/tgy

tgy -- Open Source Firmware for ATmega-based Brushless ESCs

Primary LanguageAssembly

https://github.com/sim-/tgy

Binary Downloads: http://0x.ca/tgy/downloads/

This tree contains Atmel AVR assembly code for ATmega-based 3-phase sensor-less motor electronic speed control (ESC) boards, originally for Turnigy and similar models. This work is based on Bernhard Konze's "tp-18a" software, which was a port from his earlier personal work to the TowerPro 18A and original (not current!) Turnigy Plush boards. Please see tgy.asm for Bernhard's license.

Patches and comments are always welcome! Let me know how it goes!

Features

  • 16MHz operation on most boards
  • 16-bit output PWM with full clock rate resolution (~18kHz PWM with a POWER_RANGE of 800 steps)
  • 24-bit timing and PWM pulse tracking at full clock rate resolution
  • ICP-based pulse time recording (on supported hardware) for zero PWM input control jitter
  • Immediate PWM input to PWM output for best possible multicopter response (but NOT where soft start or really any significant current limiting is needed!)
  • Accepts any PWM update rate (minimum ~5microseconds PWM low time)
  • Optimized interrupt code (very low minimum PWM and reduced full throttle bump)
  • Configurable board pin assignments by include file
  • Smooth starting in most cases
  • Forward and reverse commutation supported, including RC-car style reverse-neutral-forward PWM ranges, with optional braking

Hardware

See http://wiki.openpilot.org/display/Doc/RapidESC+Database and/or https://docs.google.com/spreadsheet/ccc?key=0AhR02IDNb7_MdEhfVjk3MkRHVzhKdjU1YzdBQkZZRlE for a complete list.
Some board pictures here: http://0x.ca/sim/esc/

Notes

  • If it breaks, you get to keep both pieces!
  • Use at your own risk, and always test first without propellers!
  • New Turnigy Plush, Basic, Sentry and Pentium boards (Hobbywing OEM) have all switched to SiLabs C8051F334, d'oh!
  • If your ESC has 6 pads and an AVR, it's probably compatible; the pads are MOSI, MISO, SCLK, GND, VCC, and RESET. If it has 4 pads, it is probably a newer SiLabs-based one, for which this code will not work. (Except HK_261000001 which has 4 pads but has an AVR.)
  • I build and maintain this in Linux with AVRA (1.3.0 or newer). Patches welcome for AVR Studio APS files, etc.
  • The TowerPro/Turnigy Plush type boards typically do not come with external oscillators, which means their frequency drifts a bit with temperature and between boards. Multicopters and RC-car/boat controllers (with a neutral deadband) would probably be better on a board with an external oscillator. The Mystery/BlueSeries boards typically have them, as well as most higher current boards.
  • This doesn't yet check temperature or battery voltage. This is not desired on multi-rotor platforms; however, people still want to use this on planes, cars, boats, etc., so I suppose I'll add it.

Building from Source

AVRA 1.3.0 or newer or avrasm2, part of the AVR Tools, should assemble this source. AVRA should also build on a Mac. "make all" will emit a .hex file for every build target; "make binary_zip" will make a release .zip file. There are some other make targets for programming.

In AVR Studio, the Makefile is not supported, and just loading tgy.asm and attempting to build it will not define the constant indicating the board type / build target. You must either edit tgy.asm or add an option for the assembler command line to define the board symbol, unless building the default "tgy" board type. For example, this option should emit the bs_nfet target: -D bs_nfet_esc=bs_nfet_esc Look near the top of tgy.asm for the includes and board information.

WARNING

Never just randomly try build targets until one works, especially not when directly powered from a LiPo! :P Many boards have inverted FET drives and different pin assignments. Toggling one wrong pin can fry multiple FETs. Some boards, like the Mystery 20A ESC, may all look the same on the outside by have different FETs as well. Be careful and check your board before flashing.

Installation

For more information, check out these sites:

http://wiki.openpilot.org/display/Doc/RapidESCs http://wiki.openpilot.org/display/Doc/Flashing+Instructions http://www.rcgroups.com/forums/showthread.php?t=1513678

See warning above! The safest arrangement is to use a current-limited bench power supply, set to a low voltage (6V-7V), for both flashing and initial testing. If the pinout is wrong and causes a short, the current limiting causes the input voltage to drop below the brown-out detection voltage of the MCU, causing all output pins to go high-impedance in hardware, and an automatic reset when the voltage comes back.

If you do not have a current-limited supply, you can improvise by using 4 AA batteries or an old NiCd pack or even a LiPo with a 12V light bulb in series to act as a current limiter. Be careful when touching the board, since it can be quite easy to turn on a FET gate with just your finger. This should be OK if you have a current-limited supply, since it should just reset.

Even if the board appears to be one tested by others, make sure yours has the expected FET pin assignments, inversions, and sense lines! The assignments for each board type can be found in the .inc files, and the actual pin mappings can be found in the first few pages of ATmega8 datasheet. This typically requires a voltmeter and, preferably, an oscilloscope.

When not powered, use a voltmeter to check the path from the MCU pins to either the FET resistors, transistors, or driver chips, and verify that they match the assignments in one of the include files. Then power up the ESC with the original firmware, and check with an oscilloscope the inversions of the same pins. A voltmeter may be used instead when the motor is stopped. Be careful not to short FET pins together! If all voltages are low when the motor is off, nothing is inverted, and the INIT_Px values in the .inc file should be 0 for all of the FET bits.

Older and smaller boards typically use P-FETs and N-FETs in the H-bridge; the P-FETs are typically driven by three NPN transistors, while the N-FETs are driven directly at TTL voltages with just a low resistor. Medium-sized and newer boards have moved to all-N-FET designs, but maintain the NPN transistors, and so have inverted P-FET pins from the MCU. Larger current (>~30A) boards typically have separate FET driver chips, to which the N-FET or P-FET pins may be inverted (but not both, since it would blow up before the MCU initializes).

PWM is usually done on the low side of the H-bridge, where high frequency driving is easiest. If the average voltage increases at the AVR pin as throttle increases, and drops to 0V when stopped, the low-side FETs are not inverted; if average voltage decreases, and rises to 5V when stopped, the low-side FETs are likely inverted. The high side FETs are only switched at every other motor commutation step, and so switch at a lower frequency, and should be off 2/3rds of the time. If at 0V when stopped, and less than 2.5V average when running, the P-FETs are not inverted. If at 5V when stopped, and more than 2.5V when running, the P-FETs are inverted. In the case of inverted an FET group, they should be listed in the INIT_Px values (to turn them ON at boot), and the "on" macros should use clear instead of set instructions. The inverse applies to the "off" macros.

There are four sense lines. The three output phases go through resistor dividers (to bring the voltage down to between 0-5V), and then are connected to ADC channels which can be accessed by the comparator with the ADC multiplexer when the ACME (Analog Comparator Multiplexer Enable) bit is enabled. Some boards use all ADC channels while others put one pin on AIN1 so that the ADC can sample voltages on other ADC channels while the comparator samples that phase. A "center tap" is established with a resistor star and connected to AIN0 on all boards, for detecting the zero-crossing from the motor back-EMF for timing. You can check which pins run to which output phases as they will have the lowest resistance from output phase to AVR pin, and all three will have a common resistance to the AIN0 pin.

Boot Loader

Since the 2012-06-01 release, the pre-built .hex files contain a boot loader that allows for flash and EEPROM reading and writing over the PWM input wire. The main purpose of this boot loader is to support software updates without having to expose the MCU or ISP pads for SPI programming, assuming this version or newer has previously been flashed.

The boot loader is available on all boards which are wired in way that permits two-way communication over the wire. Some ESCs are isolated with an opto-isolator or by components to invert the signal to be compatible with ESCs that do have an opto-isolator, such as ESCs that work with the kda.hex target. Such boards will not be able to use this protocol.

The boot loader uses an implementation of the same wire encoding as the Turnigy USB Linker. This is a simple 9600bps serial to half duplex wire converter that encodes signals in a way that should make it difficult to start a motor, even on an ESC not supporting the protocol, regardless of data sent. However, it is still possible that this could occur, so follow the same precautions as with normal ISP flashing, below.

If the Turnigy USB Linker is not available, it is possible to use an Arduino or MultiWii board as a gateway between (USB and) serial and the wire protocol used by the boot loader. See the ArduinoUSBLinker project by Chris Osgood: https://github.com/c---/ArduinoUSBLinker

Boot Loader Fuses

Since the 2012-09-30 release, long periods of high PWM input while disarmed will cause a jump to the boot loader, if present. Since the 2013-04-24 release, this became detangled from the included boot loader (BOOT_JUMP vs BOOT_LOADER), so other boot loaders may be used instead.

As a result of the automatic jumping, it is not necessary to change the fuses to enable the boot loader. It may or may not be desirable to set the BOOTRST flag depending on intended operation and the hardware in use.

If the hardware does not have any low pass filter or pull-down on the PWM input, it may easily float high by the time the boot loader checks it, which can prevent normal startup if another input (eg: I2C) is expected to be used. There should be pull-down or load present on the PWM input to prevent this if BOOTRST is enabled.

Confusion may also result if the pin floats high with nothing connected and power is connected to check for starting beeps. Likewise, touching the connector or connecting it later may drain the input and cause normal startup to occur, which may seem like the ESC is resetting while powered. This is not a problem if the PWM input is normally intended to be used.

On the other hand, with BOOTRST not set, it is not possible to recover from an interrupted flash, flash of the wrong board type, or flash of other software or versions older than 2012-09-30 without connecting to the ISP pins to do the flashing.

To enable the the boot loader on ATmega8 boards, the low nibble of the hfuse should be set to 'a'. See "make bootload_usbasp" to do this automatically. BOOTRST and BOOTSZ1 should be enabled to set 512 words, start at $0E00.

Flashing and Testing

Sort out how you want to connect an ISP programming device to the chip. Some boards have 6 pads in a row for this purpose. You can either solder wires, or make up some kind of springy-pin connector that touches the pads or chip pins without needing to solder. This is helpful when flashing more than one board. See here for some ideas and discussion: http://www.rcgroups.com/forums/showthread.php?t=1513678

Sort out which software you will use for flashing. The KKMulticopter Flashtool now supports ESC flashing and is commonly used. You may also download AVR Studio and the AVR Toolchain from www.atmel.com, or try "avrdude" on most OSes. There are plenty of resources on the web for AVR ISP programming, and many ways to do it.

Power the ESC from a current-limited supply and try to read the stock firmware (flash) and EEPROM, to use as a backup. Most are locked and will still appear to read, but the files will contain just a series of repeating/increasing digits. If you do manage to get something, consider yourself lucky!

Write down the stock fuse values, and check that they are sane. Most AVR programmers have a menu for this, but with avrdude or uisp, Google "AVR fuse calculator", select the ATmega8 target, and type in the hex values. If the "watchdog" fuse is enabled, you will want to disable it for now. The brown-out voltage should be set to 4.0V and enabled (BODEN). Leave the rest of the fuse values as shipped, and write down the new values.

Flash the desired target .hex file to the AVR, then set the fuses, if anything needs changing. If you have any errors, check the connections to and voltage at the chip. Sometimes, a weak power or signal connection can temporarily work and then fail part-way through programming, giving verification errors. This can happen particularly if the target chip is powered weakly by the programmer itself, which then back-feeds to the rest of the circuit and tries to charge the capacitor, etc. Also check that the ground pin (black wire) actually connects to a pin marked "GND", not "NC" as on some USBasp devices!

Once programming is successful, hook up a small motor without propeller and reset the power. You should hear three increasing beeps. If not, or if you hear only some beeps, the FET pinout may be incorrect or one or more FETs may be broken. Repetitive clicking can also indicate that the pinout is incorrect and causing continuous brown-out resets.

Now, if you attach a valid PWM servo pulse with low-enough pulse length, you should hear a forth beep indicating that the ESC is armed. If not, try lowering the trim as far as possible. If it still doesn't work, you may need to raise the STOP_RC_PULS value in the code.

Once armed, the ESC will try to start the motor if the pulse length (throttle) is increased. Try running up the motor slowly, and make sure everything runs smoothly. If on a variable supply, increase the voltage to the expected running voltage, and try again, with slow and rapid throttle changes.

Finally, test the ESC in the intended application, with the usual power source, without anything attached to the motor shaft/bell first. Then, attach the propeller or gear, etc., LOCK IT DOWN so it doesn't hit you in the face, and try a slow sweep from idle to full throttle. Finally, try rapid throttle changes from slow to fast. The ESC and motor should run smoothly, never lose timing, and the ESC should not reset. If using a flight control board, you can sometimes tap the board to make it output a sudden throttle increase.

For debugging reset causes, recent code has different beep sequences for each AVR reset case. See the Troubleshooting section below.

Throttle Calibration and Programming

Since the 2012-01-04 release, the input throttle range may be calibrated. This should be used whenever PWM input mode is used, including where external resonators are present, to set the usable throttle range.

The default range is set at the top of tgy.asm (stop at 1060us, full throttle at 1860us) and is not changed if no new value is saved to the EEPROM. However, boards without external oscillators (typically those which use tgy.hex) must use the internal RC oscillator on the Atmega8, which may be off by 5% - 10% between each board, and will also drift by 10% or more over a 40 degree Celsius temperature range. This can cause throttle differences between boards if not calibrated, and issues arming the ESC, particularly in cold environments.

Some flight boards may have automatic calibration procedures, or it may just be possible to set the min/max motor output to exactly match the ESC defaults listed above. ESCs with external resonators can use this method to avoid calibration. With a flight board such as the Naze32 running baseflight, the min/max can be set with CLI commands as follows:

set minthrottle=1064

set maxthrottle=1864

Otherwise, to calibrate the ESC, REMOVE ALL PROPELLERS and follow the steps that may be documented for your flight controller board. KK2 boards have a two-button procedure that emits calibration pulses automatically. With KK boards, the Yaw pot must be set to the minimum position in order to enable "pass through" mode from the RX input to the ESC output. The ESCs are then calibrated to the radio's throttle stick.

Calibration may also be done with a servo tester or with the ESC directly connected to an RX channel. The only requirement is that the input pulse has to be at or above PROGRAM_RC_PULS (default 1460us) to enter programming mode. This will differ slightly on boards with no external oscillator. If calibrating ESCs individually, try to maintain a close temperature and at least 6-7V input voltage (the ATmega8 oscillator speeds up as temperature and Vcc decrease).

With the propellers removed and the source (radio, servo tester, or flight control board) set to full throttle, power up the ESC and wait for a single beep after the typical rising initialization beeps. This indicates the high pulse length has been saved to RAM. Move the stick or knob to the lowest setting, and wait for two beeps. This indicates that the low pulse length has been saved to RAM.

If RC_PULS_NEUTRAL has been enabled (RC Car-style reverse mode), move the stick/knob to the center, and wait for three beeps. This indicates that the neutral pulse length has been saved to RAM.

If the stick is not moved after the final setting, the ESC should now write the RAM settings to EEPROM, and continue startup. The ESC should recognize the same pulse length input as a valid arming pulse, and arm and work as usual.

The input PWM pulse is measured in 24-bit space and scaled in 16.16 space to fit the number of PWM steps defined by POWER_RANGE - MIN_DUTY. There should be no measurable aliasing or quantization. Alternatively, the values may be adjusted in EEPROM directly. While calibrating, margins of 1/16th on the low end and 1/32nd on the high end are used to try to avoid problems with arming and reaching full throttle during temperature extremes. If desired, margins may be adjusted in tgy.asm between rc_prog2 and rc_prog5.

There is currently no way to reset (remove) the calibration other than by clearing the EEPROM (or reflashing without EESAVE set). This may be implemented in the future by some basic stick programming feature.

NOTE: As of 2012-03-15, throttle calibration is disabled when a brown-out reset is detected. I accidentally calibrated an ESC when testing with a NiCd pack. The pack could not supply enough current, resulting in a brown-out reset of all ESCs, excluding the flight controller. I did not lower the throttle in time, resulting in one ESC getting a stable enough signal to store a new calibration. When intentionally calibrating, be sure that you cleanly connect the power. If you don't hear the rising beeps, remove the power for a few seconds to allow the capacitors to discharge, then try again.

Troubleshooting

There are 4 main beep frequencies used at different intervals and lengths to signal the various operation and fault states. These beep frequencies are labelled here as f1 through f4, where a higher number indicates a higher frequency. Repetition indicates longer beeps.

If CHECK_HARDWARE is enabled (default on the Afro and TBS boards), the drive and sense circuitry will be tested for correct operation at boot. This check will prevent beeping if it would be unsafe to do so (for example, if beeping would end up completing a short circuit), and so may prevent further damage in some cases.

Error conditions from CHECK_HARDWARE are flashed through the LED(s), if present, or beeped (if possible) before anything else, in a looping pattern, for as long as the error condition(s) persist. All conditions cannot be checked on some hardware due to pin routing, but will be indicated by beeping or flashing a particular count followed by a pause. If more than one error condition is present, each error will be reported in sequence. The error conditions are:

1: Phase A stuck high
2: Phase B stuck high
3: Phase C stuck high
4: AIN1 stuck high
5: AIN0 stuck high
6: Phase A low-side drive broken
7: Phase B low-side drive broken
8: Phase C low-side drive broken
9: Phase A high-side drive broken
10: Phase B high-side drive broken
11: Phase C high-side drive broken

Note that if the ESC resets while the motor is still spinning (such as after a brief power cut), it is possible that the induced current will cause some transient errors that will clear once the motor stops, and then operation will continue.

During boot, the MCUCSR register is checked to see the reason for reset. For exact behaviour, see near "Check reset cause" in tgy.asm. Here are the expected beep sequences:

f1 f2 f3: Regular startup with nothing special detected

f3 f1: Voltage brown-out bit was set (MCU voltage dropped below 2.7V/4.0V)

f4: External reset (via the reset pin, as in after programming)

looping f1 f1 f3 f3: Watchdog reset (previous execution locked up)

looping beeps (8) of f2 or f4: Unknown (beeps out all MCUCSR bits, LSF)

fast repeating beeps of f4 while idle: Noise on PWM input detected

Once a valid input source is found and receiving idle throttle, f4 f4 f4 (a long f4 beep) indicates that the ESC is armed and will start the motor when throttle goes non-zero. If you are unable to start the motor and are not hearing the forth, long beep, try lowering the throttle trim, or raise it all the way to start throttle calibration (above).

If no input command is received for about 1 second, f3 f2 is beeped and the ESC returns to disarmed state, waiting for a valid arming signal.

If no valid input is received while disarmed and BEACON is enabled, an f3 beep will be emitted after about 8 seconds and then every 4 seconds.

The various beep frequencies use different FET combinations (rather than all FETs at the same time) to try to help diagnose boards with failed FETs or possible incorrect firmware pin configuration (build target). If you hear only one or two of the usual three power-up beeps, and the board worked previously, it is likely that one of the FETs has burned out. The ESC may still start and run the motor like this, but the motor will sound bad, and power and efficiency will be reduced.