The purpose of this posting will be to document the building of a video capable low vibration quad with FPV/OSD that flies for 1.5 hours.
- flight path range > 15 km (2km radius)
- max speed > 30 kph (capable of handling moderate winds)
- GoPro camera/video on gimbal
- optimal designed vibration dampened electronics platform
- fits in a suitcase
- quiet (will not disturb animal life)
It will document the following in installments:
- design validation
- frame+motor mount build
- propulsion system build
- vibration optimization method, test, and analysis
- vibration dampened EP build
- electronics build
- gimbal build
- battery optimization & build
- propulsion system optimization
- flight tuning
- flight & FPV test
- video test
Replies
Results of these analysis will be interesting. I use a cheap xcam vibration !easurememt device to tune the dampening of my gimbal. So far I had best result with the secraft silicon rings damping system.
put one on order. i'll test it too.
really liked your post on your test. will have to take a longer look and see how it was done.
Step I: Design Validation (does design satisfy the criteria)
1) Fits in suitcase - This is a plan view CAD drawing of the ship frame and VDEP support base along with approximate location of the forward GoPro and aft GPS. The ship was made as wide as possible to fit into a standard 19" wide x 26" tall suitcase.
The white area is the inside of the suitcase. With motors (or props) removed and safely placed in foam pockets, the ship can flex under high load enough to survive luggage handling. The size of the quad (between motor axles) is rectangular at 24" wide x 17.437" long.
2) Quiet - With 17.25" or 16" props (the two contenders), the spacing between props is 0.187"/6.75" and 1.437"/8" respectively. The motors are either 268KV or 380KV. The 16" props will be the quieter of the two sets, but either set produces low db at the above stated spacing.
3) Flight Path Range of > 15 km - This was demonstrated while breaking the world duration record. The most efficient flight is at approximately 10 kph. Props/motors to be similar to Rufous (the duration ship).
4) Max Speed > 30 kph - The battery chosen in the Preliminary Design Review was a 6S6P LiIon. At 5 amps max per series, the 6P battery allows about 30 amps for the ship without chemical collapse. This puts about 7 amps to each motor. Hover is at 1.8 amps per motor. So the ship will be able to accelerate at about 3 x gravity until drag overcomes the acceleration. From prior knowledge, 2g acceleration is enough to counter moderate winds.
5) GoPro Camera on Gimbal w/ 90 degree FOV - GoPro gimbals are less than 4" wide x 4" long. The worst case forward spacing between props has 2.75" to spare. According to the calculations in the Excel Custom Quad CAD, with the camera 9" in front of center as shown, the FOV = 90.8 degrees (the GoPro has an approximate 90 degree FOV).
6) Optimal Tuned Vibration Dampened Electronics Platform (VDEP) - After the frame is built, motors mounted, electronics attached, one Pixhawk will be mounted on the VDEP base and and a second Pixhawk mounted on the VDEP. Different dampeners and quantities will be placed between the two levels, brought up to different flight vibrations, with the vibrations compared for a variety of weights. The VDEP will contain the mass of the battery, gimbal, and other electronics (mass being the best damper). The optimal set/quantity of dampers will be chosen from the tests. The test will focus on 3D offset-compression (conventional and proven) style dampers. 3D Trampoline style dampers may or may not be tried. Thus while tuning will be optimal, design may or may not be optimal.
7) Other Design Criteria
- GPS/Mag located more than 2" aft of the battery to minimize magnetic interference (see box in back of drawing); longer wire cables needed between Pixhawk and GPS/Mag.
- Total ship weight less battery < 1026g (not easy but doable)
- Stiff frame (using .6" diameter carbon tubes; same as duration ship; has proved to be quite stiff/light and crash resistant)
- PixHawk bonded to frame for precise control (the area in the center of the drawing is the pocket for the PixHawk).
Here is the latest Custom Quad CAD Excel worksheet/macro. This sheet along with the one for Hexa and Octa are in different states of development and repair. But I'll try to keep revisions posted here. Make sure that your enable macros if you want the macro to work. Otherwise, it's not important.
Custom Quad CAD.xlsm
Looking forward to seeing the writeup for the finished ship.
Step V(a) - Gimbal Design
Before the ship design can be validated, the gimbal needs to be finished to see if it can fit within the constraints of the ship design.
This is an iterative process. Different motors are reviewed to see how they would fit. One was chosen with a low profile, resulting in the following preliminary design (PD). This design meets the objective of all axis of the gimbal being centered on the front plane of the CMOS in a GoPro 4 Black. This is contrary to most gimbal designs that align the:
- roll axis to the CG of pitch axis.
- the pitch axis to the CG of the camera/platform.
Yaw stability is not impacted by gravity, thus balance is less critical and can be easily accomplished by the location of the roll motor.
This is the derived design.
The next step in gimbal design verification is to see how far it is off balance. This excel worksheet calculates torque on the roll and pitch axis. These values are in inch-grams. So for example,
o the CG of the GoPro 4 is .215" forward of the pitch axis. This is because the CMOS lines up with the pitch axis to prevent axial motion coupling with the real location of the camera (see earlier explanation). The mass is 86.4 grams. Thus the torque is -19 inch-grams (7 cm-g) from the camera alone.
o the roll axis can be perfectly balanced by locating the motor to adjust for the out-of-balance GoPro (the CMOS is located on the front upper right of the camera. But as the camera rolls, the CG changes because there is a two dimensional deviation with the camera. Thus it can get out of balance up to 8 in-grams (maybe not enough to cause tremors).
The next step it to build the gimbal and test. Math will get me close, but the proof is ...
Now I'm patiently waiting for the gimbals, motors, and controllers.
The gimbal to beat is the DYS 3-axis SMART. Bought one. Took it apart as best I could. Noted weights and good design.
Weights are as follows (some weights are estimated because I'd have to pull wires and didn't want to do that).
So the package is about 265 grams actual weight and my target is 175 grams or a reduction of 90ish grams.
Interesting design notes:
- the platforms had cutout T-slots to clips wire runs.
- the dampers were different than i've tested. similar in size to the GoPro but gray, Japanese lantern shape, 2x stiffer, and a bit lighter in weight.
- the Camera fixing method is a bit heavy, but thoughtful (grips the lens, the most important part).
The realization about gimbals has caused me to reassess the design (after the frame was built, of course).
We spend a lot of test and design time trying to minimize vibrations to the camera. Later in this blog, tests are performed that accomplish huge attenuation and tells you how to do that. But then we fit the camera onto a gimbal, that induces blur. So what's the point?
When we mount the gimbal to the ship, it is never done in a manner that puts the camera CMOS/CCD sensor, photo sensor (PS), at the center of gravity (CG) of the ship. Also, it is rarely done in a manner that puts the yaw, roll, and pitch axis at the center of PS. The impact is shown in the illustration below. A -2 degree pitch in the ship is corrected by a 2ish degree reaction of the pitch axis of the gimbal. Note that:
- the camera view is still parallel with the old view after the counter move of the gimbal,
- but that the camera, when about 4" from the CG, has moved in y (0.04") and z (0.13").
- camera auto focus handles the .04" delta in y.
- a blur up to 0.13" is induced into the photo. The longer the shutter speed, the larger the impact.
- that if the PS was at CG, no such movement would take place during a pitch with everything else held constant.
This happens during a Yaw, Roll, or Pitch correction, when the PS is not perfectly aligned with the ship x/y/z CG and each gimbal yaw, roll, and pitch axis.
In design, the following exasperates the condition:
- roll axis not directly centered on the PS
- pitch axis not directly centered on the PS
- yaw axis not directly centered on the PS
- the PO not directly centered on the CG
This means that i'm going to need a bigger suitcase. Note how the PS is exactly at the x/y/z CG of the ship and the gimbal axis are directly lined up the same (on the GoPro, the PS is about .2" from the back, slightly up and left. Also note how a lens can change the direction of light but are not what needs to be lined up with the CG (the PS is).
Back to build. Oh well.
The two videos show what this dislocation coupling does when the camera's eye (CMOS/CCD) is not located at CG.
- tuning of the gimbal PIDs help the camera minimize angular follow error--camera angle in yaw, roll, and pitch.
- dislocation offset, however, deals with how yaw/roll/pitch corrections couple with camera location to dislocate the eye--side to side and up/down and fore/aft motion of the camera relative to its target.
- thus, in the following video, we only care about the side to side motion. A reference point is over-layed onto the video that remains stationary so our eyes can see the dislocation. The reference point is located in about the same spot in the camera frame in both cases.
In this first case, the eye is offset about 50 mm from CG and the ship is rolled around its CG. Few eyes are located closer than this to CG.
https://www.youtube.com/watch?v=2ndOw3M7E6Q
In this case, the eye is located less than 3 mm from CG. Practically, this is probably about as close as one can design the CG to eye relationship unless further refinement is done post build. As can be seen, with the amount of roll induced, the more centered eye had 2 to 3 times less horizontal movement.
https://www.youtube.com/watch?v=3IGWV2zGARE
The eye chart is used because a video of a far away horizon does not show this affect as well as a closeup of a eye chart. Yet the impact of the video is the same. The brain has a remarkable way of smoothing out video. Yet it is clear that shots taken from the air are no where near shots taken from a tripod. This is one more element to be solved in getting aerial shots closer to that of shots taken from a tripod. Then our videos will start to show crisp detail.
Took a look at the DYS 3-Axis SMART GoPro Gimbal to see how it lined up to the GoPro CMOS sensor (further on there is a full analysis of weights also).
Mis-alignments to the CMOS are (this is done primarily by the designers to control CG)
Yaw - 0.68" (17mm)
Roll side - 0.25" (6 mm)
Roll height - 0.26" (7 mm)
Pitch lead - 0.10" (2 mm)
Total - 0.78" (20 mm)
This equates to, per degree of total rotation, sin(1) x .78" = .0136" or
a maximum blur of 6% of the CMOS (approx unsharpening of 280 pixels).
The camera wouldn't see this much when the shutter rate is faster that 1 degree of corrective motion. More likely the unsharpening is:
- Frame rate of 1/60 per second
- Angular correction of 1 degree per 1/40 second
- or 2/3 the above
- or an unsharpening of 186 pixels
This problem is mostly with the GoPro since it's sensor is not located at it's CG. The problem appears to be serious enough to try and avert. It may be better to add weight to balance a perfectly aligned gimbal. But this too is a trade as more weight means larger vibrations. Thus try and find out.
-
1
-
2
-
3
-
4
-
5
of 5 Next