Back in early January I was tasked with the duty of delivering my recently quadrupley bypassed father to my aunty's house down on the south coast of New South Wales for some much needed rest and relaxation. As it happens, the quiet sleepy town of Moruya, in which my aunty resides, is also home to Hobby King's Australia based warehouse. Not being one to pass on an opportunity to double task efficiency, I decided after depositing my father into my aunt's capable hands that I might as well make use of the ample cargo space offered by my humble little hatchback. So I dropped in to grace to the good people at HK Aus with some good old first person custom.
To cut a long tangent short, this is what I brought home with me:
The 1500x300x400mm shipping box of the UAV-3000 was a tight fit, but no problem for the jack of all trades GTi.
Anyway, while I was there in the showroom, I noticed a few warranty return/clearance items on the bench at heavily discounted rates. Among these items was a folding multirotor platform. The IdealFly IFLY-4, with which a former customer had apparently had little success, appeared to be a nice and entirely complete airframe which simply needed a new FC. Keen for a new build, I swapping it for a single John Flynn, bid the HK team a good day and headed north.
Once at home I did a bit of research on the IFLY-4 and was surprised to find out that, due to the fact the IFLY-4 ESCs communicate via I2C, I would need to replace these as well as the FC. So, I set about ordering the following items:
* Afro Slim ESCs
* I also ordered a motor to replace one with a bit of play in it.
When these items finally arrived last week I set about assembling my new craft along with a few extra items from the parts bin including camera, RCRX, power meter, BEC, cables, fasteners, and batteries.
Introducing the IdeaFly IFLY-4 Reloaded - APM Edition.
You know how with some builds nothing seems to go right? When holes don't line up, you can't seem to find the right sized fasteners, you are forced to think and rethink, mount and remount, route and reroute and repeatedly forced to retrieve tools that are out of immediate reach? Well this build was nothing like that. It just seemed to come together beautifully.
I had originally modified the factory distribution board with the intention to recycle it for it's former purpose however, after damaging some of the PCB traces I realised it really didn't have the current carrying capacity I would like, so I just used it as a mounting platform for the APM, which was attached using soft double sided adhesive foam, choosing instead to deal with power distribution at the Power Module.
I was really pleased with my work on the Power Module. It is very very strong (much stronger than I expected, and it has been tested with great force), super compact, neat and convenient for maintenance. A Turnigy 3A BEC powers all 5VDC needs with the exception of the camera, which is left to rely on it's own on board battery (no chance of nasty noise entering camera).
Amazingly, after removing the factory I2C ESCs, I found that the Afros slotted very neatly in to the arms. Having read about others having issues with Simon K ESCs and AutoTune, I was a little worried about having problems with the Afro Slims, but I simply had to persevere due to their convenient size, affordable price and strong performance credentials. I am not exactly using them in a low kv pancake applcation either, so the "she's be right, mate" attitude prevailed. Just a dab of hot glue was required to firmly hold them in place. I was worried about using hot glue on such a heat generating component, but they don't seem to warm up much at all, as the chipset is 30A rated and they are running at less than 50% duty at max.
The GPS/Mag was purchased on ebay for about $18. It is mounted on a re-purposed FPV camera bracket which I flattened and fastened to the front of the frame using hard double sided adhesive foam and two zip ties giving a solid but nicely dampened mount for the GPS/Mag. Mounting holes were a perfect match, and I had all the right nylon M2.5 fasteners and spacers in the tool box to suit. TIP - Invest in an assortment of M2.5 and M4 nylon and SS fasteners, as it will save you so much build time.
A 433Mhz Telemetry Radio from RCTimer is mounted on a short 2-inch boom at the back of the centreplate in the same way outlined above, but with two M2.5 bolts added to ensure it remains firm and alignment stays true.
For battery duties I decided to use the 3S 2200mAh Turnigy, of which I have several. It is about 25mm too short to span the top centreplate and corresponding anchor points, so I fabricated a hold down plate to solve the problem. Keeping weight to an absolute minimum, the hold down bracket is made from Balsa wood, painted black with a texta and enclosed in clear heatshrink. The bracket is plenty strong and weight is less than 10 grams.
At this point, I had the basic hardware sorted and moved onto setting up the firmware and configuring all relevant settings in Mission Planner. This is not my first APM setup, so navigation was pretty straight forward. I did use Mission Planner instead of APMPlanner2 due to familiarity though, and I did go through some fiddling getting the right hex and character set for MinimOSD Extra (hence why I didn't tune better position for the telemetry info in the 4:3 frame - I just wanted to move on once it was working effectively). I was also surprised to learn that the only firmware setting on the Afro Slims able to be set by the user without the USB programming tool is the max and min throttle points. With the USB programming tool riced at just $6.65 this would seem an easy up sell if only mentioned in the ad.
Ready to fly, I grabbed my Nexus 7 ground station and a battery and headed out to the backyard to burn through a charge on the standard settings. As usual, the APM's default PIDs worked fine out of the box. Super pleased with myself, I headed back inside and proceeded to invest several hours sanding the props into perfect equilibrium.
Next I wanted to try AutoTune for the first time, having witnessed Ian Lions perform this operation with his TBS disco earlier this year. So I grabbed another battery and headed for the park. Ardupilot performed it's dance with effortless finesse, delivering an exciting but otherwise unremarkable result on the first attempt (only ~800mAh consumed). The improvement in flight performance was immediately recognisable and I was just overwhelmed by how precisely it could be maneuvered and positioned. What a fantastic afternoon!
That evening I headed home and began adding the requisite parts to enable FPV and aerial video capture.
For imaging duties, I chose the little Mobius Action Cam. It is a great little camera, offering 4/5ths of a gopros performance and convenience at just 1/5th of the price. Plus, it is a significant upgrade on the #26 808 I was using previously. The file format it generates is much easier to work with too, as it imports into iMovie for iPad/iPhone for fast easy editing and convenient upload to youtube.
For mechanical isolation I created two small plates from balsa wood roughly the size of two moon gel rectangles placed side by side. The top plate was cut to tightly tessellate with the exposed bolts on the underside of the GPS mount and the bottom plate was left flat. The two balsa plates were sanded and coloured with texta, moongel was placed between these two layers and the sandwich loosely wrapped in cling wrap. This gave a really nice surface to mount against and the whole thing is held in place by a relatively loose fitting velcro strap with an extra piece of moongel added below the camera for good measure. The results offered by this isolation method have proved very pleasing. First Person View of telemetry overlayed video is transmitted back to my DIY Ground Station via a DIY 5.8Ghz VTX.
The next day I jumped in the car and headed up to my sister's place in the mountains in order to give the newly updated rig a run. The whether was overcast, but relatively mild. I ran down two batteries and got some nice test shots. Unfortunately, I also lost the canopy during a particularly fast decent. The great news is that I was traveling backwards at a rate of knots when it fell so I was able to capture the descent of the canopy via the onboard camera. Using the below images the canopy was found the next day.
Smitten with my new craft, the next morning I set out at sparrow's fart for more flights. Overly eager, I was up and ready too early and cursing Siri for making me wait.
30 minutes before sun up I was parked down at Old Government House in UNESCO world heritage listed Parramatta Park. The second I saw the first few photons shoot in my direction I was away. I was starting to gain real confidence in the reliability of the craft at this point. All features and functions, like loiter precision, just spot on.
Reviewing the video from these early morning flights exposed the low light and exposure balancing limitations of the little Mobius. But there are only so many photons a tiny lens and sensor can collect in 33.3-millisecond.
I continued with more flights, moving from place to place. I drove between locations and carted my little integrated ground station so to record both telemetry (via DroidPlanner) and live video stream (via SD DVR installed into DIY ground station).
By mid morning I had been to several locations and had put about half a dozen charges through the craft.
I went home to download and review the media from the Mobius. Most flights were conducted twice, once recording 1080HD@30fps and then again recording stills at 250-millisecond intervals. It seems logical I should have realised this earlier, but the Mobius is not able to deliver live video out while simultaneously capturing and recording 3MP images 4 times a second. Mobius prioritises live output over still image capture, so I didn't pick up on this issue until getting home. Sadly, all the airtime spent capturing stills gave just gave me hundreds of images like this (the panoramas above were recovered from these corrupt files):
Then I set out again, revisiting many of the locations I had shot at earlier in the day. After getting another half a dozen or so more flights under my belt, I returned home at about noon. I put the Mobius and the installed 2200mAh Lipo on charge and grabbed a quick bite to eat for lunch. Before downloading the media from my second morning outing, and (for the first time in 2 days) without bothering to link DroidPlanner and operate the flight recorder in my ground station, I set out into the backyard to get a nice panorama of nearby Lake Parramatta and the surrounding woodland.
While performing the same pirouette maneuver (at about ~40-metres AGL), just as I had done so many time before, I noticed the craft drift noticeably beyond previously experienced limits. I immediately aborted the maneuver, and attempted to recover but failed.
What happened next is not entirely clear. I think I switched to stabilise mode, but this did not help. I then went for loiter mode, but the craft continued moving away from me at an accelerated rate. So I think I went for loiter again, but that didn't arrest the craft either. Next I remembered that RTL was associated with channel 8 so I flicked that, but this didn't help. I then remembered the the RTL failsafe I had set in the event of loss of RCTX link so I switched the transmitter off in the hope this would return it to me. Unfortunately the craft was ~300-metres away from me by this time, and it was not coming back.
In a state of fright, I jumped in the car and scanned the streets for about an hour. Nothing. It was gone. What the hell did I just do I wondered? How could the craft be so predictable and so obedient, just to go and do it's own thing all of a sudden?
Upon reflection I realised that my perception of what happened and what I was doing was probably not as accurate as I would have liked it to be.
First of all, how sure was I that I was pirouetting in loiter? Could I have been in alt hold instead? That would have explained why the craft went off alignment. I am really not sure about this, but my assumption is to blame myself and not the hardware. Also, were the mode commands I sent the craft following loss of control correct? Probably not, no, not at all actually. My channel 8 switch which has always previously been used for RTL failsafe was, at the time of the incident, still associated with AutoTune. So, when I flicked the channel 8 switch (the moment it went out of sight) I had shot myself in the foot. Next, turning off the controller would have been a good idea, but only if I had returned FTL to channel 8. In actual fact, by turning off the RCTX I had totally doomed the craft. Air Craft Investigation seems to reveal that I am completely to blame.
Ashamed, embarrassed and disappointed, not knowing what else to do and having no idea where my craft had ended up, I headed down to my local area command to file an incident report.
What a weekend!