Katie "Ket" DeSpain and Forrest Frantz teamed up to take back the flight duration world record. A Romanian group beat the previous record held by Forrest with an excellent timed flight of 2hr 7min. But, they didn't disclose how they did it. That provided the motivation to retake the record and fully disclose how we did it so industry can benefit from the research. By going lighter, performing proper analysis, and using newer technology, the record now stands at 4 1/4 hours.
The following details the build so that industry can replicate the build methods and technologies used. All aspects of the ship is reviewed in detail and shared below. While we tried to cover all new aspects of the design and build, please feel free to ask for more information if needed. When we say full disclosure, we mean full disclosure. We want to help.
I - Electronics Selection
II - Rotor Testing (Motor, Propeller, and ESC)
III - Battery Test (watt capacity at targeted draw) and Build
IV - Frame Design and Build
V - Electrical Design and Build
VI - Flight Parameters
VII - Results
Summary - Two quads using 29" props. See photo of Condor and Ket above:
- Condor (used to push the limits of technology) and
- Caracara (commercial grade)
They were (and will be) tested in various flight types with the results shown below.
- Condor--Near ground hover 4 hours 14 minutes and 34 seconds (for world record)
- Caracara--3-meter hover TBD (but expect about 3 hours; update coming soon)
- Caracara--Distance flight in full autonomous (point to point) mode TBD (but expect 3+ hours; update coming soon).
I - Electronics Selection (FCU and Radio).
While a smaller F4 FCU would be the best choice for a duration record, the PixRacer by mRobotics proved to be the best choice. The PixRacer:
- weighs only 12 grams naked
- has full autonomous capability
- stable control channels at 6S (a different mfg product that we tried kept blowing control channels at 6S but worked fine at 4S)
- has a small footprint
- uses a small Power Distribution Board (PDB)
The best radio might be a one gram satellite for max duration, but we chose a 5 gram 4XR FrSky because distance too is a reality of a good duration vessel and the 4XR is a proven Tx/Rx.
No telemetry radio antennas (didn't want to get too crazy) on the Condor but will use FPV telemetry on the CaraCara to increase its reality for the distance test. No gimbals or camera (ditto) on the Condor but will have a FPV camera on the CaraCara. To “add” equipment to a ship designed like the CaraCara, 20 grams costs about 1 minute of flight time.
Total weights for electronics (FCU, Rx, and GPS removed for the Condor), a 21 gram total, were as follows for the Condor:
Figure 1—Weight of Electronics
II - Rotor Testing (Motor, Propeller, and ESC). It's all about figuring out which rotor set can lift itself first and then start lifting for duration--hence the term net-lift efficiency (thrust in grams less the weight of the motor, propeller, the frame size that has to spread out those propellers, and the ESC relative to the watts required to perform that net-lift but exclusive of battery weigtht).
A test stand was specifically created to derive the required data for rotor efficiency and response: Amps, Volts, Thrust, Motor Temperature versus time. See Figure 2a.
Figure 2a—Test Stand
Using the test stand, we purchased a plethora of props, motors, and ESCs to test efficiency in various combinations, which resulted in a large quantity of tests.
Figure 2 shows a list of the equipment tested in various combinations. Because the system is evaluating based on net-lift efficiency, all are weighed, usually naked (void of unneeded or not-right-sized accessories and to account for mounting method, etc.).
Figure 2b –Tested Item List
It's important to compare apples to apples so all the tests used the same standard. That standard was controlled/enforced by a computer, in this case an Arduino Mega. It had the following attributes:
- allow folding props to unfold before the test begins.
- warm up before the test and cool-down after the test so the next test starts at a thermal condition that will be found in actual flight.
- hold at each PWM level long enough to get a significant average of thrust, amps, volts, and temperature.
- be able to automatically feed that data into Excel or other analysis that is standardized.
- account for differences in the thrust tare reading at the start and end of the test.
- take into account the mass of the rotor system that needs to lift itself before delivering any net-lift (usable lift).
The results in Excel, might look something like what is shown in Figure 2c - Efficiency Tests.
- The first four columns are the last test’s inputs from the test stand.
- Columns 5-7 are net-lift efficiency, net thrust, and the sample size of that latest test.
- Column 8-12 are the archives of prior test data, and column 13 is the average of columns 8-12
Figure 2c – Efficiency Tests
From that data, the following rotor system was chosen.
- Motors T-Motor Anti-Gravity MN7005-32 KV115
- Propellers T-Motor G 29x9.5
- ESCs DYS 40A MB30040 Opto Multirotor
As a general note about the selection:
- In many cases, T-Motor props are more efficient than any other prop on the market. The only general exception is with props less than 12" to 15".
- In most all cases for efficiency, response, or speed, T-Motor motors have proven superior, even when copied. This includes small, medium, and large motors and motors used for FPV and high speed.
- For efficiency, the DYS 40A MB30040 has for years proven superior even when compared to the new ESCs with braking, sync, high clock rates, and all the other latest and fancy new things (and yes, we are constantly updating our tests). For some reason, the DYS ESC has less watt loss in the electronics and gets more to the motors. The only exception on selection is when clocking for speed versus efficiency.
A peculiar note on the T-Motor props. T-Motor, in the 29” format makes two different props:
- Glossy 9.5” pitch
- Polished 8.7” pitch
The T-Motor website claims that the P29x8.7 is for longer flight time (and higher loads). Our tests showed the opposite. Were our quasi-static tests not reflective of real flight? Or, did T-Motor make a theoretical versus tested statement? We plan to test these two props again in Distance flight to find out.
The final weights for the Rotor system was as follows for the Condor.
Figure 2d—Rotor System Weights
III - Battery Test (watt capacity at targeted draw) and Build. There are three parts to this process:
- Test, and
In this explanation, the terms ‘Cell’, ‘Pack’, and ‘Battery’ will have the following meaning:
- Cell--one off the shelf battery (e.g., what people might call a single AA or 18650 size battery).
- Pack--a group of cells in parallel (e.g., in a photo, all the cells connecting to the same electrode that is facing up)
- Battery--a series of packs. Thus a 6S9P battery is composed of:
... Packs of 9 Cells each connected in parallel (+ to + and - to -)
... 6 Packs connected in series (+ to -)
... For a total of 54 cells
IIIA - Cell/Battery Analysis. Using preliminary data on the battery, we:
- Measured the mass of each Cell.
- Added the mass of connecting wires, tabs, battery structure, containment (about 2% additional weight).
- Made a table of the relationship between Pack volts & mWh (iterative as mWh is a function of draw).
- Developed an equation to convert ship weight to the minimum voltage required to keep the ship aloft.
- Used the rotor efficiency data to calculate the watts that the ship will consume at various loads.
- Used all of that data to try different battery configurations to determine the optimal theoretical battery.
Figure 3a shows what the worksheet might look like.
Figure 3a—Ship Duration Calculations
The battery analysis and test are an iterative process that will finally tell you the optimal battery configuration, the S (Series) size and the P (Parallel) size. The lower the S, the more efficient the ship. But lower S runs into the voltage limits of keeping the ship in the air. Usable battery capacity might become an issue. As seen above, a 5S10P battery can only use 10.1 Wh of the battery while a 6S9P can use 10.44 Wh. Then, as P increases, the mass of the ship increases, which requires more wattage to fly the ship, can lower rotor efficiency, and can lower the effective capacity of the battery. So, we needed to investigate the trade-offs, which the above worksheet does.
IIIB - Battery Test. The purpose of battery test is to find a battery that, at the predicted amp rate of draw, delivers the greatest watts per gram. The prerequisite for doing this test is knowing the draw rate from the analysis. This is key because battery capacity is affected by the draw rate (as draw goes up, capacity goes down). All tests for different cells are conducted at the rate appropriate for that cell. But, in general, batteries of similar chemistry have similar draw rates. For example, Li-Po, Li-Co, Li-Mn, Li (non-rechargeable) all resulted in similar draw rates. But it is important to make sure this is true for the battery used.
For the battery tests, we used a West Mountain CBA battery tester. Ideally, for ship testing, one needs to keep watts constant (the CBA can only keep amps constant). But keeping amps constant with this tester gave a close enough answer.
The West Mountain CBA battery tester comes with software for selecting:
- End voltage limit (minimum limit for the battery)
- Draw rate (the amp level drawing down the battery)
Figure 3b—Radio CBA Battery Tester
The data and graph visible during the test is shown in Figure 3c.
Figure 3c—Battery Test
IIIC - Battery Build. We used super glue to assemble the battery, as suggested by our associate Hugues. The assembly process is as follows:
- Put the battery cells into the tool (see below photo of the tools and the saddle to position the battery)
- Test each battery cell to ensure that its voltage is consistent with the others (within .3 volts)
- Mark each battery on its top edge at interface points (where it touches another battery)
- Remove the battery (put a tool under the wrap at the top edge and lift)
- Add super glue to each interface line
- Quickly reinsert the battery slightly rotated so the adhesive doesn't touch the other batteries
- once nearly completely inserted, rotate the battery correctly and push all the way down
- when cured:
... Cut away the plastic wrapper from edge (negative terminal) where the bridges will be soldered
... Lightly sand the bridge areas and positive terminals then vacuum away grit and wipe with alcohol
... Remove the assembly tools
Figure 3d—Battery Assembly Tool
Make and complete one half (3S9P) with balance charge
- Lightly sand tops and bottoms with fine sandpaper
- Clean with alcohol
- Add the large wires first
... Use multi-strand wire
... Remove the insulation
... Flatten the end to make a fan shape terminus
... Solder paste the end of the battery
... Solder the wire to the battery
... Once soldered, the battery can be accidentally shorted. To prevent this, only work with one pack at a time and cover leads not being worked on.
- Clear top edges of negative ends of wrap where contact is
- Syringe a dab of solder paste, the type with solder as an ingredient, onto each battery contact point
- Melt the dab into a pool of solder, adding additional solder if needed
- Bridge the negative dabs with solder, or solder a copper wire across the gap
- Use bare copper to bridge the positive leads
... Paste at end of copper wire
... Solder onto a dab at the battery
... Route the wire and cut to length
... Complete for the other batteries
... Add solder paste to the terminating batter and solder all lead ends at once
- File/cut off any sharp solder peaks
- Test the cell, pack, and battery voltages
Figure 3e—Battery Solder
If it checks out, put Kapton tape of other insulator to protect the ends. Then add damping pads to help keep vibrations from weakening the battery solder joints.
Figure 3f—Battery Insulation
Replicate each half and put into a series and mounted on the ship. In Figure 3d there is a saddle glued to the motor spar to keep the battery from shifting. Velcro straps are used to secure the batteries to the EP.
Figure 3h shows the battery and wires from the battery to the PDB and from the packs to the balance charge plug.
Figure 3g—Battery Model in Solidworks
Figure 3h—Battery Installed
In ongoing testing conducted after flight tests began, we came across a Chinese battery used by the growing Vaping industry (popular in China). The EFest IMR 18650 proved to be better than the Panasonic Li-Ion 18650B. In actual flight tests, the Panasonic flew for 3 hours and 10 minutes, while the Efest flew for 4 1/4 hours. Mathematics from testing does not explain this difference in flight time. We are researching to discover the cause.
Figure 3i—Capacity (Watt-Hrs) vs Time Discharged
The weight for the battery is shown in Figure 3j.
Figure 3j—Battery Weight
The EFest batteries on the Caracara look like this. Also note that on the Caracara, a battery cage (bat wings) were added to keep the batteries from moving in rough turbulence over long periods of time.
Figure 3k—Efest 6S9P Battery on Caracara
IV - Frame Design and Build. The ship used the same construction techniques discussed in Forrest’s earlier posts on building custom ships. Rather than repeat it here, please review:
Introduction - Design
Building with Round Tubes
Building Strong and Light
Gussets and Sandwich Panels
For this world record, we designed and built a simple and lightweight quad. For strength and weight, we used continuous tubes for the motor masts. To do that without adding weight, we offset the tubes in z (vertical) and welded to an Electronics Platform (EP), which is a 6mm thick carbon/Nomex floor panel. We welded one motor mast to the top of the EP and the other to the bottom of the EP. See side view in Figure 4a.
Ket did all of the Solid Works Designs.
Figure 4a—Side View of Condor Design
Figure 4b—View from Below of Electronics Platform
Figure 4c—Side View from Below of Condor Design
Figure Top View of Condor Design
Another critical aspect of ship design is designing a light-weight method of keeping the motors on the ship—motor mounts.
For a motor this large, we estimated that 1.6” weld was needed to keep the motors attached to the motor masts relative to shear forces and 9g-crash loads.
For peal forces, we employed two design methods:
- The end-cap (for keeping the tube shape stable) also has the function of holding down the motor mount.
- The motor mounts have holes in the center near each end so that the weld can seep into these holes, creating a stronger barrier against peal.
The design of the motor-mount and end-cap are shown in Figures 4e-g.
Figure 4e—Motor Mount
The holes in the end-caps are to release air pressure inside the tubes. On a future build of something similar, I’d alter the design of two of the end-caps to make the edge margin at the bottom extend farther downward, by the thickness of the motor mast tube and EP sandwich panel. This would allow the other two end-caps to also act as landing gear. The ship would be more level before power is applied. Otherwise, landing gear is not needed (in controlled take-off and landing platforms) because the motor mast tubes fully protect the electronics that are tucked up next to them. With a level landing pad, the motor masts do not touch on the CaraCara (on the Condor, the tubes are flexible enough that the battery bends the tubes in the middle so they do touch) because of the edge margin on the end-caps.
Figure 4g—Motor Mount and End Cap
Frame weights are shown in Figure 4h.
Figure 4h—Frame Weights
Caracara, AUW 4.06 kg, is heavier than Condor, AWG 3.96 kg, by about 100 grams. The primary differences in the ships are:
Feature Condor Caracara
- rigidity loose stiff
- motor mast Rockwest 45026 0.59” OD Rockwest 45001 .935” OD
- ship size 1092 mm 1060 mm
- landing gear yes no
- GPS no yes
- FPV/OSD no yes
The frame design has a huge impact on vibrations transmitted from the motors/props to the FCU, which impacts the signal to noise ratio for flight control. This following shows the results for the Condor.
Figure 4i—Vibrations Noise at the FCU
And the resulting ship control while seeing almost no signal noise. We will also publish the results of the Caracara, which is a much stiffer ship.
Figure 4j—Ship Control at Hover
V - Electrical Design and Build. This includes wire routing, tie-offs, location, and orientation or electronics.
The first photo of the ship (Figure 5a) shows the location of the:
- Props and motors at the ends of the motor masts (also Figure 5a)
- ESCs mounted on the sides of the motor masts under the prop wash for cooling
- GPS located away from the batteries but out from under the props
- EP, where the batteries are normally situated in the center of the ship
- Bolts that “hang” the PDB and FCU to the underside of the EP
- Routing of the wires from the motor to ESC and ESC to PDB
Figure 5b shows the FCU and PDB.
Figure 5b—FCU and PDB
Figure 5c shows the ESC mounted on the side of the motor masts and wires to the motors and PDB and FCU. The wires are magnetic wires for motors. These wires have an epoxy coating to keep them from shorting when in contact with another wire. They are good for zero abrasion situations like when wrapped around a stator or rotor. The yellow tape is Kapton, which is a good insulator and lightweight. On the Caracara, loops were put into the wires:
- coming out of the motor
- leading to the ESC
- coming out of the ESC
In addition, Kapton tape was placed over the wire running along the top of the motor spar to prevent handling abrasion (not shown).
Figure 5c—ESC and Motor Wire
Wire sizing was chosen from the spreadsheet that Forrest developed years ago for aerospace applications, which is available on this blog site.
Wire sizing tables for home and other ground-based environments are based on soft criteria. For example, if one wanted to maximize efficiency of wire used in a home (no heat loss), 120 VAC wires would be a meter in diameter. But that would be difficult to handle, heavy, require thick walls, would not be easy to bend around a corner, and would cost a fortune. So, wire sizing was somewhat arbitrarily set to a low, but not zero, heat loss. For example, 12 AWG is used for 10 amp 120V loads, if the run is over 30’. That results in a low heat loss and voltage drop, and the wire is fairly easy to bend, isn’t heavy, and is reasonably priced. However, one could run 20 amps through a 12 AWG wire without exceeding its ampacity limit (if exceeded, then the wire is exceeding its temperature limit).
But aerospace applications are different. The calculation for wire size can be made precisely. This is because it takes more wattage through a wire to lift a larger wire. Heat loss is the loss seen when running current through a wire. That loss can be precisely calculated in terms of watts. Loss from weight can also be precisely calculated in terms of watts. So as wire size increases, the loss in watts decreases because heat loss is less, but at the same time more wattage is required to lift the wire. Where those two lines intersect is where wire size is optimal (as long as ampacity limits are not exceeded). Figure 5d that shows the optimal wire size for the Condor between the ESC and motors (there are other sections for other wire types). These calculations can be made for all wires on the ship:
- Wires connecting a cell to a cell (on the battery pack)
- Wires connecting a battery pack to a pack in series and the battery to the PDB
- Wires connecting the ESC to the motor and PDB
- In various materials (copper, aluminum, and nickel)
Figure 5d—Wire Sizing Worksheet
VI - Flight Parameters
The following flight parameters were used.
VII - Results
Total ship weight of the Condor as flown for 4 ¼ hours is shown in Figure 7a.
Figure 7a--Condor Weight at Record Setting Configuration
Figure 7b shows the Condor being weighed. For the record flight, the Li-Co (Li-Ion) shown were replaced by the purple Li-Mn batteries and the GPS was removed.
Figure 7b--Condor Being Weighed (Li-Ion batteries & GPS)
Flight tests showed the following.
As a reference, a 300 gram camera/gimbal would reduce duration on the Condor by about 8%. Add to that 100 grams for telemetry, FPV, and other mass, the above times would decrease by about 12%.
Is this Guinness World Record beatable? Let’s hope so. Industry and mankind will greatly benefit.
These results will be updated after the Caracara flights. While the Condor was an excellent test bed for pushing the limits of technology, the Caracara will help give industry a practical example from a commercial grade ship.
To close, after the 4 ¼ hour difficult and tense manual flight, his photo was taken of the pilots, parents, and officials letting off some steam. See https://www.youtube.com/watch?v=_9rJpVSA_G0 to watch the 6 minute video of the flight. From left to right, Pat (site host/emcee), Kathy & Brad (proud parents), Ket (PIC), Bet (witness), Marc (witness), Forrest (PIC), Karen & Ed (witness).
Ket, Condor, and her proud Papa.
Ket, Condor and and her proud mentor. Well done Ket!!! I really don't know if Ket believed me when I told her that if she came an interned with me that she would break a Guinness World Record ... but then again neither of us expected to smash the old record by flying 4 1/4 hours.
Great to hear from you. Tell us about your latest exploits. Still showing the boys how to fly?
Here are our PIDs. Especially note the top row of Stabilize P values. I'll be trying higher Rate Roll/Pitch PD numbers on the CaraCara and leaving the Stabilize numbers alone. Please share what you find along with the size of your props.
I am curious about the PIDs, we usually build large drones and finding the right PIDs is always hard, auto-tune is a great help, but in some cases you have to guess a good starting number before the first flight, otherwise it's gonna end in a crash.
Wooowwww!!!!! Congratulations Guys :O . You never stop surprise Us Forrest, you are a genius :D
........and luky that I live so far or you have my visit every weekend to curious your tests, Lol.
Bustya Attila asked a really good question about finding the PIDs for a ship this large.
Condor - PIDs were important for two reasons that aren't the usual reasons. The usual reason is handling in flight/wind at above normal flight speeds. Because we were so close to the ground and that the ship was fairly flexible, we had to compromise between holding level (so it wouldn't dip and touch ground) and flight efficiency. So the PIDs we settled on were slightly loose relative to normal. One other test that I wanted to do, but didn't, was to see the effect that PIDs have on flight efficiency. Don't even know if well tuned PIDs take more or less energy.
CaraCara - I'm still in the process of tuning this ship. I found that in flight, using Condor PIDs, that the ship wavered +/- 3 degrees, which is too slow for serious flight. So I'm slowly upping them until flight behavior is solid.
Our source for PID tuning of large ships was a blog discussion that I found by searching the web for it (wish I could find it again so I can give credit to who wrote it). There was a key that can be found in the parameter list that we shared. If someone is really curious about this topic, I can dig deeper.
When you get your web site to a point where you are ready to release, I'll post it here too. It will allow easier access for more people to build smarter using a lot of the same tech used to set this world record and the speed world record set by Ryan.
Robert Giordano said:
Of course with Hugues, you never know if his sage advice and suggestions come from him or his brainiac chickens featured in some of this "how to" videos! I'm waiting for the day that his chickens escape in a drone that they built behind their auto-locking top secret doors and fly to their castle in France for a well deserved rest.
... the last 45 minutes ... we were all in shock ... no idea at all ... no mathematics to support it (we are still working on that part).
Near the end of Ket's last turn at PIC (about 2.1 hours in) we saw five battery surges that we can't yet explain. I almost shut down the flight because of safety concerns, but the surges were small enough to control, tests showed that it wouldn't get out of hand, and the judges were behind a net. We only saw one such surge when doing the 1S3P battery capacity tests. The surge was definitely a voltage increase because the ship jumped a few cm (and yawed) each time. If someone knows something about these Li-Mn batteries that explain this behavior, please let us know.
Bustya Attila said:
Congratulations, impressive engineering! This kind of performance drives the industry forward!
The question was not if but when would Forrest break some more limits ! Congratulations and very impressive achievement especially for Kate for a first internship task. Thank you also for mentioning us with our very humble contribution!