Peripherals

Peripheral Clock Gating

As usual, peripheral clock gating is still handled by driver itself, users don’t need to take care of the peripheral module clock gating.

However, for advanced users who implement their own drivers based on hal and soc components, the previous clock gating include path has been changed from driver/periph_ctrl.h to esp_private/periph_ctrl.h.

RTC Subsystem Control

RTC control APIs have been moved from driver/rtc_cntl.h to esp_private/rtc_ctrl.h.

ADC

ADC Oneshot & Continuous Mode drivers

The ADC oneshot mode driver has been redesigned.

The new driver is in esp_adc component and the include path is esp_adc/adc_oneshot.h.
The legacy driver is still available in the previous include path driver/adc.h.

The ADC continuous mode driver has been moved from driver component to esp_adc component.

The include path has been changed from driver/adc.h to esp_adc/adc_continuous.h.

Attempting to use the legacy include path driver/adc.h of either driver will trigger the build warning below by default. However, the warning can be suppressed by enabling the CONFIG_ADC_SUPPRESS_DEPRECATE_WARN Kconfig option.

legacy adc driver is deprecated, please migrate to use esp_adc/adc_oneshot.h and esp_adc/adc_continuous.h for oneshot mode and continuous mode drivers respectively

ADC Calibration Driver

The ADC calibration driver has been redesigned.

The new driver is in esp_adc component and the include path is esp_adc/adc_cali.h and esp_adc/adc_cali_scheme.h.

Legacy driver is still available by including esp_adc_cal.h. However, if users still would like to use the include path of the legacy driver, users should add esp_adc component to the list of component requirements in CMakeLists.txt.

Attempting to use the legacy include path esp_adc_cal.h will trigger the build warning below by default. However, the warning can be suppressed by enabling the CONFIG_ADC_CALI_SUPPRESS_DEPRECATE_WARN Kconfig option.

legacy adc calibration driver is deprecated, please migrate to use esp_adc/adc_cali.h and esp_adc/adc_cali_scheme.h

API Changes

The ADC power management APIs adc_power_acquire and adc_power_release have made private and moved to esp_private/adc_share_hw_ctrl.h.
- The two APIs were previously made public due to a HW errata workaround.
- Now, ADC power management is completely handled internally by drivers.
- Users who still require this API can include esp_private/adc_share_hw_ctrl.h to continue using these functions.
driver/adc2_wifi_private.h has been moved to esp_private/adc_share_hw_ctrl.h.
Enums ADC_UNIT_BOTH, ADC_UNIT_ALTER, and ADC_UNIT_MAX in adc_unit_t have been removed.
The following enumerations have been removed as some of their enumeration values are not supported on all chips. This would lead to the driver triggering a runtime error if an unsupported value is used.
- Enum ADC_CHANNEL_MAX
- Enum ADC_ATTEN_MAX
- Enum ADC_CONV_UNIT_MAX
API hall_sensor_read on ESP32 has been removed. Hall sensor is no longer supported on ESP32.
API adc_set_i2s_data_source and adc_i2s_mode_init have been deprecated. Related enum adc_i2s_source_t has been deprecated. Please migrate to use esp_adc/adc_continuous.h.

GPIO

The previous Kconfig option RTCIO_SUPPORT_RTC_GPIO_DESC has been removed, thus the rtc_gpio_desc array is unavailable. Please use rtc_io_desc array instead.
The user callback of a GPIO interrupt should no longer read the GPIO interrupt status register to get the GPIO’s pin number of triggering the interrupt. You should use the callback argument to determine the GPIO’s pin number instead.
- Previously, when a GPIO interrupt occurs, the GPIO’s interrupt status register is cleared after calling the user callbacks. Thus, it was possible for users to read the GPIO’s interrupt status register inside the callback to determine which GPIO was used to trigger the interrupt.
- However, clearing the interrupt status register after calling the user callbacks can potentially cause edge-triggered interrupts to be lost. For example, if an edge-triggered interrupt (re)is triggered while the user callbacks are being called, that interrupt will be cleared without its registered user callback being handled.
- Now, the GPIO’s interrupt status register is cleared before invoking the user callbacks. Thus, users can no longer read the GPIO interrupt status register to determine which pin has triggered the interrupt. Instead, users should use the callback argument to pass the pin number.

Sigma-Delta Modulator

The Sigma-Delta Modulator driver has been redesigned into SDM.

The new driver implements a factory pattern, where the SDM channels are managed in a pool internally, thus users don’t have to fix a SDM channel to a GPIO manually.
All SDM channels can be allocated dynamically.

Although it’s recommended to use the new driver APIs, the legacy driver is still available in the previous include path driver/sigmadelta.h. However, by default, including driver/sigmadelta.h will trigger the build warning below. The warning can be suppressed by Kconfig option CONFIG_SDM_SUPPRESS_DEPRECATE_WARN.

The legacy sigma-delta driver is deprecated, please use driver/sdm.h

The major breaking changes in concept and usage are listed as follows:

Breaking Changes in Concepts

SDM channel representation has changed from sigmadelta_channel_t to sdm_channel_handle_t, which is an opaque pointer.
SDM channel configurations are stored in sdm_config_t now, instead the previous sigmadelta_config_t.
In the legacy driver, users don’t have to set the clock source for SDM channel. But in the new driver, users need to set a proper one in the sdm_config_t::clk_src. The available clock sources are listed in the soc_periph_sdm_clk_src_t.
In the legacy driver, users need to set a prescale for the channel, which reflects the frequency in which the modulator outputs a pulse. In the new driver, users should use sdm_config_t::sample_rate_hz.

Breaking Changes in Usage

Channel configuration was done by channel allocation, in sdm_new_channel(). In the new driver, only the duty can be changed at runtime, by sdm_channel_set_duty(). Other parameters like gpio number and prescale are only allowed to set during channel allocation.
Before further channel operations, users should enable the channel in advance, by calling sdm_channel_enable(). This function will help to manage some system level services, like Power Management.

Timer Group Driver

Timer Group driver has been redesigned into GPTimer, which aims to unify and simplify the usage of general purpose timer.

Although it’s recommended to use the the new driver APIs, the legacy driver is still available in the previous include path driver/timer.h. However, by default, including driver/timer.h will trigger the build warning below. The warning can be suppressed by the Kconfig option CONFIG_GPTIMER_SUPPRESS_DEPRECATE_WARN.

legacy timer group driver is deprecated, please migrate to driver/gptimer.h

The major breaking changes in concept and usage are listed as follows:

Breaking Changes in Concepts

timer_group_t and timer_idx_t which used to identify the hardware timer are removed from user’s code. In the new driver, a timer is represented by gptimer_handle_t.
Definition of timer clock source is moved to gptimer_clock_source_t, the previous timer_src_clk_t is not used.
Definition of timer count direction is moved to gptimer_count_direction_t, the previous timer_count_dir_t is not used.
Only level interrupt is supported, timer_intr_t and timer_intr_mode_t are not used.
Auto-reload is enabled by set the gptimer_alarm_config_t::auto_reload_on_alarm flag. timer_autoreload_t is not used.

Breaking Changes in Usage

Timer initialization is done by creating a timer instance from gptimer_new_timer(). Basic configurations like clock source, resolution and direction should be set in gptimer_config_t. Note that, specific configurations of alarm events are not needed during the installation stage of the driver.
Alarm event is configured by gptimer_set_alarm_action(), with parameters set in the gptimer_alarm_config_t.
Setting and getting count value are done by gptimer_get_raw_count() and gptimer_set_raw_count(). The driver doesn’t help convert the raw value into UTC time-stamp. Instead, the conversion should be done from user’s side as the timer resolution is also known to the user.
The driver will install the interrupt service as well if gptimer_event_callbacks_t::on_alarm is set to a valid callback function. In the callback, users do not have to deal with the low level registers (like “clear interrupt status”, “re-enable alarm event” and so on). So functions like timer_group_get_intr_status_in_isr and timer_group_get_auto_reload_in_isr are not used anymore.
To update the alarm configurations when alarm event happens, one can call gptimer_set_alarm_action() in the interrupt callback, then the alarm will be re-enabled again.
Alarm will always be re-enabled by the driver if gptimer_alarm_config_t::auto_reload_on_alarm is set to true.

UART

Removed/Deprecated items	Replacement	Remarks
`uart_isr_register()`	None	UART interrupt handling is implemented by driver itself.
`uart_isr_free()`	None	UART interrupt handling is implemented by driver itself.
`use_ref_tick` in `uart_config_t`	`uart_config_t::source_clk`	Select the clock source.
`uart_enable_pattern_det_intr()`	`uart_enable_pattern_det_baud_intr()`	Enable pattern detection interrupt.

I2C

Removed/Deprecated items	Replacement	Remarks
`i2c_isr_register()`	None	I2C interrupt handling is implemented by driver itself.
`i2c_isr_register()`	None	I2C interrupt handling is implemented by driver itself.
`i2c_opmode_t`	None	It’s not used anywhere in esp-idf.

SPI

Removed/Deprecated items	Replacement	Remarks
`spi_cal_clock()`	`spi_get_actual_clock()`	Get SPI real working frequency.

The internal header file spi_common_internal.h has been moved to esp_private/spi_common_internal.h.

LEDC

Removed/Deprecated items	Replacement	Remarks
`bit_num` in `ledc_timer_config_t`	`ledc_timer_config_t::duty_resolution`	Set resolution of the duty cycle.

Temperature Sensor Driver

The temperature sensor driver has been redesigned and it is recommended to use the new driver. However, the old driver is still available but cannot be used with the new driver simultaneously.

The new driver can be included via driver/temperature_sensor.h. The old driver is still available in the previous include path driver/temp_sensor.h. However, including driver/temp_sensor.h will trigger the build warning below by default. The warning can be suppressed by enabling the menuconfig option CONFIG_TEMP_SENSOR_SUPPRESS_DEPRECATE_WARN.

legacy temperature sensor driver is deprecated, please migrate to driver/temperature_sensor.h

Configuration contents has been changed. In the old version, users need to configure clk_div and dac_offset. While in the new version, users only need to choose tsens_range.

The process of using temperature sensor has been changed. In the old version, users can use config->start->read_celsius to get value. In the new version, users should install the temperature sensor driver firstly, by temperature_sensor_install and uninstall it when finished. For more information, please refer to Temperature Sensor .

RMT Driver

RMT driver has been redesigned (see RMT transceiver), which aims to unify and extend the usage of RMT peripheral.

Although it’s recommended to use the new driver APIs, the legacy driver is still available in the previous include path driver/rmt.h. However, including driver/rmt.h will trigger the build warning below by default. The warning can be suppressed by the Kconfig option CONFIG_RMT_SUPPRESS_DEPRECATE_WARN.

The legacy RMT driver is deprecated, please use driver/rmt_tx.h and/or driver/rmt_rx.h

The major breaking changes in concept and usage are listed as follows:

Breaking Changes in Concepts

rmt_channel_t which used to identify the hardware channel are removed from user space. In the new driver, RMT channel is represented by rmt_channel_handle_t. The channel is dynamically allocated by the driver, instead of designated by user.
rmt_item32_t is replaced by rmt_symbol_word_t, which avoids a nested union inside a struct.
rmt_mem_t is removed, as we don’t allow users to access RMT memory block (a.k.an RMTMEM) directly. Direct access to RMTMEM doesn’t make sense but make mistakes, especially when the RMT channel also connected with a DMA channel.
rmt_mem_owner_t is removed, as the ownership is controlled by driver, not by user anymore.
rmt_source_clk_t is replaced by rmt_clock_source_t, and note they’re not binary compatible.
rmt_data_mode_t is removed, the RMT memory access mode is configured to always use Non-FIFO and DMA mode.
rmt_mode_t is removed, as the driver has stand alone install functions for TX and RX channels.
rmt_idle_level_t is removed, setting IDLE level for TX channel is available in rmt_transmit_config_t::eot_level.
rmt_carrier_level_t is removed, setting carrier polarity is available in rmt_carrier_config_t::polarity_active_low.
rmt_channel_status_t and rmt_channel_status_result_t are removed, they’re not used anywhere.
Transmitting by RMT channel doesn’t expect user to prepare the RMT symbols, instead, user needs to provide an RMT Encoder to tell the driver how to convert user data into RMT symbols.

Breaking Changes in Usage

Channel installation has been separated for TX and RX channels into rmt_new_tx_channel() and rmt_new_rx_channel().
rmt_set_clk_div and rmt_get_clk_div are removed. Channel clock configuration can only be done during channel installation.
rmt_set_rx_idle_thresh and rmt_get_rx_idle_thresh are removed. In the new driver, the RX channel IDLE threshold is redesigned into a new concept rmt_receive_config_t::signal_range_max_ns.
rmt_set_mem_block_num and rmt_get_mem_block_num are removed. In the new driver, the memory block number is determined by rmt_tx_channel_config_t::mem_block_symbols and rmt_rx_channel_config_t::mem_block_symbols.
rmt_set_tx_carrier is removed, the new driver uses rmt_apply_carrier() to set carrier behavior.
rmt_set_mem_pd and rmt_get_mem_pd are removed. The memory power is managed by the driver automatically.
rmt_memory_rw_rst, rmt_tx_memory_reset and rmt_rx_memory_reset are removed. Memory reset is managed by the driver automatically.
rmt_tx_start and rmt_rx_start are merged into a single function rmt_enable(), for both TX and RX channels.
rmt_tx_stop and rmt_rx_stop are merged into a single function rmt_disable(), for both TX and RX channels.
rmt_set_memory_owner and rmt_get_memory_owner are removed. RMT memory owner guard is added automatically by the driver.
rmt_set_tx_loop_mode and rmt_get_tx_loop_mode are removed. In the new driver, the loop mode is configured in rmt_transmit_config_t::loop_count.
rmt_set_source_clk and rmt_get_source_clk are removed. Configuring clock source is only possible during channel installation by rmt_tx_channel_config_t::clk_src and rmt_rx_channel_config_t::clk_src.
rmt_set_rx_filter is removed. In the new driver, the filter threshold is redesigned into a new concept rmt_receive_config_t::signal_range_min_ns.
rmt_set_idle_level and rmt_get_idle_level are removed. Setting IDLE level for TX channel is available in rmt_transmit_config_t::eot_level.
rmt_set_rx_intr_en, rmt_set_err_intr_en, rmt_set_tx_intr_en, rmt_set_tx_thr_intr_en and rmt_set_rx_thr_intr_en are removed. The new driver doesn’t allow user to turn on/off interrupt from user space. Instead, it provides callback functions.
rmt_set_gpio and rmt_set_pin are removed. The new driver doesn’t support to switch GPIO dynamically at runtime.
rmt_config is removed. In the new driver, basic configuration is done during the channel installation stage.
rmt_isr_register and rmt_isr_deregister are removed, the interrupt is allocated by the driver itself.
rmt_driver_install is replaced by rmt_new_tx_channel() and rmt_new_rx_channel().
rmt_driver_uninstall is replaced by rmt_del_channel().
rmt_fill_tx_items, rmt_write_items and rmt_write_sample are removed. In the new driver, user needs to provide an encoder to “translate” the user data into RMT symbols.
rmt_get_counter_clock is removed, as the channel clock resolution is configured by user from rmt_tx_channel_config_t::resolution_hz.
rmt_wait_tx_done is replaced by rmt_tx_wait_all_done().
rmt_translator_init, rmt_translator_set_context and rmt_translator_get_context are removed. In the new driver, the translator has been replaced by the RMT encoder.
rmt_get_ringbuf_handle is removed. The new driver doesn’t use Ringbuffer to save RMT symbols. Instead, the incoming data are saved to the user provided buffer directly. The user buffer can even be mounted to DMA link internally.
rmt_register_tx_end_callback is replaced by rmt_tx_register_event_callbacks(), where user can register rmt_tx_event_callbacks_t::on_trans_done event callback.
rmt_set_intr_enable_mask and rmt_clr_intr_enable_mask are removed, as the interrupt is handled by the driver, user doesn’t need to take care of it.
rmt_add_channel_to_group and rmt_remove_channel_from_group are replaced by RMT sync manager. Please refer to rmt_new_sync_manager().
rmt_set_tx_loop_count is removed. The loop count in the new driver is configured in rmt_transmit_config_t::loop_count.
rmt_enable_tx_loop_autostop is removed. In the new driver, TX loop auto stop is always enabled if available, it’s not configurable anymore.

LCD

The LCD panel initialization flow is slightly changed. Now the esp_lcd_panel_init() won’t turn on the display automatically. User needs to call esp_lcd_panel_disp_on_off() to manually turn on the display. Note, this is different from turning on backlight. With this breaking change, user can flash a predefined pattern to the screen before turning on the screen. This can help avoid random noise on the screen after a power on reset.
esp_lcd_panel_disp_off() is deprecated, please use esp_lcd_panel_disp_on_off() instead.
dc_as_cmd_phase is removed. The SPI LCD driver currently doesn’t support a 9-bit SPI LCD. Please always use a dedicated GPIO to control the LCD D/C line.
The way to register RGB panel event callbacks has been moved from the esp_lcd_rgb_panel_config_t into a separate API esp_lcd_rgb_panel_register_event_callbacks(). However, the event callback signature is not changed.
Previous relax_on_idle flag in esp_lcd_rgb_panel_config_t has been renamed into esp_lcd_rgb_panel_config_t::refresh_on_demand, which expresses the same meaning but with a clear name.
If the RGB LCD is created with the refresh_on_demand flag enabled, the driver won’t start a refresh in the esp_lcd_panel_draw_bitmap(). Now users have to call esp_lcd_rgb_panel_refresh() to refresh the screen by themselves.
esp_lcd_color_space_t is deprecated, please use lcd_color_space_t to describe the color space, and use lcd_color_rgb_endian_t to describe the data order of RGB color.

Dedicated GPIO Driver

All of the dedicated GPIO related Low Level (LL) functions in cpu_ll.h have been moved to dedic_gpio_cpu_ll.h and renamed.

I2S driver

The I2S driver has been redesigned (see I2S Driver), which aims to rectify the shortcomings of the driver that were exposed when supporting all the new features of ESP32-C3 & ESP32-S3. The new driver’s APIs are available by including corresponding I2S mode’s header files driver/include/driver/i2s_std.h, driver/include/driver/i2s_pdm.h, or driver/include/driver/i2s_tdm.h.

Meanwhile, the old driver’s APIs in driver/deprecated/driver/i2s.h are still supported for backward compatibility. But there will be warnings if users keep using the old APIs in their projects, these warnings can be suppressed by the Kconfig option CONFIG_I2S_SUPPRESS_DEPRECATE_WARN.

Here is the general overview of the current I2S files:

Breaking changes in Concepts

Independent TX/RX channels

The minimum control unit in new I2S driver are now individual TX/RX channels instead of an entire I2S controller (that consistes of multiple channels).

The TX and RX channels of the same I2S controller can be controlled separately, meaning that they are configured such that they can be started or stopped separately.
The c:type:i2s_chan_handle_t handle type is used to uniquely identify I2S channels. All the APIs will require the channel handle and users need to maintain the channel handles by themselves.
On the ESP32-C3 and ESP32-S3, TX and RX channels in the same controller can be configured to different clocks or modes.
However, on the ESP32 and ESP32-S2, the TX and RX channels of the same controller still share some hardware resources. Thus, configurations may cause one channel to affect another channel in the same controller.
The channels can be registered to an available I2S controller automatically by setting i2s_port_t::I2S_NUM_AUTO as I2S port ID which will cause the driver to search for the available TX/RX channels. However, the driver also supports registering channels to a specific port.
In order to distinguish between TX/RX channels and sound channels, the term ‘channel’ in the context of the I2S driver will only refer to TX/RX channels. Meanwhile, sound channels will be referred to as “slots”.

I2S Mode Categorization

I2S communication modes are categorized into the following three modes. Note that:

Standard mode: Standard mode always has two slots, it can support Philips, MSB, and PCM (short frame sync) formats. Please refer to driver/include/driver/i2s_std.h for more details.
PDM mode: PDM mode only supports two slots with 16-bit data width, but the configurations of PDM TX and PDM RX are slightly different. For PDM TX, the sample rate can be set by i2s_pdm_tx_clk_config_t::sample_rate, and its clock frequency depends on the up-sampling configuration. For PDM RX, the sample rate can be set by i2s_pdm_rx_clk_config_t::sample_rate, and its clock frequency depends on the down-sampling configuration. Please refer to driver/include/driver/i2s_pdm.h for details.
TDM mode: TDM mode can support up to 16 slots. It can work in Philips, MSB, PCM (short frame sync), and PCM (long frame sync) formats. Please refer to driver/include/driver/i2s_tdm.h for details.

When allocating a new channel in a specific mode, users should initialize that channel by its corresponding function. It is strongly recommended to use the helper macros to generate the default configurations in case the default values are changed in the future.

Independent Slot and Clock Configuration

The slot configurations and clock configurations can be configured separately.

Call i2s_channel_init_std_mode(), i2s_channel_init_pdm_rx_mode(), i2s_channel_init_pdm_tx_mode(), or i2s_channel_init_tdm_mode() to initialize the slot/clock/gpio_pin configurations.
Calling i2s_channel_reconfig_std_slot(), i2s_channel_reconfig_pdm_rx_slot(), i2s_channel_reconfig_pdm_tx_slot(), or i2s_channel_reconfig_tdm_slot() can change the slot configurations after initialization.
Calling i2s_channel_reconfig_std_clock(), i2s_channel_reconfig_pdm_rx_clock(), i2s_channel_reconfig_pdm_tx_clock(), or i2s_channel_reconfig_tdm_clock() can change the clock configurations after initialization.
Calling i2s_channel_reconfig_std_gpio(), i2s_channel_reconfig_pdm_rx_gpio(), i2s_channel_reconfig_pdm_tx_gpio(), or i2s_channel_reconfig_tdm_gpio() can change the GPIO configurations after initialization.

Misc

States and state-machine are adopted in the new I2S driver to avoid APIs called in wrong state.
ADC and DAC modes are removed. They will only be supported in their own drivers and the legacy I2S driver.

Breaking Changes in Usage

To use the new I2S driver, please follow these steps:

Call i2s_new_channel() to acquire channel handles. We should specify the work role and I2S port in this step. Besides, the TX or RX channel handle will be generated by the driver. Inputting both two TX and RX channel handles is not necessary but at least one handle is needed. In the case of inputting both two handles, the driver will work at the duplex mode. Both TX and RX channels will be avaliable on a same port, and they will share the MCLK, BCLK and WS signal. But if only one of the TX or RX channel handle is inputted, this channel will only work in the simplex mode.
Call i2s_channel_init_std_mode(), i2s_channel_init_pdm_rx_mode(), i2s_channel_init_pdm_tx_mode() or i2s_channel_init_tdm_mode() to initialize the channel to the specified mode. Corresponding slot, clock and GPIO configurations are needed in this step.
(Optional) Call i2s_channel_register_event_callback() to register the ISR event callback functions. I2S events now can be received by the callback function synchronously, instead of from the event queue asynchronously.
Call i2s_channel_enable() to start the hardware of I2S channel. In the new driver, I2S won’t start automatically after installed, and users are supposed to know clearly whether the channel has started or not.
Read or write data by i2s_channel_read() or i2s_channel_write(). Certainly, only the RX channel handle is suppoesd to be inputted in i2s_channel_read() and the TX channel handle in i2s_channel_write().
(Optional) The slot, clock and GPIO configurations can be changed by corresponding ‘reconfig’ functions, but i2s_channel_disable() must be called before updating the configurations.
Call i2s_channel_disable() to stop the hardware of I2S channel.
Call i2s_del_channel() to delete and release the resources of the channel if it is not needed any more, but the channel must be disabled before deleting it.

TWAI Driver

The deprecated CAN peripheral driver is removed. Please use TWAI driver instead (i.e. include driver/twai.h in your application).

Register Access Macros

Previously, all register access macros could be used as expressions, so the following was allowed:

uint32_t val = REG_SET_BITS(reg, mask);

In ESP-IDF v5.0, register access macros which write or read-modify-write the register can no longer be used as expressions, and can only be used as statements. This applies to the following macros: REG_WRITE, REG_SET_BIT, REG_CLR_BIT, REG_SET_BITS, REG_SET_FIELD, WRITE_PERI_REG, CLEAR_PERI_REG_MASK, SET_PERI_REG_MASK, SET_PERI_REG_BITS.

To store the value which would have been written into the register, split the operation as follows:

uint32_t new_val = REG_READ(reg) | mask;
REG_WRITE(reg, new_val);

To get the value of the register after modification (which may be different from the value written), add an explicit read:

REG_SET_BITS(reg, mask);
uint32_t new_val = REG_READ(reg);

Provide feedback about this document