1h:15min of flight time for Airbotservices'X4 equipped with brushless gimbal & fullHD camera on its 12mm carbon fiber payload rail system.

Attempts have been made reaching more or less 2 hours of flight time on "empty" frames: such crafts carry the absolute bare minimum required to minimize weight, meaning: they do not carry payload(s), and/or do not have extra electronics for autonomous/waypoint capability and/or do not carry optional bi-directional command & monitoring telemetry.

"X4" is an Airbotservices experiment UAV attempting to push the limits of flight autonomy of a "practical" drone : we tried thus something different with a fully equipped drone (BG & full HD camera), in addition having all APM/Arducopter capabilities. It carries the following hardware equipment:

  • Airbot's payload rail system composed of two parallel 12 mm carbon fiber masts
  • Airbot's 2-axis brushless gimbal (Airbot's world's lightest at 45g only) stabilized using an Alexmos controller
  • full HD Mobius Camera
  • 3DRobotics' Pixhawk (APM Arducopter)
  • 3DR Robotics real-time telemetry (bi-directional TX/RX)
  • GPS (M8N) and second external magnetometer (in addition to internal Pixhawk's magnetometer)
  • 3DR power module powering Pixhawk and used for current and voltage measurements
  • Airbot's LiIon 23100 mah 4S battery
  • Power circuit/cabling for the BG

The frame base is built inspired by Forrest Frantz's "CF round tube" constructions techniques (for an extensive discussion and how-to guide on these innovative techniques, read post here). 

Some Observations , lessons learned and potential improvements:

  • Weather status: 50% of time without wind, 50% with ~20km/h breezes

  • Initial battery state : charged at 4.18V at beginning

  • Brief Loiter (un)stability hickups. APM version 3.2.1 was used , loiter mode for 95% of the flight. The weather conditions were relatively calm but still with wind bursts once in a while. Even during "no wind" periods of time, "loiter" was having some unexplained, brief (10 seconds or so) strong twitches (or spasms) as if the quad was fighting very strong (but unexistent) winds. I repeated the same experience on Airbotservices X8, totally different craft, same observations. The three factors that impact loiter's performance (low vibrations, low compass interferences, good GPS performance) were all excellent on X4 & X8:
    • (Vibrations less than +-0,2g;
    • External Compass with very low offsets and far from EMI;
    • More than 8 satellites and HDOP<2). M8N on X4, 3DR LEA on X8 (no GPS HW suspicion)
    • EKF was enabled.

         P Loiter and rate P Loiter were tested with lower values (0,7 and 0,8 and 0,9)  than defaults (=1) with no sensible improvement on loiter's stability. 0,7 value resulted in a bigger unstability versus the default value.

           Any observations on what could be wrong, analyzing the log file posted below, are welcome....

  • Pixhawk and/or APM Arducopter's altitude management : should be way better than what it is. I already posted a few times about this, like here, and other people have confirmed my observations. This 1h15 flight really shows how bad altitude management is, with measured altitudes (displayed in MP) that are two or three times higher than real altitude (after a few minutes). Altitutde drifts is a nasty problem for (long) auto missions and I hope developers will now be convinced this should be set in the high priorities in the list of change requests for future releases (further, according to a user who got some feedback from 3DR, it seems 3DR does not consider an altitude drift issue even exists; cf support forum link above). I know Randy has confirmed an altitude EKF improvement about this in 3.3 but I'll let him chime here as I might be wrong. There are two potential different drift sources (or both may also even be concomitant), and I do not know which ones are at play : either an hardware issue due to Pixhawk's electronics (baro chip and/or how it is electronically exploited); either a software issue (different sw were used in this test:  APM /w EKF , Mission planner). To finish, with a detail on altitude problems, APM documentation wiki needs a small correction as it states that altitude is calibrated to zero when plugging the battery when it fact, in version 3.2.1 anyway, it is calibrated (and reset to zero) when arming.

  • Mechanical vibrations fatigue. As can be observed in the attached video of this blog, Mobius camera detached itself (pitch mount) due to mechanical vibration fatigue. Mobius cam was simply snapped on with a plastic cap. Similarly to loctite to maintain screws tight against vibrations, adhoc glue should have been used to secure the plastic cap "snap on" mechanism (careful, never to use loctite on plastic parts as it make it brittle).

  • A very positive and conclusive result for "screwless" frames. Airbotservices X4 is built screwless, only using glue between parts, including the 12mm carbon fiber rail system that carries the heavy battery. Although this craft has been flown many hours (each time more than one hour at a time), none of the glued parts have budged nor cracked. Use of glue is key to get a light craft and get such performance (after all, Airbus and Boeing evolved too from rivets to glue in order to lighten their planes). I did not do any structural resistance measurements but I would bet that such frame is way stronger than any screws & bolts alternatives (Forrest might chime here as he 's the expert on the subject).

  • Final battery status : voltage =3.3V per cell (plugged in, motors stopped) and =3.1V (under charge, motors running). A LiIon cell is fully empty at 2.5V per cell. Stopping at 3.1V, 20% of energy remained in the battery. It is OK to drain a LiIon battery down to 2.75V per cell under charge (=11V for a 4S battery). Extending the flight with a few more minutes was thus possible (better be careful not to damage the battery).

Mission planner's log file is downloadable here.

There you have it, a fun journey in extreme flight duration territory,




E-mail me when people leave their comments –

You need to be a member of diydrones to add comments!

Join diydrones


  • This is how i joined tubes on a different frame.  Once joined and bonded I mounted it to a plate with four non-ferrous screws.  Nylon bolds would work just as well.  


  • MR60

    Well done Todd.

    Reminds me of trying something similar with square tubing using a 3-axis router.


    Hugues - Did you join the carbon tubes in the middle using some variant of this? And do you remember the frame weight to be able to get to 1+ hours?


  • Great job! Testament to how much the tech has developed in the last couple of years. Few years ago I was fixated on going screw-less as well.   I tinkered around with a few ideas, but the hardware tech wasn't where its at today (obvious from hardware in pict).  I got it down to to the bare minimum of non-ferrous screws.  Even went so far as to bore out the arms, lol!


  • MR60

    Great to see a real duration flight.

    What size props? And any guess as to what size prop is required to achieve 1 hour?

  • Hughes, great post, thank you for sharing. I have a couple of questions if you don't mind taking a look:

    I was hoping to find a little more out about the battery. It seems to have an excellent energy density. I presume it's made of 18650 cells? Was this made or bought? If you could provide a link to store or any details on how to replicate your battery that would be much appreciated.

    The other questions relates to the choice of 4S. Did you calculate that this would be the most efficient set up? I've been under the impression that higher voltages and lower KV motors provide higher efficiency. Would you have been better off using 6S for example?

    Thanks, Tom
  • T3


    unfortunately I had no time to look into the details so far. Not sure about Paul's patch but Holger's patch implements some progressive long term integration of GPS alt to correct baro alt. Hi hope to find some time next week for testing. I'll report.

  • MR60

    Hi Ravi, 380KV and about 14-15 amps.

  • what was the current at loiter.

  • what is the KV of the motors used. T-motor 3508 comes in many KV. thanx

  • MR60

    @Thorsten, I like your simplicity approach. Does the patch you mention (Holger and/or Paul Riseborough's patch) implements the logic that Randy suggests (progressive long term integration of GPS alt to correct baro alt) ?

    I also agree that we do not care about absolute altitudes measured by the baro chip (so we do not care if it varies by 15 meters); it is rather the relative altitude (for ex v.s. the home alt) that is important (in my applications anyway.I might be wrong, but I believe it is also the case in most other applications. A ground offset can be defined and terrain's height adjustments can work on relative altitudes).

    Felix's idea of using the baro chip in the pixhawk used in an antenna tracker ground station could be used to get a more precise relative altitude, without any extra hardware or complexity ont he craft (but obviously would require extra processing software routines on the fc board).

    Thorsten, please keep posting here about your Holgers patch shall be interesting (and if I may abuse, if it really works, a tutorial to install/use it ?).

This reply was deleted.