Difference between revisions of "Category:Autopilots"

Paparazzi Autopilots

• STM32 series microcontrollers
• LPC21xx series microcontrollers

There are active and current autopilots designs using both architectures. Not all autopilots have the same capabilities, peripherals or features, but each has advantages in different applications.

The STM32 architecture is relatively new and still has bugs to be worked out. This is more suited for developers who are comfortable with joining in and working out the issues. Currently, boards are designed around the STM32F1 series, but there is future upgrade path capabilities to the F2 and F4 series, giving way to feature rich processors with a variety of peripherals and speeds. Architecture-dependent firmware code is supported in part by libopencm3. The Lisa an Krooz autopilots use the STM32.

The LPC21xx based boards use the LPC2148 and have been flying fixed wing and multi-rotors for many years. This architecture is more mature and stable but at the expense of speed and extra ports available on the newer STM32 series processors. The Tiny series, Booz, TWOG, YAPA, Umarim and NavGo autopilots all use the LPC2148.

Some autopilots have also been designed for close integration with small single-board computers, particularly those based on OMAP processors such as the Gumstix Overo series. The Lisa/L and Classix boards are designed with this in mind, though other autopilots can be easily interfaced. Further information can be found here.

A basic feature comparison table is presented to help in the autopilot hardware selection process. Stable well tested and used LPC or more cutting edge STM32 that requires some debugging.

For information regarding architecture and firmware compatibility of various subsystems and modules, please see the appropriate Subsystems overview and Modules List pages.

NOTE: The accuracy of this table may not be 100% correct, the best resource is always hardware and software source files and individual autopilot pages.

Notes:

1. Only the newest revisions of the more commonly used autopilots are listed

2. The extra 8kB of RAM on the LPC2148 shared with the USB DMA

3. The onboard sensors are almost always supplemented with external sensors. For example, TWOG can use an external IMU or IR sensors, and also needs an external GPS.

4. Input/Outputs listed are generally those easily accessible on regular autopilot connectors, customization/hacks can modify available I/O, for example free an extra I2C on Tiny and TWOG

5., 6. Some features use shared resources - denoted by X + Y where Y is shared - and cannot be used simultaneously

5. Lisa/M v2.0 shared resources include: one I2C is shared with 2 PWM outputs, two ADCs are shared with LEDs, one RX only UART is shared with the PPM capture

6. Umarim v1.0 shared resources include: 4 ADCs are shared with 4 GPIOs

7. Usually other unused pins can be used for additional GPIO with some code modifications

Schematics, CAD files, Gerber files, BOM release strategy

About the hardware development and release process.

Files needed to create the hardware have been released and can be found here. It is always good to remind oneself of the email Antoine once wrote in the mailing list before you want to start producing your own PCB's.

8 June 2011 13:25:47 Antoine Drouin wrote on the mailing list:

I've started this project together with Pascal 8 years ago and since then I have dedicated my time to try and make it successful. I'm utterly convinced of the benefits of open source, but observing how Paparazzi grew over time, I came to the conclusion that hardware is a bit different than software... "gcc tiny.brd" is not going to make a board magically appear on your desktop.

I'll list here some of my arguments in favor of releasing CAD files after the board is mature.

1. Unlike software, where an unskilled user can type make and get a piece of complex software to successfully build, assembling hardware requires tools and skills. Providing gerbers and BOM have lured a bunch of new users into believing otherwise and has created tons of frustration. I've myself fixed a number of badly assembled boards and I even recall that while helping debugging a board (so after assembly), discovering that the person had manufactured two layers PCBs instead of four layers. As the technology of the autopilot increases, this problem becomes more and more important.
2. The success of the project depends on the availability of affordable hardware. The price of hardware is directly and exponentially dependent on the number of manufactured units. If ten persons manufacture 10 boards each, the cost will be much higher than if one person manufactures 100.
3. Last and not least, the quality of assembly also depends very much on the number of manufactured units. Good quality can only be achieved through the use of automated placing and soldering, and those processes can only be used if the number of units reach a certain amount.