Thursday, December 28, 2017

Converged IoT Developer Platform

I installed a popular utility and a popular racing drone motor on a helicopter for IOT/MQTT/COAP research to see what would happen. It flipped, proximity-dived, and then cruised for 35 minutes straight, beating the next best commercial platform of 250 grams by 50+% extra endurance! The fuselage is a Blade Fusion 180 body; the main motor size is 2806 of utility inspector drones. The flight computer is taken from a real estate video shooting drone HGHGLRC Sector25cr (250 grams). The tail motor is from the racing quadcopter HGLRC Petrel 120x.  Is it nice if highways are patrolled by camera drones to broadcast the video of traffic accidents, and an AI server phones motorists to get out of the road at the immediate exit before lining up in a traffic jam?


The build and procedure notes are

The part listing is in the following table. Combo products are entries across different spreadsheets of this project, but split combo pricing to individual entries. Shipping discounts for multiple items from a vendor have been applied. No volume discounts.
The Horizon (Blade brand) components are priced the same on eBay and Amazon both with free shipping, as shown in the eBay and Amazon checkouts.
The entire craft is built without a machine shop and can be repaired by professionals in a car.  Only light hand tools and light electric tools powered by car batteries are needed. This creates jobs.


The World Is The Supply Chain Even For Specialty Parts - Cusped Airfoil

Cusped airfoil 10.5% thickness (joukowsky10.5 for short) rotor blades are interchangeable with conventional parts: The production build's cusped airfoil on the rotor blades is known to extend the efficient angle of attack range for thin blades, discussed in https://estebanhufstedler.com/2020/09/24/joukowski-airfoil-l-d/ and https://estebanhufstedler.com/2020/07/14/improving-the-naca-00xx-airfoil/.  The Fusion 180mm blade is modified from the NACA0010.5 thin airfoil with a cusped trailing edge. The result is that its fuel economy is nearly identical to that of the NACA0012 medium-thickness airfoil under the same operating parameters (blade speed and alpha). The reduced drag for the high angle of attack also keeps the stall angle of attack outside the operating angle of 10 degrees, similar to NACA0012. One video of the endurance test (6 degrees collective pitch), 3 videos of performance tests (8 degrees), and 2 videos of an extraneous test (10 degrees) are below in order. The NACA0012 blades operate with optimal efficiency at 5 degrees of alpha in static tests, and due to the quirk of vertical rotor fluid dynamics, the optimal angle of attack for cruising is 1 degree higher than that of the static test. Since the cusped blades perform nearly identically to NACA0012 blades, the endurance cruising test was done with 5+1=6 degrees of collective pitch.




The World Is The Prototype Lab

The Align 300x has the same diameter and nominal length as those in the Blade Fusion 180. Align uses metric length 38 mm; Blade uses imperial length 1.5 inches. However, all suppliers use metric bolt threads and rod diameters.

Newcomer brands, not in parts list, Skystars and SParkhobby are just as good, if not better: Four hardware bugs of the HGLRC Zeus flight computer prevent it from setting up ESCs without re-soldering. 
1) Betaflight Software version 4.2.5 with the Zeus does not allow a reboot sequence to enter boot loader mode.
2) Not allowing entering boot loader mode means the hardware boot loader button needs to be used.
3) Using the hardware boot loader button means UART1 connecting to RF control receiver blocks USB flashing betaflight software version 4.5.2 unless UART1 is desoldered.
4) LED signal wire reassigned as S5 signal wire cannot flash FVT ESC, which needs resoldering to S4 pin to do the flashing.
Skystar Mini F4 solves all 5 hardware bugs. 
The flight software is here hex .
The dump all configuration is here dump_all  .

The only difference between Skystars F7 and HGLRC F7's config is the pinout assignments and physical chip orientations.
$ diff dump_all_converged_IoT_HGLRCF722_6pin.txt dump_all_converged_
IoT_SKYSTARS_MINIF7.txt
6,7c6,7
< # Betaflight / STM32F7X2 (S7X2) 4.5.2 Mar 28 2025 / 21:47:17 (024f8e13d) MSP API: 1.46
< # config rev: 48588b7
---
> # Betaflight / STM32F7X2 (S7X2) 4.5.2 Aug  6 2025 / 03:55:33 (024f8e13d) MSP API: 1.46
> # config rev: 21f88d7
240,241c223,224
< serial 0 2 115200 115200 0 115200
< serial 1 64 115200 57600 0 115200
---
> serial 0 64 115200 57600 0 115200
> serial 1 0 115200 57600 0 115200
243,244c226,228
 
A summary of the dimensions of all component options is in the following table. And the reference component diagram is beneath the table. 

Blade 180CFX Linear Bushing DFC And Flycolor Fairy ESC Need More Research Before Mass Use

The Blade 180cfx rotor (1.69g hub + 2.24g grips +0.82g thrust +0.90g radial + 0.53g blades bolt +  0.25g dfc = 6.43g) weighs 0.45 grams less than production build (1.4g hub + 3.7g grips + 0.1g washers + 0.9g dfc +0.8g dfc bolt = 6.9g). The grips are nearly identical (4.48 grams) in weight to production build when blade and DFC boltings are included; the DFC links are 0.4 grams lighter when the radial swasharms are included as part of DFC links.

The dyed rods don't use the common metric bolting pitches. In the last picture above, the M2 bolt's common thread pitch is 0.4 mm. The dyed rod is thinner than the M2 bolt, either M1.6 or M1.4, with a common thread pitch of 0.35 mm and 0.3 mm, respectively. However, as the picture showed, the thread teeth perfectly align under light. This is likely so, to reduce the turning needed when adjusting the DFC or servo link. The production servo link's die thread pitch is measured at 0.35 mm.

This Blade 180cfx rotor has thrust bearings and has zero slop of the feathering shaft, as is, without flanged washers. When you look up the rotor dimensions chart, you see a 0.25 mm hub-grip gap, but this gap is filled by the rubber grommet fully. As shown in the dimension chart, the 180cfx's DFC link is (11.3+11.2)-(10.8+10.7)=1 mm longer than the production build's Infusion180 DFC link. So, the servo link needs tightening 1/0.35=3 full turns, or 6 half turns, or equivalently increasing 90us PWM. 

The installation is more complex with thrust bearings than the production build without thrust bearings. Production can not use the thrust bearings unless an assembly line is trained for it. 

The Flycolor Fairy 20A main rotor ESC's Startup setting can not be the same as the production build's FVT ESC. The 1.25 setting produces oscillation and cannot spool up, and the lower the Startup setting, the less oscillation. The 0.5 setting is insufficient when the battery is half discharged, as shown in the following test. The first punch-up is good. All other punch-ups after the first one have fatigue and momentary failures indicated by the non-increasing altimeter reading. 
The solution is 1.0 Startup for the main motor and 0.5 Startup (same as production build) for the tail motor.

However, the 1.0 Startup setting can still produce oscillation during spool-up, and the workaround is to spool up to -93 from 97.8, as pictured, stop there for a few seconds, and then spool up to -50 very slowly without any sudden movement. The successful video is here on the right, with the very same battery and the same starting voltage as the previous test with 0.5 Startup. It shows consecutive 4 punch-ups over the course of the mission of nearly 10 minutes.


The last test video had a tremor from 8:36 (12 seconds after the start of diving) to 8:41, if you pay attention to the tree top line while not looking directly at it for the vibration with the full-size video. It is well known that human peripheral vision is more sensitive to light intensity and flickering.

When I reduced the linear bushing gap with 2 layers of clear tape on the DFC links (4 layers filling for each link), the tremor persisted, raw video on the left, post-stabilized on the right. The tremor is between 0:12 and 0:17 during punch-out in the video, and the tremor frequency is faster than what PID vibration can produce.
The tremor can be diagnosed on tethered hovering with non-zero collective pitch when I feel like my hand is holding a sonic cavitation machine when it is operating.

The same Flycolor Fairy 20A ESC works with the production DFC rotor. Pictured here is the same hub that produced the tremor video, but with the production DFC link and grips.
Video verification is here, raw video on the left, post-stabilized on the right, taken immediately before taking the above 2 pictures. There is no tremor between 0:12 and 0:17 during punch-out. There was no modification to the craft other than the 90us PWM points compensating for the DFC link length difference, as evident by the consecutive video sequence number S0001, S0002 (landing), S0003 (takeoff), to S0004. Correct RPMs for temperature 74F for S0001 and 64F for S0004.






4 comments:


  1. Awesome blog.I find it interesting and is pretty fascinating, By reading your blog i got very useful information.Thanks for sharing this blog.
    Salesforce Consulting Services

    ReplyDelete
  2. Thanks for sharing this useful information. Keep it up.

    healthcare app development
    mobile app development company

    ReplyDelete
  3. Are you a small business owner requiring a website to reach out to a large number of customers? Do you have a website that is not giving the desired results? No matter what your requirements, we will provide customized solutions to you through our website development services. We are Promanage IT Solutions, the most popular and reliable company specializing in website development.

    ReplyDelete
  4. This article on the Converged IoT Developer Platform is insightful. As a leading Mobile App Development Company in Atlanta, we see the growing demand for integrated IoT solutions and how crucial it is to have a unified development environment.

    ReplyDelete