Stuff Made Here breaks down Automatic Bow

0:00 welcome to my live stream i'm going to be

0:02 talking about my last video which was the automatic bow

0:06 so i'm going to play through the video and i'm

0:09 going to pause whenever i think there's something interesting

0:11 or more more light i can shut on a subject

0:14 let's play the video no looking william tell oh

0:21 my wife did archery growing up i did not

0:25 but being good at stuff is so 20th century put

0:29 this on and you don't have to aim anymore

0:32 so i guess i made an aimbot when you wield

0:38 the aimbot so there's one cool little detail that i

0:41 really wanted to talk about in the video but i

0:42 didn't get to this sort of metal foil looking

0:46 thing on top of the bow this is a spring

0:50 it's it's called a constant force spring and it's like

0:53 a tape measure no matter how far you pull it

0:55 out it always pulls with the same force what

0:58 this does is it offsets the weight of the bow

1:00 for the motor that lifts up and down without

1:03 it it's really unreliable and this basically makes it so

1:07 that it's only having to accelerate or decelerate the mass

1:09 it's not having to lift it when you wield

1:12 the aimbot you almost have a superpower it tracks targets

1:16 really well almost as well as my dog contract treats

1:20 it moves the bow to correct for your lousy

1:22 aim and when everything is perfectly lined up going

1:30 into this i wanted to shoot bulls eyes lame multiple

1:34 bulls eyes still lame which led to bull's eyes flying

1:38 through the air so we're gonna try all

1:40 that and maybe hit the world's smallest william tell sweet so

1:47 a lot of people have asked me if that's staged

1:50 or scripted but she just did that the original plan

1:53 for this i've been thinking about this for at least

1:57 a year probably longer than that what i wanted

1:59 to do was just have like a bow with cameras

2:02 in it or something so that i could just

2:04 hit bullseyes that are static the idea of a bullseye

2:07 flying through the air is super cool but it's

2:10 also really really hard because if you only have

2:14 a bow that has sensors that are in the bow like

2:17 cameras and whatever you have a really hard time

2:19 figuring out where something actually is so if you're if

2:23 you have a target flying through the air it's

2:26 it's pretty easy to tell if you're pointing at it

2:28 but it's hard to tell how far away it is

2:31 and what path and space it's actually taking which is

2:34 information that you need to delete it correctly and and all

2:37 of that so i was going to do static

2:40 bullseyes i got these opti-track cameras which can do

2:44 the external tracking really well that's what pushed me over

2:46 the edge to do flying targets they became the limitation

2:50 though because they don't work outside the sun overpowers them

2:54 and so it had to be done indoors

2:56 and that basically dictated the parameters of the project this project

3:04 started when i realized there's a hole in my heart

3:06 that money just couldn't fill it's 20 21 and i

3:09 can't buy a self-aiming bow so i built it

3:12 it works now but it took me thousands of misses

3:15 to get here there was also some collateral damage

3:19 the the progression of my backstop for the arrows is something

3:22 that i wanted to talk about in the video

3:23 but it didn't it just didn't really fit the amount

3:26 of bins and other things that i just destroyed is

3:30 really really high i ultimately ended up with a super

3:35 thick moving blanket with a target behind it if

3:39 i had an idea of where i might be missing

3:42 and then sheets of plywood and i still had

3:44 arrows go all the way through that entire stack up

3:47 this is actually also part of the reason why

3:50 a lot of people commented on how bad my form

3:53 was and if i pulled the bow back the proper

3:55 distance i wouldn't punch myself in the face you want

3:58 to do what's called a cheek weld where you

4:00 stabilize your back hand against your cheek but i was

4:03 trying to keep the draw force of the bow

4:05 down so i would destroy less things when i missed

4:08 and so for the compound bow you can adjust

4:11 the draw length and you can adjust the the draw

4:14 of force which tells you how fast the arrow is

4:16 going to go but they kind of go together so if you want it to be lower force so

4:19 that you're not shooting as fast you're pulling it like

4:22 a child and that's why i was pulling it

4:24 to here it's not that i didn't know it's just

4:26 that i'd rather do that and hit myself than

4:28 destroy a bunch of stuff although my nose probably took

4:32 the biggest beating so let me show you why

4:35 an aimbot is hard hey what are we gonna do

4:41 this tv is cnc machined on my router it's

4:45 made out of a sheet of mdf some people thought

4:48 that i 3d printed it or something but i didn't

4:51 i had to machine it because i could not find

4:54 a tv i drove around to a bunch of thrift stores none of them had a tv and so

4:58 this is some model off of grab cad and the knobs

5:02 are sla printed on the form 3 and then

5:05 the machine is done on the 24r this sequence i

5:09 had a vision it's like okay i want to make

5:13 it like this this old timey tv thing it'd

5:15 be kind of funny and a neat challenge this took

5:17 me so long to get convincing because if you

5:22 just take the video and scale it down and put

5:24 in a tv like it just looks completely fake

5:27 and i ended up i found some plugin that was

5:30 advertised as turn your amazing hd footage into analog garbage

5:34 so that helped a lot but i had to add

5:37 the glow from the footage on the bezel and reflection

5:40 of the the glass on the screen and without that it just it was so fake it was

5:45 just stupid and this took me this probably added like

5:49 a day and a half to two days to the whole

5:51 editing process i don't know it probably wasn't worth

5:54 it but i i had a lot of fun making it which is i guess the that's what happens

5:59 with all these projects they're not really worth it

6:00 but they're a lot of fun archery all right a lot

6:04 of stuff just happened most importantly the arrow flew

6:07 in a curved path one thing i was really concerned

6:09 about with this sequence is how can i make

6:12 it like if i'm gonna do any kind of illustration

6:16 in the video how can i make it not

6:17 seem like a weird anachronism like if i animate

6:21 on top of the screen of the tv it like

6:24 completely takes you out of it i did some tests

6:28 and i was able to basically do the animation

6:30 and then crustify that and then it made it look like

6:33 it was some old-timey like 1940s department of energy

6:37 or department of war video when they teach you about transmissions

6:40 or something and so that worked out really nicely

6:43 aiming above the target so when the arrow curved down

6:45 it would hit the right spot the arrow curves

6:48 more or less depending on how fast it's going if

6:51 her hands are off by even a couple of millimeters

6:53 she will miss three two one [Music] this is

7:01 and she really did that you can't aim for where

7:03 the target is because it won't be there when the arrow

7:05 reaches it you have to shoot where the target

7:07 will be which is called leading but what's really hard

7:11 is the timing this arrow was fired 50 milliseconds

7:15 too late isn't that crazy 50 milliseconds 50 milliseconds is

7:20 way missed it's actually probably more like plus or minus

7:24 20 if you want to hit the target maybe

7:26 a little tighter than that it's it's shockingly narrow

7:29 the window that's one third of the time that a blink

7:32 takes let me show you 50 milliseconds that was

7:36 actually 20 milliseconds i didn't realize that i was

7:40 exporting the video in 60 frames per second and uh

7:44 apparently a lot of people on youtube actually counted

7:47 the frames and called me out on my bs

7:49 timing i didn't mean to make it wrong but oops

7:54 did you even see that let me do it again

7:56 she has to fire within that window of time if

7:58 she wants to hit the target it's crazy and i

8:01 want to hit moving targets too so my aimbot

8:03 is going to have to do this this is going to be hard nice job i mean two out

8:09 of so yeah people are asking how long did

8:12 it take me to perfect my mid-90s look uh it

8:15 was self-evident it was pretty clear that you needed

8:18 a mullet and a stash we have lots of home videos

8:21 with my dad so it was like all right well i just got to do what he had

8:25 and it'll it'll work perfectly and and it did

8:29 in high res 4k it looks really ridiculous but when you

8:32 add the tape it's just it just seems right i

8:34 don't know it's amazing all right that's enough my plan

8:40 is to make a little robot that goes between

8:42 my hand and the bow it will move the bow

8:44 so that everything is lined up just right and there

8:46 will be an even tinier robot in my other

8:48 hand which can really the brainstorming phase for this thing

8:52 was actually pretty long i probably spent a good week

8:56 trying to figure out how i could possibly make

8:59 this device so that i could hold it so that it

9:02 could handle the forces and the big problem in all

9:06 of this if we go back to this view

9:09 one of the things i was really concerned about

9:11 which was actually a big limitation is that even when

9:15 you put the the gauntlet on when you pull

9:20 way out here your arm is not made to withstand

9:24 that kind of torque it really messes up your elbow

9:29 and so when i went into this i was

9:31 i was concerned that i was gonna like literally pull

9:34 my elbow apart or something i went on the router

9:37 and i made a not origami but i made

9:40 a out of sheets i assembled a thingy which was

9:43 a gauntlet that would allow me to hold a bow

9:45 offset in a variety of different distances from my arm

9:48 so i could test it before i went and built

9:50 this whole thing and i found that i could go

9:54 maybe three inches three and a half beyond that it

9:57 just completely was like it was really painful and even

10:02 the the three inches if i did it a lot

10:04 like my elbow would be really sore the next

10:06 day that's a big limitation of this and then

10:09 the other thing is that even when you have the gauntlet

10:12 and your elbow can survive it it wants to pull

10:16 your arm in it's a weird force that you

10:18 don't normally have on your arm and so it's

10:20 challenging to hold it for a long time if it's

10:23 fully extended so the robot tries to not extend

10:26 over there unless it has to because you just can't

10:28 bear it for very long the bow it will move

10:31 the bow so that everything is lined up just right

10:33 and there will be an even tinier robot

10:34 in my other hand which can release the string to fire

10:37 the bow one other little note on all this the little

10:41 squeaky release that turned out to be a big

10:45 problem with this bow something that i did not

10:47 appreciate is how incredibly important any motion that happens remember

10:54 i don't know how many milliseconds probably on the order

10:56 of 10 to 50 milliseconds after you release completely influences

11:01 where the arrow goes so if i flinch after it

11:05 releases i can actually make the arrow hit a different

11:07 spot and even if the arrow's not touching the bow

11:10 if i tweak the bow it makes the string

11:13 pushing on the arrow push on a different vector

11:15 and it'll whip the tail of the arrow around

11:17 and do all kinds of weird stuff i was using

11:19 this little servo that took on the order of 200

11:23 milliseconds to open my body would hear it starting

11:27 to open i guess i would know it was coming

11:30 and i would flinch and then the arrow would just

11:32 completely miss it actually took me a while to teach

11:35 myself not to pay any attention and that slo-mo

11:40 series where i shoot like 10 arrows into the bullseye

11:43 that was me not paying any attention i was

11:46 looking back and like talking to the wall well it

11:49 did its thing and it would shoot so i

11:51 wasn't gonna do anything to mess it up it's kind

11:54 of interesting oh the core idea behind this is

11:56 pretty straightforward but actually designing and making this is going

12:00 to be a challenge the engineering tunes that was made

12:08 on my wife's cricket if you haven't seen a cricket

12:12 they are actually an amazing tool i usually can

12:17 do what i need on the laser cutter but they

12:19 they can cut balsa they can cut vinyl they

12:22 can cut plastic sheeting fabrics all kinds of stuff they're

12:27 super easy to use i recommend them they're really nice

12:38 i don't think i talked about this in my last

12:40 video so the parts come out of the fuse

12:42 those are the that's the sls nylon printer and they come

12:47 out of it you can usually get all the powder

12:49 off of most of the stuff just with a brush

12:52 but if you have any kind of crevasses or cracks

12:55 or anything bead blasting it is 10 times easier

12:58 than trying to do anything with a brush so

13:02 this is a little tiny bead blasting cabinet and most

13:05 of the parts that come out of it i'll bead

13:06 blast them you can get them cleaner with a really

13:09 nice surface finish this way [Music] basically all

13:19 the parts that i was machining in this project were

13:22 to adapt these pancake bldc's to do what i

13:27 needed to do this is a motor for a large

13:30 drone these motors are designed to have a propeller bolted

13:33 directly to them so they don't even have a shaft

13:35 coming out of them and so i needed a shaft

13:38 that i could mount a pinion on and also that was long enough to go to an encoder

13:43 for the feedback that's how these are controlled i was

13:45 trying to use an o drive which is a open

13:48 source brushless motor to close loop the ldc conversion

13:52 thingy it's really good super fast super powerful and i

13:59 couldn't get it working but it shows great promise [Music]

14:07 so the o drive is down here and then this is just uh one of the newer arduinos

14:14 i think it's an arduino metro all i needed

14:16 it to do was to trigger the servo and listen

14:20 to the button so i didn't need much there and then

14:26 i guess this was another since we're on the top

14:30 of the design still this rail was probably the hardest

14:35 part of this entire project because if you think

14:38 about it i have this rail which can extend

14:41 out several inches out and that's trying to bend

14:46 this entire assembly like this nylon arm just wants

14:50 to just snap in half and or at least deflect really

14:53 bad that made design in this nylon arms that it

14:56 was strong enough a challenge but then you you

14:59 run into issues with these rails as well so when

15:02 you when you twist this whole assembly and crank it

15:05 down the surface that these little rails are mounted

15:08 to will distort and then they'll start to bind up

15:11 they really need to not not move relative to each

15:14 other so this whole part of the gauntlet is wildly

15:18 overbuilt it is super stiff it's ultra ribbed this is

15:22 full of ribs and this part here is really

15:26 really stiff too to try to avoid twisting the blocks

15:28 relative to each other like i don't want the arm

15:31 to form an s shape it's really important that it

15:34 doesn't happen it's probably beefier than it needs to be

15:37 but i didn't want to have to remake it

15:39 and then all of the important surfaces on the nylon

15:42 parts are machined so i print it and then

15:44 i post machined them so they'd be flat and parallel

15:47 and everything and that's a really good technique for 3d

15:49 printed parts you can print the net thing kind

15:52 of like a casting and then machine to get

15:54 things really square or whatever works on uh fdm parts

15:58 sla parts and sls parts that was basically the representative

16:09 issue with the o drive i spent at least

16:13 a week on this i was having issues where i

16:16 was effectively stalling the motors for a bunch of different

16:19 reasons and these motors are made for a quadcopter

16:23 with a propeller blowing air over them constantly which isn't happening

16:27 here so when i run them at their max rated

16:30 current they can really only do that for a little

16:32 bit before they overheat i could do it

16:35 for a spike to move the bow and fire because

16:38 that's a really short event but if they would

16:40 stall they would burn up i just kind of gave up on them i want to go back to it

16:44 it's a really great board and it can way outperform

16:49 the the stepper motors that i used which was

16:51 sad but i just had to get this thing done

16:58 i spent over a week on the spontaneous combustion issue

17:01 we're not even going to talk about it other than

17:04 to say it's fixed so you can see this view

17:11 like this yeah you can see my square peg

17:15 in the round hole this is actually i was able

17:17 to completely rework this on the mill so this is

17:20 super duper drilled a bunch of holes and machined some

17:24 pockets and stuff to fit the stepper motor this one

17:28 i just printed this part and is i was able to get it working it can aim the bow

17:32 up and down with this linear axis you can

17:36 actually oops sorry i'm not great at these controls so

17:42 that spring that i was talking about the constant

17:44 force spring you can see it working here pretty well

17:48 and left and right with this one there's a little

17:50 kevlar line that's going to the top of this assembly

17:54 this is actually very luckily i didn't design it

17:57 this way but this is basically the center of mass

18:00 so i'm pulling right through the center of mass

18:02 so nothing is twisting or anything here's the little robot

18:08 that releases the string it does this with a little

18:10 servo motor it is super important that the bow always

18:14 points at my back hand if it doesn't then

18:16 the arrows just don't fly right at all [Music]

18:21 the arrows with this cheapo bow the arrows look like

18:25 that when you launch them normally not quite as bad

18:28 but these bows don't have enough of a cut out

18:31 for the arrows to be in line with the string

18:33 and boy do they they flex and go crazy

18:37 when you fire them the aimbot hardware is done

18:41 but at this point it's pretty much just a crappy

18:43 heavy bow i need some kind of sensor that tells

18:46 me where the bow is relative to the target

18:48 and a lot of other stuff there's eight cameras throughout

18:51 my shop that see everything they're made by a company

18:54 called optitrack here's i cannot get over how amazingly good

18:59 these cameras are the amount of time that i've spent

19:03 trying to make crappy tracking systems for with the connect

19:06 or with the intel realsense or 2d cameras

19:11 these things you literally just stick them anywhere put them

19:15 on a tripod screw them to the ceiling it doesn't

19:17 matter you walk around with the wand and they work

19:21 like that's it sometimes you have to do the calibration

19:24 a couple times if you don't have them set

19:25 up right but it is so much easier it is

19:29 really a shame that this technology is so expensive because

19:33 i if more people had it i think there's a lot of really cool things you could do

19:35 with it here's how it works if i hold up

19:38 this little reflective ball it's seen by all the cameras

19:41 at the same time imagine this is the ball

19:43 that i'm holding up and a camera is looking at it

19:45 from this angle the camera will take a picture

19:48 that looks like this if you project an imaginary tube

19:51 out from the ball in the image we know

19:53 that the ball must be in that tube we just

19:55 don't know how far away it is from the image

19:57 if we add another camera looking at the ball

19:59 from this angle we get another tube that the ball

20:01 could be in and the intersection of these tubes gives

20:04 us the location of the ball and these cameras

20:06 the detail that i'm glossing over in all of this is

20:09 that you need to know where the cameras are

20:12 that's what the the wanding is you have a stick

20:17 with balls that are known locations relative to each

20:20 other and when you wave that all around you get

20:23 a ton of pictures from from all the different

20:25 cameras and then they can look at basically the correlation

20:29 between the different views of the thing of known

20:32 geometry and then it does basically an optimization where it

20:38 figures out the locations and space for the cameras

20:42 that would make what they're seeing make sense and so

20:46 you can actually see it when you calibrated it

20:48 it renders it in real time the cameras are like

20:51 the they're like nowhere and you can see them

20:53 iteratively move to the right spots as the um calculation

20:57 converges so it's pretty cool and uh rya's asking if

21:02 the balls are retro-reflective yeah

21:04 so basically anything that's retro-reflective

21:07 works i shot some of these balls out

21:11 of my catapults in the last video and those were actually

21:14 steel ball bearings but i bought this i think it's

21:17 made by 3m it's a retro-reflective spray that you can

21:20 i think people put on their license plates so you

21:22 can't see them with the license plate cameras but they

21:25 make something you can spray on your clothes that it's

21:28 basically invisible i think it's some kind of glass

21:30 bead but when you take a picture of it

21:32 with a flash you glow like crazy the other really

21:35 cool thing about this software that i didn't really

21:37 talk about in the video is you can just say

21:40 if you see three markers in this relative positions

21:43 to each other that's this object so you can stick

21:46 them on something in any arrangement and then in the software

21:49 select them and say like this is the bow

21:52 and then the software will just wait until it

21:54 sees something where it matches that arrangement and they'll say

21:56 all right here's your bow and then it'll tell

21:58 you the orientation of it and it's all very auto-magic

22:02 the whole time i was using this i just was like man this is so good if only i

22:05 had this five years ago cameras do this super fast

22:09 and the time that it takes to blink they'll give

22:11 me 50 updates on the location of this ball

22:14 it's bananas you call these balls markers if you

22:17 want to sound like you know what you're talking

22:19 about i have those markers are i think ten dollars

22:23 a piece so you don't want to break them

22:25 tracking balls on the front of the bow the little

22:28 grip robot and on the target this lets it know

22:30 where the bow is pointing how far it's drawn back

22:33 and where the target is i wrote a really simple

22:36 program to track everything and shoot at stationary targets it's

22:41 this program was basically just doing the simplest possible thing

22:47 it was not taking into account any of the dynamics

22:51 it was just saying here's the target my bow is

22:54 pointing here if i want the vector of my arrow

23:00 to intersect with that point i need to move

23:03 it here and then it would move it there and then it just did that in a loop so

23:07 it continually adapt where it was moving to and then

23:10 it had some gains it was basically a pid

23:13 it was accounting for the drop it's time to see

23:16 what this thing can do [Music] yep that hurt

23:29 [Music] i actually cut my face i i don't know

23:39 if it's visible in the video there's actually a another

23:43 big problem the part that moves next to my knuckles

23:47 i designed it like one millimeter too close

23:50 to my knuckles and so sometimes it would just like cheese

23:54 grater my knuckles and so my knuckle got really

23:58 messed up i tried to file it down and stuff

24:00 but i just i couldn't take off enough material

24:03 and so throughout the project my knuckle got increasingly bloody

24:08 and nasty and there was a couple shots where it's

24:11 a close-up of my hand and it's like really disgusting looking

24:15 through the power of uh computer wizardry and special

24:19 effects i removed the gore because it was too disgusting

24:24 i'm putting all that force into the string and then

24:27 it releases without any warning and i punch myself

24:30 i'm trying really hard not to punch myself but i

24:33 punch myself this is this was interesting you can

24:35 say like i'm not going to punch myself but this happens

24:39 faster than you can react by the time you

24:42 even notice that it's fired you've already punched yourself

24:45 so there's nothing you can do about it oh yeah

24:48 and it's also totally missing i really should probably

24:51 move these out of the way i wanted to communicate

24:54 my struggle with this problem so i made a movie

24:57 trailer i don't even know anymore some say he's

25:04 still in there please tell me that was fluke do

25:12 you ever plan to let me go [Music] why won't

25:17 you just work i don't know exactly how i

25:21 ended up making that i i think there was

25:23 a confluence of a few different things i discovered

25:26 a pack of sound effects from all the movie trailers like

25:30 the ridiculous bass drop sound and the the cuckoo

25:36 and i really wanted to use those and then i

25:38 was thinking about how i had i know what you

25:41 did last week joke in there which i thought was

25:43 kind of funny because that's basically i spent a week

25:46 on that problem and then uh i was thinking

25:50 about how hellish it was and it just kind

25:51 of all came together and it was like oh well i

25:53 can maybe cast a mood with a movie trailer

25:57 or something it was a fun little fun little diversion

26:01 i don't know how effective it was but a lot

26:03 of the stuff in these videos just you know it's

26:06 interesting and fun and hopefully entertaining that was

26:11 my week i was stuck in integration hell where all

26:14 the pieces work by themselves but you put them together

26:17 and they try to kill each other just like children

26:19 the biggest problem is that it just won't shoot

26:22 the right spot i haven't proven conclusively why this happens

26:27 i'm pretty sure that this was the archers paradox i

26:31 recorded a bunch of data and it basically i show

26:35 a visualization of it later in fact it might

26:37 even be here let's see shooting up in the left

26:40 almost every time i could shift things over in software

26:43 but if i do that i'm gonna i guess i don't show it till later i'll point it out

26:47 but there's a visualization that shows the data i collected

26:50 where it shows where the arrow's pointing where the target

26:53 is the bow is aiming right at the target everything's

26:57 perfect but it goes to the left so i'm

26:59 pretty sure that this is basically the arrow whipping around

27:04 the bow and getting shoved to the left and just

27:07 hitting up into the left and just hide a bug

27:10 that's going to bite me later but i won't

27:12 tell if you don't all right let's put this thing

27:15 to the test hey white autobow versus wife three

27:20 shots each best shot gets the point come on she's

27:25 very pregnant so she had some difficulty uh with the bow

27:30 and then the other the other thing that was

27:32 actually really hard for her is that archers never

27:36 ever shoot anything this close you know this is like

27:40 20 feet or something that is just not something

27:44 you do and so i think it's outside of intuition

27:48 and other things that that they're used to but she

27:51 still did pretty well and here we go i need

28:03 my binoculars oh i'm closer it depends on how

28:07 you measure it i had a valid point there so

28:14 i cut it out because it wasn't wasn't interesting

28:17 but i'll say it now which is i was claiming

28:20 that if you use the manhattan distance i might

28:23 be closer if you're not familiar with the manhattan distance

28:25 you can look it up there are other ways

28:27 of measuring it that might make me the winner just saying

28:31 darn i thought i was closer i was hoping

28:33 for like a tie or something but wait if i

28:37 hit this i win everything did you go around

28:39 looking for the biggest apple you could possibly find no

28:44 [Music] so the way this setup worked is so

28:51 this table has markers on the four corners of it

28:54 here here and then you can't see the other two

28:57 from this i know what direction up is and then

29:00 i know where the center of the table is so if i put the head in the center

29:04 of the table and i know how tall my head

29:06 is i can deduce where the apple must be and so

29:10 that's how i'm doing it what i was originally

29:12 going to do is put markers on the head

29:14 and then maybe like around the apple or something

29:19 and what i really wanted to do was launch the head

29:21 through the air and hit the apple off it

29:24 i didn't end up doing that because i didn't have

29:27 a head launcher so i was gonna have to have

29:29 my wife down there throwing the head over and over

29:32 again while i tuned my uh code and got it

29:35 working or i was gonna have to build a head

29:36 launcher which i didn't feel like doing and so

29:39 i just went with the static head and called it

29:41 a day [Music] there's obviously some room for improvement

29:46 here you can see the marker's leaking yeah maybe is

29:50 it gonna make me miss no it's not to hurt

29:53 your aiming grip error check hand ow i was

29:59 really disappointed i was hoping she'd really get clunked

30:02 right in the face and and actually the reason she

30:06 said ouch wasn't from punching herself in the neck it

30:10 was she was holding the grip with her finger directly

30:14 behind the servo horn for the servo and there's

30:17 a piece of music wire that connects the uh the release

30:22 to the servo and that stabbed her directly

30:24 in the finger just went like is really sharp from cutting

30:27 it with the plier or with the diagonal cutter

30:30 so just went straight into her finger like a knife

30:32 kind of a double whammy i guess quit hitting yourself

30:36 quit hitting yourself quit hitting yours we got a bit

30:38 sidetracked there let's get back to making this work

30:41 the bow is just not shooting the right point i'm

30:44 pretty sure it has something to do with the software

30:48 so there's a third monitor joke here that i

30:50 really enjoyed someone was asking how much joy i

30:53 get out of thinking of those third monitor gags it

30:58 is extremely satisfying when you have just the perfect thing

31:02 to put on your monitor that is related to what

31:05 you're doing is is a source of great satisfaction

31:09 it's one of those things where you can spend

31:10 a lot of time just like thinking about it i

31:12 try not to usually i'll be working on the project

31:15 and thinking about what i'm going to show and what

31:18 the story is and everything and then i'll i'll be

31:20 like oh that would be perfect put on the monitor

31:22 i write it down and then i do it later

31:24 that's usually how it happens and then michael is

31:27 asking where do i rank this project in my list

31:29 of most dangerous creations i don't think this was

31:33 particularly dangerous the reason why i say that is things

31:37 like the bat they can be armed or they

31:40 can have energy in them like they can have

31:41 a shell that didn't fire or something like that and you

31:44 don't know it and you can walk in front

31:46 of it or put your hand on it or something

31:48 and then it can blow and unexpectedly maul you

31:52 a bow you have to draw it back there's no

31:54 way to leave the bow drawn like i can't draw it and then go walk in front of it

31:59 or mess with the code or something like that i'm

32:00 holding it and so the only way that it can

32:03 really hurt me is if like the bow exploded now

32:07 i could shoot someone else if you are treating

32:11 the range as a range and you just never have

32:14 anyone down range when anything is active it's pretty easy

32:17 to be safe that way you know even a project

32:19 where i use the table saw a lot compared

32:22 to this one is a lot more dangerous than

32:23 this project but i'm not going to debug this and try

32:26 to figure it out i still need to make

32:28 the bow track moving targets and all of that so

32:30 i'm going to just nuke all of this code

32:33 and hopefully replace all of these old problems with new

32:36 and exciting problems so this was actually an intentional choice

32:40 i had never interfaced with the obvi track before

32:43 when there's a lot of unknowns when uh when you

32:46 go into a project whether it's cad or or software

32:48 or something else you don't know enough to write it

32:51 in a sensible way and unless you just continually

32:55 rewrite it as you go and so what i did

32:57 here and what i do with a lot of projects is i just go for it and try to get

33:02 a prototype working and then i'll throw it away

33:06 and redo it and then i know enough to actually

33:09 do it in a pretty good way this is a good technique most of the stuff i cad i

33:14 will just go in and do the worst solid

33:17 modeling you've ever seen just to figure out the shapes

33:19 and sizes of everything and then i'll redo it

33:22 once i have that and then i can do it a lot better one week later and i have

33:27 nothing to show you because software development is very boring

33:30 but it's ready to test and i have a super

33:33 awesome voice activated target launcher all right pool in the next

33:39 500 milliseconds a lot is going to happen

33:43 the bow is drawn and ready to fire i'm pressing

33:45 this button which tells the computer to fire at will

33:48 i really like how this sequence turned out i was

33:50 trying to explain what was happening without just drawing

33:53 it on the ipad so i was trying to communicate

33:56 it all with with a series of slow-mo shots

33:58 of what goes on and i i think it mostly works

34:01 although i was limited how deep i could go

34:04 just because it's hard to illustrate some of these things

34:07 the tracking system sends the computer an update on where

34:09 everything is every three milliseconds when the computer sees

34:13 a target it checks to see if it's moving

34:14 in a parabolic motion so what you want to do is

34:17 you have a bunch of locations that the target has

34:20 been in at some time and you want to say if i fit a parabolic shape to this how

34:27 well does that parabola fit in particular does the curvature

34:31 that the parabola has is it consistent with gravity

34:36 like that would indicate it's following a ballistic path what

34:39 you don't want to do is collect a bunch

34:41 of points where it's sitting still before it's been launched

34:45 it starts to get launched you try to fit

34:47 to it but there's a bunch of points that aren't

34:51 a parabolic trajectory connected to points that are so

34:54 you want to detect it moving in a parabolic trajectory

34:58 immediately when that happens and reject all that other stuff

35:02 so there's a number of heuristics where if it's not

35:04 moving if the fit is terrible it will just

35:08 throw away points that it's collected and then it'll look

35:12 at a portion of it and see if that looks

35:14 like a parabola and then at some point that fit

35:18 is good it usually takes a few points like

35:20 four or five and then it'll chop off everything before

35:23 that and it'll kind of lock on to that and then

35:26 start broadcasting forward this works really well especially

35:33 with these cameras because they're so accurate they have

35:34 really good time resolution really good spatial resolution i also

35:38 can calibrate it so that gravity is actually aligned

35:41 with the z-axis of the coordinate system from that you can

35:45 start doing predictions of where it's going to go

35:47 that's what the red line is and then it does

35:50 a tedious little calculation to figure out where it should

35:52 move the bow to intercept the target taking into account

35:55 the time to move there release the arrow reach

35:58 the target and the curved path of the arrow

36:01 this this was a really annoying computation i actually couldn't

36:07 find a reasonable way to compute this other than to sample

36:11 if i look at the point one millimeter in front

36:14 of the target i can't physically move the bow

36:17 so that it's pointing there before the target gets

36:19 there so the first thing i'm trying to figure out

36:21 is what's the first point that i could possibly move

36:25 the bow in amat before the target passes that point

36:30 so i try to calculate that then i start

36:32 at that point and then i march forward every like

36:35 few millimeters and i say okay if i wanted

36:37 to hit the target when it was here where would

36:40 i need the bow to be pointing and then

36:44 i just keep going along with the curve and i

36:46 try to find the point that is the slowest

36:49 velocity of the target so if i can get up

36:52 to like apogee the target goes up and slows down

36:56 and then comes back down if i can shoot it

36:57 there that's the easiest point to hit it so

37:01 that's that's its first preference sometimes it's just too high

37:05 or that's really far away and it can't reach it

37:08 i actually shot over my backstop a number of times

37:11 which is pretty funny and so so it does

37:14 that then it says okay if i can't hit apogee

37:19 what's the closest point so rather than trying to move

37:23 up and intercept it just after apogee or something

37:27 like that if it's going to go up and over can i just move to the left and then

37:30 intercept it if it can do that it'll prefer

37:32 to do that and then calculates based on the draw

37:36 of the bow how fast the arrow is going

37:37 to be going and then how long it takes the little

37:40 servo thing to release and then how fast the target's

37:45 going and uh i guess those are the main things

37:50 and that's how how much it has to lead

37:52 how much time it has to lead and so getting

37:55 this all dialed especially getting the calculation for how

37:59 fast the arrow is going to go versus draw was

38:01 critical i had some approximations for the release time

38:05 and the flight speed and it would always miss i

38:08 had to measure these and once i got those dialed

38:12 it was really quite good this takes about a thousandth

38:17 of a second and then it starts moving the bow

38:20 and then it repeats this over and over again

38:22 as more tracking data comes in which allows it

38:24 to adapt for things like the shaking of my hands

38:28 oh yeah so that's one other aspect of this so

38:30 there's a point where it's trying to go to intercept

38:33 the target as more data comes in from the target

38:36 moving it refines the prediction of where the target's

38:39 going to be and then also my hands move especially

38:42 when it when it launches the bow to the side

38:45 with the motors it pushes off my hand and it

38:48 often moves like 75 of the distance that it

38:50 was trying to go because it pushes my uh

38:52 my hand away every iteration every 360th of a second

38:57 it will re-update the target and then it was

39:01 effectively running a pid there was a little bit more

39:05 going on in the controller in particular it was

39:07 looking at the motion of the bow if i know

39:10 that the bow is moving sideways at 0.5 meters per

39:15 second after i release it will actually be in a different

39:18 spot than i was when i went to release

39:21 it and so it will account for that it'll look

39:24 forward not only to where the target is the leading

39:27 i'll look forward to where the bow will be

39:29 based on the motion of my hand to zero

39:32 that out as well and so that works pretty well

39:34 as long as i'm not like vibrating my hand

39:37 if if i'm moving the bow or changing speed faster

39:40 than a couple hundred milliseconds which i don't really

39:42 do then it breaks down but that was a nice

39:46 little addition when it thinks it's amy at just

39:48 the right point it waits until the timing is just right

39:53 and then completely misses [Music] i don't know i

40:04 love that joke so good i've probably seen that 500

40:07 times it's still hilarious um someone's asking do i

40:12 take the pull force into account when calculating the speed

40:15 of the arrow yes so what i did is i

40:19 had a fish scale and i measured the pull force

40:22 of the recurve bow at a bunch of different locations

40:27 and it was extremely linear it was much linear than

40:29 i was expecting it to be and then i

40:32 used that to calculate the stored energy in the bow

40:35 and then i looked up online how much

40:38 of that transfers the arrow as a correction factor and then

40:42 i computed the speed knowing the weight of the arrow

40:45 i was able to compute the speed and then

40:48 comparing my prediction of the speed to the actual speed

40:52 with the high speed camera it was really accurate it

40:55 was within like half a meter per second roughly

40:59 and then the thing that i wanted to do which

41:02 i just never got around to doing is that the cameras

41:05 are fast enough to track the arrow like you

41:08 can put a tracking ball on the arrow and not

41:11 only get its trajectory but it's speed and everything

41:13 else the reason i i actually did this a couple

41:16 times it was hard to make an arrow where you

41:19 wouldn't destroy the tracking ball and then of course

41:21 the tracking ball changes your results so i just never

41:25 really did it but that would have been a cool

41:27 way to collect a bunch of data from a bunch

41:30 of shots and get a really accurate prediction

41:32 of the speed life catapult was getting pretty tired of waiting

41:35 an hour for me to fix a bug that i said would only take a minute so i built

41:39 this automatic catapult that's gonna let me test as long

41:41 as i want and i'm the only one i i built like nine or ten catapults right before

41:49 this this thing was a this thing was a champ

41:51 this thing must have fired thousands of times i

41:54 went back and forth and back and forth i had

41:56 it down to a science what honestly might have been

41:59 worth doing given the amount of time i spent like

42:02 loading this and going back to the bow and resetting

42:04 it and loading it is actually building a magazine

42:07 that holds 10 targets or something and then you'll

42:10 notice there's some unnatural motion of the target what i

42:15 actually had was the target had a a thin

42:18 kevlar line attached to it that stopped it a few

42:22 inches above the floor and then that went up

42:24 to the ceiling there was a spring and so that it

42:27 would launch them when they would come down they

42:28 would not smash into the floor and break the target

42:31 or mess up the balls the retroflective markers and that that's

42:35 why you see it bouncing around you could maybe

42:38 use retro-reflective tape on the arrows i didn't try

42:40 that i it should work one that's gonna suffer as expected

42:46 the new code base is chock full of exciting

42:48 problems really cool looking when it hits the why is

42:51 it too early why is it firing too late

42:53 why won't it hit the target when it's sitting still

42:56 why did it one hit ko my microcontroller it

42:59 would just like track the target no matter what i

43:02 was pretty permissive with my fits so when the target's

43:06 like bouncing around and doing stuff it would like

43:08 be trying to hit it and that's why it hit

43:11 my microcontroller i eventually just put a if it was

43:16 below a certain height it just wouldn't track it

43:19 and that allowed me to be aggressive with my tracking

43:22 and tuning but not do that stupid stuff so

43:26 many bugs but it's only a matter of time we're

43:31 knocking those bugs out we fix the microcontroller we

43:34 destroy targets sitting still now we're getting really close

43:40 at this point i'm pretty sure this is a view

43:42 of one piece of the data that i was talking about

43:44 earlier these three balls represent the arrow and the bow

43:50 these are the trackers on the bow this is

43:52 the grip and then this is the prediction of where

43:55 the arrow will go and then this is the target

43:58 that it's trying to hit when it would fire

44:00 everything was was great and it would still miss

44:03 and that's why i have concluded that it probably was

44:05 the archer's paradox and then one other interesting thing maybe

44:11 that's worth mentioning is that some people have messaged

44:15 me saying well no wonder it's not aiming the right

44:17 point your tracking markers are up here but the arrow's

44:19 down here and what i actually had and this was

44:24 really important is i had a special arrow that had

44:27 a marker on the butt and the tip and i would put that into the bow and then i

44:32 would use that to calculate basically the rigid body transform

44:36 from these three points to where the knock

44:39 of the arrow is and then from these points to where

44:42 the center line of the arrow is that was much

44:44 more accurate than just trying to measure it with tools

44:49 or basing off just the the geometry this measured

44:52 it very accurately and so if i change the grip

44:56 or mess things up you would just put the calibration

44:58 arrow in it and then recalibrate it and that worked

45:01 really well i'm pretty sure the software and the robot

45:04 are doing the right thing but the arrows aren't

45:07 going where they're supposed to go but i think

45:09 i know why i've been using a recurve bow which

45:12 comes out of the box with a fundamental issue

45:14 you have to fire the arrow if you don't know

45:17 about this effect there's a there's a video

45:19 from a smarter everyday talking about this the basic idea is

45:23 that the string is pushing the arrow so that it

45:28 wants to go through the bow but for old

45:32 school bows the arrow has to go around the bow

45:35 and what actually happens is that somehow the arrow bends

45:39 around the bow and can shoot like in line

45:44 with the the bow where it shouldn't be able to shoot

45:47 and it causes all kinds of crazy stuff people

45:51 spend a lot of time tuning arrows and all sorts

45:53 of stuff to try to make the thing very accurate

45:56 at known distances and i just didn't want to be

46:00 bothered with that and that's why i moved

46:01 to the compound there is some of this with a compound

46:04 but it is much much less around the bow

46:08 this makes the arrow do crazy things and even crazier

46:11 things if you're using the cheapest bow that money can

46:13 buy theoretically i could calculate what's gonna happen and correct

46:17 for this but we're just gonna buy our way

46:19 out of this problem this is a compound bow

46:22 it's very powerful oh that shot right there i

46:26 had been using the recurve with a rubber safety tip

46:30 on the arrows which would just bounce off the foam

46:34 i put a rubber safe tip on the arrow

46:38 in the compound i really lowered the draw force you

46:41 can see the the bolts if you're familiar with compound bows

46:45 these bolts are what set the draw force it's way

46:47 down and it just went straight through through the bin

46:51 and out to another bin and destroyed a bunch

46:54 of stuff it didn't actually hurt any of my my actual

46:57 things would have destroyed a bunch of bins it

46:59 was pretty funny but it was i was pretty

47:01 annoyed with myself for i don't know it's just

47:03 dumb why did i do that and it shoots arrows

47:06 really straight it's pretty much a drop in replacement

47:09 for the other bow except for one problem this thing

47:13 is really heavy you can see my retro reflectors

47:18 so these are nominally centered around the arrow there's actually

47:22 a little calibration disk it goes around the arrow

47:25 and you put it in here and then you can

47:26 adjust this ring and bolt it down so that it's

47:28 really close but then even with that you can get

47:32 more accurate with the calibration arrow so we're going

47:35 to do something about that if this isn't a name

47:40 bot i don't know what is i was really hesitant

47:44 to put this thing on i don't know why it

47:47 really felt like cheating when i make these projects

47:51 i i construct a set of rules in my mind

47:54 like i'm gonna hold the bow you know it's

47:58 a bow that you can hold or it's a golf club

48:00 that's self-contained or something like that and having

48:04 this thing it felt like it was taking me away

48:08 from my original goal and sort of undermined it using

48:13 the steady rest but i was okay with it because

48:17 one i just i had to have it the thing was too heavy the only way i could not

48:21 have it is to completely redesign it to be lighter

48:23 which wasn't worth it and then it wasn't actually really

48:28 helping me be steady so much that just helped

48:31 me hold the bow like the steady rest is amazing

48:35 it is so smooth and floppy and so you

48:40 can shake your arm like crazy with it this thing

48:42 supports all the weight of the bow so that all i have to do is hold it in place

48:48 this thing shoots so much better than the recurve so

48:52 you can actually see this kind of feels like cheating

48:55 you can see the uh bow locking on the target

48:58 and firing if you watch closely things to the right

49:03 locks fires and in the precision mode that it's

49:06 doing there it could have fired a little bit earlier

49:10 but it waits for the velocity of everything to settle

49:15 so the grip the bow it wants those to be

49:18 very close to zero it can do the forecasting

49:21 that i talked about before where it looks at how

49:23 the bow is moving but it is more accurate

49:26 when you wait for the velocities to settle so that's

49:29 what it's doing there curve this kind of feels

49:32 like cheating but we're way past that point i think

49:36 it's time for another round of beau versus wife

49:38 hey wife whoa what are you wearing it's engineering what

49:43 does it look like body armor with a snake you

49:47 ever seen an engineer before same thing as before whoever's

49:52 closest gets the point i brought this there's one

49:56 other neat little thing with this i guess you can't

49:59 really see it i had to modify this steady rest

50:03 uh there's a big problem when you take the bow

50:06 off of it when there's no weight on it it just like bing shoots up into the air

50:12 and it's really bad you can nail the bow

50:15 with it and so there's this part i had to make

50:20 maybe you can see it in some other shots

50:21 you can look for it if it's interesting but it

50:23 basically allows me to lock and hold the steady

50:27 rest in a lowered position so that i can load

50:29 and unload the bow and that worked pretty well

50:33 it did there's a couple times where it didn't lock

50:35 and then the whole thing like exploded like a clock

50:39 in a cartoon you know where all the gears come out

50:41 and everything it's just like and like punched me

50:44 in the arm but overall it's pretty good be down

50:47 and have it running in sharpshooter mode [Music] that's

50:54 three points for the aimbot i didn't actually do anything

50:57 but it feels uh this you can look up

51:00 this reference if you would like to see some true

51:05 culture i think it's big rigs over the road trucking

51:08 is the game there's a very famous uh engrish misspelling

51:13 where there's a trophy with three handles which i

51:15 couldn't find footage of and whenever you win it tells

51:19 you that you're a winner that's what i'm referencing here

51:22 feels like i won all right it works let's do

51:26 stuff we're gonna try william tell again i have

51:29 a little apple hopefully this is acceptable size to you

51:32 that's not an apple it's an apple and we're gonna

51:35 shoot it's actually a plum all right don't move no

51:39 looking william tell oh boom [Music] let's do that again

51:52 when these things finally work it is so satisfying

51:56 and this this type of stuff is where i

51:58 was just blown away by the optitrack because this is

52:01 the type of thing that if there's any inaccuracy

52:03 it just doesn't work i was moving the cameras around

52:06 i was taking it to the range and it just

52:08 worked again i need more [Music] okay that's pretty cool

52:18 it's time to show what this thing can really

52:20 do moving targets [Music] so you can see it going

52:43 for apogee there it pushes the bow up because

52:47 it thinks it can intercept and get it it spanked

53:04 the vinyl decals off the the target i kept destroying

53:08 these targets there was an evolution the original target was

53:12 laser cut multiple rings that were painted and then glued

53:14 to a backing sheet and i kept destroying them so

53:17 they got simpler and simpler so i could make

53:19 them really easily ultimately it was just some thin plywood

53:24 that i laser cut with vinyl decals to give

53:26 it the rings normally i wouldn't bother with the rings

53:28 but it it seemed important [Music] [Applause] [Music] oh

53:38 nice face there so these lights back here i got

53:43 these used i can't remember what brand they are it's

53:46 basically a giant heat sink with some leds on it

53:51 and i found these used for like a third

53:54 of the normal price but these were the only reason

53:56 that i could capture that high speed footage the the high

54:00 speed camera wants basically all the light it really

54:03 wants sunlight normally when i'm taking high speed footage

54:06 i have two lights like six inches from whatever it

54:09 is i'm getting footage of and the camera stuck

54:13 with the lights right next to it and that's the only

54:15 way i can get enough light these are so

54:18 bright i can actually illuminate the target from behind me

54:22 they're so powerful and they have some really cool holographic

54:26 lenses so you can put a holographic lens in front

54:28 of it it's just a thin sheet of plastic and you can make it be a wide or narrow

54:33 beam or whatever they're really really quite amazing and they

54:38 came just in time for this project i wouldn't

54:40 been able to get any of the high speed

54:41 without them [Music] [Applause] [Music] see the high-speed camera going

54:56 nuts [Music] that is so awesome oh my goodness

55:05 i just love seeing it compute where the target's gonna

55:07 go and then intercepting it so cool so that was

55:12 a case where it couldn't reach apogee so it

55:15 just shortcuts it to the side there is one more

55:18 thing i really want to see he has no idea

55:21 what's coming maybe he does there's nothing to worry

55:24 about we're just going to shoot this tiny apple off

55:27 his head to hit william tell nano we have

55:30 to be able to hit not only bullseyes we have

55:32 to hit the exact same spot on the bull's

55:35 eye every time this is the best that i can

55:37 do with the bow so we're not gonna be able

55:39 to do it every time but we should be able

55:40 to do it sometimes look how close we're getting

55:43 to splitting another arrow it wants to real bad all

55:49 right he's got his safety goggles on terrible shot

55:58 this is just embarrassing that's what i'm talking about looks like

56:03 he got him right in the heart that actually

56:06 really messed up my target it shoved the lego man

56:09 into the target and took a chunk out of it

56:11 but i yeah it was worth it kill shot 100.

56:16 oh yeah you did everything important we removed the apple

56:20 but it wasn't quite the surgical removal that i hoped for let's

56:24 try it again [Music] i couldn't believe that this worked so

56:34 i actually put these out of order this was the first one

56:37 that i did i nailed it and then i did it

56:39 again trying to get some higher speed slow-mo because the air is

56:43 going so fast it's like three frames it did remove his arms

56:46 and legs again but i think this is about as surgical

56:49 as we're gonna get this bow is really cool but it

56:56 could be a lot better in fact i'm already working

56:58 on a v2 which i'll talk about in just a second before

57:01 i get to that i want to talk about this video sponsor

57:03 i would normally skip the sponsor but there's maybe some neat

57:06 stuff to talk about there so we'll just go through it opportunity

57:09 that could literally change your life and i know that sounds

57:12 really hyperbolic but i did the thing i'm about to tell you

57:15 about nine years ago and it completely altered the trajectory

57:18 of my life so check it out and you just might change

57:21 your life this is a very unusual spot so this is

57:25 actually up in boston this is at the headquarters for form labs

57:29 this is the print production room so they make all

57:33 these sample parts and things for people they make a crazy amount

57:36 of them sponsorship it's actually for an opportunity and it's with formlabs

57:40 it's a really really cool room the machines over here these are

57:46 one of these days i'll get my hands on one

57:48 of these these are a huge sla printer that it's called the form3l

57:52 they're really cool you can print like computers and monitors

57:55 and really big stuff on them but for me really big functional

57:59 parts i think they'd be really sweet around me all those really

58:03 nice parts that you see on my projects those are made

58:05 on these machines they're 3d printers they're some of the best

58:09 so i worked on the design of these and the form 3.

58:11 so i was running that program when we were

58:15 designing the form 3 we made the optical module

58:19 in it there's a removable thing that has all

58:22 the high performance optical crap in it this machine uses

58:25 the exact same optical module as the form

58:27 three there's just two of them instead of one

58:30 and the advantage of that is you get a module

58:33 that's been designed and tested for like thousands and thousands

58:36 of hours and you can just put it in the big machine you don't have to remake it

58:40 and retest it and if you improve it it

58:42 improves both machines so it's pretty neat strategy to get

58:46 kind of two machines for one and someone's asking

58:50 how much resin to fill a tank that large

58:53 i don't know exactly where it landed but i

58:55 think it's about a liter in the world i've

58:58 used one on literally every project that i've

59:00 done i actually worked here for eight years developing

59:03 these and other machines i'm up here right now

59:06 for the annual hackathon where employees build basically whatever they

59:09 want this machine is actually it's moving a basically

59:14 a sheet of glow-in-the-dark film up and down at like

59:19 30 or 40 hertz and this is this is actually inside of a form 2 printer so it's

59:25 the z-axis of a form 2 except for up

59:27 top there's an o drive with a really overpowered

59:30 motor that just moves this thing up and down

59:32 super super fast and then there's a laser galvo scanning

59:36 out frames at different points in time to draw

59:40 3d objects it's sort of the equivalent of that uh

59:46 jarrow beam fenderson if you've ever seen the guy

59:48 that draws on an oscilloscope i think he had

59:50 a kickstarter about it same thing just in 3d

59:53 really really cool effect to see in real life

59:55 you can check out all the jobs at 4max.com okay this bow is cool but i want more

1:00:01 i want to be competitive with real archers someone

1:00:05 was asking about this in the comments for the uh

1:00:09 patreon post and [Music] i did set up all

1:00:13 the dragon cameras here they have an incredible range

1:00:17 they can see basically from where i'm shooting

1:00:20 to the target the tricky thing with setting up all

1:00:23 the cameras is that they they all have to overlap

1:00:26 so i need cameras looking at me i need

1:00:28 cameras looking at the target and they need

1:00:30 to overlap quite a bit so that they can knit

1:00:33 together into one homogenous volume that i'm tracking and so

1:00:38 it was a bit of a pain to figure out the right arrangement but once i had

1:00:42 the right arrangement it worked no problem like i was

1:00:44 saying before once they're in the right position you

1:00:47 just put them anywhere so they're just on tripods throw

1:00:49 them out wherever wave the wand and you're good

1:00:52 to go the downside of this this is 60

1:00:55 odd feet and i want to go longer than that and i there's no way that uh i can

1:01:02 go to like 200 feet or something like

1:01:04 that with this setup without a ton of cameras so

1:01:07 i have some thoughts for v2 that i can

1:01:11 do this with the cameras there's also some other ways

1:01:13 that i can do it without the cameras haven't

1:01:15 decided yet what i am going to be doing

1:01:18 is improving the bow so i've i've ordered a much

1:01:23 better compound bow it's a target bow the key

1:01:27 for the really long distance is extreme extreme repeatability

1:01:32 for its pointing i'm not going to get into what

1:01:34 that would look like here but it's a completely

1:01:38 different design than what i have with this bow

1:01:42 and i think we'll see i have to build

1:01:44 it i think it can be substantially better accuracy

1:01:48 than this one i actually took this bow

1:01:51 to a local range but that's a story the first shot

1:01:54 that i did at the range you know like

1:01:56 all the people are there and they're like oh

1:01:57 this is so exciting this is so cool wow

1:02:00 the technology it's so good and the the cable

1:02:04 for the release got caught in the cams

1:02:07 for the compound bow and it popped the string off

1:02:10 and everyone's like oh no because it's that's like you

1:02:15 know bow destroying event sometimes it was pretty funny it

1:02:19 was like the hugest air coming out of the room

1:02:22 ever but they're able to fix it it's no big deal sorry for another day what you

1:02:27 need to know is that william tell nano is

1:02:29 the key to everything when you push the target

1:02:32 really far away small errors become big errors phone hit

1:02:36 bulls eyes at 60 feet or 150 feet i

1:02:39 need to be able to hit this apple every single

1:02:41 time at this that was actually the origin

1:02:43 of william tell nano is backing out the accuracy

1:02:47 that i need at a shorter distance to hit

1:02:51 very accurately bull's-eyes at a long distance it turned out

1:02:54 to be a little bit bigger than the apple

1:02:56 for william tell nano and so that became my little

1:02:59 target to see how well i was doing and so

1:03:02 you can see that i can't hit that every

1:03:03 time i can't hit a bull's-eye at distance every

1:03:05 time although at the range i was i was

1:03:09 doing pretty reasonable but not good enough i want

1:03:13 to like be competitive ideally with like an olympic archer

1:03:18 but we'll see if i can do that this distance

1:03:20 plus it's awesome i think i need to make

1:03:22 a version too and i kind of have a design in my head what it would look like

1:03:27 i'm probably going to do that get ready legos

1:03:29 you're ready legos watch out if that or any

1:03:33 other crazy things i build sound cool please consider

1:03:36 subscribing it helps me out and you'll get i was

1:03:38 trying to figure out what the constant force spring

1:03:40 should be this is a string that goes up to the ceiling and back down to a weight

1:03:44 and that acts as sort of a constant force spring

1:03:46 although it's weird because it pulls on it

1:03:49 pulls my arm up too that caused some problems

1:03:52 and i guess the final thing get notified when

1:03:55 i post new videos here if you'd like to support

1:03:57 these projects yeah i've had a number of people

1:04:00 telling me that the whisker biscuit is bad

1:04:03 and that a dropper rest is better probably

1:04:06 at the longer distance but i just can't imagine that factoring

1:04:09 in at the distances i was doing here even

1:04:12 for the uh superstitious value i will not have

1:04:14 a whisker biscuit next time i will have a dropper

1:04:18 rest check out my patreon that's pretty cool thanks there's

1:04:24 my that's the little latching mechanism to remove the bow

1:04:28 so uh otherwise thanks tuning in and i will

1:04:33 see y'all later have a good night or wherever

1:04:36 your time zone is for me it's night see ya