Files
ardupilot/libraries/AP_HAL_ChibiOS/RCOutput_iofirmware.cpp
T
Andy Piper 2e0ee77b4d AP_HAL_ChibiOS: share GCR telemetry decode with the IOMCU F1 path
RCOutput_iofirmware.cpp carried a second copy of the decode table, used by
the reversed-channel decoder on STM32F1. That copy is live: PB8 and PA1 are
declared BIDIR on iomcu-f103-dshot, so those channels kept the old
one-in-sixteen false accept rate while the main path was fixed.

Hoist the table, the quintet check and the checksum into one helper so the
two cannot drift again. Removing the duplicate saves 172 bytes on the F103,
which has 8k of flash left.
2026-09-02 19:52:59 +01:00

420 lines
17 KiB
C++

/*
* This file is free software: you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This file is distributed in the hope that it will be useful, but
* WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
* See the GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License along
* with this program. If not, see <http://www.gnu.org/licenses/>.
*
* Code by Andy Piper and Siddharth Bharat Purohit
*
* There really is no dshot reference. For information try these resources:
* https://blck.mn/2016/11/dshot-the-new-kid-on-the-block/
* https://www.swallenhardware.io/battlebots/2019/4/20/a-developers-guide-to-dshot-escs
*/
#include <hal.h>
#if defined(IOMCU_FW) && HAL_DSHOT_ENABLED
// need to give the little guy as much help as possible
#pragma GCC optimize("O2")
#include "RCOutput.h"
#include <AP_Math/AP_Math.h>
#include "GPIO.h"
#include "Scheduler.h"
#if HAL_USE_PWM == TRUE
using namespace ChibiOS;
extern const AP_HAL::HAL& hal;
#ifdef HAL_WITH_BIDIR_DSHOT
THD_WORKING_AREA(dshot_thread_wa, 512);
#else
THD_WORKING_AREA(dshot_thread_wa, 64);
#endif
static const char* rcout_thread_name = "rcout";
void RCOutput::timer_tick()
{
if (dshot_timer_setup) {
return;
}
uint32_t dshot_mask;
if (is_dshot_protocol(get_output_mode(dshot_mask))) {
chThdCreateStatic(dshot_thread_wa, sizeof(dshot_thread_wa),
APM_RCOUT_PRIORITY, &RCOutput::dshot_send_trampoline, this);
dshot_timer_setup = true;
}
}
void RCOutput::dshot_send_trampoline(void *p)
{
RCOutput *rcout = (RCOutput *)p;
rcout->rcout_thread();
}
/*
thread for handling RCOutput send on IOMCU
*/
void RCOutput::rcout_thread() {
// don't start outputting until fully configured
while (!hal.scheduler->is_system_initialized()) {
hal.scheduler->delay_microseconds(1000);
}
rcout_thread_ctx = chThdGetSelfX();
chRegSetThreadNameX(rcout_thread_ctx, rcout_thread_name);
rcout_timer_t last_cycle_run_us = 0;
while (true) {
chEvtWaitOne(EVT_PWM_SEND | EVT_PWM_SYNTHETIC_SEND);
// start the clock
const rcout_timer_t last_thread_run_us = rcout_micros();
// this is when the cycle is supposed to start
if (_dshot_cycle == 0) {
last_cycle_run_us = rcout_micros();
// register a timer for the next tick if push() will not be providing it
if (_dshot_rate != 1) {
chVTSet(&_dshot_rate_timer, chTimeUS2I(_dshot_period_us), dshot_update_tick, this);
}
}
// if DMA sharing is in effect there can be quite a delay between the request to begin the cycle and
// actually sending out data - thus we need to work out how much time we have left to collect the locks
const rcout_timer_t timeout_period_us = _dshot_rate ? (_dshot_cycle + 1) * _dshot_period_us : _dshot_period_us;
// timeout is measured from the beginning of the push() that initiated it to preserve periodicity
const rcout_timer_t cycle_start_us = _dshot_rate ? last_cycle_run_us : last_thread_run_us;
// DMA channel sharing on F10x is complicated. The allocations are
// TIM2_UP - (1,2)
// TIM4_UP - (1,7)
// TIM3_UP - (1,3)
// TIM2_CH2 - (1,7) - F103 only
// TIM4_CH3 - (1,5) - F103 only
// TIM3_CH4 - (1,3) - F103 only
// and (1,7) is also shared with USART2_TX
// locks have to be unlocked in reverse order, and shared CH locks do not need to be taken so the
// ordering that will work follows. This relies on recursive lock behaviour that allows us to relock
// a mutex without releasing it first:
// TIM4_UP - lock (shared)
// TIM4 - dshot send
// TIM4_CH3 - lock
// TIM2_UP - lock
// TIM2_CH2 - lock recursive (shared)
// TIM2 - dshot send
// TIM3_UP - lock
// [TIM3_CH4 - shared lock]
// TIM3 - dshot send
// [TIM3_CH4 - shared unlock]
// TIM3_UP - unlock
// TIM2_CH2 - unlock recursive (shared)
// TIM2_UP - unlock
// TIM4_CH3 - unlock
// TIM4_UP - unlock
dshot_send_groups(cycle_start_us, timeout_period_us);
#if AP_HAL_SHARED_DMA_ENABLED
dshot_collect_dma_locks(cycle_start_us, timeout_period_us);
#endif
if (_dshot_rate > 0) {
_dshot_cycle = (_dshot_cycle + 1) % _dshot_rate;
}
}
}
#if defined(STM32F1)
void RCOutput::bdshot_disable_pwm_f1(pwm_group& group)
{
stm32_tim_t* TIMx = group.pwm_drv->tim;
// pwmStop sets these
TIMx->CR1 = 0; /* Timer disabled. */
TIMx->DIER = 0; /* All IRQs disabled. */
TIMx->SR = 0; /* Clear eventual pending IRQs. */
TIMx->CNT = 0;
TIMx->CCR[0] = 0; /* Comparator 1 disabled. */
TIMx->CCR[1] = 0; /* Comparator 2 disabled. */
TIMx->CCR[2] = 0; /* Comparator 3 disabled. */
TIMx->CCR[3] = 0; /* Comparator 4 disabled. */
}
#endif
#if defined(HAL_WITH_BIDIR_DSHOT) && defined(STM32F1)
// reset pwm driver to output mode without resetting the clock or the peripheral
// the code here is the equivalent of pwmStart()/pwmStop()
void RCOutput::bdshot_reset_pwm_f1(pwm_group& group, uint8_t telem_channel)
{
osalSysLock();
stm32_tim_t* TIMx = group.pwm_drv->tim;
bdshot_disable_pwm_f1(group);
// at the point this is called we will have done input capture on two CC channels
// we need to switch those channels back to output and the default settings
// all other channels will not have been modified
switch (group.bdshot.telem_tim_ch[telem_channel]) {
case 0: // CC1
case 1: // CC2
MODIFY_REG(TIMx->CCER, TIM_CCER_CC2E | TIM_CCER_CC1E, 0); // disable CC so that it can be modified
MODIFY_REG(TIMx->CCMR1, (TIM_CCMR1_CC1S | TIM_CCMR1_IC1F | TIM_CCMR1_IC1PSC),
STM32_TIM_CCMR1_OC1M(6) | STM32_TIM_CCMR1_OC1PE);
MODIFY_REG(TIMx->CCMR1, (TIM_CCMR1_CC2S | TIM_CCMR1_IC2F | TIM_CCMR1_IC2PSC),
STM32_TIM_CCMR1_OC2M(6) | STM32_TIM_CCMR1_OC2PE);
MODIFY_REG(TIMx->CCER, (TIM_CCER_CC1P | TIM_CCER_CC2P),
(TIM_CCER_CC1P | TIM_CCER_CC2P | TIM_CCER_CC1E | TIM_CCER_CC2E));
break;
case 2: // CC3
case 3: // CC4
MODIFY_REG(TIMx->CCER, TIM_CCER_CC3E | TIM_CCER_CC4E, 0); // disable CC so that it can be modified
MODIFY_REG(TIMx->CCMR2, (TIM_CCMR2_CC3S | TIM_CCMR2_IC3F | TIM_CCMR2_IC3PSC),
STM32_TIM_CCMR2_OC3M(6) | STM32_TIM_CCMR2_OC3PE);
MODIFY_REG(TIMx->CCMR2, (TIM_CCMR2_CC4S | TIM_CCMR2_IC4F | TIM_CCMR2_IC4PSC),
STM32_TIM_CCMR2_OC4M(6) | STM32_TIM_CCMR2_OC4PE);
MODIFY_REG(TIMx->CCER, (TIM_CCER_CC3P | TIM_CCER_CC4P),
(TIM_CCER_CC3P | TIM_CCER_CC4P | TIM_CCER_CC3E | TIM_CCER_CC4E));
break;
default:
break;
}
// pwmStart sets these
uint32_t psc = (group.pwm_drv->clock / group.pwm_drv->config->frequency) - 1;
TIMx->PSC = psc;
TIMx->ARR = group.pwm_drv->period - 1;
TIMx->CR2 = group.pwm_drv->config->cr2;
TIMx->EGR = STM32_TIM_EGR_UG; /* Update event. */
TIMx->SR = 0; /* Clear pending IRQs. */
TIMx->DIER = group.pwm_drv->config->dier & /* DMA-related DIER settings. */
~STM32_TIM_DIER_IRQ_MASK;
if (group.pwm_drv->has_bdtr) {
TIMx->BDTR = group.pwm_drv->config->bdtr | STM32_TIM_BDTR_MOE;
}
// we need to switch every output on the same input channel to avoid
// spurious line changes
for (uint8_t i = 0; i<4; i++) {
if (!group.is_chan_enabled(i)) {
continue;
}
if (group.bdshot.telem_tim_ch[telem_channel] == group.bdshot.telem_tim_ch[i]) {
palSetLineMode(group.pal_lines[i], PAL_MODE_STM32_ALTERNATE_PUSHPULL);
}
}
/* Timer configured and started.*/
TIMx->CR1 = STM32_TIM_CR1_ARPE | STM32_TIM_CR1_URS | STM32_TIM_CR1_CEN;
osalSysUnlock();
}
// see https://github.com/betaflight/betaflight/pull/8554#issuecomment-512507625
// called from the interrupt
void RCOutput::bdshot_receive_pulses_DMAR_f1(pwm_group* group)
{
// make sure the transaction finishes or times out, this function takes a little time to run so the most
// accurate timing is from the beginning. the pulse time is slightly longer than we need so an extra 10U
// should be plenty
chVTSetI(&group->dma_timeout, chTimeUS2I(group->dshot_pulse_send_time_us + 30U + 10U),
bdshot_finish_dshot_gcr_transaction, group);
group->pwm_drv->tim->CR1 = 0;
// Configure Timer
group->pwm_drv->tim->SR = 0;
// do NOT set CCER to 0 here - this pulls the line low on F103 (at least)
// and since we are already doing bdshot the relevant options that are set for output
// also apply to input and bdshot_config_icu_dshot() will disable any channels that need
// disabling
group->pwm_drv->tim->DIER = 0;
group->pwm_drv->tim->CR2 = 0;
group->pwm_drv->tim->PSC = group->bdshot.telempsc;
group->dshot_state = DshotState::RECV_START;
//TOGGLE_PIN_CH_DEBUG(54, curr_ch);
group->pwm_drv->tim->ARR = 0xFFFF; // count forever
group->pwm_drv->tim->CNT = 0;
uint8_t curr_ch = group->bdshot.curr_telem_chan;
// we need to switch every input on the same input channel to allow
// the ESCs to drive the lines
for (uint8_t i = 0; i<4; i++) {
if (!group->is_chan_enabled(i)) {
continue;
}
if (group->bdshot.telem_tim_ch[curr_ch] == group->bdshot.telem_tim_ch[i]) {
palSetLineMode(group->pal_lines[i], PAL_MODE_INPUT_PULLUP);
}
}
// Initialise ICU channels
bdshot_config_icu_dshot_f1(group->pwm_drv->tim, curr_ch, group->bdshot.telem_tim_ch[curr_ch]);
const stm32_dma_stream_t *ic_dma =
group->has_shared_ic_up_dma() ? group->dma : group->bdshot.ic_dma[curr_ch];
// Configure DMA
dmaStreamSetPeripheral(ic_dma, &(group->pwm_drv->tim->DMAR));
dmaStreamSetMemory0(ic_dma, group->dma_buffer);
dmaStreamSetTransactionSize(ic_dma, GCR_TELEMETRY_BIT_LEN);
dmaStreamSetMode(ic_dma,
STM32_DMA_CR_CHSEL(group->dma_ch[curr_ch].channel) |
STM32_DMA_CR_DIR_P2M |
STM32_DMA_CR_PSIZE_HWORD |
STM32_DMA_CR_MSIZE_HWORD |
STM32_DMA_CR_MINC | STM32_DMA_CR_PL(3) |
STM32_DMA_CR_TEIE | STM32_DMA_CR_TCIE);
// setup for transfers. 0x0D is the register
// address offset of the CCR registers in the timer peripheral
uint8_t telem_ch_pair = group->bdshot.telem_tim_ch[curr_ch] & ~1U; // round to the lowest of the channel pair
const uint8_t ccr_ofs = offsetof(stm32_tim_t, CCR)/4 + telem_ch_pair;
group->pwm_drv->tim->DCR = STM32_TIM_DCR_DBA(ccr_ofs) | STM32_TIM_DCR_DBL(1); // read two registers at a time
// Start Timer
group->pwm_drv->tim->EGR |= STM32_TIM_EGR_UG;
group->pwm_drv->tim->SR = 0;
group->pwm_drv->tim->CR1 = TIM_CR1_ARPE | STM32_TIM_CR1_URS | STM32_TIM_CR1_UDIS | STM32_TIM_CR1_CEN;
dmaStreamEnable(ic_dma);
}
void RCOutput::bdshot_config_icu_dshot_f1(stm32_tim_t* TIMx, uint8_t chan, uint8_t ccr_ch)
{
// F103 does not support both edges input capture so we need to set up two channels
// both pointing at the same input to capture the data. The triggered channel
// needs to handle the second edge - so rising or falling - so that we get an
// even number of half-words in the DMA buffer
switch(ccr_ch) {
case 0:
case 1: {
// Disable the IC1 and IC2: Reset the CCxE Bit
MODIFY_REG(TIMx->CCER, TIM_CCER_CC1E | TIM_CCER_CC2E, 0);
// Select the Input and set the filter and the prescaler value
if (chan == 0) { // TI1
MODIFY_REG(TIMx->CCMR1,
(TIM_CCMR1_CC1S | TIM_CCMR1_IC1F | TIM_CCMR1_IC1PSC),
(TIM_CCMR1_CC1S_0 | TIM_CCMR1_IC1F_1));// 4 samples per output transition
MODIFY_REG(TIMx->CCMR1,
(TIM_CCMR1_CC2S | TIM_CCMR1_IC2F | TIM_CCMR1_IC2PSC),
(TIM_CCMR1_CC2S_1 | TIM_CCMR1_IC2F_1));
} else { // TI2
MODIFY_REG(TIMx->CCMR1,
(TIM_CCMR1_CC1S | TIM_CCMR1_IC1F | TIM_CCMR1_IC1PSC),
(TIM_CCMR1_CC1S_1 | TIM_CCMR1_IC1F_1));
MODIFY_REG(TIMx->CCMR1,
(TIM_CCMR1_CC2S | TIM_CCMR1_IC2F | TIM_CCMR1_IC2PSC),
(TIM_CCMR1_CC2S_0 | TIM_CCMR1_IC2F_1));
}
if (ccr_ch == 0) {
// Select the Polarity as falling on IC2 and rising on IC1
MODIFY_REG(TIMx->CCER, TIM_CCER_CC1P | TIM_CCER_CC2P, TIM_CCER_CC2P | TIM_CCER_CC1E | TIM_CCER_CC2E);
MODIFY_REG(TIMx->DIER, TIM_DIER_CC1DE | TIM_DIER_CC2DE, TIM_DIER_CC1DE);
} else {
// Select the Polarity as falling on IC1 and rising on IC2
MODIFY_REG(TIMx->CCER, TIM_CCER_CC1P | TIM_CCER_CC2P, TIM_CCER_CC1P | TIM_CCER_CC1E | TIM_CCER_CC2E);
MODIFY_REG(TIMx->DIER, TIM_DIER_CC1DE | TIM_DIER_CC2DE, TIM_DIER_CC2DE);
}
break;
}
case 2:
case 3: {
MODIFY_REG(TIMx->CCER, TIM_CCER_CC3E | TIM_CCER_CC4E, 0);
// Select the Input and set the filter and the prescaler value
if (chan == 2) { // TI3
MODIFY_REG(TIMx->CCMR2,
(TIM_CCMR2_CC3S | TIM_CCMR2_IC3F | TIM_CCMR2_IC3PSC),
(TIM_CCMR2_CC3S_0 | TIM_CCMR2_IC3F_1));
MODIFY_REG(TIMx->CCMR2,
(TIM_CCMR2_CC4S | TIM_CCMR2_IC4F | TIM_CCMR2_IC4PSC),
(TIM_CCMR2_CC4S_1 | TIM_CCMR2_IC4F_1));
} else { // TI4
MODIFY_REG(TIMx->CCMR2,
(TIM_CCMR2_CC3S | TIM_CCMR2_IC3F | TIM_CCMR2_IC3PSC),
(TIM_CCMR2_CC3S_1 | TIM_CCMR2_IC3F_1));
MODIFY_REG(TIMx->CCMR2,
(TIM_CCMR2_CC4S | TIM_CCMR2_IC4F | TIM_CCMR2_IC4PSC),
(TIM_CCMR2_CC4S_0 | TIM_CCMR2_IC4F_1));
}
if (ccr_ch == 2) {
// Select the Polarity as falling on IC4 and rising on IC3
MODIFY_REG(TIMx->CCER, TIM_CCER_CC3P | TIM_CCER_CC4P, TIM_CCER_CC4P | TIM_CCER_CC3E | TIM_CCER_CC4E);
MODIFY_REG(TIMx->DIER, TIM_DIER_CC3DE | TIM_DIER_CC4DE, TIM_DIER_CC3DE);
} else {
// Select the Polarity as falling on IC3 and rising on IC4
MODIFY_REG(TIMx->CCER, TIM_CCER_CC3P | TIM_CCER_CC4P, TIM_CCER_CC3P | TIM_CCER_CC3E | TIM_CCER_CC4E);
MODIFY_REG(TIMx->DIER, TIM_DIER_CC3DE | TIM_DIER_CC4DE, TIM_DIER_CC4DE);
}
break;
}
default:
break;
}
}
// decode a telemetry packet from a GCR encoded stride buffer, take from betaflight decodeTelemetryPacket
// see https://github.com/betaflight/betaflight/pull/8554#issuecomment-512507625 for a description of the protocol
uint32_t RCOutput::bdshot_decode_telemetry_packet_f1(dmar_uint_t* buffer, uint32_t count, bool reversed)
{
if (!reversed) {
return bdshot_decode_telemetry_packet(buffer, count);
}
uint32_t value = 0;
uint32_t bits = 0;
uint32_t len;
// on F103 we are reading one edge with ICn and the other with ICn+1, the DMA architecture only
// allows to trigger on a single register dictated by the DMA input capture channel being used.
// even though we are reading multiple registers per transfer we always cannot trigger on one or other
// of the registers and if the one we trigger on is the one that is numerically first each register
// pair that we read will be swapped in time. in this case we trigger on ICn and then read CCRn and CCRn+1
// giving us the new value of ICn and the old value of ICn+1. in order to avoid reading garbage on the
// first read we trigger ICn on the rising edge. this gives us all the data but with each pair of bytes
// transposed. we thus need to untranspose as we decode
dmar_uint_t oldValue = buffer[1];
for (int32_t i = 0; i <= count+1; ) {
if (i < count) {
dmar_int_t diff = buffer[i] - oldValue;
if (bits >= 21U) {
break;
}
len = (diff + TELEM_IC_SAMPLE/2U) / TELEM_IC_SAMPLE;
} else {
len = 21U - bits;
}
value <<= len;
value |= 1U << (len - 1U);
oldValue = buffer[i];
bits += len;
i += (i%2 ? -1 : 3);
}
if (bits != 21U) {
return INVALID_ERPM;
}
return bdshot_decode_gcr_erpm(value);
}
#endif // HAL_WITH_BIDIR_DSHOT && STM32F1
#endif // HAL_USE_PWM
#endif // IOMCU_FW && HAL_DSHOT_ENABLED