I purchased a PX4 Vision 1.5 Quadcopter from Holybro with the idea of using it as a platform for a challenge.gov competition I was going to walk on for. The competition involved creating a sUAS platform for indoor searching of buildings in an unknown state. (Partially collapsed, on fire, ect) The big requirement here being that the environment is GPS denied and the pilot will not have visual contact with the drone.
The PX4 Vision platform seemed like a good place to start for this. It included an Optical Flow/Lidar that should work for GPS replacement. The Structure Core Depth camera should make avoiding obstacles while flying very easy. According to the marketing material, the software was already pre-setup, just needed some calibration for our use. In reality- this was far from the truth.
Ultimately the project wasnāt able to meet the timeline for the walk on competition- but valuable information was gained and Iām going to take some time to write down and share what I learned along the way.
Hardware
Quadcopter-
Propellers:
The quad came with 6030x3 gemfan props from Holybro. Holybro claims these props are optimized for this system. The problem is that they recommend a 4S battery. The motors say they should be used with a 6S @ the 1750KV sizing. Needless to say- battery life was lower than I wanted with these props. I was barely getting 11 minutes of just slow flying around/hover.
So I loaded my model into ecalc.ch and played with props sizing until I got max flight time. Then I looked at which prop sizes were actually available near that sizing played with them all until I figured out which is best.
The best I was able to do was 7"x2.4"x2 props but I could only find them in 4MM motor stud size. The next best prop size I could find was 7"x3.5"x2 props by gemfan.
Old Props 11.1Min (6030)
New Props 13.3Min (7035x2) while maintaining a 2.2 thrust to weight ratio and keeping under the 20amp ESC limit.
Please note that the 7" props are VERY tight on the PX4 Vision Frame. 1-2mm. But in my 40 flights so far have not had an issue as long as things are tightly ziptied or glued into place.
Battery:
The battery life with the Holybro recommended battery was awful. A 3000mah 4S netted 7 minutes battery life with no camera onboard and the stock props.
Somehow DJI was able to get a 4S 5000mah battery to weight 320grams. I was on the hunt.
I was able to find a battery that had a higher mah vs weight than DJI using a Energy Dark Lithium V2 4S 8400mAh Li-ion Battery that weighed 556 grams. These new Li-ion cells are larger and much higher amperage than the old Li-ion cells of past. 90 Amp max- we were averaging 24 amps at hover.
This new battery netted a hover time of 15 mins and 30 seconds (After the above prop change)
PM07:
The power distribution board worked well for my needs. I was able to power both my FLIR camera and the SIYI HM30 by soldering to extra ESC pads that the board had. These ports gave battery voltage.
I also had to adjust how the Flight Controller (6C) was connected to the PM07. It was originally hooked up so that it went from FMU PWM Out on the flight controller to IO PWM IN on the PM07. This means that the last 4 AUX pins on the FMU PWM out were tied to solder points on the PM07. To use the servo connections on the side of the PM07- I had to split the ground cable coming from the flight controller and take the last 4 aux pins to a new 10 PIN GH connector. This allowed me to connect to the FMU PWM IN port on the PM07 and then was able to get the servo pins working on FMU PWM Out (PM07) while still allowing the motors to be controlled over the first 4 pins hooked into IO PWM IN.
Please reference included wiring diagram for more information. (Coming soon)
Why didnāt Holybro hook up the motors to IO PWM Out from the Flight controller instead of from FMU PWM Out? This would have allowed for me to use the full FMU PWM aux pins and would have allowed me to just hook up another cable to get it working vs having to split my ground. I understand I could have changed the software myself but I had assumed that Holybro/Auterion set it up like this for a reason. Anyone have insight here??? Why use FMU PWM Out for motors instead of IO PWM Out?
Companion Computer:
No changes made to the companion computer.
Optical/Flow Lidar:
Because we were working in a GPS denied environment- the optical flow and lidar sensor are very important pieces to a stable system. Sometimes these environments are dark and this poses extra challenges for these sensors.
The PX4 Vision 1.5 comes with 1x PMW3901 optical flow sensor (Thone ThoneFlow-3901U, connected via Telem3 Serial Port) and 1x TOF Infrared distance sensor (PSKāCM8JL65āCC5, connected via GPS 2 Port)
At night- these 2 sensors are almost useless and because we have no GPS- were causing problems with our position holding ability. Ultimately these 2 sensors need much more testing to see if there is anything we can do to make them perform better.
We opted at this point to replace the sensors with an Ark Flow unit. Once setup we were able to hold position much tighter. Dark environment testing was much better as well. We never tested in a completely dark room but performance was better as light decreased vs the Thone.
Ultimately I donāt think the Thone will work well for many people but would like to be proved wrong here.
Structure Core Depth Camera:
No changes to the depth camera. At one point I did try to update the firmware on the camera so I could use the latest SDK from Structure. This broke the camera integration on the companion computer though. This was because it requires a different driver. I was able to obtain a firmware downgrade tool from Structure which fixed my issue.
Camera System:
FLIR DUO PRO R hooked into a SIYI HDMI>LAN which then goes to the SIYI HM30 Air Unitās Video port. (This is just a 4 pin ethernet port)
The FLIR is getting power from a PM07 ESC port and is also connected to FMU PWM Out on PM07 for camera control.
See Camera System Diagram for more information. (Coming soon)
GPS/Compass:
One of the first things we did was remove the GPS from the PX4 Vision. This had an unexpected result of making yaw and z control very flaky and jumpy. (As the quad is yawing back and forth the Z has trouble staying up) Upon further looking I found that the internal compass on the 6C was often at times 180 degrees out of sync with reality. I looked for stand alone compasses but was unable to find anything that would work out of box with PX4. I ultimately ended up hooking the GPS back up and then disabling only the GPS in PX4 parameters. This way the external compass can still be used.
We plan on testing a Rm3100 and Bosch compass and possibly releasing how to make these work on PX4 so that stand alone options are available.
Radio System:
SIYI HM30 Air Unit has 4 connections on the quadcopter. Telemetry, Control (SBUS), Ethernet (Video), and Power.
This system worked flawlessly for our needs and is very flexible. Most of the time I left the ground station inside while I was flying outside and never had any issues with video or control quality.
See wiring diagrams for more information. (Coming soon)
Flight Controller:
Vibration Mounts- Holybro opted to just sticky tape the flight controller to the carbon fiber panel. We removed this and put on gel mounts ) [Kyosho Zeal 5mm ] We also had to add a nut to the top carbon fiber battery plate because the gel moved the FC up a few MM.
More information to come!