3 Axis Stabilized Rig

edited May 2011 in Control Systems
I

Comments

  • Mike, I hope you get good weather! I'd love to see test flights with this.

    One test in particular I'd like to see, if it's not too much to ask: Is it possible to turn the system on and off during a flight? I'd love to see a video made with your camera both with the system on and with it off. By matching features frame-to-frame you should be able to get an idea of the magnitude of oscillation as well as a frequency. Heck, if you can break it down for X and Y and capture the frame-to-frame offsets over several seconds of video, you should be able to get a power spectrum and see where your residuals are. Sorry, I don't have software that'll do this, but I'm working on it. Video Deshaker will do this in a grid pattern across the frame, but I haven't figured out how to grab its offsets and throw it into a data file. Even without the analysis, just having those two videos, one stabilized and one not, would be great.

    Sorry for being pokey like this. As you saw in the other thread on rig stabilization, I'm trying to figure out how to get as close to a Cineflex as I can, using a kite rather than a full-sized heli. The gyro stabilized rigs you and Scott Armitage have made are leaps and bounds in that direction. I'd like to see how far.

    Ok, another test: I know a lot of the thrust for stabilized rigs is for doing kite aerial video, but I'm a still photography kinda guy. My whole reason for wanting to build a stabilized rig is so I can put a camera up on a nice moonlit night and do KAP of city skylines (ok, in my case small town skylines) and get nice clear images coming off my camera. If the wind and weather cooperates with you, I'd love to see a KAP set done just after sunset as the sky is getting dark. I already have a good feel for what it's like without stabilization (bad). It'd be encouraging to see how slow a shutter speed you can use and still get a good percentage of sharp images coming off.

    This is cool cool stuff!

    Tom
  • Wonderful summary Mike. You have been busy! Proof will be in the experimental results. I agree with Tom. It would be good to see a KAV shot with and without the stabilization. The S95 can be set up with CHDK / SDM scripts to cycle through stills and video in flight....not certain about turning off an on active stabilization while in flight....

    Love the work! Could be a neat KAP product and bring pendulums back into vogue.

    WW
  • edited May 2011
    Beyond Awesome as usual, Mike!

    Now...adapt the pendulum to the collapsible spars idea (another recent thread) and lengthen the pendulum by 2x or even 3x to increase the period/slow it down maybe...plus keep it easy to transport like your current one; just keep the pendumlum center axis rigid (use square tubing and ferrules?, and invent a reliable joint).

    I can hardly wait for this to be available by plans, kit, or whatever ...way to go!
  • edited May 2011
    Tom, Jim,
    I have tried a couple videos with and without stabilization prior to adding the pan-hold and acceleration compensation. The stabilized videos kept the subjects well framed as compared with the unstabilized videos. Horizons generally stayed stable, but there was still some high frequency jitter present. I was using a gyro only with high frequency washout, so the horizions were stable but tilted due to the wind bias. On the un-stabilized video, the horizon whipped wildly and was much harder to watch. I will do some more tests with the pan-hold and accelerometer compensation and post the results later.

    I work with pendulums primarily because I like their simplicity and the technical challenge they pose for getting clear, well framed shots. They also enable the use of AeroKAP which would be difficult using a picovet. In general a short, rigid pendulum like I use (~1.5 ft) will be subject to higher frequency oscillations and line vibrations as compared to a picovet which oscillates at lower frequencies and damps out line vibrations more effectively. The update rate of my servos is high enough to support well framed still shots with a pendulum, but will not provide super smooth videos like a Cineflex. I'm working on increasing my servo update rate to see how that works. A higher speed servo like the GS-1 may work better for videos, or, ultimately the rotating mass gyro's that are being discussed in your "Thoughts on Rig Stability" discussion thread.

    I do intend to try my rig with longer shutter speeds in low light situations. I'll keep you posted.....
  • edited May 2011
    Phil,
    I like the colapsible spar approach for the pendulum, very cool! This would work well to launch the rig with the pendulum retracted and then allow it to lengthen once it's up in the air.... dang, just when I thought I was done working on this... :). Scott Armitage raised a somewhat similar idea with Single Line Suspensions.
  • Mike, something I utterly forgot about in that rig stability thread is that you're timing your shutter to the apex of your swings. That's something else I'd really love to see a side-by-side of, too: A set without the apex sync, and a set with. I'm betting that lowers your usable shutter speed by a LOT. It doesn't help with video, but for stills it buys a lot of ground.

    Tom
  • edited May 2011
    Tom,
    (... copied from a recent response on your "Thoughts on Rig Stability" thread...) Great suggestion, I will make a S/W build that triggers a set of pictures at the apex of the swing and one at the trough and set the shutter speed to something like 1/100 and 1/15 sec to see how they turn out. Other setting suggestions are welcome. I'll also take some videos with and without the stabilization (traveling up the line, sitting on the line, and returning). Should be fun. I can't guarantee that the subjects will be very interesting (flying at the test field can get pretty boring once you've seen a few thousand pictures of the same stuff) but it should provide some good data to evaluate.

    I programmed my controller to trigger the shutter when roll rate is less than 2 deg/sec. If you listen closely in the "Roll-Shutter" video above you can hear the shutter activating at or very near the apex of the swing. I have not measured the delay from USB voltage removal to shutter trip on my S95, but I assume it is on the order of 10 msec. I still need to check my system to make sure three are no other delays on the syncronization... I will get to that as time permits.

    Assuming that the shutter is being activated when the rig is not moving much (not including high frequency line vibrations), do you think that the camera's built-in IS system will help achieve better pictures?
  • I do.

    We've got image stabilizers on two of our cameras at work. They work similarly to how IS lenses work, by moving a lens group or a flat plate. In our case it's flat plates. (The big one is more than half a meter across. Big 'n slow.) Something we're constantly worrying about is total throw of the system. Because of the nature of what we do, we can't let the system run out of throw and peg at one extreme of travel. So we're constantly off-loading the integrated error of the IS system to our pointing system.

    This is something I don't ever see discussed with digital camera IS systems, but it's just as valid: If you're moving so fast you run out of travel in the IS system while the shutter is open, it's like the IS isn't there at all. In some cases it can be worse. If a rig is bouncing all over the place, I honestly don't think IS does much good. But if you can take most of the motion out of the camera before the shutter is tripped, it gives the IS system a much better chance of operating as designed.

    The one thing no IS can take care of is roll. Pitch and yaw it can handle. Roll is another matter. This opens a whole 'nuther can of worms when it comes to rig stabilization: which axes to favor with your stabilization system. For an active system, there's probably some benefit to making roll the fastest axis with the least jitter.

    Tom
  • edited May 2011
    I''ve posted a set of contact sheets on Flickr from a flight today that shows general results with and without active stabilization enabled. The rig was programmed to take a 24 photo, 360 deg panoroma with a fixed tilt. Contact sheets seemed like a decent way to compare what's going on as far as consecutive picture compositions. All shots were taken with an S95 at a shutter speed of approx. 1/1000 sec (camera optical stabilization enabled in continuous mode) and are sharp with and without stabilization. The stabilized photos are oriented much more consistently as indicated by the generally level horizon as compared to the unstabilized photos.

    These are best viewed at full size on Flickr.

    I've had limited time to work on this the past couple of weeks, but have confirmed the obvious, that short un-damped pendulums are a challenge because they tend to be more prone to high frequency movements and vibrations as compared to longer pendulums or Picovets. I'm finding that my system works well for keeping the camera situated correctly for still shots and panaromas, but may not be ideal for very low shutter speeds or video due to to the high frequency movement and vibration content unless you use higher bandwidth servos, additional aerodynamic / mechanical dampening or de-shaker post-processing. More results to follow as the season progresses.

    Stabilizer off, Photos 1-12: Stabilizer Off 1001

    Stabilizer off, Photos 13-24: Stabilizer Off 1002

    Stabilizer on, Photos 1-12: Stabilizer On 1001

    Stabilizer on, Photos 13-24: Stabilizer On 1002

    I was flying with a Jones 7ft Rok and the wind was light at 8 - 10 MPH, so the rig wasn't moving around too much. I plan to do additional tests with and without stabilization in windier conditions to further evaluate the stabilization effectiveness.

    I also have a video from a few days ago that I will post to YouTube (it's rather large at 1GB so we'll see if it uploads ok). It provideds a pretty good comparison of with and without stabilization.
  • Mike, Great job!

    I'm still working with the "Batman" format....(chuckle).
  • edited May 2011
    I
  • Mike - When do you anticipate putting together the recipe and sources for this fantastic rig?

    Many of us have replicated your shuttles and had greater success with photos than we otherwise would have done (I really like my LeDuc reel shuttle and its super-stability traveling up while the Aurico does its thing, and have all the parts to constuct the propeller-powered rig).

    Is there a chance you could specify what is needed? Your experiments and photos/KAV are fantastic. Does it bother you much that you are a genius? ;-]
  • AerialLensGuy, thanks for bringing this thread back up, because I utterly missed all the updates from Mike!

    The differences are very VERY striking between stabilized and not stabilized! Tons of food for thought here.I honestly don't know which I'm more excited by, the three-axis stabilization, or that you have the provision to trip your shutter at the apex of a swing when the motion is the least.

    Mike, I'm glad you differentiated between the low-frequency, high amplitude oscillation and the high-frequency, low amplitude vibration. I really think these are separate issues and can be treated individually. Some stabilization systems appear to take energy out of one and put it into the other, but I also think there are some that can dump energy from one without inducing much more in the other. It appears your active stabilization falls into that latter category. It's taking out the low-frequency oscillation, but not attacking the high frequency stuff. For active stabilization I think this would be seriously hard to do since it would require very fast sensors, very fast servos, and a very fast servo loop. If you look at the frequencies you're trying to take out, you'd need a servo loop running at at least four times that speed in order to deal with them. That's asking a lot from RC servos.

    But I think there are comparable techniques that can attack the high frequency stuff. I'm curious if you've looked at the design of George Lawrence's Captive Airship camera he used to photograph San Francisco after the 1906 earthquake. It's very different from the modern Picavet or pendulum suspensions, but it's closer to the pendulum than the Picavet, so it might be right up your alley. One of the key features was a weight that was suspended under the camera by three cables, running up to three arms attached to the camera. This made the entire thing one extremely long, extremely low-frequency pendulum. If the framework is sufficiently rigid, almost all the energy in the high frequency vibrations goes away, leaving only low-frequency oscillations. Mr. Lawrence didn't have a solution for low frequency oscillations, but you do. Put the two together and you may wind up with the proverbial rock in the sky.

    Seriously SERIOUSLY cool stuff.

    Tom
  • edited July 2011
    Phil,
    Thanks for the kind words.

    I've posted details of the pan servo modification and mechanical construction of the rig on my Flickr site. I'm still refining the electronics and software portions of the rig and am considering ways to make this available to others in a practical way down the road.

    Here's a basic list of the rig components (excluding mechanical building materials) to get an idea of the costs involved:

    Rig Servos: HS-65MG
    . Roll: Standard servo
    . Pan: Modified for continuous rotation with proportional control
    . Tilt: Modified to add dynamic DC breaking

    Rig Controller: PIC18F14K22 Micocontroller (wired on handmade circuit board)

    Programming Language: C18 C Compiler

    Intertial Sensors: IMU-3000EVB (Evaluation Board)
    . 3 axis gyro
    . 3 axis accelerometer (use 2 axis)
    . I2C interface to PIC

    Power Source: LIPO 7.4 VDC 400MAH, regulated down to 5 VDC and 3.0 VDC

    I can work on a more detailed description of how it all hooks together as time permits.
    Mike
  • edited July 2011
    Tom,
    I agree, this is pretty fun stuff! Concerning addressing the high frequency components of movement, I'm not very familiar with the George Lawrence design, but I did find this site which discusses it in some detail and will look into it further. Christian Becot has some information on pendulum suspensions ("Thwarting Pendulum") that may also be worth investigating.

    I've noticed that quite a bit of the unwanted higher frequency rig oscillations are in the pitch axis. My pendulum connection to the shuttle body is via a single screw and has relatively little friction or dampening to keep it from swinging freely in the pitch axis (parallel with the kite line). I've noticed a tendency for it to oscillate (kind of a "pumping" type of resonance for lack of better words) in the pitch axis with small changes in the line tension. It's interesting that on lower wind days, motion in the pitch axis and is often worse than motion in the roll axis. The active stabilization does a good job of compensating for this, but it does ocassionally get a bit too fast for the servos to handle. A simple hydraulic damper used in model race cars may help quite a bit in the pitch axis to slow things down and doesn't add too much weight (~8gr). Troy Raymond (Envision) presented a nice design using a hydraulic damper in this thread. (lots of other interesting things discussed in this thread as well).

    Another intriguing way to reduce unwanted high frequency movement to obtain very low shutter speeds for single shots is to put the rig into a free fall before taking the picture. This is somewhat related to a slack line condition which is highly favorable to low shutter speeds. As seen in the AeroKAP videos the scene gets significantly more clear and stable as the shuttle is released and free falls down the line (see rig decent at 5:38 Video 1 (Moderate Winds)). It would be interesting to add some additional tilt servo compensation while the rig is falling to hold a fixed horizion. The free fall can be induced by releasing the shuttle line brake prior to taking the picture. The picture can be triggered when the fall reaches "steady state" while holding a steady aim on the horizion. Once the picture has been taken, the brake can be re-engaged and the rig flown back up to position with the propeller for the next shot. This would require a standard AeroKAP shuttle. An entire session could be accomplished this way by playing the rig up and down ~50 feet of line. ... may be worth trying.

    Mike
  • Hi Mike

    I know this thread is a little old, but I really need a little assistance on the pan-axis compensation on my rig. Currently I have the problem of my gyro attempting to "compensate my manual commands" and even trying to "compensate its own compensation", resulting in the rig rocking back and forth in the pan axis. Also, i am purely using the gyro readings for my pan axis (accelerometer readings has no aid in pan axis), thus I am facing the problem of gyro drift as well.

    You mentioned that you made use of the absolute position of the gyro to compensate for the gyro drift, and also the controller is programmed in such a way that the gyro knows when it needs to compensate and when it does not. Could you share how do you actually achieve the above mentioned (maybe in terms of algorithm)?

    Thank you so much for your help!!
  • edited March 2012
    Hi Jason,
    Pan-hold is a bit tricky to implement. I'm using an approach that works reasonably well, and will post a description of it in the next few days.

    Mike
  • edited March 2012
    Jason, I'm familiar with the oscillation issues you described above and have experienced them myself while trying to get this to work. I'm assuming you have a working knowledge of closed loop control systems (or you wouldn't have gotten this far with it :). It had been a while since I have used this stuff before getting back into KAP but I did manage to wake up enough brain cells to piece it together. This entry got a little more involved than I had hoped, but here goes.....

    Following are some general notes for the Pan-Hold function I have implemented on my rig.

    Source of Unwanted Pan Motion:
    In an ideal situation, no additional pan motion will occur after the camera is moved to a desired pan position. However, in the real world, unwanted additional pan motion occurs when the rig oscillates side to side around the axis of the inclined kite line. The pan motion that we are trying to compensate for with Pan-Hold is the result of the rig rotating about the kite line axis when the line is inclined at an angle greater than 0 deg. This unwanted pan motion generally increases as the kite line angle increases. The goal of Pan-Hold is to eliminate unwanted pan motion once the camera has been moved to its desired pan position.

    Pan Servo:
    The Pan-Hold function requires a pan servo that has positional control and a very wide angular rotation range. Positional control requires some type of position feedback be mounted to the servo output shaft (i.e. a potentiometer) so that it can be driven to a commanded absolute position. It would be best to be able to control the servo from 0
  • edited August 2012
    This is a little off-topic, but here is a really cool, albeit PASSIVE way to add stabilization to a picavet.

    Oftentimes redundancy such as pictured in the link can remove that pesky remaining instability in a camera. I like this concept.
  • edited August 2012
    Getting HEAV - YYYY !! For all the rig damping you do, you're still just hanging from a lively line.
  • edited August 2012
    Point well taken, Paul. But these little cylinders look like a pretty lightweight addition. On the other hand, heavy is sometimes a good thing (e.g., inertia). A bad Analogy: Shooting images from a mushmobile (Lincoln Continental) tends to come out better than from a BMX (lightweight bicycle) when going down the same road. 8^)

    Nevertheless, you are right -- it is a battle to anticipate the changes and absorb the energy of the string.
  • And in strong wind, you probably can suspend a BMX from a kite line, but not sure about a Lincoln! ;)
Sign In or Register to comment.

Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

In this Discussion