The one thing that stuck to my mind while watching this video was that no matter how complicated and technical the build gets there is always a role that "Hot glue" can play in it. 😀
Or, to paraphrase a paraphrase by Adam Savage: “If you want something that is moving to not move, use duct tape (or hot glue). If you want something that is not moving to move, use WD40.”
@@zanderwohl you could just write a program to only take/record frames when the rocket is pointing a certain degree instead of this contraption trying to counter rotate with the rocket itself, which will probably make it really unstable. which he pointed out, and he wants to fix it by adding another spinning reaction wheel which goes the opposite way.... insanity. my idea way more practical, no unnecessary weight, and dont worry it wont be blurry, especially when ure recording at 120fps like he is. at that point spinning it faster would be beneficial for the video, and stability. but that wont bring in the views, so i can understand why go the unpractical asf route. its actually a very good strategy to do something in a bad way so alot of people comment their opinion on why its bad and the algorithm sees it as very good viewer engagement. and viewer engagement pays itself in weight of gold. its what makes your videos go viral. thats the truth unfortunately gotta get used to it regardless if you like it or not
Me too, I watched it and saw the fins rotating on those two seconds of stable video. I'd call that a success (at a Proof Of Concept level, which this has to be)!
@@MrPaxio your idea is practical but you lose a lot in video quality and continuity. the video will be just a compilation with images with no connection to one another. since the guy's a youtuber who makes videos, his idea makes way more sense, mr. know it all.
The guidance subsystem uses deviations to generate corrective commands to drive the missile from a position where it is to a position where it isn't, and arriving at a position where it wasn't, it now is.
Consequently, the position where it is, is now the position that it wasn't, and it follows that the position that it was, is now the position that it isn't.
What I find awesome is that till now you tried to get your rockets "just" to fly, but now you've passed that level where you can start using them to try your own experiences... that means, building flying rockets is now "usual" for you, and that's an extraordinary step forward
I think the results are a lot better than you give yourself credit for! Unfortunately we all know getting that last 20% of the way there takes so much effort. I was going to say have a deadband to disable spinning if the roll rate is under the lower limit of the motor, but a stepper should be even better for butter smooth video! Keep it up!
@@polyscient Dunno the specs of the degree steps but he could utilize half steps so it wont be as jarring, gearing can help too but will lose some top speed.
He should also put a limit to the degree of roll error so after one full rotation of error or so it doesnt counter rotate a lot and just goes back to the heading it started at
Conservation of angular momentum... Your camera ring is a reaction wheel as others have stated. Look at adding a counter-rotating ring as well to maintain the angular momentum of the airframe. You could also use that combined system to de-spin the airframe like... 1 second before chute deployment to mitigate chute entanglement.
While I agree that the despin rig is acting like a reaction wheel, I don't think adding a second wheel is the best idea. It adds mass and complexity to an already existing rocket, which could be more of a headache than it is helpful. I think the best move is to use high rpm slip rings to get all of the electronics for this sub assembly into the avionics bay, lowering it's mass, dropping it's moment of inertia to lower the torque roll it's exerting on the vehicle. It doesn't solve the problem completely, but it could have the potential to greatly reduce the impact.
I’m not saying he doesn’t need to optimize the mass of the camera ring, that is obvious. However fin induced spin is completely GONE and reversed only 5 seconds after the camera ring kicked in. The oscillation this would induce would be worse than not spinning at all. Optimization will only go so far in extending that time, I doubt by a power of magnitude, and certainly not to apogee. Counter rotating wheel is the best option IMO. He already explored a fairly mechanical execution, it’s not too much to mechanically couple a second ring to the same stepper in v2.
@@jerry3790 I think this misses the point though. He’s trying to continuously stabilize the camera against a spin stabilized airframe. My secondary point of despinning at Apo should really be taken as like a secondary beneficial side effect if he decided to implement that in the same system. My point of the second ring is to counter spin the camera ring ONLY.
I have mentioned this in the last video, but something you could do rather than spinning the whole assembly, is spinning a mirror around the central point and then putting the camera statically in the body of the rocket. This would allow for a lot less load to be spun up and down. Similar concept to what you have already done, just less stuff spinning
I also suggested this when he had a gopro sticking halfway out the side of the rocket body causing unbalanced drag. Unfortunately this fix wont work for this use case, as the center of the viewpoint will remain static but the entire image will still be spinning around that central point, expecting no blur at the edges of the frame would be naïve.
if you stick a rotating periscope onto a cam, you'll have to put the cam in the middle. that will lose you the easy cable passthrough. which might be acceptable.
@@chrisjacobsen1659 : Yep, shutter speed isn't infinite, so you'll get blur everywhere outside of the exact rotation axis, and in this case it could be pretty bad.
It's not an engineer's job to overcomplicate things. In fact it is exactly the opposite. But it is in the engineering spirit to take something (even apparently stupid) as a challenge and do anything to make it work. Well done Joe, amazing video as always!!
I think that’s what (don’t over engineer your design) my machine design professor (TRW experienced) told me on a team design project we proposed back in ‘75. “Back to the drawing board fellas” he said with a chuckle.
I have worked for engineers most of my life....thay may not be meant to over complicate things, but oh man they sure seem to think that they are that's for sure!
yup! its not the motor, its actually the driver that isn't good at slow speeds, but he would need to add some kind of encoder to keep track of where the motor is for field oriented control(FOC).
Agreed! He needs a rotary encoder so he can keep track of the motor position while being able to supply more current to allow for torque holding. A cheap sensored ESC and motor should solve this problem, but I'm worried he'll go down the stepper path bc he's lost hope for brushless. IMHO, steppers are the wrong move. They're meant for precision and low rpm applications, not for high speed despinning like Joe wants to use them for. Oh well, fingers crossed he reads our comments!
As a proof of concept I’d call that a huge success. I have to admit some skepticism about destining the camera at first, but the footage looked pretty darn good when the rig worked. The stepper motor idea seems solid. Those little brushless motors don’t always have the resolution for fine work like this. I really enjoy watching your thought process through these videos. You’re solving ridiculously hard technical challenges and have the good sense to make fun of yourself when appropriate. The fact that you’ve been able to do as much as you have is incredibly impressive, and the way you learn from your failures is inspirational.
3:00 the other problem with using a 360 camera is that, ultimately, it is still a camera, and if you spin it like that, you'll get a ton of motion blur. You can stabilize the image in post and get a very stable video... of motion blur.
Dude! It's been years watching your channel and this is still some of the best content on the internet. I hope your having fun and enjoying it still because I know we are!
Despinning the camera on your crazy prop-copter would be an interesting challenge. Slightly easier than this though because you don’t have to sandwich in the structure like this one does.
When it works it works pretty well. I hope you iterate on it with FOC and a brushless gimbal motor actually designed for this kind of thing, what you already have is very promising!
Joe, you are way too hard on yourself - that footage (for a first or second-generation design/build) was absolutely AMAZING! You are an incredible engineer, machinist, and rocketeer!!
“The camera spinner didn’t work…..great” Joe, I’m not sure you saw the same video I did. For a *first try prototype* that was amazing. Give yourself more credit dude, you are an inspiration to those of us whose engineering consists of making parts to fix something at home.
Seems like a success to me, most of the issues outlined you predicted or take a little programming effort to fix. It's wonderful to have such good footage that pretty much outlines every aspect of the problems.
honestly, that was a flawless first attempt. i know you engineer for expensive rockets you can't rapidly iterate, but that is far better than i expected from a sensorless brushless motor
I know they're uncool these days, but a little brushed gear motor would also solve this problem as you don't need the power density of brushless here, and they're very smooth
Might want to research rollerons in your fins, like some air-to-air missiles. Those use passive rollerons spun up by the slipstream of the carrying aircraft. I was working on a space shot where we thought about using hard disk drive platters and motors inside the fins for the gyroscopic fin stabilization. Regardless, keep up the superb work!
My take on this comment is: 1) probably not what you are intending and 2) that you're saying to change the camera spinner so instead of being de-spun by a motor based on calculations, to instead attach fins with rollerons that de-spin using inertia... but I'm not even sure that would make sense.
Came down to say something like this, fins are pretty excessive for high mach vehicles, *especially* low mass rockets. Something like rollerons or tiny flap stabilizers makes more sense. Rollerrons are pretty heavy and are probably reliant on the mass for a strong flex/vibration resistant housing. But the idea is probably interesting to look into. Especially passive self-corrective stabilization.
To those that might not know, rollerons are an enclosed gyroscope inside the trailing edge of a tailfin that is spun up by airflow over a small exposed portion. (Very common on military rockets/missiles, it's that toothy little wheel on the tailtin corners) They are connected to the static portion of the tailfin with a simple (but sometimes damped) hinge and function as a standard control surface but are passive and have only 2 moving parts. If tuned correctly, the gyroscopic forces automatically tilt the control surface the perfect amount to counteract any roll and they can be surprising precise/accurate/responsive bc gyroscopes respond instantly and proportionally to even the smallest/slowest rates of change while not overcorrecting for the largest/fastest. It can essentially be thought of as a "wireless" equivalent to a mechanical bevel gear, neat.
Honestly the brief moments where the camera DOES counter-rotate correctly make it pretty clear that this is a fantastic approach if you can iterate it over the finish line.
Several others have mentioned my concern, that of the spinning camera assembly acting as a gyro and possibly screwing with the spin stabilization of the rocket itself. Would it be possible to mount the batteries closer to the center of the assembly? It looks like there is room. Since those are probably the greater part of the weight, moving them closer to the axis of rotation should help limit the effects, might even make it easier for the motor to spin the assembly up.Edit: Personally I think the results were pretty darn good for a prototype, we'll wait for the second flight for it to be perfect!
@@tjm2212 Yes, though I think you mean the camera assembly wouldn't have any angular momentum. If it were somehow levitated inside the rocket in a vacuum then as the rocket starts to rotate the camera would maintain its orientation due to its nonzero rotational inertia, and the fact that the rocket body would not be exerting a torque on it. If the motors could perfectly oppose the movement of the rocket then the camera subsystem would essentially be decoupled from the angular momentum of the rocket and would only contribute its mass to the downward force due to gravity. But for now I think the implementation leaves a bit to be desired 😂
Others have said this but I want to add as well, the camera system is working as a reaction wheel and is changing the angular momentum of the rocket. When the camera is spun up, this is causing the rocket body to spin faster and also potentially some weird gyroscopic effects. Two things can be done that I know of: 1. Try to lower the moment of inertia of the camera system, or 2. Add a wheel with the same moment of inertia as the camera system with opposite angular velocity and acceleration. This video was awesome, can’t wait to see what you’re doing next.
Dude, i have watched your videos on and off since you had 1000 subs. And what an amazing improvement you've made in all areas. But especially your ability to be a natural when presenting in front of a camera. You've gone from nigh unwatchable to among the very best. Greetings from Denmark
This is sick. Excited to see how this comes along in the future. I'm quite impressed you are doing so many disciples of engineering concurrently: mechanical, aero, electrical, controls, software, manufacturing!
Even adding to that is the filming, voice-over, and editing. People rarely mention that, but that isn't as easy as counting to four. Massive amount of skill all around. Definitely a good role model and someone to look up to!
Have you considered using a more passive stabilization using gyroscopes? Get them spun up before launch, and they should resist any roll. They could even double as steering control in the future.
He did experiment with gyros but they saturate very quickly. They need to be either very fast or quite heavy to be effective. Idk what RPMs he got up to, so maybe a hard drive motor would be fast enough.
I've just started construction on my L3 rocket and have started mixing my own propellants after going to the balls launch this year. I hope to eventually be on a space shot team and I know that is ambitious, but compared to what you do I am just a child in a playground. I love watching you videos and I learn a lot. Keep it up!
I think you might need a sensored brushless motor and something like a VESC compatible ESC so you can get higher torque at low rpms. A good source of info on this might turn out to be the Onewheel Community, or people who diy their own eboards. Do some research on VESC and see if it's what you need! I do also have some hesitations about stepper motors. Stepper motors aren't traditionally great at high speed applications, but it would absolutely have the ability to despin those slower rotations near apogee. If you do go this route I recommend the Trinamic stepper drivers, specifically the TMC2130-TA (SPI interface so should be Ava compatible, up to 1.4A phase current, passive breaking, and load dependant speed control). Some sort of clutched assembly to allow for the use of both a brushless motor for high speeds and a stepper for low speeds could be an interesting trajectory, but would add several layers of complexity. I've also seen a lot of other people address the reaction wheel issue, and while I agree with them, I don't think the best answer is to add another reaction wheel to counter the spin. I think the best idea is to get the mass of your assembly down by off loading as much of the electronics for the despin rig to the avionics bay. Have you thought of using high rpm slip rings for all the circuitry, and leaving essentially only the cameras on the rig? I'd assume this would get the mass of the rig down a considerable amount, making the it's moment of inertia much lower, exerting less torque roll on the rocket. No matter what, I can't wait to see where you go from here Joe! This space shot is closer than you think!
Don’t beat yourself up. I think that was a good starting place. It’s funny at 1340 I was thinking of stepper motors, and then that’s what you started talking about. My field was in marine, electronics and satellite communication. The earlier equipment is gyros and then they went to stepper motors. I am fascinated, watching you make your own rockets and rocket nozzles. I would like to see a walk-through of your shop and your tools subscriber here South Florida I often watch, the falcon nines takeoff and sometimes see the boosters return.
@@hund4440 My thought was, to de-spin the camera to an acceptable angular velocity, where the wouldn't be any motion blur. But I don't actually know how slow it would need to be. You're right, maybe it would still be too fast, although the part that was semi-stabelized looked clear enough.
I really like what you've done here. Like the stepper motor idea too. It's much simpler and adaptable. Presuming of course that you're balancing everything, it's a pretty clean solution. I know there's also a lot of even smaller stepper motors that might also do the trick. You might find that a pancake stepper works nicely too by angling your drive gear 90 degrees. I made a similar modification to my high-speed 3D printer to decrease the weight of the printhead. My situation needed high torque though, not high speed. I only mentioned it, because some of those pancake steppers are really small.
That's a really cool idea, when you initially started the video, I thought about a similar solution, the only difference is that I thought you may end up turning the entire "turntable" into a brushless DC motor. I love this and you should definitely keep trying!
There is an easy way to accomplish this. Add counter rotating fins just for your camera. Your camera has a bearing, so it will rotate as fast as rocket is rotating but in opposite direction. Let the wind gust to do the rest.
I think the footage looks already good! Sure, there are some parts where it fails. But at the sweet point where it works, it works pretty good! Nice Job!
And it is great footage to present the problems. The whole analysis shows an admirable level of control. Might not be enough for the builder, but it certainly makes an exceptional RUclips video.
Damn I love watching this. Truly the Engineering way: do I need this? No. Do I want it? Yes. Also you should give yourself more credit. You did the hard part. You converted an idea into a working prototype. All you need to do now is get rid of the gremlins that make the thing not do exactly what you want. Also, I don't know if it is what you meant about the small fins when explaining why it might not've worked but it seems the rocket isn't in constant spin. It seems to spin, then stabilize, then spin, then stabilize. Good luck and can't wait to see the next step in development :)
for the stabilization, you should stabilize based on the delta of the roll, since its not actually important that the camera holds a steady roll direction, but just that it dampens roll rate. this would solve the windup issue and it might also make it easier to get the motor to just not bother correcting if the roll rate is lower than its minimum starting speed, and just those changes alone i think would be a huge improvement for the viewability additionally, although it probably doesnt matter at the current RPMs youre experiencing, you could always have it take into account the framerate of the camera and find the nearest multiple, so if your camera is 60hz and the rocket is rolling at 55rps, you could spin the camera to +5rps rather than -55rps and the framerate will hopefully turn it into a smooth video, although i guess that might end up producing some artifacts with shutter speed producing warped video... but its something to consider nonetheless
A a few others have mentioned you can take the ideas and tech used in high speed cameras following very fast moving objects and use moving mirrors. This solves a couple of problems also making it easier to mount bigger cameras(the gopro) inside the rocket and use a spinning periscope to capture beautiful footage of the outside. Great work it is really interesting watching the progression!
Those are mutually exclusive. BLDC and FOC are two diffrent methods of commutation control for a motor. There are no BLDC vs FOC motors they are just brushless motors being controlled by a BLDC or an FOC esc. Still I agree and FOC system would solve all the brushless motor issues.
FOC motor control is technically 3 phase AC control. BLDC is as you said is direct current. A brushless motor can be used as both a DC motor or an AC motor depending on the commutation method. So saying a BLDC motor makes no sense as the motor does not require a DC control scheme it can use an AC scheme makeing it a BLAC motor. Therefore its inaccurate to call it a BLDC motor instead its a BL motor using BLDC commutation.
@@nocare BLDC just the common terminology. If I'm correct, the PWM signal is sinusoidally pulsed DC, so technically they could be still DC motors. Edit: I'm not looking for an argument, was just trying to be helpful as a hobbyist. If I google BLAC gimbal motor there are no results. Googling BLDC gimbal motors gives the exact result of my suggestion. There is no more to the whole thing from my side.
@@zsigmondkara hmmm how about this. What then is the name of the commutation method used by a standard brushless ESC. SInce we both agree FOC is not the standard method most ESCs employ.
Half a million subs is nothing to sneeze at, and it is tremendous accomplishment, but how the howly-f does this channel only have half a million subs? Absolutely first rate in every way!
Love it! BPS Space videos always get the design vibes rolling! Question: 1.) Why don't you write a code to calculate roll rate vs shutter timing and frame grab and "stitch" a roll stabilized video together? I'm not a coder so this may be more difficult than I realize with mixing video encoding or is camera sensor and shutter hardware the bottleneck? 2.) I was also wondering, have you considered integrating a coaxial system that incorporates a reaction wheel? If power and data connections through rotating electrical slip ring connector are stable, the mass of the batteries could be leveraged as reaction mass? The additional mass of the rotating camera system seems very high unless the quality of video is essentially "mission priority" over usable payload mass. Edit :BPS Space contained a . and converted to hyperlink removed.
cameras can do 60fps usually, and you know the spin rate exactly from the sensor, so you probably can stitch anything together under 60 rotations per second. You would get some weird artifacts because of rolling shutter delay, there are CV cameras that can do more fps and instant shutter but that's a little more expensive. But this is a cool idea, never seen anyone try it before
1) Wouldn't the frame rate of the stabilized video be limited to the rate of rotation? If the camera is rotating at 1 RPS you would only get 1 FPS. Also if the rotation is fast enough each frame will be distorted by the CCD's scanning pattern.
@@polyscient I think you would if it was a single camera set up. However with dual opposing cameras (which i assume was for balance) the field of view could be optimized to meat at a central plane or known good overlap. Once again this is just assumption but then the rotation rate would adjust a "variable" in a function that chooses which video frame to grab based the camera settings for shutter speed. I think you will lose a significant amount of "off target data" so a higher FPS capture/sensor speed is desirable. Is the sensor chip scanning pattern distortion accountable enough as a known calculable constant? If so wouldn't using a lower resolution sensor potentially yield greater scanning speed as a relative function to surface area? Or as simple as rotating the unit so the distortions are less obtrusive or don't align with horizon or other desired targets?
@@ffoska The rolling shutter delay effect is almost exactly what we are trying to accomplish. By determining where the target field of view is(from the flight computer or other hardware), the algorithm picks individual frames to stitch a new video from. At high rocket roll rates is where the algorithm I think would really take off in that it would begin "shedding" erroneous frames off the target field of view in order to "stitch" a new stabilized video. I think this could be implemented downstream as a "fix it in post" or at a ground station with sufficient graphical compute power to re-encode/stitch on the fly. Implementing it on the rocket flight hardware seems unnecessary and detrimental to the rockets flight performance which we can see is nominal even with the addition of moving payloads (keepin' the pointy end up and hula-hooping).
@@TAR-D yes, by 'rolling shutter' I meant the CCD scanning pattern. You would just record the frames, but you also record the telemetry, so you can also use that later - but it's probably not even needed, because the sampling is fast enough to get multiple images of the same region and do keypoint matching. And I also meant using multiple cameras to get a 360 view, so it's just a question of unspinning the image.
Only just found your channel. You could take a hint from T-PODs they use an angled mirror with a stepper motor for small corrections, the head of the pod has the main motor, de-rotation on other axes is done in software when you reach the gimbal limit. I am thinking you have a transparent section for the mirror which does the de-rotation while the the camera assembly just rotates within the core. Or you could also have a pyramid or cone for the mirror (if using software image correction) again inside a transparent ring section for the rocket core.
This could probably be stabilized passively using fins on the rotating camera rig. The fins on the righ could be mounted such that they can swivel freely in pitch axis; this will make them automatically align themselves with the flow and dampen camera rig rotation.
Honestly, that footage looks really good! I'd call the flight a success, everything you were worried about actually worked and you've proven all your hypotheses.
We need more of this. Although you don't consider this a success looking at it from the outside it worked very well. I wasn't sure if this was possible but now you have me convinced.
I loved As an engineer it is my job to overcomplicate things, lool Speaking of which, you should keep the whole control assembly, batteries speed controller chips motor, everything stationary above, then only spin a ring with the camera. Then slipring if you need any wires. This way it can be shorter section of inner tube for strength, easier pid tuning, less reaction wheel weight, everything. And can make the tube 3 inch. Have the drive gear on the inner tube be above the bearing so the motor can be mounted to the bulkhead. Or inside the tube and motor goes in for better strength Yeah for the motor you should use a sensored brushless with and tune the esc pids for speed, similar to how electric longboards can accelerate smoothly from the known weight and accel they want, and rc cars want fast accel. Or use a stepper motor with a at least an 8 bit driver like a printer, if you need real speed higher, or a servo motor, or a continuous rotation servo. Or a 775 brushed motor, or more power any .3w brushed motor, Then you’ll be good Good video and looks fun to learn
This one time in school we made rockets and launched them with compressed air, and I did not have enough fins for proper stability so I folded the ends of the fins to around 30-45 degrees so it was stable due to gyroscopic stability (I think that's the name?), and its cool how it works so well. Great video!
While I can understand your disappointment in the footage you got but it is still probably the best 'looking down' footage you have ever got. It is also way easier to watch! Don't be too down hearted and keep making those little changes.
What you need is an AC servo. There are cheap & quick DIY solutions like SimpleFOC. Stick an encoder onto your motor, replace the ESC with a SimpleFOC board, and that's all you need.
Brother, I got to tell you I live outside of Nashville and I cheer for you every time you go out and do a successful lunch. Even though I can't be there. I'm also an engineer and I'm glad that you went with this solution because this seems like the most fun one to implement IIBH. That's secretly why you're really doing it, isn't it?
I'm so excited for the ultimate spaceshot, Joe, but I'm equally thrilled with every step and iteration along the way! I love these components that are technical but explained in such a way that my non-engineer brain can understand them, and I wholly endorse this pursuit of non-spinning footage from a spinning rocket because that sounds AWESOME 😁I'm also immensely pleased that SEND IT! is serving as such a reliable testbed for all this tech development - keep up the great work Joe, and I'll be here to follow along on your journey!
Ive seen on yt some video of a rocket that had lower bottom rotating and upper "stage" was stationary whole the time. They talked in a video, that its pretty common fix for this problem. Also stepper in higher rotating speeds... gl man, you can do it :P
1. May put a reaction wheel for stability so you don't need a rotating rocket. The camera can be outside without this complicated stuff. 2. If you keep with the rotating camera idea, I think the motor is fine. What I suggest is change on the programming: DO NOT trying to compensate when the rotation of the rocket is so slow. It will look actually cool! Just my suggestions, your work is really great!
Stepper motors: You're going to need a stepper motor controller. At work, I've been using Pololu Tic TIC825, a Stepper Motor Controller/driver that offers USB, UART, I2C, RC pulses, step and direction, and analog voltage interfaces. The controller can be used to control position OR speed. The other Tic controllers seem to use the same microcontroller, and differ mostly in the range of motor currents and supply voltages. I've been greatly impressed by their performance. The T825 goes to 50k steps/second, and most stepper motors are 200 steps/revolution, so up to 50k/200 = 250 rev/second. In your application, you're counter-spinning the camera to minimize the instantaneous roll-rate. Maintaining a fixed counter-spun *position* is not actually necessary (fixed positioning would be the extended feature target, but you'd probably need a better gyro.) I've assumed the camera ring has it's own roll rate gyro, but it likely has a little bit of roll-rate drift, even when when stationary. Capture rate drift while sitting on the pad, then after launch, offset measured roll rate by the captured drift rate. Having the camera ring doing a slow roll during flight shouldn't be a problem.
You wanted video stability.....I'd say you accomplished it for the most part. You can't always get perfection on your first attempt. Now comes the fun of the refining process.
- the big main fins overpowered the small fin intended to spin the rocket. - a BLDC motor can be operated as a stepper, and microstepped for smaller spin intervals. This will require a reprogramming of the speed controller.
The motor isn't necessarily the problem, it is the ESC. The hobby ESC's use back emf to synchronize the commutation of the phases. A better ESC / better ESC tuning could decrease the back emf voltage (which is proportional to rpm) that is required to synchronize the servo and motor. A way to avoid the need for the esc to rely on b-EMF to syncroise is to use a motor sensor i.e. hall-effect sensor / rotary encoder / magnetic encoder... this turns it into more of a servo drive for positon / velocity control. The next step from there is an ESC that produces sine wave phase ouputs rather than bang bang outputs.... then you will have arrived a a field oriented control (FOC) servo! Check out simplefoc or Tinymovr
I hope you were being cheeky. The job of an engineer is not to overcomplicate things, but to make them as simple as possible while meeting the requirements (spec/cost/schedule...). ie: No more complex than necessary. Using the lathe as a test driver was very cool - and was a simple approach!
In my opinion the brushless motor is the good aproach to solve this problem, the issues that you had with the “low resolution” of the motor it’s caused because you are using the wrong speed controler, not the optimal motor for the problem. Try looking for some Brushless FOC control and you will realize that is what you need to solve that. If you need help with it or some more starter friendly info just contact me! Keep doing what you do!
I think, like others have said, for a 1st flight test, albeit of a design you had already done multiple revisions to statically, this was the complete opposite of your conclusion....... factoring in all the things you KNEW before flight (concerns over the motor choice, low rotation speed performance, etc). This was a GREAT success as the bit of stabilised footage did have the exact result you wanted.... you wanted to test if you COULD stabalise footage in this manner and you proved you can, but with further refinements to get the outcome you want across the range of spin speeds. Other have said you may need to do something to balance the angular momentum as what you intend is for the rocket to behave as if the spining section wasn't there, but the spinning section - as this scale - is acting as a reaction wheel affecting the rockets inherent spin. Adding a counter rotating element of the similar mass should go some way to balancing those angular momentums. Tl;dr I genuinely think your conclusion is wrong considering the constraints on your ability to test systems on the ground (as you say, wind tunnel and lot and lots and lots of CFD would help, and a big company would have access to that stuff, but you as a small group cannot). You proved it is actually possible to stabalise the footage in this way, and know areas you can improve from the get go, and using the retrieved footage to guide further improvements.....
Hey, stepper motors have a pretty slow maximum speed. See if it is sufficient for you. Make sure to not gear it down much, since they can do very precise microstepping and you need all the speed you can get. Other than that, the solution that will work (but is overkill) is something like an ODrive.
Same rig - Get a bit chonkier, slower motor, and use an ODrive S1 for precise low speed control instead of a cheap ESC. Won't go to sleep on you and should work great.
Careful with the Stepper motor idea. Before you get to deep in to it, test it's torque.. I tried a stepper for a focuser on my telescope and the micro stepper motor (looks similar to yours) lacked the needed torque. and since your plate will be jerked quite heavily, just be sure it and work under that condition. I would try putting a Brass ring around your brushless outrunner motor and mount it so that the rotation axis is perpendicular to your rocket's long axis.. spin up the motor pre-launch and use as a gyro with a free moving rotator plate.
Great progress Joe, Im sure you will get better results from the Stepper Motor. My only question/suggestion is to ensure the payload is rotationally balanced as it could be affecting the spin rate and influencing the IMU, looking forward to the next launch.. PR
How about using a gyroscope which can move in x,y,z coordinates. The outermost layer will rotate along with the rocket. You can rotate the internal wheel without bothering about speed and timing, this will create a gyro motion and stabilize the cameras attached to the pivot point.
Crazy good. Just smooth it out, put a pair of control surfaces on that spinamajig and voiala, You can steer that thing straight out of atmosphere. Thanks for the ride, have fun.
I have found that stepper motors benefit greatly with reduction drives. Giving greater power and accuracy without loosing steps due to reduced torque of direct drive. Experience from designing and building CNC equipment. Love your concept and testing.
Well done, nearly there! As a specialist imager (photographer), I would have taken another route. I would have have run the camera at, may be, 120 or 240 frames per second and then edit every one in three or four or five frames to get a slowly rotating image. I could then stabilize the image or could accept the residual roll. Edit: Or set up a camera that takes 1 image every rotation. I suppose this supports the idea that you take the route that you are most familiar with, neither is wrong. One can be more effective in the end. ( I think your idea would be more effective because the roll is not constant.) Good luck!
Awesome Joe ! Yes using BLDC motors won't give better angular precision stepper servo is comparative much better that ordinary stepper motors. I hope you saw Vikram - S launch by Skyroot Aerospace as well !
What you need is not gears and an ESC, its a ring type USM motor like those used by Nikon and Canon for their professional (D)SLR lenses. They are direct drive (immediate), high torque, low inertia and completely hollow!
I'm a retired engineer and I'm impressed. I've had a lot of projects that didn't work nearly that well on the first try. And don't forget the engineers motto: If it ain't broke, it needs more features. 😉
Try to mount a step motor (like old capstan from old VCR) to move all the mecha. This kind of step motors are strong and react to slow speeds with very precision. They are controlled by PWM and uses a hall sensor to read back the pulse from the motor… to me is the best option.
This is more my area of expertise. I'd strongly advise against using a stepper. Stepper motors are synchronous motors. If they lose sync they stall and won't regain sync until you reduce the drive speed to below their 'pull in' speed. You only need to exceed the maximum torque for a few milliseconds to make the motor stall. The high vibration environment of a rocket is pretty much worst case scenario. Additionally the available torque drops of sharply with speed so you have to over rate the motor considerably to get enough overhead. My first choice would be a humble DC brushed motor. Simple, cheap and reliable. Decent quality brushed motors turn smoothly at low speed and they can be driven with a simple H-bridge driver. Just keep the PWM frequency reasonably high, say 3kHz upwards. For a little better performance and a slight reduction in weight, field oriented control of a brushless DC motor works but it's a lot more complexity and you will need to spend a fair amount of time tuning it for stability. Look into drivers and firmware designed for camera gimbals. I'd recommend using the motor with hall sensors at a minimum. While FOC can be used without sensors you run into a similar problem as you have with steppers. If you momentarily exceed the maximum torque of the motor it will lose position.
Couple of thoughts from a former missile engineer: rather than trying to stabilize to a specific angle, just compensate on rate. If you're determined to stabilize on an angle, account for rollover at +/- 360
Something like the Odrive 13.2 Module from m5stack may fit your needs here a bit better than a stock hobby ESC. BLDCs operating in sensorless mode typically rely on back emf sensing to figure out what coils to activate. So startup on them typically means kicking the motor over hard to get it to move so it can know where it is. Sensored controllers have a calibration to map sensor angle to the coils in the motor and can get reasonably low speed control. It would require a rotational sensor on the motor but those are fairly simple these days. I don't know if the mentioned module meets your weight constraints, but the concept is similar and there's a large number chips that can drive small BLDCs in a sensored way if you wanted to go down that route. May be easier on your power budget than the stepper.
I get that you are an engineer, but you should not be so hard on yourself. That test went well, and more importantly it gave you lots of info on what version 2 needs to do. That's a success in my book.
The one thing that stuck to my mind while watching this video was that no matter how complicated and technical the build gets there is always a role that "Hot glue" can play in it. 😀
Yo what if you did a 360 cam and a spinning cam all in one
Or, to paraphrase a paraphrase by Adam Savage:
“If you want something that is moving to not move, use duct tape (or hot glue). If you want something that is not moving to move, use WD40.”
I bought a 150w Hotglue gun and 15×.5in glue sticks. I approve this comment. Lol
He’s going to one day sign his name on a rocket using hot glue.
Potting only hurts the guy who has to strip everything apart after!
I was shocked when you said it wasn't a success! It was a massive improvement in video stabilization, even if it wasn't perfect.
almost there doesnt mean it crossed the finish line
@@MrPaxio Bad way to think of this - you can still see much more than before.
@@zanderwohl you could just write a program to only take/record frames when the rocket is pointing a certain degree instead of this contraption trying to counter rotate with the rocket itself, which will probably make it really unstable. which he pointed out, and he wants to fix it by adding another spinning reaction wheel which goes the opposite way.... insanity. my idea way more practical, no unnecessary weight, and dont worry it wont be blurry, especially when ure recording at 120fps like he is. at that point spinning it faster would be beneficial for the video, and stability. but that wont bring in the views, so i can understand why go the unpractical asf route. its actually a very good strategy to do something in a bad way so alot of people comment their opinion on why its bad and the algorithm sees it as very good viewer engagement. and viewer engagement pays itself in weight of gold. its what makes your videos go viral. thats the truth unfortunately gotta get used to it regardless if you like it or not
Me too, I watched it and saw the fins rotating on those two seconds of stable video. I'd call that a success (at a Proof Of Concept level, which this has to be)!
@@MrPaxio your idea is practical but you lose a lot in video quality and continuity. the video will be just a compilation with images with no connection to one another.
since the guy's a youtuber who makes videos, his idea makes way more sense, mr. know it all.
The missile knows where it is, it knows this by knowing where it isnt.
And substracting everything from there it is
By subtracting where it is from where it isn't, or where it isn't from where it is (whichever is greater), it obtains a difference, or deviation
The guidance subsystem uses deviations to generate corrective commands to drive the missile from a position where it is to a position where it isn't, and arriving at a position where it wasn't, it now is.
Consequently, the position where it is, is now the position that it wasn't, and it follows that the position that it was, is now the position that it isn't.
Brilliant reference 👍🏻
What I find awesome is that till now you tried to get your rockets "just" to fly, but now you've passed that level where you can start using them to try your own experiences... that means, building flying rockets is now "usual" for you, and that's an extraordinary step forward
I think the results are a lot better than you give yourself credit for! Unfortunately we all know getting that last 20% of the way there takes so much effort. I was going to say have a deadband to disable spinning if the roll rate is under the lower limit of the motor, but a stepper should be even better for butter smooth video! Keep it up!
He should still have a deadband because the steps of the motor will probably be shaky looking below a certain speed.
@@polyscient Dunno the specs of the degree steps but he could utilize half steps so it wont be as jarring, gearing can help too but will lose some top speed.
I was thinking the same. Not perfect, but once it woke up, it tried hard.
He should also put a limit to the degree of roll error so after one full rotation of error or so it doesnt counter rotate a lot and just goes back to the heading it started at
I think it was a bloody amazing first attempt
Conservation of angular momentum... Your camera ring is a reaction wheel as others have stated. Look at adding a counter-rotating ring as well to maintain the angular momentum of the airframe. You could also use that combined system to de-spin the airframe like... 1 second before chute deployment to mitigate chute entanglement.
While I agree that the despin rig is acting like a reaction wheel, I don't think adding a second wheel is the best idea. It adds mass and complexity to an already existing rocket, which could be more of a headache than it is helpful. I think the best move is to use high rpm slip rings to get all of the electronics for this sub assembly into the avionics bay, lowering it's mass, dropping it's moment of inertia to lower the torque roll it's exerting on the vehicle. It doesn't solve the problem completely, but it could have the potential to greatly reduce the impact.
I’m not saying he doesn’t need to optimize the mass of the camera ring, that is obvious. However fin induced spin is completely GONE and reversed only 5 seconds after the camera ring kicked in. The oscillation this would induce would be worse than not spinning at all. Optimization will only go so far in extending that time, I doubt by a power of magnitude, and certainly not to apogee. Counter rotating wheel is the best option IMO. He already explored a fairly mechanical execution, it’s not too much to mechanically couple a second ring to the same stepper in v2.
Or just give up on the camera and go for the 360 one. Or give up on it altogether
There are easier ways to despin though, such as yoyo despinning.
@@jerry3790 I think this misses the point though. He’s trying to continuously stabilize the camera against a spin stabilized airframe. My secondary point of despinning at Apo should really be taken as like a secondary beneficial side effect if he decided to implement that in the same system. My point of the second ring is to counter spin the camera ring ONLY.
I have mentioned this in the last video, but something you could do rather than spinning the whole assembly, is spinning a mirror around the central point and then putting the camera statically in the body of the rocket. This would allow for a lot less load to be spun up and down. Similar concept to what you have already done, just less stuff spinning
I also suggested this when he had a gopro sticking halfway out the side of the rocket body causing unbalanced drag. Unfortunately this fix wont work for this use case, as the center of the viewpoint will remain static but the entire image will still be spinning around that central point, expecting no blur at the edges of the frame would be naïve.
if you stick a rotating periscope onto a cam, you'll have to put the cam in the middle. that will lose you the easy cable passthrough. which might be acceptable.
@@chrisjacobsen1659 : Yep, shutter speed isn't infinite, so you'll get blur everywhere outside of the exact rotation axis, and in this case it could be pretty bad.
That's sort of how IR seeker heads work.
@@chrisjacobsen1659 Assuming the camera isn't global shutter, you might also get some weird rolling shutter effects.
It's not an engineer's job to overcomplicate things. In fact it is exactly the opposite. But it is in the engineering spirit to take something (even apparently stupid) as a challenge and do anything to make it work. Well done Joe, amazing video as always!!
I think it is just a meme.
... you obviously have not met a car engine they are a whole different breed
I think that’s what (don’t over engineer your design) my machine design professor (TRW experienced) told me on a team design project we proposed back in ‘75. “Back to the drawing board fellas” he said with a chuckle.
I have worked for engineers most of my life....thay may not be meant to over complicate things, but oh man they sure seem to think that they are that's for sure!
tell that to a modern mechanic, watch him laugh in your face..!
Good Flight! In electric skateboards you get better slow rpm performance and jerkiness in brushless motors by using a sensored ESC and motor.
yup! its not the motor, its actually the driver that isn't good at slow speeds, but he would need to add some kind of encoder to keep track of where the motor is for field oriented control(FOC).
CD/DVD motors have sensors and drivers integrated in a single board. They're quite generic and you can find online how to use them...
Ok, I was wrong about integration part, here's info: akashkumar4u.blogspot.com/2013/05/cddvd-motor-driver-hack-to-run-its-own.html?m=1
Agreed! He needs a rotary encoder so he can keep track of the motor position while being able to supply more current to allow for torque holding. A cheap sensored ESC and motor should solve this problem, but I'm worried he'll go down the stepper path bc he's lost hope for brushless. IMHO, steppers are the wrong move. They're meant for precision and low rpm applications, not for high speed despinning like Joe wants to use them for. Oh well, fingers crossed he reads our comments!
What I wanted to comment on. And as Seth states, stepper motors probably won't be fast enough.
As a proof of concept I’d call that a huge success. I have to admit some skepticism about destining the camera at first, but the footage looked pretty darn good when the rig worked. The stepper motor idea seems solid. Those little brushless motors don’t always have the resolution for fine work like this.
I really enjoy watching your thought process through these videos. You’re solving ridiculously hard technical challenges and have the good sense to make fun of yourself when appropriate. The fact that you’ve been able to do as much as you have is incredibly impressive, and the way you learn from your failures is inspirational.
3:00 the other problem with using a 360 camera is that, ultimately, it is still a camera, and if you spin it like that, you'll get a ton of motion blur. You can stabilize the image in post and get a very stable video... of motion blur.
thats a very good point. I havent thought about that
Dude! It's been years watching your channel and this is still some of the best content on the internet. I hope your having fun and enjoying it still because I know we are!
man it is so great to be a BPS patron so you can see the videos EARLY
So true omg!!
@@BPSspace wouldn't it be easier to spin only a mirror to correct the rotation rather than the entire camera system?
Spinny cameras have been the bane of my existence lately too😉 Excited to “borrow” your final design once the issues are ironed out
Despinning the camera on your crazy prop-copter would be an interesting challenge. Slightly easier than this though because you don’t have to sandwich in the structure like this one does.
kim jong un will be proud of your work
I love how you used the lathe as a test rig for a spinning rocket. Genius!
When it works it works pretty well. I hope you iterate on it with FOC and a brushless gimbal motor actually designed for this kind of thing, what you already have is very promising!
Joe, you are way too hard on yourself - that footage (for a first or second-generation design/build) was absolutely AMAZING! You are an incredible engineer, machinist, and rocketeer!!
“The camera spinner didn’t work…..great”
Joe, I’m not sure you saw the same video I did.
For a *first try prototype* that was amazing.
Give yourself more credit dude, you are an inspiration to those of us whose engineering consists of making parts to fix something at home.
Seems like a success to me, most of the issues outlined you predicted or take a little programming effort to fix. It's wonderful to have such good footage that pretty much outlines every aspect of the problems.
honestly, that was a flawless first attempt. i know you engineer for expensive rockets you can't rapidly iterate, but that is far better than i expected from a sensorless brushless motor
I know they're uncool these days, but a little brushed gear motor would also solve this problem as you don't need the power density of brushless here, and they're very smooth
would go with DC instead of stepper, because they're usually lighter
What are you talking about, that looks incredibly. From a “where we were to where we are now” it’s an incredibly upgrade. That was bad ass.
Might want to research rollerons in your fins, like some air-to-air missiles. Those use passive rollerons spun up by the slipstream of the carrying aircraft.
I was working on a space shot where we thought about using hard disk drive platters and motors inside the fins for the gyroscopic fin stabilization.
Regardless, keep up the superb work!
My take on this comment is:
1) probably not what you are intending and
2) that you're saying to change the camera spinner so instead of being de-spun by a motor based on calculations, to instead attach fins with rollerons that de-spin using inertia... but I'm not even sure that would make sense.
Came down to say something like this,
fins are pretty excessive for high mach vehicles, *especially* low mass rockets.
Something like rollerons or tiny flap stabilizers makes more sense.
Rollerrons are pretty heavy and are probably reliant on the mass for a strong flex/vibration resistant housing.
But the idea is probably interesting to look into.
Especially passive self-corrective stabilization.
To those that might not know, rollerons are an enclosed gyroscope inside the trailing edge of a tailfin that is spun up by airflow over a small exposed portion. (Very common on military rockets/missiles, it's that toothy little wheel on the tailtin corners) They are connected to the static portion of the tailfin with a simple (but sometimes damped) hinge and function as a standard control surface but are passive and have only 2 moving parts.
If tuned correctly, the gyroscopic forces automatically tilt the control surface the perfect amount to counteract any roll and they can be surprising precise/accurate/responsive bc gyroscopes respond instantly and proportionally to even the smallest/slowest rates of change while not overcorrecting for the largest/fastest. It can essentially be thought of as a "wireless" equivalent to a mechanical bevel gear, neat.
Honestly the brief moments where the camera DOES counter-rotate correctly make it pretty clear that this is a fantastic approach if you can iterate it over the finish line.
Several others have mentioned my concern, that of the spinning camera assembly acting as a gyro and possibly screwing with the spin stabilization of the rocket itself. Would it be possible to mount the batteries closer to the center of the assembly? It looks like there is room. Since those are probably the greater part of the weight, moving them closer to the axis of rotation should help limit the effects, might even make it easier for the motor to spin the assembly up.Edit: Personally I think the results were pretty darn good for a prototype, we'll wait for the second flight for it to be perfect!
But the camera assembly should always have a rotational inertia of zero. There is no spinning up. Just opposing what the rocket wants to do.
@@tjm2212 Yes, though I think you mean the camera assembly wouldn't have any angular momentum. If it were somehow levitated inside the rocket in a vacuum then as the rocket starts to rotate the camera would maintain its orientation due to its nonzero rotational inertia, and the fact that the rocket body would not be exerting a torque on it. If the motors could perfectly oppose the movement of the rocket then the camera subsystem would essentially be decoupled from the angular momentum of the rocket and would only contribute its mass to the downward force due to gravity. But for now I think the implementation leaves a bit to be desired 😂
Others have said this but I want to add as well, the camera system is working as a reaction wheel and is changing the angular momentum of the rocket. When the camera is spun up, this is causing the rocket body to spin faster and also potentially some weird gyroscopic effects. Two things can be done that I know of:
1. Try to lower the moment of inertia of the camera system, or
2. Add a wheel with the same moment of inertia as the camera system with opposite angular velocity and acceleration.
This video was awesome, can’t wait to see what you’re doing next.
Dude, i have watched your videos on and off since you had 1000 subs.
And what an amazing improvement you've made in all areas. But especially your ability to be a natural when presenting in front of a camera. You've gone from nigh unwatchable to among the very best.
Greetings from Denmark
This is sick. Excited to see how this comes along in the future. I'm quite impressed you are doing so many disciples of engineering concurrently: mechanical, aero, electrical, controls, software, manufacturing!
Even adding to that is the filming, voice-over, and editing. People rarely mention that, but that isn't as easy as counting to four. Massive amount of skill all around. Definitely a good role model and someone to look up to!
Have you considered using a more passive stabilization using gyroscopes? Get them spun up before launch, and they should resist any roll. They could even double as steering control in the future.
I think it would work, but it's not as much fun 🙂
He did experiment with gyros but they saturate very quickly. They need to be either very fast or quite heavy to be effective. Idk what RPMs he got up to, so maybe a hard drive motor would be fast enough.
I've just started construction on my L3 rocket and have started mixing my own propellants after going to the balls launch this year. I hope to eventually be on a space shot team and I know that is ambitious, but compared to what you do I am just a child in a playground. I love watching you videos and I learn a lot. Keep it up!
I think you might need a sensored brushless motor and something like a VESC compatible ESC so you can get higher torque at low rpms. A good source of info on this might turn out to be the Onewheel Community, or people who diy their own eboards. Do some research on VESC and see if it's what you need! I do also have some hesitations about stepper motors. Stepper motors aren't traditionally great at high speed applications, but it would absolutely have the ability to despin those slower rotations near apogee. If you do go this route I recommend the Trinamic stepper drivers, specifically the TMC2130-TA (SPI interface so should be Ava compatible, up to 1.4A phase current, passive breaking, and load dependant speed control). Some sort of clutched assembly to allow for the use of both a brushless motor for high speeds and a stepper for low speeds could be an interesting trajectory, but would add several layers of complexity. I've also seen a lot of other people address the reaction wheel issue, and while I agree with them, I don't think the best answer is to add another reaction wheel to counter the spin. I think the best idea is to get the mass of your assembly down by off loading as much of the electronics for the despin rig to the avionics bay. Have you thought of using high rpm slip rings for all the circuitry, and leaving essentially only the cameras on the rig? I'd assume this would get the mass of the rig down a considerable amount, making the it's moment of inertia much lower, exerting less torque roll on the rocket. No matter what, I can't wait to see where you go from here Joe! This space shot is closer than you think!
VESC will not help, lower kV motors will do.
@@igor_misic Aw crap. I guess you're right. I obviously have no idea what I'm talking about.
But a sensored motor is probably a key thing it needs.
Don’t beat yourself up. I think that was a good starting place. It’s funny at 1340 I was thinking of stepper motors, and then that’s what you started talking about. My field was in marine, electronics and satellite communication. The earlier equipment is gyros and then they went to stepper motors. I am fascinated, watching you make your own rockets and rocket nozzles. I would like to see a walk-through of your shop and your tools subscriber here South Florida I often watch, the falcon nines takeoff and sometimes see the boosters return.
Have you thought about de-spinning a 360-Camera? It would combine the de-spinning with post processing stabilisation to remove the remaining jitter.
My first thought, but there might be too much motion blur
@@hund4440 My thought was, to de-spin the camera to an acceptable angular velocity, where the wouldn't be any motion blur. But I don't actually know how slow it would need to be.
You're right, maybe it would still be too fast, although the part that was semi-stabelized looked clear enough.
I really like what you've done here. Like the stepper motor idea too. It's much simpler and adaptable. Presuming of course that you're balancing everything, it's a pretty clean solution. I know there's also a lot of even smaller stepper motors that might also do the trick. You might find that a pancake stepper works nicely too by angling your drive gear 90 degrees. I made a similar modification to my high-speed 3D printer to decrease the weight of the printhead. My situation needed high torque though, not high speed. I only mentioned it, because some of those pancake steppers are really small.
I'm now trying to make a rocket too!
Thanks for the inspiration :) 💙
Me too
well now youre trying to get on a watchlist
That's a really cool idea, when you initially started the video, I thought about a similar solution, the only difference is that I thought you may end up turning the entire "turntable" into a brushless DC motor. I love this and you should definitely keep trying!
I love rockets so I decided to study Audio Engineering ✌
There is an easy way to accomplish this. Add counter rotating fins just for your camera. Your camera has a bearing, so it will rotate as fast as rocket is rotating but in opposite direction. Let the wind gust to do the rest.
I think the footage looks already good! Sure, there are some parts where it fails. But at the sweet point where it works, it works pretty good! Nice Job!
And it is great footage to present the problems. The whole analysis shows an admirable level of control. Might not be enough for the builder, but it certainly makes an exceptional RUclips video.
Damn I love watching this. Truly the Engineering way: do I need this? No. Do I want it? Yes. Also you should give yourself more credit. You did the hard part. You converted an idea into a working prototype. All you need to do now is get rid of the gremlins that make the thing not do exactly what you want.
Also, I don't know if it is what you meant about the small fins when explaining why it might not've worked but it seems the rocket isn't in constant spin. It seems to spin, then stabilize, then spin, then stabilize. Good luck and can't wait to see the next step in development :)
for the stabilization, you should stabilize based on the delta of the roll, since its not actually important that the camera holds a steady roll direction, but just that it dampens roll rate. this would solve the windup issue and it might also make it easier to get the motor to just not bother correcting if the roll rate is lower than its minimum starting speed, and just those changes alone i think would be a huge improvement for the viewability
additionally, although it probably doesnt matter at the current RPMs youre experiencing, you could always have it take into account the framerate of the camera and find the nearest multiple, so if your camera is 60hz and the rocket is rolling at 55rps, you could spin the camera to +5rps rather than -55rps and the framerate will hopefully turn it into a smooth video, although i guess that might end up producing some artifacts with shutter speed producing warped video... but its something to consider nonetheless
I was thinking about frame rate shenanigans as well, definitely something to keep in mind.
A a few others have mentioned you can take the ideas and tech used in high speed cameras following very fast moving objects and use moving mirrors. This solves a couple of problems also making it easier to mount bigger cameras(the gopro) inside the rocket and use a spinning periscope to capture beautiful footage of the outside. Great work it is really interesting watching the progression!
You could try a bldc gimbal motor with FOC. Awesome video, as always!
Those are mutually exclusive.
BLDC and FOC are two diffrent methods of commutation control for a motor. There are no BLDC vs FOC motors they are just brushless motors being controlled by a BLDC or an FOC esc.
Still I agree and FOC system would solve all the brushless motor issues.
BLDC means Brush Less Direct Current motor. They can be controlled by Field Oriented Control (FOC). SimpleFOC is an open source implementation.
FOC motor control is technically 3 phase AC control.
BLDC is as you said is direct current.
A brushless motor can be used as both a DC motor or an AC motor depending on the commutation method.
So saying a BLDC motor makes no sense as the motor does not require a DC control scheme it can use an AC scheme makeing it a BLAC motor.
Therefore its inaccurate to call it a BLDC motor instead its a BL motor using BLDC commutation.
@@nocare BLDC just the common terminology.
If I'm correct, the PWM signal is sinusoidally pulsed DC, so technically they could be still DC motors.
Edit: I'm not looking for an argument, was just trying to be helpful as a hobbyist. If I google BLAC gimbal motor there are no results. Googling BLDC gimbal motors gives the exact result of my suggestion. There is no more to the whole thing from my side.
@@zsigmondkara hmmm how about this.
What then is the name of the commutation method used by a standard brushless ESC.
SInce we both agree FOC is not the standard method most ESCs employ.
Half a million subs is nothing to sneeze at, and it is tremendous accomplishment, but how the howly-f does this channel only have half a million subs? Absolutely first rate in every way!
Love it! BPS Space videos always get the design vibes rolling!
Question:
1.) Why don't you write a code to calculate roll rate vs shutter timing and frame grab and "stitch" a roll stabilized video together? I'm not a coder so this may be more difficult than I realize with mixing video encoding or is camera sensor and shutter hardware the bottleneck?
2.) I was also wondering, have you considered integrating a coaxial system that incorporates a reaction wheel? If power and data connections through rotating electrical slip ring connector are stable, the mass of the batteries could be leveraged as reaction mass? The additional mass of the rotating camera system seems very high unless the quality of video is essentially "mission priority" over usable payload mass.
Edit :BPS Space contained a . and converted to hyperlink removed.
cameras can do 60fps usually, and you know the spin rate exactly from the sensor, so you probably can stitch anything together under 60 rotations per second. You would get some weird artifacts because of rolling shutter delay, there are CV cameras that can do more fps and instant shutter but that's a little more expensive.
But this is a cool idea, never seen anyone try it before
1) Wouldn't the frame rate of the stabilized video be limited to the rate of rotation? If the camera is rotating at 1 RPS you would only get 1 FPS.
Also if the rotation is fast enough each frame will be distorted by the CCD's scanning pattern.
@@polyscient I think you would if it was a single camera set up. However with dual opposing cameras (which i assume was for balance) the field of view could be optimized to meat at a central plane or known good overlap. Once again this is just assumption but then the rotation rate would adjust a "variable" in a function that chooses which video frame to grab based the camera settings for shutter speed. I think you will lose a significant amount of "off target data" so a higher FPS capture/sensor speed is desirable.
Is the sensor chip scanning pattern distortion accountable enough as a known calculable constant? If so wouldn't using a lower resolution sensor potentially yield greater scanning speed as a relative function to surface area? Or as simple as rotating the unit so the distortions are less obtrusive or don't align with horizon or other desired targets?
@@ffoska The rolling shutter delay effect is almost exactly what we are trying to accomplish. By determining where the target field of view is(from the flight computer or other hardware), the algorithm picks individual frames to stitch a new video from. At high rocket roll rates is where the algorithm I think would really take off in that it would begin "shedding" erroneous frames off the target field of view in order to "stitch" a new stabilized video.
I think this could be implemented downstream as a "fix it in post" or at a ground station with sufficient graphical compute power to re-encode/stitch on the fly. Implementing it on the rocket flight hardware seems unnecessary and detrimental to the rockets flight performance which we can see is nominal even with the addition of moving payloads (keepin' the pointy end up and hula-hooping).
@@TAR-D yes, by 'rolling shutter' I meant the CCD scanning pattern. You would just record the frames, but you also record the telemetry, so you can also use that later - but it's probably not even needed, because the sampling is fast enough to get multiple images of the same region and do keypoint matching. And I also meant using multiple cameras to get a 360 view, so it's just a question of unspinning the image.
Only just found your channel. You could take a hint from T-PODs they use an angled mirror with a stepper motor for small corrections, the head of the pod has the main motor, de-rotation on other axes is done in software when you reach the gimbal limit. I am thinking you have a transparent section for the mirror which does the de-rotation while the the camera assembly just rotates within the core.
Or you could also have a pyramid or cone for the mirror (if using software image correction) again inside a transparent ring section for the rocket core.
This could probably be stabilized passively using fins on the rotating camera rig. The fins on the righ could be mounted such that they can swivel freely in pitch axis; this will make them automatically align themselves with the flow and dampen camera rig rotation.
I bet this would be super hard to do, but it might be the coolest possible solution.
Honestly, that footage looks really good! I'd call the flight a success, everything you were worried about actually worked and you've proven all your hypotheses.
Rockets are cool!
Yes!
I was waiting for this!! It looks great at works awesome!
4 cameras, 60-120fps and unspin the video in post!
He's tryna avoid that tho
@@breadcat7332 yea I get it, engineering is fun, I let the scope creep happen too. But it’s worth remembering the primary goal sometimes.
We need more of this. Although you don't consider this a success looking at it from the outside it worked very well. I wasn't sure if this was possible but now you have me convinced.
I really enjoy your uploads, so enjoyable to see things fail and then watch you work out how to do it better. Super inspiring
I loved As an engineer it is my job to overcomplicate things, lool
Speaking of which, you should keep the whole control assembly, batteries speed controller chips motor, everything stationary above, then only spin a ring with the camera. Then slipring if you need any wires.
This way it can be shorter section of inner tube for strength, easier pid tuning, less reaction wheel weight, everything. And can make the tube 3 inch.
Have the drive gear on the inner tube be above the bearing so the motor can be mounted to the bulkhead. Or inside the tube and motor goes in for better strength
Yeah for the motor you should use a sensored brushless with and tune the esc pids for speed, similar to how electric longboards can accelerate smoothly from the known weight and accel they want, and rc cars want fast accel.
Or use a stepper motor with a at least an 8 bit driver like a printer, if you need real speed higher, or a servo motor, or a continuous rotation servo. Or a 775 brushed motor, or more power any .3w brushed motor,
Then you’ll be good
Good video and looks fun to learn
slipring only needed if the camera didnt have built in power or record signal. Or graphite brush and ring rial
I think it looked pretty good. Use the data and build a spinning rocket rig to simulate a launch.
This one time in school we made rockets and launched them with compressed air, and I did not have enough fins for proper stability so I folded the ends of the fins to around 30-45 degrees so it was stable due to gyroscopic stability (I think that's the name?), and its cool how it works so well. Great video!
I think your experiment was tremendously successful. Once the ESC started to work your mechanism and control worked beautifully. Compliments.
Man, some of the coolest videos on yt...
While I can understand your disappointment in the footage you got but it is still probably the best 'looking down' footage you have ever got. It is also way easier to watch! Don't be too down hearted and keep making those little changes.
What you need is an AC servo. There are cheap & quick DIY solutions like SimpleFOC. Stick an encoder onto your motor, replace the ESC with a SimpleFOC board, and that's all you need.
Brother, I got to tell you I live outside of Nashville and I cheer for you every time you go out and do a successful lunch. Even though I can't be there. I'm also an engineer and I'm glad that you went with this solution because this seems like the most fun one to implement IIBH. That's secretly why you're really doing it, isn't it?
I'm so excited for the ultimate spaceshot, Joe, but I'm equally thrilled with every step and iteration along the way! I love these components that are technical but explained in such a way that my non-engineer brain can understand them, and I wholly endorse this pursuit of non-spinning footage from a spinning rocket because that sounds AWESOME 😁I'm also immensely pleased that SEND IT! is serving as such a reliable testbed for all this tech development - keep up the great work Joe, and I'll be here to follow along on your journey!
Ive seen on yt some video of a rocket that had lower bottom rotating and upper "stage" was stationary whole the time. They talked in a video, that its pretty common fix for this problem. Also stepper in higher rotating speeds... gl man, you can do it :P
I think you're being pretty hard on yourself. I really think it did amazing for a first attempt. Significantly better than I even expected!
1. May put a reaction wheel for stability so you don't need a rotating rocket. The camera can be outside without this complicated stuff.
2. If you keep with the rotating camera idea, I think the motor is fine. What I suggest is change on the programming: DO NOT trying to compensate when the rotation of the rocket is so slow. It will look actually cool!
Just my suggestions, your work is really great!
Oh man, as someone with a computer vision background, what an amazing problem. Loved your doing this for the sake of it!
Stepper motors: You're going to need a stepper motor controller. At work, I've been using Pololu Tic TIC825, a Stepper Motor Controller/driver that offers USB, UART, I2C, RC pulses, step and direction, and analog voltage interfaces. The controller can be used to control position OR speed. The other Tic controllers seem to use the same microcontroller, and differ mostly in the range of motor currents and supply voltages. I've been greatly impressed by their performance. The T825 goes to 50k steps/second, and most stepper motors are 200 steps/revolution, so up to 50k/200 = 250 rev/second.
In your application, you're counter-spinning the camera to minimize the instantaneous roll-rate. Maintaining a fixed counter-spun *position* is not actually necessary (fixed positioning would be the extended feature target, but you'd probably need a better gyro.) I've assumed the camera ring has it's own roll rate gyro, but it likely has a little bit of roll-rate drift, even when when stationary. Capture rate drift while sitting on the pad, then after launch, offset measured roll rate by the captured drift rate.
Having the camera ring doing a slow roll during flight shouldn't be a problem.
This is very good work.
With gyroscopes you can directly try to stop the rocket from spinning. I'm looking forward to seeing this.
You wanted video stability.....I'd say you accomplished it for the most part. You can't always get perfection on your first attempt. Now comes the fun of the refining process.
- the big main fins overpowered the small fin intended to spin the rocket.
- a BLDC motor can be operated as a stepper, and microstepped for smaller spin intervals. This will require a reprogramming of the speed controller.
What an absolute masterpiece of rocket, defo my favourite channel of all RUclips is urs
The motor isn't necessarily the problem, it is the ESC. The hobby ESC's use back emf to synchronize the commutation of the phases. A better ESC / better ESC tuning could decrease the back emf voltage (which is proportional to rpm) that is required to synchronize the servo and motor. A way to avoid the need for the esc to rely on b-EMF to syncroise is to use a motor sensor i.e. hall-effect sensor / rotary encoder / magnetic encoder... this turns it into more of a servo drive for positon / velocity control.
The next step from there is an ESC that produces sine wave phase ouputs rather than bang bang outputs.... then you will have arrived a a field oriented control (FOC) servo!
Check out simplefoc or Tinymovr
Yes, I would even say rocket engineer. Your vids are the biggest motivation to me solving engineering problems!
I hope you were being cheeky. The job of an engineer is not to overcomplicate things, but to make them as simple as possible while meeting the requirements (spec/cost/schedule...). ie: No more complex than necessary. Using the lathe as a test driver was very cool - and was a simple approach!
In my opinion the brushless motor is the good aproach to solve this problem, the issues that you had with the “low resolution” of the motor it’s caused because you are using the wrong speed controler, not the optimal motor for the problem. Try looking for some Brushless FOC control and you will realize that is what you need to solve that. If you need help with it or some more starter friendly info just contact me! Keep doing what you do!
I think, like others have said, for a 1st flight test, albeit of a design you had already done multiple revisions to statically, this was the complete opposite of your conclusion....... factoring in all the things you KNEW before flight (concerns over the motor choice, low rotation speed performance, etc). This was a GREAT success as the bit of stabilised footage did have the exact result you wanted.... you wanted to test if you COULD stabalise footage in this manner and you proved you can, but with further refinements to get the outcome you want across the range of spin speeds.
Other have said you may need to do something to balance the angular momentum as what you intend is for the rocket to behave as if the spining section wasn't there, but the spinning section - as this scale - is acting as a reaction wheel affecting the rockets inherent spin. Adding a counter rotating element of the similar mass should go some way to balancing those angular momentums.
Tl;dr I genuinely think your conclusion is wrong considering the constraints on your ability to test systems on the ground (as you say, wind tunnel and lot and lots and lots of CFD would help, and a big company would have access to that stuff, but you as a small group cannot). You proved it is actually possible to stabalise the footage in this way, and know areas you can improve from the get go, and using the retrieved footage to guide further improvements.....
I thought it was a wonderful attempt Joe and you learn so much about ideas only when you try them out. Great work. 👍😊
Hey, stepper motors have a pretty slow maximum speed. See if it is sufficient for you. Make sure to not gear it down much, since they can do very precise microstepping and you need all the speed you can get.
Other than that, the solution that will work (but is overkill) is something like an ODrive.
Same rig - Get a bit chonkier, slower motor, and use an ODrive S1 for precise low speed control instead of a cheap ESC. Won't go to sleep on you and should work great.
Careful with the Stepper motor idea. Before you get to deep in to it, test it's torque.. I tried a stepper for a focuser on my telescope and the micro stepper motor (looks similar to yours) lacked the needed torque. and since your plate will be jerked quite heavily, just be sure it and work under that condition.
I would try putting a Brass ring around your brushless outrunner motor and mount it so that the rotation axis is perpendicular to your rocket's long axis.. spin up the motor pre-launch and use as a gyro with a free moving rotator plate.
Great progress Joe, Im sure you will get better results from the Stepper Motor. My only question/suggestion is to ensure the payload is rotationally balanced as it could be affecting the spin rate and influencing the IMU, looking forward to the next launch.. PR
So happy to see you back on screen in such a good mood 😍
What a creative mind!
The footage isnt so bad. It is really great to see it.
For someone who isn’t an engineer, the first try was also very impressive 😅 Great Job, Very interesting videos ❤
How about using a gyroscope which can move in x,y,z coordinates. The outermost layer will rotate along with the rocket. You can rotate the internal wheel without bothering about speed and timing, this will create a gyro motion and stabilize the cameras attached to the pivot point.
Just watched the Mark Rober collab! WOW! How cool was that! Glad to see you are finally getting some well deserved recognition for your insane skills!
Crazy good. Just smooth it out, put a pair of control surfaces on that spinamajig and voiala, You can steer that thing straight out of atmosphere. Thanks for the ride, have fun.
I have found that stepper motors benefit greatly with reduction drives. Giving greater power and accuracy without loosing steps due to reduced torque of direct drive. Experience from designing and building CNC equipment. Love your concept and testing.
Well done, nearly there!
As a specialist imager (photographer), I would have taken another route. I would have have run the camera at, may be, 120 or 240 frames per second and then edit every one in three or four or five frames to get a slowly rotating image. I could then stabilize the image or could accept the residual roll. Edit: Or set up a camera that takes 1 image every rotation.
I suppose this supports the idea that you take the route that you are most familiar with, neither is wrong. One can be more effective in the end. ( I think your idea would be more effective because the roll is not constant.) Good luck!
Awesome Joe ! Yes using BLDC motors won't give better angular precision stepper servo is comparative much better that ordinary stepper motors.
I hope you saw Vikram - S launch by Skyroot Aerospace as well !
What you need is not gears and an ESC, its a ring type USM motor like those used by Nikon and Canon for their professional (D)SLR lenses. They are direct drive (immediate), high torque, low inertia and completely hollow!
I'm a retired engineer and I'm impressed.
I've had a lot of projects that didn't work nearly that well on the first try.
And don't forget the engineers motto:
If it ain't broke, it needs more features. 😉
What do you mean it didn't work? It worked great. The parachute deployment footage is also breathtaking. Can't wait for the space shot.
Try to mount a step motor (like old capstan from old VCR) to move all the mecha. This kind of step motors are strong and react to slow speeds with very precision. They are controlled by PWM and uses a hall sensor to read back the pulse from the motor… to me is the best option.
This is more my area of expertise. I'd strongly advise against using a stepper. Stepper motors are synchronous motors. If they lose sync they stall and won't regain sync until you reduce the drive speed to below their 'pull in' speed. You only need to exceed the maximum torque for a few milliseconds to make the motor stall. The high vibration environment of a rocket is pretty much worst case scenario. Additionally the available torque drops of sharply with speed so you have to over rate the motor considerably to get enough overhead.
My first choice would be a humble DC brushed motor. Simple, cheap and reliable. Decent quality brushed motors turn smoothly at low speed and they can be driven with a simple H-bridge driver. Just keep the PWM frequency reasonably high, say 3kHz upwards.
For a little better performance and a slight reduction in weight, field oriented control of a brushless DC motor works but it's a lot more complexity and you will need to spend a fair amount of time tuning it for stability. Look into drivers and firmware designed for camera gimbals. I'd recommend using the motor with hall sensors at a minimum. While FOC can be used without sensors you run into a similar problem as you have with steppers. If you momentarily exceed the maximum torque of the motor it will lose position.
Couple of thoughts from a former missile engineer: rather than trying to stabilize to a specific angle, just compensate on rate. If you're determined to stabilize on an angle, account for rollover at +/- 360
Something like the Odrive 13.2 Module from m5stack may fit your needs here a bit better than a stock hobby ESC. BLDCs operating in sensorless mode typically rely on back emf sensing to figure out what coils to activate. So startup on them typically means kicking the motor over hard to get it to move so it can know where it is. Sensored controllers have a calibration to map sensor angle to the coils in the motor and can get reasonably low speed control. It would require a rotational sensor on the motor but those are fairly simple these days. I don't know if the mentioned module meets your weight constraints, but the concept is similar and there's a large number chips that can drive small BLDCs in a sensored way if you wanted to go down that route. May be easier on your power budget than the stepper.
I get that you are an engineer, but you should not be so hard on yourself. That test went well, and more importantly it gave you lots of info on what version 2 needs to do. That's a success in my book.
Genius idea spinning the camera. Cant wait for version 2 :)