Reveal the secrets of Skytrak in the dark

[David Shen]

One day it occurred to me to look at Skytrak with my infrared night vision camera. Then I took the photos below. All photos below were taken without any visible light.

Through these photos, you will understand that the Skytrak has its own light source, even if there is no light in the room, the skytrak can see your ball clearly. I even played a few balls in this completely dark room and Skytrak read very well.

The last picture is what I'm guessing Skytrak might see, completely black and white since it works on the infrared spectrum.

The description in the image below has been modified based on Morini's comments.

[Morini]

The line is the laser curtain. The reflected signal generated when the ball crosses this line is collected by the receivers just underneath the emitter. The waveform is analyzed and ball speed is estimated. The IR strobes are then fired at a very precise time so that the ball is in view of the respective imagers.

IR cameras are perfect for inspecting strength of laser curtain. This is by far the most common part of Skytrak to wear/burn out.

[Morini]

I should add that 1 & 4 in the first image are not cameras, but rather the IR receivers that pick up the reflected light from the laser curtains (6 & 7). Those are basic (but cusom) 850 nm line lasers. Skytrak/Rapsodo call the receivers "receiver optical subassembly", ROSA for short, and similarly the transmitters are referrd to as TOSA.

2 & 3 are indeed cameras, counterintuitively, they are not high speed. Just cheap imagers (CMOS presumably).

5 is the IR strobe array. My understanding is that they don't all fire simultaneously, but rather selectively, depending on where the ball is expected to be in the image.

8 is actually two visible lasers, the left and right ball placement lasers.

The patents located here are very interesting:

https://patents.google.com/patent/US20140028838

Also, one of the pioneers investigating the internals of Skytrak was GSF poster davray666 who shared some great info here some years ago (7 to be exact):
https://golfsimulatorforum.com/forum...-technical-faq

Here is what the ball looks like to the Skytrak (actual image):

[David Shen]

Hi Morini, very nice comments. Tomorrow I will revise the description on the picture.

[David Shen]

Hi Morini, I'm curious how skytrak calculates the spin of the ball. Is the spin rate measured by skytrak accurate?

[Morini]

Yes, quite accurate. Take a look at the patents. Two pictures taken exactly 4.14 milliseconds apart. I believe it first flattens the spherical representation into a "flattened" 2d image for each picture. It then looks for overlapping areas between the two images, and measures how these rotated/translated to determine side (up and down translation) and backspin (counter clockwise rotation). Since time is known, the rotation is converted to RPM. Once you have side spin and back spin, you could of course convert that to spin axis and total spin, since these are mathematically linked.

Spin is accurate if it finds that overlap. The more the better obviously. And since we are dealing with 100 pixels diameter (give or take), there are minor measurement errors

[David Shen]

It's good to know that Skytrak measures spin so accurately. Your explanation gave me more confidence on Skytrak. 👍

[Morini]

It helps to think of Skytrak as simply a camera. It takes two pictures of the ball 4.14 milliseconds apart. That is really all the hardware does. The rest happens in software. If I gave you those two pictures, and asked you to calculate side spin and back spin? As mentioned, if I find overlapping areas, and used some image software to make measurements, I'd be pretty confident. Just physics. There is a known issue where 20 degree 130 mph+ balls have their spin halved or quartered. I don't know why, but it must a blind spot of sorts for the unit where it makes assumptions and gets them wrong.

[David Shen]

Morini I have modified the text description on the first picture, please check, if there is any incorrect place, I will modify it again.

[Morini]

It is accurate (to the best of my knowledge)

[David Shen]

I also added a video after the last photo above.

[Morini]

Here is a suggestion for you, David, since you have an IR capable camera (supposedly phone selfie cams don't have IR filter, but mine didn't work).

Turn off most (not all as you need to see the ball obviously) external light sources (visible and IR) and actually hit a ball. You should see the ball show up twice in one of the frames. My original idea was to have an overhead or down the line cam get a second opinion on HLA ... a VERY accurate one.

[David Shen]

I will try that this night and video it. Actually, I've tried hitting the ball in very dark conditions and it's been very bad, but skytrak read very well.

[David Shen]

Morini I made a very interesting video according to your suggestion. You will see three balls in one video frame as you mentioned. https://youtu.be/z_XbD_Nh2aM

[Morini]

Wow, I knew there were three strobes, but expected the final one later. I might actually do this myself. As mentioned, I planned to do this years ago. I really want to see a putt, as and a drive if you have time. It would be great if you could film towards skytrak so I can see where the ball is relative to the unit.

[Morini]

Just for fun, it looks to be about a 1/4 ball diameter gap between the first two frozen balls. That means the ball moved 1.25 ball diameters, or 2.1 inches. 2.1 inches in 0.00412 seconds is 510 inches per second, which is 29 mph ball speed. Was I close?

[David Shen]

Yes, very close.👍

[GungHoGolf]

Just for fun, it looks to be about a 1/4 ball diameter gap between the first two frozen balls. That means the ball moved 1.25 ball diameters, or 2.1 inches. 2.1 inches in 0.00412 seconds is 510 inches per second, which is 29 mph ball speed. Was I close?

Curious why the gap is smaller between the first two ball positions - would have expected a consistent interval between IR flashes.

Guessing they take 3 pics so a wide range of ball speeds will be covered with at least two ball positions in view. Using the outermost balls should give more accurate calculations.

The picture also illuminates (pun intended) the importance of not having IR-reflective objects behind the balls as viewed from the camera, like those bright pants that a ball could theoretically disappear into.

[Morini]

The timing of the three strobes are:
0
4.14 milliseconds
10.4 milliseconds.

For everything but putts and chips, the ball is out of view when the third strobe fires. It is likely a check to see if a ball is visible in order to eliminate false triggers. That is why putting has a different mode. It likely turns off or ignores the final strobe. And also why short chips don't read ... the ball is still in the image and the shot is cancelled.

Yes, reflective objects in background would be bad. But David's camera is likely more sensitive than the skytrak imagers (as you can see, the ball is quite washed out). Since the strobe intensity decreases by the square of the distance, a wall 8 feet from the skytrak would get 1/64th the amount of light that the ball gets, which means that non-reflective white is probably fine. But a reflective strip on the pant leg 4 feet away would be bad news.

[cmatt]

Thanks guys - this is very informative stuff! I'm always looking for ways to reduce mis-reads with my SkyTrak and having a better understanding of how it works will certainly help. My question is: despite the SkyTrak's use of the IR spectrum, is it reasonable to presume that more light (within reason) on the hitting area is better and will help the unit track and read? I currently have a 6W LED pin light over my hitting area which seems to work pretty well but I saw a youtube video that suggested LEDs can flicker and potentially inhibit the performance of the unit. The suggestion was to used incandescent lighting to illuminate the hitting area. Thoughts? Does it even matter with the IR and self-illuminating nature of the SkyTrak? Thanks again for your posts on this - very cool.

[GungHoGolf]

Thanks guys - this is very informative stuff! I'm always looking for ways to reduce mis-reads with my SkyTrak and having a better understanding of how it works will certainly help. My question is: despite the SkyTrak's use of the IR spectrum, is it reasonable to presume that more light (within reason) on the hitting area is better and will help the unit track and read? I currently have a 6W LED pin light over my hitting area which seems to work pretty well but I saw a youtube video that suggested LEDs can flicker and potentially inhibit the performance of the unit. The suggestion was to used incandescent lighting to illuminate the hitting area. Thoughts? Does it even matter with the IR and self-illuminating nature of the SkyTrak? Thanks again for your posts on this - very cool.

Lighting doesn't matter. SkyTrak uses IR flash photography. Other special lightning can't help, and may hurt. Avoid fluorescent and flickering LEDs (most modern leds don't flicker).

[cmatt]

Appreciate the insight - thanks Ron.

[Morini]

Ron is absolutely correct. Light can only hurt because the skytrak "freezes" the ball with two very brief IR strobes. Any continuous external light would only serve to interfere with this process.