Servers by jDrones

Building a Better Quad Frame


(Brad Kauffman) #117

Hi Rob,
I took a look at the tool box you have so far and I have a couple comments.

  1. In the tube comparison you have J=2I which is correct for polar moment of inertia but it is incorrect for calculating torsional stress or deflection in a thin walled section. J for a thin walled closed section is given by: J=4A^2t/S where A is the cross sectional area at the mid line of the wall, t is the wall thickness, and S is the perimeter measured at the mid line of the wall. For the example of an internal dimension of 13mm and outside dimension of 16mm as in the example:

for square:
A=1414=196
t=1
S=14
4=56
J=4196^21/56=2744

for round:
A=14^2pi/4=153.94
t=1
S=14
pi=43.98
J=4153.94^21/43.98=2155.3

This gives a stiffness improvement of 2744/2155.3=1.273 or 27.3%, not 38.2%

It should also be noted that this applies to stiffness only. For strength calculation the sharp corners on a square section will change the failure mode and in composite may well be lower than the circular section.

  1. In the battery calculator it shows a linear relationship. The useful capacity of a Lipo battery drops as the current increases. You can increase the useful energy in a battery by discharging it slower. This is traditionally calculated with the Peukert formula. Here is a paper on the Peukert formula applied to Lipo batteries:
    http://www.mdpi.com/1996-1073/6/11/5625/pdf

The Peukert formula is C=T*I^k where C=capacity, T=discharge time, I= discharge current and k is the Peukert coefficient. From this paper it can vary from 1 to 1.3 for a Lipo. At 1.3 it is a huge effect.

Unfortunately suppliers don’t provide k (unless they are very good batteries). If you know the capacity of a battery pack at 2 different discharge rates you can calculate k. This is why you see a different size and weight for the same capacity battery pack rated at 10C versus 65C.


(Brad Kauffman) #118

Typo- in the tube example it is 15mm outside dimension, not 16mm.


(Robert Giordano) #119

Thank you for checking my math! I encourage everyone to do that. I’m using these formulas:
For round tubes: https://www.amesweb.info/SectionalPropertiesTabs/SectionalPropertiesHollowCircle.aspx

For square tubes: https://www.amesweb.info/SectionalPropertiesTabs/SectionalPropertiesHollowRectangle.aspx

If you don’t find an error on those pages, then I will look at my code and figure out where I made an error. Please let me know. Thanks!

Next, thanks for the Peukert formula! I will definitely look into that!! What’s interesting is, the discharge of the battery (as shown by time vs voltage) in the last drone I built was fairly linear. But, it was only using about 20% throttle and 9-10amps to hover.


(Brad Kauffman) #120

Hi Rob,
Both of those sites have the same error. J=Ixx+Iyy only works for solid sections and thick walled hollow sections. The equation for J for thin walled closed sections such as these is as I quoted. For reference I have attached page 210 from “Theory and Analysis of Flight Structures” by Robert Rivello. See equation 8-71. Thin walled open sections such as I beams or C channels is another type of calculation all together. If you are interested I can provide that as well.

When pulling low currents on a battery you are well down the Peukert curve and get maximum capacity from the battery. On my 700mm hex I run 30V (8S) to keep the current low. I can use a 10C battery. This helps get the most out of the battery and I can reduce wire gauges and save some weight. I get 40 minutes on it and I am carrying a lot of weight.


(cala2) #121

Rob: how do you fix the motors directly to the square tube? with only two bolts? you use some kind of filling material not to fatigue the tube?, thank´s.
Following this interesting discussion.


(Robert Giordano) #122

Hi Cala,

It’s nice to see you over here! I remember you from the DIYdrones forum. Anyway, I just went outside to take a picture with my phone so it is not the best but it shows the arrangement…

As you can see this is a special case where the bolt holes in the 3508 motor just clear the outside of the 15mm square tubing so there are no bolt holes needed in the tubing and no filler material is needed inside the tube. For now I just used a motor mount plate I had laying around but I will replace it with a smaller plate when I actually put it all together. I had to drill one hole in the top of the square tubing for the motor shaft and retaining clip.

I will be building two identical frames, one with 16mm round tubing and motor mounts, the other with this setup. I’ll be comparing the two for weight, performance, and ease of build.


(cala2) #123

Nice Rob, thank´s, yes, I learn a lot with your and Forrest discussions, I´m with vtol´s now and mount the motors like Greg do it with two motor mounts cross but looks heavier than that tecnique.


(Brad Kauffman) #124

My leaving was delayed by a day so I did an experiment today. My hex has round tubes for arms so I printed off a fairing to stick on the underside of the arm that would make it into an airfoil. I put the fairings on two of the arms on one side and flew it to see if it would effect the motor output to those arms. If it improved the drag it would show up with lower output to those 2 motors and higher output on the other side. In an undisturbed flow this would reduce the drag by a factor of 10. So far looking at the data it didn’t make any difference to the power required and this should be way more pronounced than the difference in square to round which is only a factor of 2. I want to play with the angle of the fairing a bit and run a couple more tests but it doesn’t look like this issue is worth worrying about much.


(Brad Kauffman) #125

I ran a FEM of the torque box I 3D printed the other day. It showed a torsional stiffness of 0.5 deg/ft lb and that it should fail somewhere around 8 ft lb if it didn’t buckle first. I took it out to the shop and torqued it to 6 ft lb which it held fine but was beginning to show buckling in the compression members. It certainly shows that printing a whole frame from flat panels and gluing them together will make an efficient frame. Using plywood would be even better since the modulus is higher than ABS.


(Robert Giordano) #126

Here’s my torsion box, built to the same dimensions of 6" x 3.5" x 1". It’s not the prettiest box in the world but I was surprised how stiff it is.

There are no cutouts yet because I wanted to weigh it and test it before I made any cuts. Before I even started making the box, I entered the pieces into dCal and the app showed it would weigh 44g. When I actually built it, that’s what it weighed.


(cala2) #127

My robocat has something similar cf box and is really stiffnes, love that design.


(Brian Riskas) #128

We used airfoil shaped carbon tubes on our big crop dusting multi with good results. Good flight times and very low vibrations


(Robert Giordano) #129

Tubing Calculator UPDATED to v1.2

I have updated the Tubing Calculator so it uses the torsional constant for a thin-walled closed section instead of the polar moment of inertia, as @Ragbagger suggested. The new results are not wildly different from the previous version, but they are more accurate now. Weight comparisons and deflection are the same as before. Only the torsional strength comparisons have been updated.

This example uses the same size tubes that @Ragbagger used in his example.
“J” is now the torsional constant for a thin-walled closed section (tubing).


(Chris Albertson) #130

Rob,

I was thinking the same thoughts about the arms, that the best design would be a faired cross section. But the problem is the quad can fly at various airspeeds. I wonder if I have to worry about relative wind from forward motion. I guess no on a camera drone that only hovers but what about a racing drone?

A racing drone might cruise with a 35 degree nose down angle. How to fair the arms?

At 35 depress we have high body drag and maybe the props interact

I think at high speeds the goal is the lowest possible drag from the body and antenna andbarms and at the same time the projection of the prop disks in the forward direction needs to such that the “rear” props get the cleanest air


(Chris Albertson) #131

Yes. and THAT is where it matters. If the drone mostly hovers there is not much room to improve it. OK maybe add 2% to the flight time but the goal is to fly 60 miles per hour 4 feet off the ground in trees there is HUGE room to improve the current designs. Racing drones are actual more popular now then are camera drones.

I have a 3D printer and would be happy to work with some one more expert then me in aerodynamic

One more idea: If you own a drone you have a “free” wind tunnel. Mount the drone on a stand so the four pros are on a vertical plan and run the engines up. I have a wind meter, the kind on weather stations and I can calibrate the wind.

My goal is to make a mono coup 5 inch racing drone with minimum drag and all the parts inside. I’d put the design file online for anyone to print.


(Robert Giordano) #132

Hi Chris,

As you mentioned, drones that mostly hover won’t benefit too much from aerodynamic improvements and I agree with the 2% figure. The end goal is going to be the determining factor and much of my work involves either still photography or film production.

The initial goal of this blog is to design a frame for a 650 class quadcopter that:

  • is both light and stiff, two of the most important factors for a camera ship.
  • can stay in the air for at least 30 minutes fully loaded.
  • can be built by anyone, especially students, with no special tools, no 3D printers, etc.
  • can be built without exotic, difficult to source, or expensive materials.

I picked a 650 class drone because it seems the most practical for a wide range of action cameras and gimbals. This size drone with 15" props has a similar footprint to the DJI Inspire, which is what many people are using in film production these days.

I am very close (maybe another couple of weeks) to building a working prototype of my latest frame and posting the plans online. I will be building two identical frames, one with arms made from round tubing and the other with square tubing. I will be testing them against each other to see the difference in forward flight.

There is definitely a need for fast, forward flight beyond racing drones and this includes film production. Whether you’re shooting a car commercial, following a boat pulling a wakeboarder, or following dirt bikes on a trail, there is a need for camera drones that can fly at 25 to 50mph.

If you look at the DJI Inspire, some effort went into its aerodynamics. As it leans forward at different angles, the profile remains almost the same. But the Inspire fails when it comes to weight. The Inspire 2 is 3690g AUW with the X4S, their smaller camera, and 3900g with the X5S, their better camera. Even if its kind of aerodynamic, its still a flying brick imo.

So, I’ll definitely be getting into all of these things because they all need to be addressed.


(Chris Albertson) #133

I looked at the DragonPlate web site and see they offer streamlined, airfoil shape tube. Would be good but each stick costs $125. Perhaps we look at how RC airplanes are made and use a hot-wire cut foam fairing over square or rectangular tube?

An interesting experiment after you have data on the two frames you are making would be to fair the square tubes and see if they outperform the round ones. For a quick test balsa sheet would work

One other reply odd-ball sounding test method, but it works. Black box recording of ESC telemetry. The drone will actual fly with DIFFERENT props on each arm. You hover the drone with one of the four props swapped out then look at the motor current and rpm for each motor. In hover the computer is making 100% that each motor is making the same thrust and lift. Examine the log and see if current is reduced or increased for the one non-mating prop. Could to the same by fairing one one arm. Then you have data under identical conditions.

With on matched props the drone’s performance is limited to the lowest performing prop, the computer will do it’s job and keep the drone from tumbling out of the sky. Some people think it can’t work but try it. Hover one foot off the ground over soft grass so there is low risk. The log gives a good A/B test.

This kind of testing is best done with Cleanflight or Betaflight because you only need to fly it line of sight manually and you are only collecting data on the SD card.

With e smaller size drones I experiment with I can 3D print as set of arms with very complex shapes and it costs me 1.9 cents per gram for the materials. Frames are coming out under 200 grams. PLA is not nearly as strong as carbon with this small size it does not matter.


(Robert Giordano) #134

That sounds like a really cool idea! I will definitely experiment with that!! First, I have to build this frame and I’m pleased to say its very, very close. I know I’ll probably make something even better in the future but I want to at least build my current design, fly it, and document its performance. I’m pretty sure it will outperform any commercially available, 650-class drone or drone kit on the market. =)


(Forrest) #135

Interesting idea indeed. The problem (the FCU, with the exception of the PID setting, wouldn’t care about the difference) would be trying to baseline the result.

  • on a well-built (most aren’t) test stand, you can precisely control each test
  • but, if looking at relative performance (the odd out rotor versus the other three), as long as the flights were quite similar, that might work.
    … so not absolute numbers
    … but relative numbers using
    i. Motor channel PWM values (e.g., for a quad, RCOUT CH1, CH2, CH3, CH4)
    ii. not sure if there is anything else in the logs
    … would not be able to accurately derive efficiency numbers from PWM, but would be able to get ordinal if not close

(Robert Giordano) #136

Plywood Box Torsion Tests - Solid vs Cutouts

I’ve confirmed @Ragbagger’s design by testing my 2mm (1/16") birch plywood box, both with and without any cutouts. On my torsion test stand there was no perceptible difference in stiffness between the solid box and the same box with cutouts, even though the cutouts resulted in 20% less weight.

The next step will be to make some longer boxes and test those as well. The frame design I’m working on will use a box that is 310mm in length. I will be testing a 25mm height (same as the box in the video) but I also want to compare 20mm and 30mm heights at the full 310mm length.

My next frame design will be flexible and efficient enough so if you don’t have a Dremel, router, or an easy way to do the cutouts, you will still have a very efficient frame with longer flight times that most commercially available drones.