Goliath - A Gas Powered Quadcopter

In the Pacific Northwest and want to see Goliath Mk. II or the #EVPR: Electric Variable Pitch Rotor in person? Come out to OMSI's Robot Weekend on June 16th and 17th. This will be the first time Goliath is displayed with all of the controls integrated into the vehicle. The last of the variable pitch rotors has been assembled and is ready to go on the vehicle.

Below is a photo with the first two variable pitch rotors mounted.

Progress is being made the flight controls and the hub for the first prototype was attached to Goliath and spun up to make sure it held together (see #EVPR: Electric Variable Pitch Rotor for more details). There were no issues with the EVPR, there was an issue with the flight controller. When the vehicle was activated, the controller didn't power up properly. The Pixhawk consists on an FMU and an IO board. The IO board power was the only light coming on, nothing else. As a work around for this test, the flight controller was removed and I went back to controller the throttle with just a standard RC receiver directly connected.

While I can do a little bit more testing without the controller, it's not going to be too long before I need a new controller to start testing the interface between it and the EVPR. However, I'm a little hesitant to get another Pixhawk as I'm not sure what went wrong with the old one. It was about 3 years old, but there was only a handful of hours on it. There were a few rough tests, before I nailed down the isolation on the avionics tray and Goliath had two solenoids go bad, most likely due to vibration.

I'm familiar with the PX4 flight stack and at least know conceptually how to proceed with modifying the software to work with the EVPR. However there a few controllers that use the PX4 stack including the newer Pixhawk 2.1. Of course it uses different connectors, so the stack of DF13 connectors I have laying around as well as the GPS would be worthless, but maybe it makes sense to upgrade.

If anyone has any thoughts I'd love to hear them.

Up till now the work on Goliath has concentrated the drive train and the structure. The controls were put on the back burner until the other design problems were addressed. The other items aren't done, but the project has progressed to the point that having a control system would be helpful.

When Goliath was originally conceived three years ago, the default control scheme was to use vanes underneath the rotors to direct the airflow for control. This was chosen because it was the simplest hardware setup to implement and it's been demonstrated to work for hovercraft. Back in October I started doing some basic calculations to size the control vanes and determine the required servo sizes. Turns out assuming that if it works for a hovercraft, it'll work for Goliath was a bad assumption.

The issue with using vanes is the rotor downwash velocity. Goliath has a similar amount of horsepower as a hovercraft, but instead of 1 fan, there are 4 rotors, so the power per area is reduced by a factor of 4. Additionally the equation for the force generated by a vane is:

So if the downwash is reduced by a factor of 2, the force created by the vane is decreased by a factor of four. The end result is that at full thrust, a single vane would have generated only two pounds of force. Which would be grossly inadequate. More force can be generated using multiple vanes in parallel, but the forces would still be low.

I was discussing this issue with @Benchoff at the OSHW summit, and he suggested using grid fins instead. Doing some back of the envelope calculations show that grid fins should generate enough force. The downside is that the the grid fins have much higher drag, which would reduce the payload or flight time.

Ignoring their complexity, variable pitch rotors would be the ideal control scheme. Variable pitch rotors would be able to generate larger moment torques than either vanes or grid fins. However, the increased complexity and the fact that Goliath is already a complex project, convinced me not to pursue this.

However, it's been three years and I really want to see Goliath fly, so I've decided to start building both grid fins and a variable pitch rotor. If I pick one scheme and it doesn't work, then it'll be that much longer before it can fly. So I'll incrementally develop both and see which one works out better.

Grid Fins

The grid fins I'll document as part of Goliath as they are relatively straight forward. I have sourced some material to create the fins. The fins will be made from aluminum louvers for florescent lighting. It was difficult finding sheets big enough to make a 36" disc from, but I finally found some 4'x4' sheets (shown below).

The next step will be cutting out a test disc and placing it under a rotor to determine the control forces generated.

Variable Pitch Rotors

The variable pitch rotors are a different story. I had decided not to pursue this until I came across some research that made me realize that it may be possible to create an electrically actuated variable pitch rotor with the servos contained inside the rotor hub.

I've created a separate project, #EVPR: Electric Variable Pitch Rotor, and I'll be documenting the progress there. I'll be populating more of the design details there, be sure to follow the project if you're interested and want to get updates. Additionally, I think that the project can be useful for other multi-rotors and even conventional aircraft, so I'm entering #EVPR: Electric Variable Pitch Rotor in the 2017 Hackaday Prize. If you think it's worthwhile, but sure to give it a like.

Goliath hovered for the first time in September of 2016. The hover performance was less than desirable since it required a higher throttle setting than hoped and the vehicle did not rise evenly. It tended to favor the port side or the aft. Even more puzzling, was that it tended to lift off first on the side that had the most weight. Ballast could fix the issue, but understanding why is also important. Testing has continued to evaluate the aerodynamics of the setup. Below is a video compilation of some of those tests.

Test 12 was a simple flow visualization of the rotor downwash. Tufts of yarn were added to the frame to show the flow direction along the radius of the rotor and into the frame. The tufts behaved as expected, with the tufts under the rotor mostly steady. Inside the frame, the tufts indicated the flow reversed and flowed upward due to ground effects. While the tufts wiggled, there did not appear to be anything that suggested any unsteady flow phenomena.

Tests were also conducted outside to see if the shop walls and ceiling were effecting the aerodynamics. Occasionally in the past, loose debris had been ingested into the rotors and the debris recirculated inside the wake as the flow got turned around by the walls and got re-ingested. Testing outside reduced the re-circulation.

Test 16 nearly ended up with the vehicle getting damaged. There were four hold-downs, intended to allow the vehicle to move slightly upward, yet remain captive. They weren't made long enough and the hold-downs failed on the aft end of the vehicle. Fortunately, the throttle was reduced in time and the vehicle settled back on the stand (albeit precariously).

The hold-downs were fixed and the testing continued. During Test 17, the vehicle again lifted up, favoring the port side, but at a reduced throttle setting. However, the test stand didn't allow enough movement for a full hover to be achieved. The test showed that the asymmetries were present, regardless.

In theory, the rotors themselves should have been out of ground effect as they were at least one diameter above the ground. However, for quadcopters, it may be that the ground effect is dependent on the length scale of the four rotors together and not the length scale of a single rotor. If that is true, then perhaps the port rotors are experiencing higher thrust since they are slightly closer to the ground. It's difficult to tell exactly. This may be why the Mallory Hoverbike has the offset rotors catty-corner from each other.

I'd hoped to be well into working on the controls on Goliath by now, but the shorter days and colder weather mean less time in the shop. I'm still nailing down some lingering issues with the drive train. The new pulleys are weeping grease because the bearings are getting too hot. I suspect it's because I'm using all thread axles and nuts to keep the bearings in place. I'm working on building the proper axles and axles mounts to go with the new pulleys.

Meanwhile I wanted to document the progress made on mitigating the vibrations that the avionics experience. This was accomplished by better isolating the engine from the frame and the avionics tray from the frame. The new engine mounts are made primarily of rubber, but are built such that if the rubber fails, the bolts are still captive. Stainless steel bolts are used to attach the mounts.

The avionics tray was switched from aluminum to steel. This was to add mass to help reduce the displacement of the avionics tray. Below is the new tray with some of the avionics populated.

The tray is mounted to the frame using four Expansion Nuts. I forgot to take a picture of them before I installed them, so here is a link. Below is a shot showing the flange on the expansion nut between the tray and the frame.

So how much did all the changes help? Data from the Pixhawk shows a huge reduction in the pitch rates down by a factor of 5 to 10. This means that the Pixhawk should be able to control Goliath once the rest of the hardware is complete.

Hopefully the next log update in the not too distant future will be about fixing the bearing issues.

It's been a little over two months since Goliath hovered for the first time. Here are some of things that have gone on since then.

First was a trip to the Portland Maker Faire, September 11 and 12th at OMSI. It was a lot of fun showing off the vehicle, especially now that it hovers. Thanks to everyone who came out and checked out the project. I had a lot of great conversations that weekend.

After getting back to the shop and making sure everything was in working order, it was time to get back to work. The first order of business was to finalize some of the hardware. While Goliath hovered, the center of gravity didn't seem to be in the most ideal location. In order to remedy this, all of the hardware needs to be finalized with the flight weight components, particularly the remaining steel pulleys.

This was the same process as before, but since these are idlers and tensioners, no holes are needed for bolts. The last of of these components are complete, with each saving around a pound of weight. There is now only one one steel pulley left on the vehicle, the main engine pulley.

There was also a hiccup with the battery and solenoid. The battery died, likely because it was undersized and was being deeply discharged. A new bigger battery was added with 18Ah, 80% more capacity than the previous battery, but with only 4 more lbs. Some thought was given to switching from the lead AGM type to a Lithium based battery, which would save about 10 lbs. However, the AGM has been demonstrated to work with the high vibration environments.

The solenoid failed again, the second on that's failed on Goliath. It's obvious that the stock lawnmower ones aren't built for the vibrations that Goliath creates. It was replaced with another stock solenoid as a temporary fix, but a heavy duty one needs to be sourced for the future.

Lastly, the temporary Avionics Tray was replaced with a permanent one. This time it was made from 16 gauge steel. It weighs a little more than a pound. The idea is that the added mass will help to dampen the vibrations. The tray is isolated from the frame. using rubber expansion nuts.

I'll have more details in a future log post, but the great news is that between the new engine mounts and the avionics tray have reduced the vibrations that the Pixhawk experiences low enough for it to work now.

Stay tuned for more details.

Over Labor Day weekend the assembly of the Mk. II vehicle was completed. The new vehicle was weighed and the current weight is 170 lbs, 50 lbs lighter than the previous vehicle. The first couple of tests were conducted and Goliath has hovered for the first time!

There is obviously a lot more to be done, but hovering is a big step in the right direction.

Also, if you're in the Portland area this weekend, come see Goliath at the Maker Faire!

Work has continued on the Mk II vehicle and it's nearly complete. The latest work has been attaching the idler pulleys, tensioners and the battery.

The ilder pulleys were mounted to same 1/8" plate used for the engine mounting plate. Aluminum angle was riveted around the edges and the plates were bolted to the frame.

Below, the two idler pulleys for the single sided belt are attached.

The center idler for the double sided belt required a different approach. The upper attachment was already part of the engine mounting plate. The lower attachment was made from a smaller piece of plate and riveted to two cross members that were added to the frame. Below is a shot looking at the underside of the frame where the cross members were added.

The other idler for the double sided belt was attached using two different methods. On the lower side, the aluminum plate with angle was used and bolted to the frame. For the upper side just a plate was used was bolted to the engine mounting plate, using the same bolts for the engine mount.

With the idler pulleys all attached and the tensioner attachment method decided upon, some additional gussets were added to reinforce the aft end of the vehicle.

With the aft gussets attached, the exhaust pipes could mounted.

Not shown are the flexible mounts scavenged from the original frame.

Next a tray for the battery was created from angle pieces.

A rubber liner was added to the bottom of the tray.

To keep the battery in place an additional piece of angle was bolted in place on top of the battery.
Now all that remains to complete the Mk. II vehicle is to add to the electronics and the fuel tank. With some luck, the vehicle could be complete and running by Labor Day.

Stay tuned...

Goliath Mk. II will be at the Portland Mini Maker Faire on Saturday and Sunday, September 10th and 11th at OMSI. This will be the first time the Mk. II vehicle will be shown in public. I'm working on completing the vehicle and having its first test before the Maker Faire. The vehicle will not be run at the Faire. Look for it at the McCloud Aero Corp booth.

More details can be found at:

https://www.facebook.com/makerfairepdx/

This last weekend was productive and the major portions of the frame are now complete.

After attaching the engine mounting plate to the frame, the next step was assembling the lower frame. The lower frame is similar to the lower ring of the upper frame, so it was possible to reuse the jig for the upper frame with some minor modifications. This was simply making new wood rotor shaft mounts for the lower frame and swapping them out in place of the old ones. Then all of the lower frame elements were cut and placed in the jig.

With all of the elements in place, the same process for riveting the gussets was followed.

Next the lower rotor shaft mounts were machined out of metal. There was a couple of hiccups due to the double sided tape not holding and bit breaking.

Once the mounts were completed, the shaft mounts were attached with bolts to the lower frame in the same manner of the upper frame.
In the future the gussets would be attached to the top side of the lower frame. However, at this stage the method of attaching the idler pulleys and tensioners hadn't been finalized. Therefore the upper gussets were left off at this point.

With the two halves completed, it was time to start attaching them together. Both halves of the frame were attached to the rotor axles to hold the two halves relative to each other. It was easiest to work with the frame flipped over since most of the riveting was on the bottom.

Four lengths of tubing were initially cut and attached on the sides of the frame along with the gussets for the lower frame. Clecos were used to hold the parts together prior to riveting.

Next the lower cross members were added to the lower frame

With all of the new elements attached, it was time to riveting the parts and finalized the attachment between the frames. At that point the frame was flipped back over to see the progress.

At this point about 80% of the frame is complete. the frame members at the front of back need to be added and some smaller cross members need to be added to stiffen the frame. However, before completing the frame elements, the attachment hardware needs to be completed to make sure there aren't any issues with the remaining structure. Once the attachment hardware has been finalized, then the frame can be finalized.