"Phoenix" - Rebuilding a unique airborne sensor package

RazorWind

Supreme [H]ardness
Joined
Feb 11, 2001
Messages
4,646
This thread will be about a project much different from standard case mod forum fare. It will make more sense if you read the back story, but if you don't care to do that, just skip this first post - the build process will begin in my replies.

Back Story:

I work for The University of Texas at Austin, in a department that does geologic research. We have a group within our department that does airborne surveying to support this research. This group is one of a very few in academia that actually owns its own aerial survey equipment, including an airborne LIDAR scanner (which is pretty freakin' cool), and our newest acquisition, a custom-made hyperspectral camera rig that includes four cameras that, together, can "see" a wide swath of the electromagnetic spectrum, with a high degree of spectral resolution. In other words, if a standard JPEG file has three bands - red, green and blue - the data produced by this system has several hundred, with red green and blue being among them. Part of my job, in the department's IT group, is to operate, maintain and support the aerial survey equipment, which is as much an information technology (and mechanical) challenge as it is a scientific one.

When it was delivered to us, the original system looked like this:
earthscan.jpg


Tragically, in late 2014, the airplane that the aforementioned camera system was installed in crashed, killing the pilot, destroying the airplane, and severely damaging the system. I got really lucky - the plane was on its way to pick me and a coworker up, so we could perform the system's first test flight after we took delivery of it. If it had crashed an hour later, I'd have been riding in it.

Here's the news article about the crash:
http://kxan.com/2014/11/23/dps-working-possible-plane-crash-in-lee-county/
If you look at the photo of the wreckage, the original camera system can be seen; it's the grey box right in the center of the frame. It actually did surprisingly well, given that it suffered a near-vertical impact with the ground at 90+ knots. Nevertheless, it suffered some pretty horrific looking damage; the housing, and many of its internal components, are a total loss.

For various reasons, it is not practical for us to send the system back to its manufacturer for repairs. Instead, it's fallen to our aerial survey team to salvage what remains of this system and construct a new sensor package from any still usable parts. Luckily, only three of the system's four cameras were installed in it at the time of the crash, and none of them were irreparably damaged.
 
Rebuilding the "Phoenix"

When the system was delivered to us, it had several glaring shortcomings that I'm going to address during the process. They are:

  • Weight. The system's original weight was nearly 250 pounds for JUST the sensor box itself. This doesn't include any of the external parts, such as the operator display, power inverter, external UPS, or cables. Because it weighs so much, it's very nearly too heavy to be used in the Cessna 206 it was originally designed for. Whether or not the crew had a heavy breakfast could conceivably make the difference between being over weight or not.

  • AC Power - This is an airborne system, meant to be installed in an aircraft, and used to take photos of the ground below. As such, just like any AC-powered device installed in a vehicle, it required a bulky inverter to convert the 28VDC supplied by the aircraft to 120VAC. Every single component inside it runs on DC power, though, and many of the components require 24V anyway, so as part of this process, I'm going to replace all the wall warts inside it with DC-DC power supplies, which should save weight, power and complexity.
  • Excessive complexity - Originally, this system used two standard Windows PCs to control its four cameras, with one computer driving two cameras each. The time stamps for every pixel it recorded had to very precisely synchronized, necessitating a pretty clever, but clumsy and complex GPS-driven mechanism for synchronizing the system clocks between the two computers.

    After testing the system on the ground, and looking at our actual anticipated use, it was determined that in reality, the need to use more than two of the cameras simultaneously was unlikely to ever arise, and that operations could be simplified significantly by just using one PC to drive all of the cameras.
  • Consumer-grade RAID Storage - This system produces A LOT of data. Continuous data rate from just one of the cameras is in the neighborhood of 120+ MB per second, and a typical day of surveying might consist of four or five hours of actual surveying.

    As it was delivered to us, the storage mechanism for all this data was a pair of Mediasonic external RAID enclosures that were bolted into the housing. Getting the data out of them required removing the drives from the boxes, and plugging them into an indentical box! Surprisingly, this actually did seem to work about 90% of the time, but it was pretty flaky, and basically required us to rely on shortcomings of the cheap RAID boxes in there. I argued with one of the vendor's engineers about this, but he swore his company had been using this setup successfully for years. I was never convinced, but I will conceed that it probably would have been OK, assuming we never made any mistakes. The RAID boxes were destroyed in the crash anyway.

    I'm going to just replace these with SSDs. My own testing has shown that newer SSDs are fast enough to handle the data rate required, and the cost of SSDs is pretty small in comparison to how much we spend on fuel for the airplane in a single day.

The ultimate reason for all these shortcomings, I think, was probably cost. We got a smokin' deal on this thing originally, and that meant the folks that built it had to cut some corners in terms of how thoroughly integrated it all is in order to meet their required budget.
 
The first step in the process was to dissemble the damaged housing, and remove all the components.

Here's the system as it was given back to us, after recovery from the wreckage of the airplane (Photo Credit: Dr. Jeffrey Paine):

disassembly0.0.jpg

disassembly0.1.jpg


This is the mounting plate that attaches the base of the system to the floor of the plane. Because every airplane is a little different, you effectively need one of these for every airplane/sensor combination. As you can see in the photos, this piece is pretty stout - that's part of the airplane still attached to it, and the other three bolts tore themselves out of the floor.

disassembly1.jpg


This piece makes up the bottom of the sensor box itself. It's mounted using the bolt holes you see to the mounting plate pictured above using six to eight 10mm bolts, depending on the mounting arrangement. It's about 3/8" thick, and supposed to be flat, but it obviously isn't anymore. You can also see that vibration isolation springs that attach it to the box itself got hyperextended, and no longer return to center.

disassembly2.jpg

disassembly2.5.jpg


These two show a couple of other shots mid-disassembly. Initially, I hadn't planned on disassembling everything, but it quickly became clear that I'd have to as more and more parts came out bent. In the second photo, you can see one of the four cameras, connected to one of the control PCs, which I'm testing for basic function.
disassembly3.jpg

disassembly4.jpg
 
Here's another photo of the exterior of the system, after disassembly. This panel made up the top of the system; I'm not sure what punched the hole in the top of it, but that's oil and dirt smeared on it, so I'm guessing it collided with the ceiling of the cabin, at some point during the crash.

rebuild015.jpg


On a brighter note, most of the rebuild parts have arrived! Everyone loves parts porn, right?

The hardest part of the project will be replacing all the AC-DC wall warts inside the case with DC-DC converters. The most exotic one is this trick 24VDC ATX PSU, made by Powerstream. It's a "700W" unit, which is probably way more than I actually need, but I wanted to be sure I have enough amps on the 12V rail, and it seems to be heavy on the +5V and +3.3V for some reason.
rebuild001.jpg

rebuild002.jpg


This is one of the original PCs that came with the system. It has a ridiculously stout aluminum M-ATX chassis and weighs about 25 pounds with the system installed. Some of the panels are over 1/4" thick!
rebuild003.jpg

rebuild004.jpg


The waterjet cut side panels arrived as well. They're currently just raw aluminum, but I'm waiting on a price quote to have them anodized in black, so they don't just look unfinished.
rebuild005.jpg


The frame kit arrived as well. This is made out of 80/20 T-channel stock, cut to our custom lengths, and then fastened together with spiffy aluminum brackets and cap screws. In the photo, you can see it assembled on top of the keel plate, which was salvaged from the original system.
rebuild006.jpg


The frame is composed of two box sections, which are hinged on one side. This will allow access to the internal components without the need to disassemble it - a feature the original system did not have. The finished version will have a gas charged piston to hold the lid up when it's open.
rebuild007.jpg


A close-up of one end of the keel plate. It's a big thick piece of aluminum with threaded holes drilled for mounting all the various components. The gold colored component in the photo is an inertial measurement unit, used to record the system's position while data is being recorded, so that the data can be referenced to a real location later.
rebuild008.jpg

rebuild009.jpg
 
Here, we constructed a test rig for performing spectral calibration on one of the cameras.
rebuild010.jpg

rebuild011.jpg


The camera in the photo is sensitive to VNIR radiation or "very near infrared." Wavelengths range from about 350nm up to about 1000 nm, which includes the visible portion of the spectrum.

The lamp in front of the camera is a fluorescent krypton tube that emits light at very specific wavelengths. We take pictures of it with the camera, and use that information to relate measured values to real world wavelengths, so that we know exactly what type of radiation the camera is receiving.
 
Here, I'm test fitting the floor panel to the upper portion of the system. You can that it has an opening on the hinged side for cables to pass through. In addition to providing a fastening point for various hardware, this panel serves to separate the upper compartment from the lower, keeping heat from the PC from affecting the cameras in the compartment below.
rebuild012.jpg


Test fitting some of the external ports to their panels.
rebuild016.jpg


I salvaged this pretty tamper-resistant switch from the old system; it can be seen in some of the photos of the wreckage in this thread. The retaining nut fought me tooth and nail all the way off, but I eventually got it. If I hadn't been at work, liberating this switch probably would have been at least a one beer job.
rebuild017.jpg

rebuild018.jpg


It appears to be the illuminated type, but the LED part wasn't hooked up before. I'm sort of curious what color it is.

The rest of the power supplies. These slick DC-DC converters will make up the rest of the power subsystem. There are two of the large ones, which condition the dirty 28ish VDC power from the aircraft to 24VDC for the cameras and PC, and one smaller 12V one to power the remaining two cameras. There's also a small 5V one, which powers an ethernet switch, used to connect some of the cameras to the computer.
rebuild019.jpg

rebuild020.jpg

rebuild023.jpg


Here, I'm test fitting the power supplies to their mounting bracket, which will go along one side of the lower portion of the system.
rebuild024.jpg


Some various other connectors. The big one is an amphenol cannon plug that will be used to connect the system's power lead to the aircraft. The smaller two will be mounted on the system's front panel, for connecting the IMU to a GPS antenna mounted on top of the airplane.
rebuild024.jpg

rebuild025.jpg


One of the cooling fans. This is a 120mm 24VDC fan that moves a freakin' crapload of air. The system will have a total of four of these cooling it - two on the top and two on the bottom. The typical airplane is not air conditioned, and at least one of the cameras is very sensitive to temperature, so good airflow through the enclosure will be important, especially in the summer in Texas.
rebuild026.jpg
 
Another one of the cameras, a Sofradir EC Atom-1024 Microbolometer. That name is a fancy way saying "predator vision," which is exactly what it produces. It's sensitive to thermal radiation in the 7-14 micrometer range, and is super fun to take photos of your junk with.*

After the crash, this camera would respond to commands, but not produce an image. I had to crack the housing open to figure out why, and after repairing a tiny connector inside there, I got it kind of working, with a ton of noise in the image.
rebuild027.jpg


After a trip back to its manufacturer, the microbolometer is now right as rain.
rebuild0036.jpg

rebuild037.jpg


Here it is being test fit back into the keel plate;
rebuild038.jpg


*Obviously, we do not use university hardware to take photos of our junk, unless it's for imporant scientific research.
 
The "very near infrared" (VNIR) camera that I mentioned before:
rebuild029.jpg

rebuild030.jpg


This camera appears to have survived the crash almost entirely undamaged, except for a small nick to the rim of its lens assembly. Nevertheless, in order to ensure its accuracy (and thus our own scientific rigor) we have to do check its calibration. Thus, it's still in the test bench, and hasn't been fitted back into the system.

This component drives the camera's shutter mechanism, seen in the second photo above, with the cable sticking out. It connects to the PC via good old RS232 serial.
rebuild032.jpg
 
Camera number 3 is a short wave infrared unit, sensitive to 1000 to 2500nm. To be honest, I'm not really sure what this is used to measure but to be sure, it's one of the more interesting cameras, technologically.
rebuild034.jpg


Most of what you see in the photo is an enclosure that contains an elaborate cooling mechanism. The camera itself looks similar to the blue cube-shaped component of the VNIR camera. The cooling mechanism is required because the actual sensor inside this camera must be kept at about 71K (that's about -200 degrees celsius) in order to produce a clear image. It consists of a tiny refrigeration unit, the "hot" side of which is then cooled by a big meaty peltier thermocouple. The heatsink you see on top of it cools the hot side of the peltier.

This cooling system has been somewhat problematic, as the refrigerant inside it has a tendency to leak out. As a result, it was away for repair at the time of the crash, and was not damaged. I seriously doubt that it would have fared well in the crash.

Here it is reinstalled on the keel plate. The set screws you see in its brackets are used for fine adjustment of the cameras' alignment. I'm not looking forward to that part of the process.
rebuild035.jpg
 
Subbed. This is a very interesting (to me) project. I really enjoy seeing old or damaged equipment restored to proper working order.
 
Subbed. This is a very interesting (to me) project. I really enjoy seeing old or damaged equipment restored to proper working order.

Thanks!

The original system actually isn't old at all. We had just taken delivery of it, although I think we originally ordered it almost three years ago now, and it just took that long for the company we bought it from to get everything working.

I need to take some more photos of the damaged panels. The ones I've posted here are from the side that didn't impact the ground. The ones on the opposite side are, suffice to say, no longer flat.

Anyway, back to the story!
Earlier this week, some more parts of the power subsystem arrived. These circuit breakers are pretty common in aviation, and pretty much every airborne sensor I've encountered uses them. No seems to actually stock them, though, so when you order them, the lead time is like nine weeks.

From left to right, the 30 amp ones will be used to switch the components that require 24V power: the SWIR and MWIR camera will be one circuit, and the control PC will be on the other. The 50 amp breaker is the master, and controls power to the entire system.
rebuild043.jpg


These springs are vibration isolation mounts. They connect the keel plate to the base plate that attaches to the floor of the plane, without rigidly mounting the system. They serve as sort of a crude image stabilization mechanism, but more importantly, prevent vibration from the aircraft's engines from shaking the system apart.
rebuild044.jpg

rebuild045.jpg


Here, I'm crafting a new power cable. It consists of four 10 gauge wires, and the Amphenol cannon plug I posted above. The connectors for the other end of it haven't arrived yet, but that end consists of a big meaty LEMO connector.
rebuild046.jpg

rebuild047.jpg

rebuild048.jpg

rebuild049.jpg


This is the inside of the lower "rear" panel, to which the circuit breakers and fan switch are attached. I'm still missing a couple of components that get mounted to this - namely the socket for the power cable, which goes in the big round hole on the left, and the ammeter, which goes in the rectangular hole above the breakers.
rebuild051.jpg


Today, I moved all the hardware for controlling the SWIR camera from the extra PC into the one that's going back in, swapped out the power supply for the DC-DC unit, and began installing all the cables for it.
rebuild052.jpg

rebuild053.jpg

rebuild055.jpg

rebuild057.jpg

You'll notice that the PSU in the last one has a series of screw terminals in lieu of the regular chevron shaped socket. :cool:

I also reinstalled the VNIR camera (the blue one) back into its bracket in the lower part of the case. Recalibration of it is apparently complete, and went well. The manufacturer says it wasn't damaged in the crash.
rebuild058.jpg

rebuild059.jpg


The system and one of several tables full of parts waiting to go into it. You can also see the test stand I built for it back before the crash. The cameras can look through a hole in the stand, and capture images of whatever we put underneath - a much more elegant solution that just turning it on its side, for several reasons.
rebuild060.jpg
 
What is the anticipated weight savings for the new build compared to the old one?
 
What is the anticipated weight savings for the new build compared to the old one?

Weight savings in the box itself will probably be roughly 20 to 30 pounds, by eliminating one of the control PCs, the RAID boxes and eight hard drives. I've also reduced the physical height of the housing, which should save a couple pounds' worth of external panels and the frame.

Installed weight - that is, the weight of the sensor package plus any external components - is probably more like 80 to 100 pounds, by virtue of eliminating an enormous battery backup and a gigantic inverter, that weighed about 50 and 20 pounds respectively.

I've been struggling with getting Windows installed on the control PC in the last few days. Something about the motherboards that came with the system is flaky, and Windows - even the installer - refuses to boot with AHCI enabled. I unfortunately need AHCI, as the data from the cameras may come in at a rate faster than can be handled via legacy IDE.

I seriously struggled for so long that I considered having the purchasing folks downstairs just buy me a new motherboard, but that would have required a new CPU as well. Also, because I need four PCI express slots plus a VGA port, the selection of m-atx motherboards that might work isn't that big. At length, I tried flashing one of the mobos with Gigabyte's beta UEFI BIOS, and that, miraculously, seems to have fixed the problem (knock on wood).

Anyway, here's where the system's at now:

These are the only particularly exotic PC hardware in the system. They're Camera Link video capture cards. The smaller one on the left serves the SWIR camera, and the one on the right serves the VNIR camera.
rebuild061.jpg


The VNIR camera requires an external shutter controller. Basically, it has a little solenoid on the side of it for opening and closing the shutter, which is driven by a little extra microcontroller. The microcontroller gets its commands via a plain RS232 serial port. This bracket came in the original system for mounting that controller, but I need the spot it once occupied for part of the power supply. I drilled a couple of holes in it so I could mount it directly to the frame instead.
rebuild062.jpg

rebuild063.jpg

rebuild064.jpg


This is the best I could do one-handed of the space between the power supply and the cameras. That's the SWIR camera on right, and one of the 24VDC converters on the left. You can see the rat's nest of power wiring on the far end, in the background.
rebuild069.jpg


Here's the system with the top portion opened, and my high tech prop rod. Eventually, that will be replaced with a gas strut, but that's low on the priority list at the moment.
rebuild072.jpg

rebuild073.jpg


A better look at the MWIR camera, and the nightmare of power supply wiring. The wiring is actually pretty simple, but because I'm going to have to disassemble this to send the exterior panels out for finishing, I haven't taken time to neaten it up yet. It just has to work and not catch fire for now.
rebuild074.jpg


The back side of the system. The DC-DC power supply I'd ordered for the PC was dead as a doornail for some reason, so it's away being RMA'd. Unfortunately, because it comes from Taiwan, the turnaround on that is like 5 weeks. In the mean time, I'm using this Antec I had lying around. In this photo, all the expansion cards have been removed from the PC while I troubleshot the AHCI issue.
rebuild075.jpg


A shot of the front of the system with the lid open.
rebuild076.jpg


Here's one of the control PC cases, with one end removed. You can see how much aluminum there is here.
rebuild077.jpg


This is the part missing from the last photo. It appears to have been milled from a solid 10mm thick piece of stock, and probably weighs at least 5 pounds on its own.

rebuild078.jpg
 
Alright!

I've been working on this, but there's presently not a whole lot to see. I've moved the system into an empty office for now, as I now have to test all the various components, and the power supply for it is LOUD. I have to wear ear plugs when it's running, or I get a headache after an hour or so.

Here's the system where it's sitting now:
rebuild079.jpg

I'm a little disappointed that you can't see the indicators are lit up in the resized version of this image, but it's actually powered up there.

I had all the cameras installed for a while, so I took the opportunity to take this photo of the underside. With the exception that airplanes are not generally made out of plywood, this is what it looks like from below when it's installed in an aircraft. Flight direction toward the bottom of the image.
rebuild080.jpg


The next part of the process is to get all the software configured. I struggled for a couple of days to just get all the hardware drivers installed, but I eventually got that done. The next step is to configure the software that actually handles the cameras. This is easier said than done, however, as three of the four cameras require a piece of software called the "Pleora eBus Runtime." Unfortunately, each one requires a different version of it, and only one version can be installed on the system at a time.

The solution to this problem is to write my own software to control at least two of the cameras. This will be beneficial anyway, for reasons I'll be explaining in my next post. Stay tuned.
 
I'm digging a worklog on a system that reminds us not all computers sit in a box in an office (or our mom's basement, for some of our members).

I also dig the nostalgia of building a "complete" system, back to a time when everything you needed couldn't always be found on a shelf at a retailer. Keep up the good work, sir.
 
I'm digging a worklog on a system that reminds us not all computers sit in a box in an office (or our mom's basement, for some of our members).

I also dig the nostalgia of building a "complete" system, back to a time when everything you needed couldn't always be found on a shelf at a retailer. Keep up the good work, sir.

Thanks!

Here's a shot of my (very alpha) camera control application. It's not much to look at, but I think it will get the job done. Because I'm compiling this against a version of the Pleora eBus drivers that I picked out, I can make it able to control at least three of the four cameras, whereas the original arrangement required four different pieces of software that were barely able to play nice together on two pcs. The extra space on the right of the window will get filled with the display for the other cameras eventually.
rebuild082.jpg


Here's a screenshot of the same frame being viewed in Exelis ENVI, an application for interpreting spatial raster data (such as aerial photography). That's a cheap shop light with an incandescent light bulb in the picture, as seen by the MWIR camera.
rebuild083.jpg


Controlling all four cameras from one piece of software will be a huge improvement over having to use four separate ones, but there is another really critical reason to make our own control software. That reason is that not even one of the original control applications supported any sort of external triggering.

Aerial surveying is a precise business. Most survey flights make use of some sort of GPS-based navigation system, where waypoints are figured out before the flight, and the cameras are triggered by a computer when the airplane reaches those waypoints. The folks that ordered our system did not know this, and thus did not specify that it had to have this in the contract. As a result, there was originally zero provision for it, and triggering was going to be entirely manual. I'm sure most of you can imagine how much potential there is for the human operator to screw that up.

Enter: Trackair!
rebuild087.jpg

rebuild089.jpg


This is an off-the-shelf airborne surveying navigation system. It has its own GPS, and the large black component is a Microsoft Surface Pro. It runs its own unique software that keeps track of the aircraft's position, and sends out a signal via the serial port(s) when it needs to tell the cameras (or some other sensor, like a LIDAR) to start shooting.

Previously, we were going to just let this suggest to the operator when he should start the cameras recording. In my own software, I can just listen for the start command on a serial port, and trigger automatically. :cool:

I mentioned the power supply yesterday night. Here's a couple of shots of it. This produces up to 100 amps ( :eek: ) at 28VDC. It's originally intended function is to supply power to an airplane on the ground in lieu of its batteries. Most are much bigger than this, and contain batteries of their own, but this one just gets its juice from the wall.
rebuild085.jpg

rebuild086.jpg
 
This is an amazing project. Are you doing it solo, or is there a team working in it?
 
This is an amazing project. Are you doing it solo, or is there a team working in it?

It's pretty much just me. There's a woman in our group who is an expert at using ENVI who worked on the spectral calibration of the VNIR camera, and there are other folks in the department who do stuff like purchasing parts, but I'm the only person actually turning wrenches or writing code on this system at the moment. It's worth noting that I've had a considerable amount of support from folks at the manufacturers of all the various components, but that mostly consists of them sending me the drivers for whatever component they made.

Once the system is mission ready, we have a team of five or six people who will work to collect and analyze data with it.
 
rebuild090.jpg


Here's an updated screenshot of the control software, which I'm calling "BevoVision" (A play on the original vendor's "Hypervision") until someone comes up with a real name for it.

As you can see, I've now got two cameras working. That's the thermal camera on the left, and the VNIR on the right. The thermal image is pretty self explanatory - that's my arm holding a shop light with an incandescent bulb.

The VNIR image, on the other hand, is two dimensional, but unlike a normal photograph, each scanline represents a different wavelength, as seen at a very narrow strip of the floor or ground underneath the sensor. Frames are captured at about 30fps as the system is passed over the ground, and a 2D image showing a given wavelength can then be constructed later by taking say, scanline number 375 from every frame, and stacking them together, to make a new, more conventional looking, photograph.
 
Updated screenshot:

rebuild091.jpg


The biggest change from a few days ago is that I got support for the shutters implemented, along with the IMU.

I posted a couple of photos of one of the shutter control mechanisms earlier in this thread. This system's cameras work like ordinary digital cameras in the sense that they can record continuously without opening and closing a shutter for every frame. They nevertheless need shutters, however, in order to be able to capture totally dark frames. These dark frames are used to measure what value at each pixel on the sensor constitutes a "zero" reading, as even in total darkness these types of sensors still register a small amount of light hitting them. This effect is non-uniform, meaning you have to take a picture of the inside of the shutter, and then subtract it from every real image to get accurate results.

The IMU took some doing mostly because required a user interface to configure it, and that always takes a ton of boilerplate "plumbing" code. Also, the IMU communicates using a very terse protocol that, even though I had the manual, took about a day and a half for me to fully understand. It's working now, though - those are real values shown at the bottom of the window.
 
Minor success!

rebuild092.jpg


At last, I got all four cameras working at the same time. The SWIR camera (center top) fought me tooth and nail, but once I figured out what the trick was, getting it working as seen here was actually surprisingly simple. The image appears as just a gray blur here because I had just turned it on, and the cooling unit had not yet been able to cool the detector down to operating temperature, which is 71K, or -200 degrees Celsius.

I also added a normalize feature to the image previews, so they don't seem unnaturally dark.

There's still a lot more to do than I'd really like, but at this point, every important piece of hardware is now at least kind of supported, and I think we could go out and collect usable data. To get full support for all of the advanced features of the cameras, some of which we really need, I think I'm going to have to arrange for us to buy the SDKs for some of the cameras from their manufacturers.
 
A couple more parts showed up yesterday, so I did a little more assembly today.

So many baggies!
rebuild_094.jpg


Out of the bags, we have two mated pairs of LEMO connectors. These LEMO connectors are amazing, but you'd think they're made out of unicorn bones or something. These took six weeks to get from the one vendor that would even respond to me, and four of them cost something like $400. :eek: On the bright side, they're pretty tough and very secure, which is important for this application.

rebuild_095.jpg

rebuild_096.jpg


You begin to see where all the money goes when you take one apart. This is the system end of the power cable prior to assembly. It's obviously more complex than your typical four pin Molex.
rebuild_099.jpg

rebuild_100.jpg


And this one is for the power supply for the operator display.
rebuild_101.jpg

rebuild_104.jpg
 
And we're back from the dead!

I've been so busy working on other projects that I haven't had tons of time to spend working on this, but now that those are out of the way, I've had a little time to concentrate on it.

Here, the system is turned on its side for the installation of the lower mounting plate. You can see the plate in the second photo. I posted some photos of the original that got bent up in the crash earlier in this thread, but this is what it's supposed to look like.

rebuild_106.jpg

rebuild_107.jpg


As you can see here, the clearance available to get the mounting screws in is pretty small - maybe an inch. I had to trim an an allen wrench down to a nub to get get these tightened down. Also visible here are the vibration isolator springs.
rebuild_108.jpg

rebuild_110.jpg


And here it is sitting upright, with the weight supported by springs and mounting plate. The device in the background is a piece of surveying equipment that I'm using to measure the relative locations of the corners of the camera housings inside the box to that of the IMU and corners of the housing. This is needed to improve the precision of our finished data product.
rebuild_120.jpg

rebuild_121.jpg


I've got some more photos to post of the final assembly and the later revisions of the capture software. The system is, as I type this fully assembled and loaded onto a dolly. We'll be taking it down to the airport today and installing it in the aircraft. I'll be doing some ground testing today, but the weather is terrible, so we can't fly immediately. Weather permitting, we'll take the system for its first test flight on Monday of next week.
 
what is your final operational weight for the unit now?
 
what is your final operational weight for the unit now?

I actually forgot to weigh it, but it's definitely lighter than our airborne LiDAR, which weighs around 100kg (230ish pounds).


I'm thinking installed weight is around 200 lbs., vs. close to 300 lbs. in the original configuration. Most of this is due to the elimination of the UPS and inverter, but the box itself is also quite a bit lighter, as I eliminated an entire PC from the design.

Here's the finished box in the hangar, after we unloaded it from the truck. I wish I'd had the panels anodized - the bare aluminum looks pretty cool, but it shows fingerprints like crazy, especially in humid environments.
rebuild_123.jpg


There's not really much to see of the installation process. We install the mounting plate pictured below into the aircraft using whatever fastener arrangement that aircraft uses, and then we bolt the instrument to the the mounting plate. Oddly, there isn't any obvious standard for these things, so for every different aircraft we might install a particular system in, we have a different mounting plate.
rebuild_124.jpg


And here is the system installed and bolted onto the mounting plate.
rebuild_125.jpg


The operator display, on its mounting arm.
rebuild_126.jpg


Looking forward from the operator seat.
rebuild_127.jpg


Another view of the operator seat. It's pretty cozy in there - you get maybe a foot of leg room on the left side. I'm about 6' tall, so I have to stick my legs around the sensor, to the right.
rebuild_129.jpg


The clearance between the sensor and the pilot's seat. Not a whole lot of space to spare, and this is actually an improvement over some other systems I've used in this airplane.
rebuild_130.jpg


The airplane we'll be using for the test flights and first field campaign is a ~1978 Cessna 206/G. As airplanes go, it's pretty pedestrian, but it's still super fun to ride in. I think I'm in the minority of aerial surveyors who actually like small planes, but this is about as small an aircraft as we can realistically use. As you can see, the equipment is a relatively tight fit in there, and you start running into problems where you can't fill the fuel tanks all the way up if any of the crew are fat, due to weight limitations. This system is thankfully light enough that that shouldn't be a major problem.

rebuild_128.jpg


Another couple of shots of the outside of the airplane. My coworkers pictured here are in the process of taking measurements of the location of the box inside the aircraft, relative to a very precise GPS antenna mounted on top of the wing (flat thing between the two radio antennas in the second photo).
rebuild_133.jpg

rebuild_134.jpg


We need that measurement so we can adjust the position reported by the GPS a little when we process the data - without it, everything seen in our photography would appear to be about 5 feet off from its actual position. This is less of an issue with photography, but a very critical thing when capturing airborne LiDAR, as the accuracy of LiDAR data is measured in millimeters.
 
How are your employers with the time/resources it ended up taking to do this?

Looks to me like you did a great job, but having done some custom chassis and system integration I know how labor intensive it can be.
 
How are your employers with the time/resources it ended up taking to do this?

Looks to me like you did a great job, but having done some custom chassis and system integration I know how labor intensive it can be.

They're pretty supportive. It was an expensive process, but the capability this gives us is, I'm told, unique among civilian outfits. Furthermore, our actual product is peer reviewed literature, and I've now got tons to write about just from the process of building this, not even considering the science we can produce by actually using it.

One of the things I'm hoping to do in the future is publish the control software as an open source project. Nothing like it exists as an open source solution (because there's buckets of money to be made by insisting on proprietary ones), and that could be useful for this industry. My software is considerably more flexible than some others I've seen already, and I'd describe it as "mid-alpha quality."
 
Back
Top