which nobody can deny!
     
    #40
  41. drakulie

    drakulie Talk Tennis Guru

    Joined:
    May 3, 2004
    Messages:
    24,465
    Location:
    FT. Lauderdale, Florida
    The pros serves are measured at contact.

    I don't think the radar will measure the serve doing it from the angle you state. it has to be moving towards the radar.

    When I called the company that sold me the radar, they stated if I wanted to get the most accurate, as they call it - "out of hands" reading i should place the radar with in 5 feet. However, to do this while serving I would have to buy a tripod to mount the radar on, and there is the chance of hitting the radar, unless I placed the radar on the tripod and on the other side of a fence. But this would be useless as there would always be speculation as to whether "the serve" went in or not.
     
    #41
  42. onehandbh

    onehandbh Hall of Fame

    Joined:
    Dec 19, 2005
    Messages:
    3,071
    maverick1, I was just kidding regarding all those additional variables. Just the
    largest contributors should be take into account. The most difficult part
    still seems to be estimating exactly when the point of contact took place.:)

    drakulie, i don't recommend you put the radar on a tripod right in front
    of you given what happened last time it was on the fence:eek:

    Perhaps this will help: If we assume that in drakulie's serve the ball struck
    the sweetspot, we might be able to gestimate how much further the
    ball traveled down (due to gravity & the toss) before the racquet struck it.
    It probably could only have traveled about the distance of a ball or two
    before the racquet, which is moving very fast at this point, strikes it.
    Maybe even less b/c the racquet, which is moving very fast would have
    struck it. Can you figure this out maverick1? This might give a closer
    estimate as to how many frames (13.x?) it took.
    Drakulie seems to have a sampras-like high toss.
     
    #42
  43. maverick1

    maverick1 Semi-Pro

    Joined:
    Jul 23, 2006
    Messages:
    573
    1hbh, I only vaguely understand what you are hinting. It just seems too hard to do any more.
    drakulie, I work as a programmer in the financial indsutry. My work has nothing to do with this stuff, but I studied mechanical engineering in college, so that helps. (holy crap, I didn't realize I entered college more than 27 years ago! My degree is older than most people on this forum)
     
    #43
  44. The Gorilla

    The Gorilla Banned

    Joined:
    Nov 21, 2006
    Messages:
    2,495
    #44
  45. GRANITECHIEF

    GRANITECHIEF Hall of Fame

    Joined:
    Oct 4, 2004
    Messages:
    3,750
    Location:
    Santa Barbara
    I've done this before, but there is huge human error involved. If you look at the results, one tenth of a second difference = huge:
    .5 = 106mpm, .4=133mph, and of course one tenth of a second would be pretty good degree of accuracy for most humans.

    Better Solution, use an oscilloscope with a microphone attached that is rigged to actually start and stop a timer on the spikes. ie on the contact of the serve and contact against a piece of plywood placed at the baseline.
     
    #45
  46. Koensayr

    Koensayr Guest

    Granite, I think there's actually a very workable method in there.

    However, I don't think using a plywood is a good idea because that will take into account of the time it took from the ball hitting the court, and then bouncing back up and into the board.

    How about this idea:

    Take a portable digital camcorder and put it on the side so that you can get a shot of the whole court. Then, the person take a serve, and record it with the camcorder. Now, by analyzing the wave file (make an audio dump of the audio from the video, this is why you need a digital camcorder to make it much easier), we can calculate the time between the racquet hitting the ball, and the ball first hitting the court. The problem with using video is the frame rate of the video, however, the typical sampling rate for audio with a digital camcorder I think is 32kHz, which is 32000 samples per second, WAY better than the 13 frames per second or so of video.

    Now, why use a camcorder instead of just an audio recorder? Well, we now need to calculate the distance from the 2 points of contact. We have the time between those 2 points from the audio file. However, without a video, we have to make assumptions about the path that the ball took, which could mess up the actual distance the ball traveled. Now, from the video, there is probably a way to combine all the frames, since the background should be the same, the ball in each frame should trace out the path that the ball took. From this we have the distance the ball traveled between contact with racquet and contact with the court.

    Now, we all know what happens when you divide that distance by the time we calculated earlier :-D


    (edit: hmm, maybe I should have patented the idea before I posted it :))
     
    Last edited by a moderator: Jan 4, 2007
    #46
  47. Cavallino

    Cavallino Rookie

    Joined:
    Oct 9, 2006
    Messages:
    140
    My Video (With music) :D

    sorry...wrong place
     
    #47
  48. maverick1

    maverick1 Semi-Pro

    Joined:
    Jul 23, 2006
    Messages:
    573
    [Edit, I realized that Granitechief was referring to a different method]
    1/10th second error is pretty hard to make when you have 30 frames/second. How can you miscount by 3 whole frames?

    In my experience, you can easily get an accuracy of better than 1/3rd of a frame. The ball moves quite a bit in each frame, so it is pretty easy to guess time to within a small fraction of a frame. Say you have a frame just before contact and the next frame shows the ball has left the racket, but the amount the ball has moved is only about a 3rd of the amount it moves in the next frame. Then I would count only 1/3rd of the first frame.

    BTW, MikeCottrill measured a recent Australian Open serve of Roddick that was shown in slow motion at 180 fps. The difference between the formula and the Radar gun's speed was about 1 mph (less than 2).


    Koensayr's method seems to be a refinement of this idea. It may work. The only thing is you have to be careful about the time it takes the sound to travel to your mike. If you record it from behind the server, the 60 ft difference distance would cause an eror of .06 sec(sound travels at about 1000 ft/sec); that error would be equal to 2 whole frames of video.
    You can either compensate for it or make sure your mike is equidistant from the server and the point of bounce.
     
    Last edited: Jan 22, 2007
    #48
  49. 35ft6

    35ft6 Legend

    Joined:
    Feb 18, 2005
    Messages:
    6,557
    #49
  50. ewcrider

    ewcrider New User

    Joined:
    Aug 10, 2007
    Messages:
    87
    Location:
    Western MA/Eastern NY
    #50

Share This Page