Well the gimbal controller finally showed up, it spent a week in New York, presumably making its way through customs, it was shipped from Latvia. It seems to be working pretty good, The one thing that I would definitely recommend are gimbal motors with hollow axel shafts, this would make wiring the gimbals and IMU much easier and clean. I’m now ready to mount it on the hex copter and see what happens, stay tuned…
I haven’t been doing much with my drones or my photography lately, but with more time on my hands that will soon be changing. Here is another incarnation of a DIY three axis gimbal using Turnigy 5208 BLDC motors. The rest of the parts were fabricated from $5 worth of 1.5″ x 1/8″ aluminum stock. A Basecam controller is on it’s way. Total cost of about $300, it should easily perform at least as well as one costing more than a grand, and probably more durable than a carbon fiber gimbal as well, depending, of course, on the quality of the Chinese motor bearings.
This is actually just a prototype for an ultimate build intending to cary my Nikon D800. Unfortunately my DJI “Tuned Propulsion” motor/prop combo’s don’t produce their advertised rated thrust, therefore, they can’t cary my D800. Im designing an new larger hex with higher quality motors, ESC’s, and props and it is also in the works. Stay tuned…
Since I’ve had some time in-between jobs I dusted off my franked-drone and mounted up a GoPro Hero Black and played around with it some. I didn’t spend the time to re adjust the feedback parameters for the lighter weight and it shows in the videos. The first is kind of a cool one, since I accidentally captured a deer running away.
In the second video I tried to show off and, wouldn’t you know, I flew too close to the Sun. The remarkable thing is, with this hard crash at about twenty-five MPH, my structural wood franken-drone came out unscathed with only a slightly bent landing gear. Oops, I forgot to take into account that the battery was low and therefore the copter did not gain any altitude at full throttle with full forward stick position, live and learn.
If this would have been any carbon fiber composite or aluminum metal frame it would most certainly now be worthless junk. As it stands, it bounced off the ground, I made a full recovery, and brought it safely back home in one piece. Any landing that you walk away from is a good landing!
The only damage was this bent landing gear. OK, for you engineers; How much force is required to bend a 3/8 threaded rod? This was no light crash 🙂
Since the DJI motors did not produce their rated thrust I will be building another drone, probably with T-Motors that will be sure to make rated thrust along with, most likely, seventeen inch props that will certainly be able to easily lift my D800 with plenty of reserve power.
360 degree panorama of the End of the Ash River Trail: (click on it for a larger image)
If you have been following me at all concerning my adventures into multi-rotor platforms for photography you may already know that I was pursuing a rock steady stable platform from which to take aerial photography shots from. Videography is definitely a back burner issue with me, my main interest is with still photography.
My initial disappointment was with the DJI Phantom 2 Vision in that the camera was rather poor quality, at least when measured against professional photographic equipment. The dynamic range was low, the lens was too wide an angle, the fisheye aspects were unacceptable, and worse the images from this camera were difficult at best to correct for image distortion.
I almost immediately sought out alternatives, the Go-Pro was also unacceptable to me since it also has a pronounced fisheye look to its images. Initially I adapted my Nikon J1 camera to the Phantom but found that it was simply too heavy to be safely flown with the Phantom. After exploring other avenues I found that I would have to spend over five grand to get a decent platform that could safely lift my Nikon J1, let alone my intended camera, my Nikon D800.
I was searching for a cost friendly alternative so I first built a quad using DJI’s E800 motor/ESC combination but found that these motors don’t come close to their advertised thrust in their “Tuned Propulsion System” it should, in all honesty, be called the Detuned Propulsion System. So I set out to build my hexacopter.
I was having troubles tuning the necessary vibration isolators to prevent the inevitable resolution robbing vibrations away from the camera. I was even looking to scrap my drone program entirely in favor of a balloon based system which has very little or no vibrations at all.
Then it dawned on me after a test flight. I have used tie cord as a safety backup in case the vibration isolators would become separated from the camera platform and after a previous hard landing all four indeed became detached. The next flight I forgot to reattach the isolators so the platform was hanging by the tie cord loops. The four tie cord loops were not all equal in length so the camera was flown at a ten degree, or so, tilt angle. After landing and giving my self a Homer Simpson DUH! I decided to download the flashcard and see what I got.
To my surprise the sharpest images that I have taken to this day from any aerial platform! That was the key! For the first time I was able to stitch a decent ten image panorama together! The problem with fisheye lenses is that they don’t stitch into decent quality images very well. Even though the camera was tilted and the copter wavered around due to both the normal GPS wondering and well as the ten to twenty MPH gusty breezes my stitching program, PtGui Pro, was able to easily stitch the ten images together. The other main problem with this particular panorama is that I did not take the images with enough overlap. While there are some areas of the image that certainly need improvement it is the IMAGE RESOLUTION that beats all previous images taken from any ariel platform, at least by myself!
Next to stabilize it even more I will be looking into a downward facing stabilization camera to supplement the GPS for position hold as well as an ultrasonic range finder to supplement the altitude hold function. Using vibration isolators between the camera platform and then suspending the gimbal from the airframe with tie cord is a perfect answer for total vibration isolation for still imagery, while this is obviously not a very good solution for videography – well I really don’t care that much about moving pictures anyway 😉
Update: I may have spoken too soon when suggesting that tie cord may not work for video, here is a video taken with the exact same setup as was used for my stitched pano:
You can clearly see the difficulties in taking a series of still images to stitch together for a pano, the copter is flailing around in the wind. Most of the jitteriness seen in the above video would be minimized when a two axis stabilizing gimbal is utilized. Note that the video from the Nikon J1 is not very good to begin with, compared to the D800, slow panning on a stationary tripod results in marked loss of resolution, not seen at all in this video. I am becoming convinced that a third axis gimbal motor may be necessary due to the unreliably unstable rotation control of the hexacopter is around the z-axis, especially when windy. Yup, retractable landing gear and a third axis gimbal motor may ultimately be necessary.
Well here it is, my first incarnation prototype for a HexCopter. I have to say that a hex handles so much nicer than a quad, much more stable and predictable. This version weighs about 11 lb. (5 kg) or about 830 g/motor or just 30 g above DJI’s recommended 800 g/motor. The total maximum flight time, with an 10,000 mAh battery, was just under 20 min. with the final battery voltage of 21.35 Volts. The DJI ESC’s started giving yellow LED signals with a rapid loss of altitude afterwords. Using a Return To Launch (RTL) voltage of 22.2V should give ample reserve RTL time for a useful flight time of about fifteen minutes. Update: My battery charger reported that it took 10,325 mAh to recharge the battery.
There are several areas where weight can be shaved to allow for the additional weight for my Nikon D800. I am still fine tuning the vibration isolators for my Nikon J1 camera, there is no gimbal motors used at this time. If you look closely at the video the Nikon J1 is having severe problems with the autofocus, I’m probably going to have to give up and simply set the focus manually to infinity, as well as resort to using manual or aperture exposure modes, and forget about 360 degree pano shots, especially near sunrise or sunsets.
I’m not yet sure if I need to reduce the number of isolators, I’m currently using six DJI phantom isolators. I bought isolators for DJI larger gimbals but I found that these were way too stiff for the Nikon J1, although those are probably the ones I will need to use for my D800.
For all of those paranoid morons who are chomping at the bit to shoot down a drone I added an example of just how close a typical drone used by an amateur photographer would have to be to get any meaningful images. All of you dipsticks should realize that only a multi-million dollar military grade drone will be able to count the pimples on your nude sunbathing girlfriend/wives butt-cheeks, and from an altitude higher than you will ever be able to see or hear it from, let alone, shoot it down from 😉
Here is a quick video with my first prototype two axis gimbal with my Nikon D800 with a 10mm DX fisheye lens attached. Note the heavily overloaded vibration dampers were actually being supported by zip ties, obviously not very effective. But you can see exactly when one of the motors hit maximum thrust, the ESC LED turned yellow, the drone started shaking and within a few seconds it no longer could maintain altitude. By about 30 seconds three of the four motors were maxed out and it became uncontrollable and flipped on crash landing at about 38 seconds, Nothing broke 🙂
The aircraft weighed exactly 12.0 lbs. or 5443 g. This means about 1360 g of thrust was needed per motor in order for it to hover. The DJI E800 could not supply this much thrust for more than twelve seconds, the motors were very, very warm after this 40 second flight. There was little or no control possible once the aircraft started loosing altitude. Post flight battery voltage was 24.91 V. So much for their 2100 g of thrust rating. Proof positive DJI does not live up to its own specifications.
By the way this copter worked spectacular with my lighter Nikon J1 weighing in at at only eight pounds or about 900 g thrust required per motor.
Well I finally attached my Nikon J1 onto my quad, actually I first attached it a few weeks back but the jello effect prevented decent video. Actually the jello effect was not that bad but it still was there. I used the vibration isolation platform from my Phantom Vision 2 so it is a bit overloaded. I must shop around for some heavier duty isolators.
You will notice that when pointed into the wind everything is fine, however, when pointing away from the wind the copter had to point nose high in order to offset the wind. This caused the isolator platform to rest on the battery causing a slight jello effect which then caused the camera difficulties auto focusing. Also note the tilt of the horizon when pointing perpendicular to the wind.
The battery tested well, with out any wind, the battery flight time for simply hovering was 30 minutes from full charge to 22.2 Volts, where I have programmed the PixHawk to Return To Launch (RTL) mode, providing ample reserves. But in todays wind the flight time was limited to about eighteen minutes, Remember while stationary hovering in one position it was actually fighting the wind and actually flying at an airspeed of 15-20 MPH+.
I am now ready to design and build a two axis gimbal for it as well. Here is a short video, note that I was testing this beast in a 15-20 MPH wind that was gusting to well over 30 MPH. While the video may cause air sickness I think the PixHawk handled it quite well, at least after it decided on a point on which to hold its position. I had to provide some inputs initially to prevent it from blowing into a tree.
Here is a short video showing my quad just after using the APM Autotune mode to automatically set the PID feedback values for the autopilot. I must say it did a wonderful job, albeit a it resulted in a fairly aggressive tuning. This is in contrast to the initial tuning with Qground control, the PX4 branch, which does not yet have an auto tune feature. Instead it uses an iterative process starting with holding the actual copter in your hand while PID tuning. Mind you, I wore my leather jacket and chaps along with my full face helmet while spinning the copter up to hovering power while holding it in my hand. I do hope that you realize that that these copters are little more than Cuisinarts with carbon fiber blades. 😮
Towards the end of the video you will notice that the low battery warning occurred and the autopilot entered into the RTL Return To Land mode. I tried to bring the beast down by manually overriding the RTL with the throttle but it would not respond to my commands. That is until I switched back to Stabilized (manual) mode and awe shite! The throttle was near minimum and the damn thing dropped like a rock. I had the default settings with the left most switch controlling the flight modes which means that my thumb was off of the throttle – HUGE F-ING MISTAKE! As you can see just how quickly things can turn to shite, especially when so close to the ground.
Here are the results, one broken prop, one broken landing gear dowel, and one broken quick release prop adapter, which actually saved the prop it was attached to. I’m looking into using fiberglass rods, with pads at the ground contact area, positioned at about a 60 degree cant for shock absorption rather than the 90 degree positioning in this prototype set up.
This is sooooooo freaking much fun I can’t possibly tell you. I love engineering and product testing. I don’t mind spending resources on learning experiences but I do take exception with being taken for a rube by nefarious marketing and sales practices. On the other hand DJI does actually have some decent concepts that it is putting into practice, mixed emotions.
It seems as if DJI is in the process of updating the E800 3510/350KV motor and is designating the new motor as 3511/350KV. I have to wonder if this was due to my experience with the faulty ESC and subsequently burned up motor, as well as my posts documenting the DJI failure. If so I commend DJI, if this is just a marketing ploy to redirect my posts critiquing the failure of the 3510 series then, well then DJI are simply douchebags attempting to distance themselves from their F-UP. Time will tell, I canceled my order for additional 3510’s for my hex-copter build and will wait until the new motors become available.
Stand by for an extensive review and testing of their “upgraded” 3511 motor….
While I am still waiting for DJI parts to continue flying, (have I mention DJI customer service sucks?) I don’t think that I have mentioned that their parts supply chain is also extremely poor. I thought that I would do some testing on the 620 ESC and 3510/350kV motor combination. I used the FrSky X8S receiver and their Taranus Plus transmitter to supply the servo Pulse Width signal directly to the ESC.
The motor responds to servo signals from 1145 micro-seconds to 2025 micro-seconds, which translates into motor RPM’s from 658 to 8670. At a Voltage of 24.87 this translates into 349.05 RPM/Volts which is approximately the rated value of 350kV, this was tested with the motor lightly loaded using a stubby prop that I fabricated from one of my broken DJI props that was destroyed by one of my brand new ESC’s that was defective from DJI ;-(
I measured the RPM’s using a photodiode beneath the prop and a light source above it and measured the signal with my oscilloscope.
Here is the signal at the max speed of 8760 RPM’s, note that the dips occur when the propeller shades the photodiode from the light source above it. Every other dip represents one full revolution:However, when I redid the test with the full size DJI 13.5 inch prop the ESC motor combination only produced an RPM of 7500 and failed to reach its maximum RPM by over 1100 RPM’s. This translates into an RPM/Volt rating of only 307 kV. I used a fully charged battery with a capacity of 10,000 mAh. I took the measurements within one minute of operation.
By this time the battery voltage had dropped from its initial charge of 25.17 Volts to 24.43 Volts. This means that in the real world the motor specifications as advertised by DJI are worthless. DJI rates the RPM/Volt at 25 Volts rather than the
half 15% capacity of 22.2 Volts, the point when warnings are sounded and the pilot should be thinking of landing very soon or risk an uncontrolled crash landing. Now I am not sure of the standards in the RC community, if there are any, I can assure you that these are very deceptive and misleading marketing tactics.
I also measured the thrust by placing the motors on another rig with the motor weighted down on a scale and placed high enough that it was above any ground effects. I was only able to achieve just less than 1600 grams of thrust, well shy of the rated 2100g of thrust as advertised by DJI. This is most likely due to the fact that the motor fails to reach it’s maximum RPM by over 1100 RPM’s and it’s RPM/Volt rating when actually loaded down with the DJI propeller. I used a fully recharged battery and did the thrust test within ten seconds of operation.
This is also very shy of the 200% recommended maximum thrust rating for its rated take off weight. I was deceived that its maximum thrust was well over the 200% rule of thumb. Actually there are several people who recommend 120% above this rating. Now with only a 1600g thrust I suppose that the 800 gram take off weight per motor barely fails to meet this minimum, and ONLY when the battery is fully charged 🙁
I also used this to test and calibrate the Attopilot current and Voltage sensor board. I found the board’s outputs actually comes extremely close to its rated values of 63.69 mV/Volt and 36.6 mV/A, the measured values were actually 63.66 mV/v and 36.22 mV/A. Remember that I overheated my board which may explain the extremely small discrepancy of the measured current.
Also if you look at the data, supplied in the spreadsheet link below, that the current compensation factor value has a much larger error in currents below three Amps but seems to level off at values above this. I did not test values greater than about fifteen amps so I would not expect laboratory grade signal from a current sense board that is rated for currents of 90A. And, like I’ve already said, this may be due to my ham handed soldering and repeated heatings of this board.
Update, I received a new Attopilot current/sense board and the current factor is indeed much closer for currents above about 1/2 Amp. It seems that my previous board was indeed affected by excess heat but still remains fully functional.
Update 2015 06 7:
I tested four other motors that I have and found one of them made thrust, actually 2150 grams of thrust. The remaining three made between 1800 and 1950 grams of thrust.