Telescope mount

LX200 protocol: East is negative
libnova: East is positive

Links

• A3967 Datasheet
• Easy Driver schematics
• INDI protocol
• STM32vlDiscovery schematics
• STM32F100 pin layout etc
• STM32 M3 reference manual
• Newlib stubs
• Positional astronomy
• Motion profiles why s-curves
• Telescope pointing
• Motion Profile Calculations
• ChibiOS/RT
• soundstepper lx200 emulator commands

Captain's Log

Previously

• Installed the Code sourcery arm Lite gcc compiler (only 32 bit version, had to install ia32-libs)
• Got the STM32 firmware installed (used a script i found on the internets)
• Got timer working (make sure you first enable the system clock for the device, even for IO ports during pin setup)

Stardate 20121123

• The Makefile didn't work for Newlib (it went unnoticed since all the examples didn't use newlib functions).
• You need a stub file for newlib containing target-specific functions (eg. _write).
• Another problem was that my link step didn't use gcc but ld. gcc can find the correct lib itself (given that you add -mthumb -cpu=cortex-m3). This resulted in the use of BLX calls to newlib functions without bit 0 of the destination set. On Cortex-M3, with only thumb mode available, this causes a crash.
• I tested using GDB. To use 'call' with strings as parameters, malloc and free needed to be linked. So I briefly allocated some bytes in main.
• UART uses polling at the moment.

Current status: Now, I can access the UART using standard C library functions like printf. I also gained some insight into how the Firmware libraries work. While not a big fan yet, I'm slightly more convinced than previously that it might be worthwhile, esp. for initialisation.

Stardate 20121125

• Modified stubs to work with interrupts
• Wrote a sample CLI program that can do some useless things.

Current status: Mostly done playing with UART/Newlib. Now it's time to reintegrate the Stepper motor code (using firmware lib?), and make it do things with the UART command line.

Stardate 20121127

• Expanded the CLI with some motor control commands
• reintegrated TIMER3 initialisation and cleaned it up a (tiny) bit
• Changed interrupt source from update event (timer reaches ARR) to CC3 (compare value has been reached) so I can change things after the stepper pulse is sent
• Rewrote the timer interrupt
• Discovered I can crank up the testmotor voltage to 24V, much better torque (12V unipolar motor).
• Perhaps current setting of the driver is not correct.

Current status: I can control the motor using minicom: change speed, direction, query position. It can't stop yet, though.

Stardate 20121130

• Major breakthrough: Rewritten the entire timer IRQ so that it accepts commands from main through a FIFO.
• Each command is a motion segment (accelerate, cruise, stop)
• Debug pin shows IRQ length to be ~ 3us on idle, 25us while accelerating
• The algorithm is not perfect: when calculating the position change for an acceleration from 400steps/s to 4000steps/s at 400steps/s^2, I get 17800 steps (in 9s time). However, measurements show that the actual value is 2033 ticks off. This will have to be investigated further, but it's most likely a rounding error.

Current status: Trapezium profiles can be entered by hand. Next a compiler for these things has to be written.

Stardate 20121202

• Trapezoidal profiling added
• Improved the accuracy
• Got the second timer working (on PA2, direction is PA1)
• Dissected the Polaris mount. The motors attached are PF42-4813G: 20Ω, 315mA per phase. Gearbox is 1:120.
• The motors of the SkySensor are PF42-48G: 100Ω, 90mA per phase, Gearbox is 1:300.
• For R.A I have to turn the knob 10 times to move 1h
• For DEC, 4 turns equal 10°.
• Added the second EasyDriver board
• The DMD-2 is now officially declared dead. I cut the R.A. cable and soldered a second connector on it.
• Connected the setup with the telescope mount.

Current status: Great progress today, I can now begin testing with the telescope. Movements are very slow, even at max speed (~4000 steps/sec). Maybe stall speed is higher when full stepping is selected? BTW, for normal RA tracking, I calculated 128 microsteps/s would be needed... tested this and it goes rather smooth.

Stardate 20121203

• Performed some quick tests. At vmin/vmax/a = 1200/3000/600 the dec motor works repeatably, with 50° = 921570 steps
• HA motor shows some slippage (how???) UPDATE: This is not slippage of the mount, but of the HA ring
• RA is ~ 721200 steps/h

Stardate 20121207

• libnova compiles without a problem (well, there was this gnu lib extension in the tm struct that gave a problem)
• Linker errors turned out to be caused by a wrong ordering of the libraries (libnova uses math functions so the correct order is -lnova -lm)
• a hardfault was caused by a failed malloc while trying to print a double (decreased some buffer sizes).
• 100k size, flash is getting full...
• LX200 protocol spec is full of weird shit, either bugs or badly designed.

Stardate 20121209

• Got a preliminary version of the lx200 protocol to compile on the STM32
• to install the INDI server, sudo apt-get install indi-bin

Stardate 20121214

• I turned the timer code for the stepper motor into a C++ class, cleaned it up and added full documentation. It's rather clean code by now.
• There was a slight "gulp"-moment when I compiled the code: Code size had increased dramatically, even though I ditched most of the extra code (the LX200 emulator was gone, e.g.). With everything else removed, I arrived at around 115KB, with the emulator included I even came 36KB short! Finally I managed to reduce it by including the G++ compiler options -fno-rtti -fno-exceptions. Those features are real memory hogs, turning it of reduced the code size from 115KB to 36KB. I guess I'll have to live without them ;-).
• Some errors have slipped into the code, a bug hunting session should clear this up.

Current status: A small step towards code cleanup has been taken, once I get the STM32EasyDriver class working, I can focus on the LX200 code again.

Stardate 20121215

• The EasyDriver class has been tested and works flawlessly as far as I know.
• The driver class has been tested on multiple timers and channels.
• MS1, MS2 and enable pins have been connected on the DEC motor. Functionality has been added to the class to control those pins.
• Step size can be changed on the fly without affecting the speed. While pretty cool, I discovered that step size has no effect on the stall speed. I hoped that with full steps I could achieve a higher stall speed (due to increased torque), but unfortunately this is not the case.
• I moved the CLI test code to a seperate file, cli.cpp. It has been thoroughly cleaned up and it now uses the EasyDriver class.

Goals

• To modernize the stepper controller of my telescope mount and make it interface a modern PC
• To gain an understanding in celestial coordinate systems
• To improve my electronics insights and confidence
• To better understand modern MCU design, in particular the Cortex-M3
• To dissect the INDI protocol and evaluate its capabilities and applications
• To have fun doing all of the above

Ideas

Make the controller's only interface USB, then make a dedicated remote control.

TODO

• Trapeziunprofiel motorsnelheid.
• UART driver schrijven.
• FTDI interface met STM32 bord.
• INDI protocol uitvlooien.
• Snelheid meten van de stappenmotoren met huidige controller.
• ADC driver schrijven.
• Joystick uitlezen.
• DIN connectoren zoeken voor motoren telescoop montering.
• Trigonometrie.