Thank y’all for the feedback on the camera system! I saw a few suggestions for 360 video or stitching cameras in post - the reason for avoiding those options is that while the tech to stitch cameras together is getting pretty good, I *hate* how those lil stitch lines look, and un-fisheyeing a lens is usually an imperfect thing. If we’re rolling around at like 4hz the wobble from the fisheye artifacts is gonna look a little sketchy. As for 360 cams, I haven’t found any that can do quite the resolution and frame rate I’m looking for yet :(
Won't the camera spinning act as a reaction wheel that would slow the roll rate? Might need to add a counter-rotating wheel to neutralize the effect, if the effect is significant. I bet you'll model that in Simulink if you haven't already.
There are cameras from e-consystems and the like that are global shutter, reasonable resolution + frame rate and are able to be synchronised. The issue is that they need a host computer to record the images and that is all weight and complexity I'm assuming you don't want. If you do go down that route you can sync a scaled signal from the IMU output frequency to the camera shutters or some other common clock source. You can then use the resulting data for attitude estimates based on vision alone. What about an altitude estimate based on the curvature of the Earth? :) Love the videos, keep up the great work!
This isn't a model rocket channel anymore, these are highly engineered complex vehicles! Keep up the amazing work!! I believe you'll even make it to orbit one day!
I think that this work has been “highly engineered complex vehicles “ since the beginning of Joes channel. The evolution of the engineer progressed right along with the vehicles
Beside the spinning camera mount, have you considered getting a 360° shot around the rocket using 3 mounted cameras with fish eye lenses. You could merge the video images in editing, and having 3 cameras evenly mounted would also prevent deviation caused by 1 camera sticking out casuing imbalance. Just an idea
If the rocket spins really fast it would probably cause some image tearing or similar artifacts, but I think this would be a much better aproach than a complicated spinny mechanism, as it would involve no moving parts. This would make this aproach much more reliable, but I dont know if it would solve Joe's problem. I would at least give it a shot, maybe with some high FPS camera to minimize said artifacts.
@@brunomorenomata7978 framerate isn't really important for minimizing blur/jelly. the solution to capturing fast motion without issues is a camera with a global shutter running at a very high shutter speed. for framerate to start keeping up with shutter speed you need something like 10000-20000fps, which is unreasonable in this form factor.
I think AVA-lanche is a good name. In Germany we say "never forget where you come from". And I think AVA took a huge part in your project. Big props for your videos!
In your previous videos whenever Andrew shows up I get a feeling of relief, as the grounded, level headed friend is there so all is going to be fine... THIS time however, you guys took it to Kerbal level 10 and smooshed rockets together. Well done!
You are one of the best channel on youtube ever! This is amazing what you are doing are sharing with all of us that love rockets and space...good luck!
I find it Ssooo refreshing to find a young person willing to do something this complex on there own and not hibernating in a dark, cave like setting, playing computer games.
Hello, I've done a lot of rocketry with a college team and seen you at FAR a couple times. Just wanted to let you know that we used that radio module (RFD900) for a couple tests and we determined that it kinda sucked. Just cuz its one watt doesn't mean too much. 200mW to 1W is an equivalent gain to like 5 dbi, which is easy to get with just a better receiving antenna. We swapped to the rf4463f30 at 433mhz and have been getting really long ranges during our tests (like 10km or longer). That rfd900 is hot garbage compared to this module and id recommend using an SDR to recieve instead of another radio module. Just a piece of my experience working on radio for rockets. Willing to chat more about this if you wanna meet our team Space Enterprise at Berkeley. Should be launching a liquid very soon
Yeah this sounds a lot like antenna directionality being a big part of that. If the antenna is oriented vertically, and you are "right under the rocket" in terms of the antenna radiation pattern (remember that dipoles have a torus pattern where the antenna pierces the doughnut hole), it wont really matter how good your radios are. Its going straight up (hopefully), and if you are standing 100 meters away in a blockhouse, you are essentially directly under the transmitter, at it's weakest point. If both dipoles are oriented vertically, you'll get crap reception. I used to work in the UAV industry. We spent way too much time messing with antenna orientation trying to get good reception. It might be ideal in the rocket to orient two "paper" antenna 90 degrees around the outside body of the rocket, parallel to the ground. Then match that at the ground station. this will give you two torus patterns oriented vertically in a cross pattern. The greatest reception is above and below the rocket's flight path, but still gives you two solid lobes if it rolls over on it's side. Worst case is you will always be
@@epremeaux what would you think about using a helical antenna? More in general than in his particular case even at the higher frequency of 915 MHz helical is still bigger than his bulkhead but still a short one, around the outside of the stabilizing fins, might work. And being that is chosen transmitter is a diversity type, he can use a dipole as the secondary antenna in case things *go sideways*. I'm asking cuz I was thinking of doing this myself with a much smaller rocket and 2.4 GHz telemetry(a helical at 2.4ISM fits the bulkhead perfectly), and based on transmitter receiver orientation , as you pointed out, that's the best I came up with. Also thinking of spin stabilization, the antenna is going to be circularly polarized so that shouldn't be a problem.
@@petevenuti7355 yes generally, helical antennas have superior reception in all directions and orientations. Sport drone racers prefer them over dipoles for their FPV transmitters. These things are tumbling and spinning on every axis. I'm not entirely sure about the RANGE comparisons, but signal stability in any orientation and relationship usually outweighs range considerations. And as you say, in diversity mode, having a backup dipole or patch for range is the ideal combo.
All this makes me think the REAL solution for model rocketry ranges may in fact be to: 1: adopt a common telemetry protocol and radio set. 2: set up telemetry relay shacks connected to the internet at various distances around the range. 3: the ground station connects to the common telemetry server (shared by everyone) rather than directly to the rocket. Basically this now the common system model for amateur satellite tracking stations and it works really well. Theres always multiple shacks receiving data. Basically, much of the work in software, antenna tracking etc can just be copy/pasted. Worth the investment for fixed ranges.
Regarding the control of the spinning rocket. I recently came across an image of a Sidewinder AIM-9L missile which reminded me of how they stabilise it. They use spinning flywheels on the tips of main fins using the gyroscopic effect to control the stability. The use of the gyros would help at higher altitudes in low air density. The gyros use the passing airflow to start the spinning. They are small, simple and would certainly help.
Joe, let me encourage you, you are ding the very thing I had wished I could do when I was 15 or 16 in the mid-1970s. This is the kind of technology that makes even greater and affordable advances possible. Your space shot in the future will be a radical success. Keep looking up.
Didn't expect to see Keysight as a sponsor! 😮I work in the hardware team of a fairly big nasdaq listed company and we use a bunch of their oscilloscopes. They're awesome.
Not sure how feasible this is, but what if instead of spinning the camera and taking on all the complexity that involves, fly 4 (ish) cameras with overlapping fields of view and use software stabilization in post similar to how 360 cameras work.
@@hengehogs that’s a good point, my main concerns without actually doing it myself are challenges around alignment and how well the FOVs would overlap.
For the camera, you should look into a translucent plastic tube with a camera mount located inside it. You can perhaps put a servo on it to capture some 360 degree footage too, but mostly just to get the camera out of the way of the rocket's oncoming airflow. Also don't forget to compensate for its weight, so that the rocket doesnt tumble... if the plastic turns out to be too blurry to get good footage, you can always cut a hole the size of the camera lens.
This sounds like a great suggestion, I'm going to use this idea for an upcoming build I'm working on if you don't mind. it hadn't crossed my mind to use a clear tube or pipe for the fuselage, I've seen some extremely strong thin wall polycarbonate and lexan
Omg i know just enough to realise how much work and how talented and motivated you are. I made model planes and study computer science but you are so much on another league
just a thought for the camera stabilizer. Obviously for the higher rotations using a gyroscope to measure the spin will be the most effective, but for the lower spin rates, you might look into using cross correlation between two separate frames, and using that to determine which way the camera is moving (which fun fact is the same way that laser mice work). Being on the rocket, you'd only want vertical movement between the images, and any horizontal movement would indicate the camera was rotating.
Hey Joe, Nice video I love it and I love that you explained the use a Kalman Filter. If you want to tune your Kalman Filter to obtain good parameters without lunching 15 times Avalanche : You can record raw sensor data when you launch your rocket and use them after as offline simulated data when you tune your Kalman parameters. Concerning your spinning camera thing : It looks funny, but you ablsolutely need to have a stable mass rotation (without off centered mass that adds oscillations) Looking forward to see new videos
For the camera, maybe consider keeping the camera stationary pointing straight (lens on the center of the body) and spin an angled mirror overhead to look out (periscope method). This should be much easier to balance (and lighter weight) so that spinning an off balance camera does not induce more wobble in the rocket. You will also be able to use a much smaller motor and position it more precisely as you are slinging around much less mass.
For stencils to come out better, hit the painters tape with a clear coat or your base cost colour first before applying top coat. It will remove any bleeding.
For the camera, perhaps try mounting it inside the fuselage and have an outlet with mirrors that directs the image up or down the rocket (something like a submarine periscope)? It could make it more aerodynamic and more compact. Plus, less hot glue 😂
@@nardgames this would need a transparent ring in the rocket where the camera could see through. There are other solutions (cameras) that don't need a hardware fix for the rotation problem.
Another option for the spinning camera challenge would be to try an sync up the framerate with the spinn rate. Like the wagon wheel effect but sort of opposite
Very thankful for you sharing the mistakes, problems and issues and how you are growing in knowledge and experience. . So many video productions on the Tube, show very little to no problems or issues. . I like real life videos. . Awesome work your are doing.
Joe, your work is incredible! I enjoyed every minute watching this... At first I was confused by the GoPro hanging out the side and I asked myself, wouldn't it be easier to install a MIRROR and put the camera inside the tube. When I saw the (ridiculous) spinning camera mechanism I thought this is screaming for a spinning mirror assembly. I an no rocket engineer tho
i havent watched your videos in a couple months, and its so cool that you can already make rockets like this while i still remember you make small model rockets edit: actually more than a year
Looks like an inspiring set of goals! Iteresting I used to build little model rockets in the 70's and the tube dimensions, fins, etc haven't really changed. thanks for update, cheering for ya!
Pro photographer and videographer here. Here's some thoughts on capturing video to correct for the spin. I'll break it down into two sections. The first describes the hardware. The second explains the post-processing Hardware Create a parabolic dome that has a diameter slightly larger than the body of the rocket. Make the interior of that dome a mirror using paint (yes, it exists). Split the body of the rocket into a top and bottom half. Mount the GoPro inside the bottom half of the rocket so the lens is looking towards the top half. Mount the dome to the rocket so it acts as a connection between the two parts of the body. As in, 4 (or more if needed) points of contact occur with "wires/struts/supports" between the bottom section of the rocket body and the dome. Then, the entire top of the dome can be connected all the way around to the top section of the body. The thinner the struts that connect the domes bottom edge to the bottom part of the rocket (when viewed from along the Z axis "top to bottom") the less obstruction will occur for the camera. Software/Post Processing The image from the camera will look very strange without some post processing. You'll get a black circle in the middle of the image with the rest of the frame being very stretched. To correct this, first download the GoPro Max 360 video stitcher add-on for AE and install it. It'll be free on their website. Then bring the video into Adobe After Effects. Apply the new filter for 360 video. It will take some tweaking to get the image to look correct, but there's a lot of settings to play with. If that's not working, you can use code in AE to correct it manually. Now you have a flat looking image that looks straight down with a black circle in the middle, that spins very quickly. Take your telemetry data that captures the spin of the rocket and use this to "move" the video clip. Now, inverse (make negative) the values from that telemetry. Really, AE will just add counter spin, but because it's matched perfectly to the spin of the rocket, then the image will be stable. You will need to digitally zoom the clip a little to turn it into a clean 16*9 ratio. Finally, add in a "Smart Fill" to correct the black hole in the middle. The end result will be the rocket spinning from a point in the middle of the image (this will look like the walls of the bottom half of your rocket and will be pulled into a single point) and the ground/horizon stretching off on all sides. The horizon and ground will no longer be spinning. It's a long comment, sorry. Explaining these things is difficult with just written words. It's just one idea for solving your camera problem. I feel it would also solve the aerodynamics a bit, but that could also cause other issues with aerodynamics that I haven't considered yet. Good luck, and thanks for keeping us posted on the progress and journey.
I have NO experience with rockets more sophisticated than any that said Estes on the box, but about the camera aboard the spinning rocket: how precise is the timing? Could a bot choose frames that were shot when the camera was pointed the right direction? Some misalignment would be inevitable but a bit of post-production pan-and-scan like was used to show movies on 4:3 TV screens could bring things back. You'd lose a lot of framerate in playback but you'd wind up with a steady, if somewhat stop-motion-y, shot in one direction, and of course you could choose any direction or even generate a panoramic image. And since all of this would be done on the ground with the existing video data, no further hardware - or weight - would need to be launched.
Thinking out loud, if the camera chassis was on a freely rotating frame/wheel that was prevented from spinning by an onboard gyro or two then you could avoid all the gearing and maybe even the electric motor - just spin the thing up just prior to launch (magnetic bearings to keep friction losses to a minimum perhaps). You could even use a CO2 cartridge to shoot at some vanes (think turbine) to spin up the gyro(s) - during flight or just prior to launch. Just a thought.
Could you use a small mirror to reflect the view you want onto the GoPro lens inside the rocket body? It could likely be much smaller than sticking the camera halfway out the tube.
Mounting a parabolic to a stepper powered gimbal tracking the rssi can phenomenally enhance your reception. If you get the focal aperture of the dish lined up to the target the link quality and range can be increased by an order of magnitude +. They use the same principle in satelite linked aircraft. Plus it looks super cool in action.
I have also used kalman filter and did sensor data fusion for my final year project engineering . It was a pretty good experience working on robot . Data fusion was the lidar scanner data and RGB-D camera data
Nice idea with the spinning camera. But wouldn't it be way more failsafe mounting four cameras and stitching the footage in post? Impressing work anyways!
Great work! Looking forward to following this series! Those RFD radios are really great. I've had rock solid telemetry from 30km away on a drone with these at 500mW
For your Kalman Filter tuning, would it be possible to use previously logged data as inputs in your dev simulation (I’m assuming in Simulink?) You might already be doing this but it might be a bit cheaper and faster to iterate in a test SIL environment with real flight data you already have and then chuck it on the vehicle then just trying it each time you fly? Love your work!
I have to think this is already part of the process, tweak and fly each iteration is expensive. Balancing that though is you wanna make sure you aren’t optimizing the filter for only one flight profile/data set and it will be unstable on a different flight. One method could be to fly 10 times with a base filter and then optimize SIL on that set of flights, but that means you fly the lowest performance filter most. My guess is that there are tweaks each time, and all past flight data is ran through it to validate, then another flight is made to gather another data set but also get flights on the latest code.
@@skylar7320 for sure I think I’d also be surprised if he wasn’t doing this already. The only reason I ask is because in his video it at least wasn’t immediately obvious to me that he was. One other thing we do in my workplace is we have several sets of reference missions that we will then apply dispersions to to increase our dataset size somewhat similar to a Monte-Carlo analysis. Obviously this technique still has bias inherent in the real world data but it does allow for some more filter robustness. Also, just thinking loud, but I also think it might be worth having a low-fidelity HIL setup for this particular vehicle since it does seem to be a good proving vehicle for several avionics packages. Ok I’m done being an armchair engineer 🫡
Love your incremental engineering style! Also, easy tip for drawing perfect longitudinal lines on a cylinder: hold tube against a door frame (or other 90° angle piece).
Won't spinning the camera cause gyroscope procession "more roll"? Also, could you use ToF radio signal and some math to determine altitude? Keep up the amazing work, you're inspirational!
I think you're on the right track with the "spinny camera thing"- it's effectively just a single axis camera gimbal with infinite rotation on its only axis. You get the appropriate high torque/low KV gimbal style brushless motor and the corresponding gimbal BLDC motor controller in there and I think you'll be good to go. There are a few open source gimbal controller code bases out there you could reference as well (you have enough new stuff going on without reinventing a gimbal control algorithm), likely will just need a decent gyro to track rotation.
Joe, have you heard of _Insta360?_ They build 360° Cameras that take spherical videos in up to 4k resolution and their _Insta360° One RS_ is completely modular. You can switch the objectives, detach the camera parts and could set it on opposing sides of the rocket. It could give awesome videos. Edit: Since the camera is always filming in a 360° sperical view field, you don't need a hardware solution to stabilize the video. The Insta360 Software stabilizes the video for you.
As other commenter's have said, rotating mirror is way to go vs rotating the entire camera setup. Molding composite fins into an 3d printed airfoil shape mold should to allow you to reduce drag by optimizing your airfoil for your altitudes/velocities, and allow calculation of how much lift/rotational force you're applying as well via CFD. You could also tune the airfoil design for each stage, for instance the booster having a designed low roll state so it induces the least amount of drag in the thickest part of the atmosphere, with the sustainer stage having an airfoil that will induce spin more effectively at supersonic speeds vs at subsonic speeds.
Hey, here’s an idea… always build two rockets at once… that way when you aren’t sure about something, you can test it or replace a bad test. You also force yourself to double check everything since you will check the rocket against the other rocket to make sure they’re all the same… And if you want to change one variable in the alt rocket, you can effectively A/B test without having to start from scratch each time.
Instead of slowing down the camera spin platform, you could increase the rocket spin. Get the benefit of increased stability along with not having to re-engineer the camera platform. Just a thought. Incredible work!
Awesome goal. You could use IMU data from roll rate to adjust the speed of anti-spin for the spin-ring. A key thing to consider is ring-balance, as unbalanced spin could cause jello in the camera view. Not sure if spin vibration forces could effect rocket stabilization, but probably not something that would help.
What's really cool, is maybe this new data will allow for a better revision of AVA in the future. That'd be really awesome. So that Fiberglass looks exactly like G10 material, I work with that everyday. Please make sure you're using a proper respirator, you "DO NOT" want exposure long term to this. Also, to really shape, drill, cut as the material is very strong. Make sure you use any coated tool with AiTIAN Like a Titanium Nitride, or a straight Carbide tool. Something with high abrasion resistance. They're not cheap tools either. Best of luck, I'll keep watching, love the content. We also employ Keysight formerly known as Agilent. for our machine testing and other circuit board testing parts. Pretty good stuff.
I really love seeing your process to solve problems, that camera spinning idea totally caught me off guard but that's a really sweet way to solve it! Thanks for making such good engineering vids man!
If you record the raw outputs of your sensors, you can simulate your kalman filter output on the ground in software. Also, beware that a kalman filter assumes noisy data without a bias error, which is often not the case. EKFs can help, and you can run a pre-filter that estimates the bias, then subtracts it. Good luck!
For the camera, my first thought was something like the "Patrol" camera made by Observant Innovations. A single fixed camera looking up at a downward facing parabolic mirror which produces a toroidal image which can be software unwrapped into a square image in any direction around the camera. Probably something like that could be done with whatever camera you prefer so that you can maintain the framerate and resolution.
Very cool, always wanted to try a space shot. Almost gave it a whirl many years ago with a buddy who was working with nitrous/polymer hybrid rockets at the USAFA, but schedules never worked out. Oh, well. The camera thing looks like it could get complicated and heavy in a hurry, I agree with another commenter who suggested just throwing in a stout gyro and having some good bearings, let it try to stabilize itself more or less. Probably wouldn't work perfectly, but would be crazy easier.
Great video! For a better view in the sun, consider getting an external e ink screen. The refresh rate is not as fast, but should be perfect to watch data even in the brightest sun.
You should check out the Corredor crew 360 drone video. they have played around with a bunch of 360 and spin stabilized camera techniques and I think a lot of it is transferable to what you’re working on.
For the common filter, have you tried having two sets of weights for each state? The imu will be much more accurate than the gps during powered flight through coast phase, since it a greater refresh rate. So if your pre apogee operations has the imu weighted heavier which then switches to gps heavier operations during decent, you could optimize for accuracy.
Kalman filter stegosaurus tail looking like the on-camera and voiceover sound levels in this video amirite!?!?! JK dude, your stuff is always inspiringly awesome. Can't wait to see the rest of this project!
Thank y’all for the feedback on the camera system! I saw a few suggestions for 360 video or stitching cameras in post - the reason for avoiding those options is that while the tech to stitch cameras together is getting pretty good, I *hate* how those lil stitch lines look, and un-fisheyeing a lens is usually an imperfect thing. If we’re rolling around at like 4hz the wobble from the fisheye artifacts is gonna look a little sketchy. As for 360 cams, I haven’t found any that can do quite the resolution and frame rate I’m looking for yet :(
Won't the camera spinning act as a reaction wheel that would slow the roll rate? Might need to add a counter-rotating wheel to neutralize the effect, if the effect is significant. I bet you'll model that in Simulink if you haven't already.
Use a mirror, keep the entire camera over CG
+1 vote for mirror, this was the first thing that came to my mind that how I do it.
There are cameras from e-consystems and the like that are global shutter, reasonable resolution + frame rate and are able to be synchronised. The issue is that they need a host computer to record the images and that is all weight and complexity I'm assuming you don't want.
If you do go down that route you can sync a scaled signal from the IMU output frequency to the camera shutters or some other common clock source.
You can then use the resulting data for attitude estimates based on vision alone. What about an altitude estimate based on the curvature of the Earth? :)
Love the videos, keep up the great work!
gopro hero 11 has "horizonsteady" that might be helpful in this case.
This isn't a model rocket channel anymore, these are highly engineered complex vehicles! Keep up the amazing work!! I believe you'll even make it to orbit one day!
(every mars garage will have orbital rockets)
I think that this work has been “highly engineered complex vehicles “ since the beginning of Joes channel. The evolution of the engineer progressed right along with the vehicles
@@ITWUT Completely agree, he's always blown our minds
You caved so easily 😂
this guy wrote the same comment under every video
Beside the spinning camera mount, have you considered getting a 360° shot around the rocket using 3 mounted cameras with fish eye lenses. You could merge the video images in editing, and having 3 cameras evenly mounted would also prevent deviation caused by 1 camera sticking out casuing imbalance. Just an idea
If the rocket spins really fast it would probably cause some image tearing or similar artifacts, but I think this would be a much better aproach than a complicated spinny mechanism, as it would involve no moving parts. This would make this aproach much more reliable, but I dont know if it would solve Joe's problem. I would at least give it a shot, maybe with some high FPS camera to minimize said artifacts.
I wonder if you could use the recorded telemetry in post processing to map out the roll?
Good idea that will help taking footage and stable the rocket a lot
@@brunomorenomata7978 framerate isn't really important for minimizing blur/jelly. the solution to capturing fast motion without issues is a camera with a global shutter running at a very high shutter speed. for framerate to start keeping up with shutter speed you need something like 10000-20000fps, which is unreasonable in this form factor.
@@brunomorenomata7978 and the added benafit of seeing evryting
I think AVA-lanche is a good name.
In Germany we say "never forget where you come from".
And I think AVA took a huge part in your project.
Big props for your videos!
Thanks for having me onboard for a 2-stage! And rocketpoxying the adapter when the tube was out of round 😂
Do you know what kind of Kalman filter you guys are using?
give the people what they want
Meat Rocket!
@@haph2087😳
In your previous videos whenever Andrew shows up I get a feeling of relief, as the grounded, level headed friend is there so all is going to be fine... THIS time however, you guys took it to Kerbal level 10 and smooshed rockets together. Well done!
may the kraken be with them.
all hail the holy mission of rocket-smooshing
:o)
That 2-stage flight was dope!
us: make active control
joe: spiny camera boiiiii
You are one of the best channel on youtube ever! This is amazing what you are doing are sharing with all of us that love rockets and space...good luck!
Keep up the great work joe!! You are a big inspiration!
I find it Ssooo refreshing to find a young person willing to do something this complex on there own and not hibernating in a dark, cave like setting, playing computer games.
This is elon’s clone
Hello, I've done a lot of rocketry with a college team and seen you at FAR a couple times. Just wanted to let you know that we used that radio module (RFD900) for a couple tests and we determined that it kinda sucked. Just cuz its one watt doesn't mean too much. 200mW to 1W is an equivalent gain to like 5 dbi, which is easy to get with just a better receiving antenna. We swapped to the rf4463f30 at 433mhz and have been getting really long ranges during our tests (like 10km or longer). That rfd900 is hot garbage compared to this module and id recommend using an SDR to recieve instead of another radio module. Just a piece of my experience working on radio for rockets. Willing to chat more about this if you wanna meet our team Space Enterprise at Berkeley. Should be launching a liquid very soon
Yeah this sounds a lot like antenna directionality being a big part of that. If the antenna is oriented vertically, and you are "right under the rocket" in terms of the antenna radiation pattern (remember that dipoles have a torus pattern where the antenna pierces the doughnut hole), it wont really matter how good your radios are. Its going straight up (hopefully), and if you are standing 100 meters away in a blockhouse, you are essentially directly under the transmitter, at it's weakest point. If both dipoles are oriented vertically, you'll get crap reception. I used to work in the UAV industry. We spent way too much time messing with antenna orientation trying to get good reception.
It might be ideal in the rocket to orient two "paper" antenna 90 degrees around the outside body of the rocket, parallel to the ground. Then match that at the ground station. this will give you two torus patterns oriented vertically in a cross pattern. The greatest reception is above and below the rocket's flight path, but still gives you two solid lobes if it rolls over on it's side. Worst case is you will always be
@@epremeaux what would you think about using a helical antenna? More in general than in his particular case even at the higher frequency of 915 MHz helical is still bigger than his bulkhead but still a short one, around the outside of the stabilizing fins, might work. And being that is chosen transmitter is a diversity type, he can use a dipole as the secondary antenna in case things *go sideways*.
I'm asking cuz I was thinking of doing this myself with a much smaller rocket and 2.4 GHz telemetry(a helical at 2.4ISM fits the bulkhead perfectly), and based on transmitter receiver orientation , as you pointed out, that's the best I came up with. Also thinking of spin stabilization, the antenna is going to be circularly polarized so that shouldn't be a problem.
@@petevenuti7355 yes generally, helical antennas have superior reception in all directions and orientations. Sport drone racers prefer them over dipoles for their FPV transmitters. These things are tumbling and spinning on every axis. I'm not entirely sure about the RANGE comparisons, but signal stability in any orientation and relationship usually outweighs range considerations. And as you say, in diversity mode, having a backup dipole or patch for range is the ideal combo.
I suppose a patch is not ideal on a rocket if it's spin stabilized ;)
All this makes me think the REAL solution for model rocketry ranges may in fact be to:
1: adopt a common telemetry protocol and radio set.
2: set up telemetry relay shacks connected to the internet at various distances around the range.
3: the ground station connects to the common telemetry server (shared by everyone) rather than directly to the rocket.
Basically this now the common system model for amateur satellite tracking stations and it works really well. Theres always multiple shacks receiving data.
Basically, much of the work in software, antenna tracking etc can just be copy/pasted.
Worth the investment for fixed ranges.
Regarding the control of the spinning rocket. I recently came across an image of a Sidewinder AIM-9L missile which reminded me of how they stabilise it. They use spinning flywheels on the tips of main fins using the gyroscopic effect to control the stability. The use of the gyros would help at higher altitudes in low air density. The gyros use the passing airflow to start the spinning. They are small, simple and would certainly help.
Joe, let me encourage you, you are ding the very thing I had wished I could do when I was 15 or 16 in the mid-1970s. This is the kind of technology that makes even greater and affordable advances possible.
Your space shot in the future will be a radical success. Keep looking up.
avalaunche ? exciting development and I'm amazed at the progress you've made over the years, much respect 👍
Didn't expect to see Keysight as a sponsor! 😮I work in the hardware team of a fairly big nasdaq listed company and we use a bunch of their oscilloscopes. They're awesome.
Not sure how feasible this is, but what if instead of spinning the camera and taking on all the complexity that involves, fly 4 (ish) cameras with overlapping fields of view and use software stabilization in post similar to how 360 cameras work.
With this strategy you would also get even drag around the vehicle. Likely helping with some of the wiggliness on ascent!
@@hengehogs that’s a good point, my main concerns without actually doing it myself are challenges around alignment and how well the FOVs would overlap.
I had the same idea, but you'd need really high shutter speeds or the footage would be blurry af
@@OperationDarkside yes. Shutter speed but not frame rate. They are independent (unless the equipment forces their linking)
Isn't there already a product put there that does exactly this?
If I recall correctly it's a ball that you can throw in the air to take pictures.
Been waiting for this BPS space shot video 💯
For the camera, you should look into a translucent plastic tube with a camera mount located inside it. You can perhaps put a servo on it to capture some 360 degree footage too, but mostly just to get the camera out of the way of the rocket's oncoming airflow.
Also don't forget to compensate for its weight, so that the rocket doesnt tumble... if the plastic turns out to be too blurry to get good footage, you can always cut a hole the size of the camera lens.
This sounds like a great suggestion, I'm going to use this idea for an upcoming build I'm working on if you don't mind. it hadn't crossed my mind to use a clear tube or pipe for the fuselage, I've seen some extremely strong thin wall polycarbonate and lexan
Was thinking the same thing. Lexan would be a good choice. If the thing is internally mounted, you could probably have it on a gimbal too.
Omg i know just enough to realise how much work and how talented and motivated you are. I made model planes and study computer science but you are so much on another league
Watched till end don't need chapters for this very satisfied keep going man this is really awesome Rocket 🚀 project! I like that double shot 👍
just a thought for the camera stabilizer. Obviously for the higher rotations using a gyroscope to measure the spin will be the most effective, but for the lower spin rates, you might look into using cross correlation between two separate frames, and using that to determine which way the camera is moving (which fun fact is the same way that laser mice work).
Being on the rocket, you'd only want vertical movement between the images, and any horizontal movement would indicate the camera was rotating.
Hey Joe,
Nice video I love it and I love that you explained the use a Kalman Filter.
If you want to tune your Kalman Filter to obtain good parameters without lunching 15 times Avalanche :
You can record raw sensor data when you launch your rocket and use them after as offline simulated data when you tune your Kalman parameters.
Concerning your spinning camera thing :
It looks funny, but you ablsolutely need to have a stable mass rotation (without off centered mass that adds oscillations)
Looking forward to see new videos
For the camera, maybe consider keeping the camera stationary pointing straight (lens on the center of the body) and spin an angled mirror overhead to look out (periscope method). This should be much easier to balance (and lighter weight) so that spinning an off balance camera does not induce more wobble in the rocket. You will also be able to use a much smaller motor and position it more precisely as you are slinging around much less mass.
For stencils to come out better, hit the painters tape with a clear coat or your base cost colour first before applying top coat. It will remove any bleeding.
For the camera, perhaps try mounting it inside the fuselage and have an outlet with mirrors that directs the image up or down the rocket (something like a submarine periscope)? It could make it more aerodynamic and more compact. Plus, less hot glue 😂
If the mirror were mounted on the spinning platform, it would solve the rotating video issue too.
@@nardgames this would need a transparent ring in the rocket where the camera could see through. There are other solutions (cameras) that don't need a hardware fix for the rotation problem.
A prism would hold up better to the forces involved in launching a rocket.
@@ericwolf5874 true, but a block of glass is also quite heavy.
@@Craftlngo agreed... everything is a tradeoff. A plastic prism might be lighter.
Awesome video Joe! Great job and keep flying, looking forward to your space shot development for sure.
Little by little and a little becomes a lot .
Thanks you Joe and I'm exciting to see your next projets
just respect for every thing you do
Thanks for everything your doing , mit , berkely etc
and all the equipment you use that they also use
Another option for the spinning camera challenge would be to try an sync up the framerate with the spinn rate. Like the wagon wheel effect but sort of opposite
Very thankful for you sharing the mistakes, problems and issues and how you are growing in knowledge and experience. . So many video productions on the Tube, show very little to no problems or issues. . I like real life videos. . Awesome work your are doing.
Joe, your work is incredible! I enjoyed every minute watching this... At first I was confused by the GoPro hanging out the side and I asked myself, wouldn't it be easier to install a MIRROR and put the camera inside the tube. When I saw the (ridiculous) spinning camera mechanism I thought this is screaming for a spinning mirror assembly. I an no rocket engineer tho
Super happy to see you back in form! Love your videos.
Since it's a space shot, may your skies be black and your winds be nonexistent
i havent watched your videos in a couple months, and its so cool that you can already make rockets like this while i still remember you make small model rockets
edit: actually more than a year
Wow. Amaizing job joe! Keep it up 😄
Looks like an inspiring set of goals! Iteresting I used to build little model rockets in the 70's and the tube dimensions, fins, etc haven't really changed. thanks for update, cheering for ya!
Pro photographer and videographer here. Here's some thoughts on capturing video to correct for the spin. I'll break it down into two sections. The first describes the hardware. The second explains the post-processing
Hardware
Create a parabolic dome that has a diameter slightly larger than the body of the rocket. Make the interior of that dome a mirror using paint (yes, it exists). Split the body of the rocket into a top and bottom half. Mount the GoPro inside the bottom half of the rocket so the lens is looking towards the top half. Mount the dome to the rocket so it acts as a connection between the two parts of the body. As in, 4 (or more if needed) points of contact occur with "wires/struts/supports" between the bottom section of the rocket body and the dome. Then, the entire top of the dome can be connected all the way around to the top section of the body. The thinner the struts that connect the domes bottom edge to the bottom part of the rocket (when viewed from along the Z axis "top to bottom") the less obstruction will occur for the camera.
Software/Post Processing
The image from the camera will look very strange without some post processing. You'll get a black circle in the middle of the image with the rest of the frame being very stretched. To correct this, first download the GoPro Max 360 video stitcher add-on for AE and install it. It'll be free on their website. Then bring the video into Adobe After Effects. Apply the new filter for 360 video. It will take some tweaking to get the image to look correct, but there's a lot of settings to play with. If that's not working, you can use code in AE to correct it manually.
Now you have a flat looking image that looks straight down with a black circle in the middle, that spins very quickly. Take your telemetry data that captures the spin of the rocket and use this to "move" the video clip. Now, inverse (make negative) the values from that telemetry. Really, AE will just add counter spin, but because it's matched perfectly to the spin of the rocket, then the image will be stable. You will need to digitally zoom the clip a little to turn it into a clean 16*9 ratio.
Finally, add in a "Smart Fill" to correct the black hole in the middle. The end result will be the rocket spinning from a point in the middle of the image (this will look like the walls of the bottom half of your rocket and will be pulled into a single point) and the ground/horizon stretching off on all sides. The horizon and ground will no longer be spinning.
It's a long comment, sorry. Explaining these things is difficult with just written words. It's just one idea for solving your camera problem. I feel it would also solve the aerodynamics a bit, but that could also cause other issues with aerodynamics that I haven't considered yet. Good luck, and thanks for keeping us posted on the progress and journey.
Kudos - don’t do in hardware what you can do in software 🙂.
Dude is god tier at naming things
Ava is the computer
And AVAlanche is the rocket to test it, literally the perfect name combo
I have NO experience with rockets more sophisticated than any that said Estes on the box, but about the camera aboard the spinning rocket: how precise is the timing? Could a bot choose frames that were shot when the camera was pointed the right direction? Some misalignment would be inevitable but a bit of post-production pan-and-scan like was used to show movies on 4:3 TV screens could bring things back. You'd lose a lot of framerate in playback but you'd wind up with a steady, if somewhat stop-motion-y, shot in one direction, and of course you could choose any direction or even generate a panoramic image.
And since all of this would be done on the ground with the existing video data, no further hardware - or weight - would need to be launched.
Even though I'm a pilot, I'm right there with on the Estes rockets. Glad Joe is taking it further than I did.
Thinking out loud, if the camera chassis was on a freely rotating frame/wheel that was prevented from spinning by an onboard gyro or two then you could avoid all the gearing and maybe even the electric motor - just spin the thing up just prior to launch (magnetic bearings to keep friction losses to a minimum perhaps). You could even use a CO2 cartridge to shoot at some vanes (think turbine) to spin up the gyro(s) - during flight or just prior to launch. Just a thought.
That's what I was initially thinking.
Could you use a small mirror to reflect the view you want onto the GoPro lens inside the rocket body? It could likely be much smaller than sticking the camera halfway out the tube.
You are awesome. Thank you for all of your amazing videos. These really help keep that engineer itch going.
Isn't the camera spinner basically a reaction wheel that takes away from the spin stabilization?
Mounting a parabolic to a stepper powered gimbal tracking the rssi can phenomenally enhance your reception. If you get the focal aperture of the dish lined up to the target the link quality and range can be increased by an order of magnitude +. They use the same principle in satelite linked aircraft. Plus it looks super cool in action.
I think it would be cool to see Rollerons for spin stabalization.
I have also used kalman filter and did sensor data fusion for my final year project engineering . It was a pretty good experience working on robot . Data fusion was the lidar scanner data and RGB-D camera data
Nice idea with the spinning camera. But wouldn't it be way more failsafe mounting four cameras and stitching the footage in post? Impressing work anyways!
Great idea
Great work! Looking forward to following this series! Those RFD radios are really great. I've had rock solid telemetry from 30km away on a drone with these at 500mW
For your Kalman Filter tuning, would it be possible to use previously logged data as inputs in your dev simulation (I’m assuming in Simulink?) You might already be doing this but it might be a bit cheaper and faster to iterate in a test SIL environment with real flight data you already have and then chuck it on the vehicle then just trying it each time you fly? Love your work!
I have to think this is already part of the process, tweak and fly each iteration is expensive. Balancing that though is you wanna make sure you aren’t optimizing the filter for only one flight profile/data set and it will be unstable on a different flight. One method could be to fly 10 times with a base filter and then optimize SIL on that set of flights, but that means you fly the lowest performance filter most. My guess is that there are tweaks each time, and all past flight data is ran through it to validate, then another flight is made to gather another data set but also get flights on the latest code.
@@skylar7320 for sure I think I’d also be surprised if he wasn’t doing this already. The only reason I ask is because in his video it at least wasn’t immediately obvious to me that he was. One other thing we do in my workplace is we have several sets of reference missions that we will then apply dispersions to to increase our dataset size somewhat similar to a Monte-Carlo analysis. Obviously this technique still has bias inherent in the real world data but it does allow for some more filter robustness. Also, just thinking loud, but I also think it might be worth having a low-fidelity HIL setup for this particular vehicle since it does seem to be a good proving vehicle for several avionics packages. Ok I’m done being an armchair engineer 🫡
Well done this reminds of the early stages of rockets appreciate the hardworking u put in it 😃
Hello Joe. What about using a 360° camera or 2 180° cameras to capture the video and stabilizing it in post production ?
Love your incremental engineering style! Also, easy tip for drawing perfect longitudinal lines on a cylinder: hold tube against a door frame (or other 90° angle piece).
Won't spinning the camera cause gyroscope procession "more roll"? Also, could you use ToF radio signal and some math to determine altitude?
Keep up the amazing work, you're inspirational!
The camera will only be spinning relative to the rocket's frame of reference. So… yes, but not relative to Earth.
really exciting to think of people doing this as individuals
but what happened to the ice cup
I think you're on the right track with the "spinny camera thing"- it's effectively just a single axis camera gimbal with infinite rotation on its only axis. You get the appropriate high torque/low KV gimbal style brushless motor and the corresponding gimbal BLDC motor controller in there and I think you'll be good to go. There are a few open source gimbal controller code bases out there you could reference as well (you have enough new stuff going on without reinventing a gimbal control algorithm), likely will just need a decent gyro to track rotation.
Joe, have you heard of _Insta360?_ They build 360° Cameras that take spherical videos in up to 4k resolution and their _Insta360° One RS_ is completely modular. You can switch the objectives, detach the camera parts and could set it on opposing sides of the rocket. It could give awesome videos.
Edit: Since the camera is always filming in a 360° sperical view field, you don't need a hardware solution to stabilize the video. The Insta360 Software stabilizes the video for you.
this is my favorite rocket so far from BPS for sure
boy, you look like Elon ❤ 😄
Great idea for a new mission. Good luck these next few years!👍🏼
He is really look like Elon Musk
As other commenter's have said, rotating mirror is way to go vs rotating the entire camera setup.
Molding composite fins into an 3d printed airfoil shape mold should to allow you to reduce drag by optimizing your airfoil for your altitudes/velocities, and allow calculation of how much lift/rotational force you're applying as well via CFD.
You could also tune the airfoil design for each stage, for instance the booster having a designed low roll state so it induces the least amount of drag in the thickest part of the atmosphere, with the sustainer stage having an airfoil that will induce spin more effectively at supersonic speeds vs at subsonic speeds.
i think it would be cleverer to call it Avalaunch
🏆
Im so f***ing hyped for this project. Your videos are insane keep going
Bro u look like elon musk
Hey, here’s an idea… always build two rockets at once… that way when you aren’t sure about something, you can test it or replace a bad test. You also force yourself to double check everything since you will check the rocket against the other rocket to make sure they’re all the same…
And if you want to change one variable in the alt rocket, you can effectively A/B test without having to start from scratch each time.
I'm probably the biggest fan of your channel, but could you pleeease use metric! No one knows what 20ft is
Instead of slowing down the camera spin platform, you could increase the rocket spin. Get the benefit of increased stability along with not having to re-engineer the camera platform. Just a thought. Incredible work!
You look like Elon musk
Awesome goal. You could use IMU data from roll rate to adjust the speed of anti-spin for the spin-ring.
A key thing to consider is ring-balance, as unbalanced spin could cause jello in the camera view. Not sure if spin vibration forces could effect rocket stabilization, but probably not something that would help.
As I'm bio students but my mind still like technology related to space , i love to watch your video...
What's really cool, is maybe this new data will allow for a better revision of AVA in the future. That'd be really awesome.
So that Fiberglass looks exactly like G10 material, I work with that everyday. Please make sure you're using a proper respirator, you "DO NOT" want exposure long term to this. Also, to really shape, drill, cut as the material is very strong. Make sure you use any coated tool with AiTIAN Like a Titanium Nitride, or a straight Carbide tool. Something with high abrasion resistance. They're not cheap tools either. Best of luck, I'll keep watching, love the content.
We also employ Keysight formerly known as Agilent. for our machine testing and other circuit board testing parts. Pretty good stuff.
I really love seeing your process to solve problems, that camera spinning idea totally caught me off guard but that's a really sweet way to solve it! Thanks for making such good engineering vids man!
If you record the raw outputs of your sensors, you can simulate your kalman filter output on the ground in software.
Also, beware that a kalman filter assumes noisy data without a bias error, which is often not the case. EKFs can help, and you can run a pre-filter that estimates the bias, then subtracts it.
Good luck!
Yes, I was going to suggest the same.
For the camera, my first thought was something like the "Patrol" camera made by Observant Innovations. A single fixed camera looking up at a downward facing parabolic mirror which produces a toroidal image which can be software unwrapped into a square image in any direction around the camera.
Probably something like that could be done with whatever camera you prefer so that you can maintain the framerate and resolution.
One day I’ll get my own rocket to fly!
These videos will fill my need til that day comes :D
Cool video! Looking forward to seeing more development with the space shot
I'm stoke for this new test project. The rocket has GOT to be AVAlanch!
Very cool, always wanted to try a space shot. Almost gave it a whirl many years ago with a buddy who was working with nitrous/polymer hybrid rockets at the USAFA, but schedules never worked out. Oh, well. The camera thing looks like it could get complicated and heavy in a hurry, I agree with another commenter who suggested just throwing in a stout gyro and having some good bearings, let it try to stabilize itself more or less. Probably wouldn't work perfectly, but would be crazy easier.
I love that massive amount of Amazon packages in the background
I only build pyrotechnic rockets but I love your channel. Go get them!
12:20 "When you're doing Mach 4, or Mach 5, it gets a little bit harder" **Biggest understatement of the century right there**
Glad to see Keysight, aka "Pre-Carly" Hewlett Packard, on board here. Also, try putting the GoPro in the middle and put a mirror hanging outside.
Nice video! Excited for your starship model rocket!
Feeling so excited for the new project 😁
Excellent video and documentary of your journey 👏🏼
This guy looks like is like Elon Musk and Veritasium combined; and just as smart. Keep up the good work!
Stumbled across this channel. Thank you YT algo. Subbed!
Two stage adds so much excitement haha was cool for sure
Great video! For a better view in the sun, consider getting an external e ink screen. The refresh rate is not as fast, but should be perfect to watch data even in the brightest sun.
Браво! Це все дуже круто! Удачі тобі в нових досягненнях)
You should check out the Corredor crew 360 drone video.
they have played around with a bunch of 360 and spin stabilized camera techniques and I think a lot of it is transferable to what you’re working on.
Great video, Joe, as always. Keep it up!
It relaxes me after a long day
For the common filter, have you tried having two sets of weights for each state? The imu will be much more accurate than the gps during powered flight through coast phase, since it a greater refresh rate. So if your pre apogee operations has the imu weighted heavier which then switches to gps heavier operations during decent, you could optimize for accuracy.
Kalman filter stegosaurus tail looking like the on-camera and voiceover sound levels in this video amirite!?!?!
JK dude, your stuff is always inspiringly awesome. Can't wait to see the rest of this project!
One of the best parts of this video is when you said you were having fun. Throughout you can genuinely get a sense of your excitement. Keep at it Joe!
Smashing stuff, Joe. Keep up the good work.