The RMT (Remote Control Transceiver) peripheral was designed to act as an infrared transceiver. However, due to the flexibility of its data format, RMT can be extended to a versatile and general-purpose transceiver, transmitting or receiving many other types of signals. From the perspective of network layering, the RMT hardware contains both physical and data link layers. The physical layer defines the communication media and bit signal representation. The data link layer defines the format of an RMT frame. The minimal data unit in the frame is called the **RMT symbol**, which is represented by :cpp:type:`rmt_symbol_word_t` in the driver.
{IDF_TARGET_NAME} contains multiple channels in the RMT peripheral [1]_. Each channel can be independently configured as either transmitter or receiver.
The RMT hardware defines data in its own pattern -- the **RMT symbol**. The diagram below illustrates the bit fields of an RMT symbol. Each symbol consists of two pairs of two values. The first value in the pair is a 15-bit value representing the signal's duration in units of RMT ticks. The second in the pair is a 1-bit value representing the signal's logic level, i.e., high or low.
The driver encodes the user's data into RMT data format, then the RMT transmitter can generate the waveforms according to the encoding artifacts. It is also possible to modulate a high-frequency carrier signal before being routed to a GPIO pad.
The RMT receiver can sample incoming signals into RMT data format, and store the data in memory. It is also possible to tell the receiver the basic characteristics of the incoming signal, so that the signal's stop condition can be recognized, and signal glitches and noise can be filtered out. The RMT peripheral also supports demodulating the high-frequency carrier from the base signal.
The description of the RMT functionality is divided into the following sections:
-:ref:`rmt-resource-allocation` - covers how to allocate and properly configure RMT channels. It also covers how to recycle channels and other resources when they are no longer used.
-:ref:`rmt-carrier-modulation-and demodulation` - describes how to modulate and demodulate the carrier signals for TX and RX channels respectively.
-:ref:`rmt-enable-and-disable-channel` - shows how to enable and disable the RMT channel.
-:ref:`rmt-initiate-tx-transaction` - describes the steps to initiate a transaction for a TX channel.
-:ref:`rmt-initiate-rx-transaction` - describes the steps to initiate a transaction for an RX channel.
-:ref:`rmt-multiple-channels-simultaneous-transmission` - describes how to collect multiple channels into a sync group so that their transmissions can be started simultaneously.
-:ref:`rmt-rmt-encoder` - focuses on how to write a customized encoder by combining multiple primitive encoders that are provided by the driver.
-:ref:`rmt-power-management` - describes how different clock sources affects power consumption.
-:ref:`rmt-iram-safe` - describes how disabling the cache affects the RMT driver, and tips to mitigate it.
-:ref:`rmt-thread-safety` - lists which APIs are guaranteed to be thread-safe by the driver.
-:ref:`rmt-kconfig-options` - describes the various Kconfig options supported by the RMT driver.
Both RMT TX and RX channels are represented by :cpp:type:`rmt_channel_handle_t` in the driver. The driver internally manages which channels are available and hands out a free channel on request.
To install an RMT TX channel, there is a configuration structure that needs to be given in advance :cpp:type:`rmt_tx_channel_config_t`. The following list describes each member of the configuration structure.
-:cpp:member:`rmt_tx_channel_config_t::gpio_num` sets the GPIO number used by the transmitter.
-:cpp:member:`rmt_tx_channel_config_t::clk_src` selects the source clock for the RMT channel. The available clocks are listed in :cpp:type:`rmt_clock_source_t`. Note that, the selected clock is also used by other channels, which means the user should ensure this configuration is the same when allocating other channels, regardless of TX or RX. For the effect on the power consumption of different clock sources, please refer to the :ref:`rmt-power-management` section.
-:cpp:member:`rmt_tx_channel_config_t::resolution_hz` sets the resolution of the internal tick counter. The timing parameter of the RMT signal is calculated based on this **tick**.
-:cpp:member:`rmt_tx_channel_config_t::mem_block_symbols` has a slightly different meaning based on if the DMA backend is enabled or not.
- If the DMA is enabled via :cpp:member:`rmt_tx_channel_config_t::with_dma`, then this field controls the size of the internal DMA buffer. To achieve a better throughput and smaller CPU overhead, you can set a larger value, e.g., ``1024``.
- If DMA is not used, this field controls the size of the dedicated memory block owned by the channel, which should be at least {IDF_TARGET_SOC_RMT_MEM_WORDS_PER_CHANNEL}.
-:cpp:member:`rmt_tx_channel_config_t::trans_queue_depth` sets the depth of the internal transaction queue, the deeper the queue, the more transactions can be prepared in the backlog.
-:cpp:member:`rmt_tx_channel_config_t::invert_out` is used to decide whether to invert the RMT signal before sending it to the GPIO pad.
-:cpp:member:`rmt_tx_channel_config_t::with_dma` enables the DMA backend for the channel. Using the DMA allows a significant amount of the channel's workload to be offloaded from the CPU. However, the DMA backend is not available on all ESP chips, please refer to [`TRM <{IDF_TARGET_TRM_EN_URL}#rmt>`__] before you enable this option. Or you might encounter a :c:macro:`ESP_ERR_NOT_SUPPORTED` error.
-:cpp:member:`rmt_tx_channel_config_t::io_loop_back` enables both input and output capabilities on the channel's assigned GPIO. Thus, by binding a TX and RX channel to the same GPIO, loopback can be achieved.
-:cpp:member:`rmt_tx_channel_config_t::io_od_mode` configures the channel's assigned GPIO as open-drain. When combined with :cpp:member:`rmt_tx_channel_config_t::io_loop_back`, a bi-directional bus (e.g., 1-wire) can be achieved.
-:cpp:member:`rmt_tx_channel_config_t::intr_priority` Set the priority of the interrupt. If set to ``0`` , then the driver will use a interrupt with low or medium priority (priority level may be one of 1,2 or 3), otherwise use the priority indicated by :cpp:member:`rmt_tx_channel_config_t::intr_priority`. Please use the number form (1,2,3) , not the bitmask form ((1<<1),(1<<2),(1<<3)). Please pay attention that once the interrupt priority is set, it cannot be changed until :cpp:func:`rmt_del_channel` is called.
Once the :cpp:type:`rmt_tx_channel_config_t` structure is populated with mandatory parameters, users can call :cpp:func:`rmt_new_tx_channel` to allocate and initialize a TX channel. This function returns an RMT channel handle if it runs correctly. Specifically, when there are no more free channels in the RMT resource pool, this function returns :c:macro:`ESP_ERR_NOT_FOUND` error. If some feature (e.g., DMA backend) is not supported by the hardware, it returns :c:macro:`ESP_ERR_NOT_SUPPORTED` error.
To install an RMT RX channel, there is a configuration structure that needs to be given in advance :cpp:type:`rmt_rx_channel_config_t`. The following list describes each member of the configuration structure.
-:cpp:member:`rmt_rx_channel_config_t::gpio_num` sets the GPIO number used by the receiver.
-:cpp:member:`rmt_rx_channel_config_t::clk_src` selects the source clock for the RMT channel. The available clocks are listed in :cpp:type:`rmt_clock_source_t`. Note that, the selected clock is also used by other channels, which means the user should ensure this configuration is the same when allocating other channels, regardless of TX or RX. For the effect on the power consumption of different clock sources, please refer to the :ref:`rmt-power-management` section.
-:cpp:member:`rmt_rx_channel_config_t::resolution_hz` sets the resolution of the internal tick counter. The timing parameter of the RMT signal is calculated based on this **tick**.
-:cpp:member:`rmt_rx_channel_config_t::mem_block_symbols` has a slightly different meaning based on whether the DMA backend is enabled.
- If the DMA is enabled via :cpp:member:`rmt_rx_channel_config_t::with_dma`, this field controls the maximum size of the DMA buffer.
- If DMA is not used, this field controls the size of the dedicated memory block owned by the channel, which should be at least {IDF_TARGET_SOC_RMT_MEM_WORDS_PER_CHANNEL}.
-:cpp:member:`rmt_rx_channel_config_t::invert_in` is used to invert the input signals before it is passed to the RMT receiver. The inversion is done by the GPIO matrix instead of by the RMT peripheral.
-:cpp:member:`rmt_rx_channel_config_t::with_dma` enables the DMA backend for the channel. Using the DMA allows a significant amount of the channel's workload to be offloaded from the CPU. However, the DMA backend is not available on all ESP chips, please refer to [`TRM <{IDF_TARGET_TRM_EN_URL}#rmt>`__] before you enable this option. Or you might encounter a :c:macro:`ESP_ERR_NOT_SUPPORTED` error.
-:cpp:member:`rmt_rx_channel_config_t::io_loop_back` enables both input and output capabilities on the channel's assigned GPIO. Thus, by binding a TX and RX channel to the same GPIO, loopback can be achieved.
-:cpp:member:`rmt_rx_channel_config_t::intr_priority` Set the priority of the interrupt. If set to ``0`` , then the driver will use a interrupt with low or medium priority (priority level may be one of 1,2 or 3), otherwise use the priority indicated by :cpp:member:`rmt_rx_channel_config_t::intr_priority`. Please use the number form (1,2,3) , not the bitmask form ((1<<1),(1<<2),(1<<3)). Please pay attention that once the interrupt priority is set, it cannot be changed until :cpp:func:`rmt_del_channel` is called.
Once the :cpp:type:`rmt_rx_channel_config_t` structure is populated with mandatory parameters, users can call :cpp:func:`rmt_new_rx_channel` to allocate and initialize an RX channel. This function returns an RMT channel handle if it runs correctly. Specifically, when there are no more free channels in the RMT resource pool, this function returns :c:macro:`ESP_ERR_NOT_FOUND` error. If some feature (e.g., DMA backend) is not supported by the hardware, it returns :c:macro:`ESP_ERR_NOT_SUPPORTED` error.
Due to a software limitation in the GPIO driver, when both TX and RX channels are bound to the same GPIO, ensure the RX Channel is initialized before the TX Channel. If the TX Channel was set up first, then during the RX Channel setup, the previous RMT TX Channel signal will be overridden by the GPIO control signal.
If a previously installed RMT channel is no longer needed, it is recommended to recycle the resources by calling :cpp:func:`rmt_del_channel`, which in return allows the underlying software and hardware resources to be reused for other purposes.
The RMT transmitter can generate a carrier wave and modulate it onto the message signal. Compared to the message signal, the carrier signal's frequency is significantly higher. In addition, the user can only set the frequency and duty cycle for the carrier signal. The RMT receiver can demodulate the carrier signal from the incoming signal. Note that, carrier modulation and demodulation are not supported on all ESP chips, please refer to [`TRM <{IDF_TARGET_TRM_EN_URL}#rmt>`__] before configuring the carrier, or you might encounter a :c:macro:`ESP_ERR_NOT_SUPPORTED` error.
-:cpp:member:`rmt_carrier_config_t::frequency_hz` sets the carrier frequency, in Hz.
-:cpp:member:`rmt_carrier_config_t::duty_cycle` sets the carrier duty cycle.
-:cpp:member:`rmt_carrier_config_t::polarity_active_low` sets the carrier polarity, i.e., on which level the carrier is applied.
-:cpp:member:`rmt_carrier_config_t::always_on` sets whether to output the carrier even when the data transmission has finished. This configuration is only valid for the TX channel.
For the RX channel, we should not set the carrier frequency exactly to the theoretical value. It is recommended to leave a tolerance for the carrier frequency. For example, in the snippet below, we set the frequency to 25 KHz, instead of the 38 KHz configured on the TX side. The reason is that reflection and refraction occur when a signal travels through the air, leading to distortion on the receiver side.
When an event occurs on an RMT channel (e.g., transmission or receiving is completed), the CPU is notified of this event via an interrupt. If you have some function that needs to be called when a particular events occur, you can register a callback for that event to the RMT driver's ISR (Interrupt Service Routine) by calling :cpp:func:`rmt_tx_register_event_callbacks` and :cpp:func:`rmt_rx_register_event_callbacks` for TX and RX channel respectively. Since the registered callback functions are called in the interrupt context, the user should ensure the callback function does not block, e.g., by making sure that only FreeRTOS APIs with the ``FromISR`` suffix are called from within the function. The callback function has a boolean return value used to indicate whether a higher priority task has been unblocked by the callback.
-:cpp:member:`rmt_tx_event_callbacks_t::on_trans_done` sets a callback function for the "trans-done" event. The function prototype is declared in :cpp:type:`rmt_tx_done_callback_t`.
-:cpp:member:`rmt_rx_event_callbacks_t::on_recv_done` sets a callback function for "receive-done" event. The function prototype is declared in :cpp:type:`rmt_rx_done_callback_t`.
The "receive-done" is not equivalent to "receive-finished". This callback can also be called at a "partial-receive-done" time, for many times during one receive transaction.
Users can save their own context in :cpp:func:`rmt_tx_register_event_callbacks` and :cpp:func:`rmt_rx_register_event_callbacks` as well, via the parameter ``user_data``. The user data is directly passed to each callback function.
In the callback function, users can fetch the event-specific data that is filled by the driver in the ``edata``. Note that the ``edata`` pointer is **only** valid during the callback, please do not try to save this pointer and use that outside of the callback function.
-:cpp:member:`rmt_tx_done_event_data_t::num_symbols` indicates the number of transmitted RMT symbols. This also reflects the size of the encoding artifacts. Please note, this value accounts for the ``EOF`` symbol as well, which is appended by the driver to mark the end of one transaction.
-:cpp:member:`rmt_rx_done_event_data_t::received_symbols` points to the received RMT symbols. These symbols are saved in the ``buffer`` parameter of the :cpp:func:`rmt_receive` function. Users should not free this receive buffer before the callback returns. If you also enabled the partial receive feature, then the user buffer will be used as a "second level buffer", where its content can be overwritten by data comes in afterwards. In this case, you should copy the received data to another place if you want to keep it or process it later.
-:cpp:member:`rmt_rx_done_event_data_t::num_symbols` indicates the number of received RMT symbols. This value is not larger than the ``buffer_size`` parameter of :cpp:func:`rmt_receive` function. If the ``buffer_size`` is not sufficient to accommodate all the received RMT symbols, the driver only keeps the maximum number of symbols that the buffer can hold, and excess symbols are discarded or ignored.
-:cpp:member:`rmt_rx_done_event_data_t::is_last` indicates whether the current received buffer is the last one in the transaction. This is useful when you enable the partial reception feature by :cpp:member:`rmt_receive_config_t::extra_flags::en_partial_rx`.
:cpp:func:`rmt_enable` must be called in advance before transmitting or receiving RMT symbols. For TX channels, enabling a channel enables a specific interrupt and prepares the hardware to dispatch transactions. For RX channels, enabling a channel enables an interrupt, but the receiver is not started during this time, as the characteristics of the incoming signal have yet to be specified. The receiver is started in :cpp:func:`rmt_receive`.
:cpp:func:`rmt_disable` does the opposite by disabling the interrupt and clearing any pending interrupts. The transmitter and receiver are disabled as well.
RMT is a special communication peripheral, as it is unable to transmit raw byte streams like SPI and I2C. RMT can only send data in its own format :cpp:type:`rmt_symbol_word_t`. However, the hardware does not help to convert the user data into RMT symbols, this can only be done in software by the so-called **RMT Encoder**. The encoder is responsible for encoding user data into RMT symbols and then writing to the RMT memory block or the DMA buffer. For how to create an RMT encoder, please refer to :ref:`rmt-rmt-encoder`.
Once you created an encoder, you can initiate a TX transaction by calling :cpp:func:`rmt_transmit`. This function takes several positional parameters like channel handle, encoder handle, and payload buffer. Besides, you also need to provide a transmission-specific configuration in :cpp:type:`rmt_transmit_config_t`:
-:cpp:member:`rmt_transmit_config_t::loop_count` sets the number of transmission loops. After the transmitter has finished one round of transmission, it can restart the same transmission again if this value is not set to zero. As the loop is controlled by hardware, the RMT channel can be used to generate many periodic sequences with minimal CPU intervention.
- Setting :cpp:member:`rmt_transmit_config_t::loop_count` to `-1` means an infinite loop transmission. In this case, the channel does not stop until :cpp:func:`rmt_disable` is called. The "trans-done" event is not generated as well.
- Setting :cpp:member:`rmt_transmit_config_t::loop_count` to a positive number means finite number of iterations. In this case, the "trans-done" event is when the specified number of iterations have completed.
..note::
The **loop transmit** feature is not supported on all ESP chips, please refer to [`TRM <{IDF_TARGET_TRM_EN_URL}#rmt>`__] before you configure this option, or you might encounter :c:macro:`ESP_ERR_NOT_SUPPORTED` error.
-:cpp:member:`rmt_transmit_config_t::eot_level` sets the output level when the transmitter finishes working or stops working by calling :cpp:func:`rmt_disable`.
-:cpp:member:`rmt_transmit_config_t::queue_nonblocking` sets whether to wait for a free slot in the transaction queue when it is full. If this value is set to ``true``, then the function will return with an error code :c:macro:`ESP_ERR_INVALID_STATE` when the queue is full. Otherwise, the function will block until a free slot is available in the queue.
There is a limitation in the transmission size if the :cpp:member:`rmt_transmit_config_t::loop_count` is set to non-zero, i.e., to enable the loop feature. The encoded RMT symbols should not exceed the capacity of the RMT hardware memory block size, or you might see an error message like ``encoding artifacts can't exceed hw memory block for loop transmission``. If you have to start a large transaction by loop, you can try either of the following methods.
- Increase the :cpp:member:`rmt_tx_channel_config_t::mem_block_symbols`. This approach does not work if the DMA backend is also enabled.
- Customize an encoder and construct an infinite loop in the encoding function. See also :ref:`rmt-rmt-encoder`.
Internally, :cpp:func:`rmt_transmit` constructs a transaction descriptor and sends it to a job queue, which is dispatched in the ISR. So it is possible that the transaction is not started yet when :cpp:func:`rmt_transmit` returns. You cannot recycle or modify the payload buffer until the transaction is finished. You can get the transaction completion event by registering a callback function via :cpp:func:`rmt_tx_register_event_callbacks`. To ensure all pending transactions to complete, you can also use :cpp:func:`rmt_tx_wait_all_done`.
In some real-time control applications (e.g., to make two robotic arms move simultaneously), you do not want any time drift between different channels. The RMT driver can help to manage this by creating a so-called **Sync Manager**. The sync manager is represented by :cpp:type:`rmt_sync_manager_handle_t` in the driver. The procedure of RMT sync transmission is shown as follows:
:cpp:func:`rmt_new_sync_manager` can return a manager handle on success. This function could also fail due to various errors such as invalid arguments, etc. Especially, when the sync manager has been installed before, and there are no hardware resources to create another manager, this function reports :c:macro:`ESP_ERR_NOT_FOUND` error. In addition, if the sync manager is not supported by the hardware, it reports a :c:macro:`ESP_ERR_NOT_SUPPORTED` error. Please refer to [`TRM <{IDF_TARGET_TRM_EN_URL}#rmt>`__] before using the sync manager feature.
For any managed TX channel, it does not start the machine until :cpp:func:`rmt_transmit` has been called on all channels in :cpp:member:`rmt_sync_manager_config_t::tx_channel_array`. Before that, the channel is just put in a waiting state. TX channels will usually complete their transactions at different times due to differing transactions, thus resulting in a loss of sync. So before restarting a simultaneous transmission, the user needs to call :cpp:func:`rmt_sync_reset` to synchronize all channels again.
As also discussed in the :ref:`rmt-enable-and-disable-channel`, calling :cpp:func:`rmt_enable` does not prepare an RX to receive RMT symbols. The user needs to specify the basic characteristics of the incoming signals in :cpp:type:`rmt_receive_config_t`:
-:cpp:member:`rmt_receive_config_t::signal_range_min_ns` specifies the minimal valid pulse duration in either high or low logic levels. A pulse width that is smaller than this value is treated as a glitch, and ignored by the hardware.
-:cpp:member:`rmt_receive_config_t::signal_range_max_ns` specifies the maximum valid pulse duration in either high or low logic levels. A pulse width that is bigger than this value is treated as **Stop Signal**, and the receiver generates receive-complete event immediately.
- If the incoming packet is long, that they cannot be stored in the user buffer at once, you can enable the partial reception feature by setting :cpp:member:`rmt_receive_config_t::extra_flags::en_partial_rx` to ``true``. In this case, the driver invokes :cpp:member:`rmt_rx_event_callbacks_t::on_recv_done` callback multiple times during one transaction, when the user buffer is **almost full**. You can check the value of :cpp:member::`rmt_rx_done_event_data_t::is_last` to know if the transaction is about to finish.
The RMT receiver starts the RX machine after the user calls :cpp:func:`rmt_receive` with the provided configuration above. Note that, this configuration is transaction specific, which means, to start a new round of reception, the user needs to set the :cpp:type:`rmt_receive_config_t` again. The receiver saves the incoming signals into its internal memory block or DMA buffer, in the format of :cpp:type:`rmt_symbol_word_t`.
Due to the limited size of the memory block, the RMT receiver can only save short frames whose length is not longer than the memory block capacity. Long frames are truncated by the hardware, and the driver reports an error message: ``hw buffer too small, received symbols truncated``.
The copy destination should be provided in the ``buffer`` parameter of :cpp:func:`rmt_receive` function. If this buffer overflows due to an insufficient buffer size, the receiver can continue to work, but overflowed symbols are dropped and the following error message is reported: ``user buffer too small, received symbols truncated``. Please take care of the lifecycle of the ``buffer`` parameter, ensuring that the buffer is not recycled before the receiver is finished or stopped.
The receiver is stopped by the driver when it finishes working, i.e., receive a signal whose duration is bigger than :cpp:member:`rmt_receive_config_t::signal_range_max_ns`. The user needs to call :cpp:func:`rmt_receive` again to restart the receiver, if necessary. The user can get the received data in the :cpp:member:`rmt_rx_event_callbacks_t::on_recv_done` callback. See also :ref:`rmt-register-event-callbacks` for more information.
An RMT encoder is part of the RMT TX transaction, whose responsibility is to generate and write the correct RMT symbols into hardware memory or DMA buffer at a specific time. There are some special restrictions for an encoding function:
- During a single transaction, the encoding function may be called multiple times. This is necessary because the target RMT memory block cannot hold all the artifacts at once. To overcome this limitation, the driver utilizes a **ping-pong** approach, where the encoding session is divided into multiple parts. This means that the encoder needs to **keep track of its state** to continue encoding from where it left off in the previous part.
- The encoding function is running in the ISR context. To speed up the encoding session, it is highly recommended to put the encoding function into IRAM. This can also avoid the cache miss during encoding.
To help get started with the RMT driver faster, some commonly used encoders are provided out-of-the-box. They can either work alone or be chained together into a new encoder. See also `Composite Pattern <https://en.wikipedia.org/wiki/Composite_pattern>`__ for the principle behind it. The driver has defined the encoder interface in :cpp:type:`rmt_encoder_t`, it contains the following functions:
- The function might be called multiple times within a single transaction. The encode function should return the state of the current encoding session.
- The supported states are listed in the :cpp:type:`rmt_encode_state_t`. If the result contains :cpp:enumerator:`RMT_ENCODING_COMPLETE`, it means the current encoder has finished work.
- If the result contains :cpp:enumerator:`RMT_ENCODING_MEM_FULL`, the program needs to yield from the current session, as there is no space to save more encoding artifacts.
A copy encoder is created by calling :cpp:func:`rmt_new_copy_encoder`. A copy encoder's main functionality is to copy the RMT symbols from user space into the driver layer. It is usually used to encode ``const`` data, i.e., data does not change at runtime after initialization such as the leading code in the IR protocol.
A configuration structure :cpp:type:`rmt_copy_encoder_config_t` should be provided in advance before calling :cpp:func:`rmt_new_copy_encoder`. Currently, this configuration is reserved for future expansion, and has no specific use or setting items for now.
A bytes encoder is created by calling :cpp:func:`rmt_new_bytes_encoder`. The bytes encoder's main functionality is to convert the user space byte stream into RMT symbols dynamically. It is usually used to encode dynamic data, e.g., the address and command fields in the IR protocol.
-:cpp:member:`rmt_bytes_encoder_config_t::bit0` and :cpp:member:`rmt_bytes_encoder_config_t::bit1` are necessary to specify the encoder how to represent bit zero and bit one in the format of :cpp:type:`rmt_symbol_word_t`.
-:cpp:member:`rmt_bytes_encoder_config_t::msb_first` sets the bit endianess of each byte. If it is set to true, the encoder encodes the **Most Significant Bit** first. Otherwise, it encodes the **Least Significant Bit** first.
Besides the primitive encoders provided by the driver, the user can implement his own encoder by chaining the existing encoders together. A common encoder chain is shown as follows:
This section demonstrates how to write an NEC encoder. The NEC IR protocol uses pulse distance encoding of the message bits. Each pulse burst is ``562.5 µs`` in length, logical bits are transmitted as follows. It is worth mentioning that the least significant bit of each byte is sent first.
A full sample code can be found in :example:`peripherals/rmt/ir_nec_transceiver`. In the above snippet, we use a ``switch-case`` and several ``goto`` statements to implement a `Finite-state machine <https://en.wikipedia.org/wiki/Finite-state_machine>`__ . With this pattern, users can construct much more complex IR protocols.
When power management is enabled, i.e., :ref:`CONFIG_PM_ENABLE` is on, the system adjusts the APB frequency before going into Light-sleep, thus potentially changing the resolution of the RMT internal counter.
However, the driver can prevent the system from changing APB frequency by acquiring a power management lock of type :cpp:enumerator:`ESP_PM_APB_FREQ_MAX`. Whenever the user creates an RMT channel that has selected :cpp:enumerator:`RMT_CLK_SRC_APB` as the clock source, the driver guarantees that the power management lock is acquired after the channel enabled by :cpp:func:`rmt_enable`. Likewise, the driver releases the lock after :cpp:func:`rmt_disable` is called for the same channel. This also reveals that the :cpp:func:`rmt_enable` and :cpp:func:`rmt_disable` should appear in pairs.
If the channel clock source is selected to others like :cpp:enumerator:`RMT_CLK_SRC_XTAL`, then the driver does not install a power management lock for it, which is more suitable for a low-power application as long as the source clock can still provide sufficient resolution.
By default, the RMT interrupt is deferred when the Cache is disabled for reasons like writing or erasing the main Flash. Thus the transaction-done interrupt does not get handled in time, which is not acceptable in a real-time application. What is worse, when the RMT transaction relies on **ping-pong** interrupt to successively encode or copy RMT symbols, a delayed interrupt can lead to an unpredictable result.
There is a Kconfig option :ref:`CONFIG_RMT_ISR_IRAM_SAFE` that has the following features:
This Kconfig option allows the interrupt handler to run while the cache is disabled but comes at the cost of increased IRAM consumption.
Another Kconfig option :ref:`CONFIG_RMT_RECV_FUNC_IN_IRAM` can place :cpp:func:`rmt_receive` into the IRAM as well. So that the receive function can be used even when the flash cache is disabled.
The factory function :cpp:func:`rmt_new_tx_channel`, :cpp:func:`rmt_new_rx_channel` and :cpp:func:`rmt_new_sync_manager` are guaranteed to be thread-safe by the driver, which means, user can call them from different RTOS tasks without protection by extra locks.
Other functions that take the :cpp:type:`rmt_channel_handle_t` and :cpp:type:`rmt_sync_manager_handle_t` as the first positional parameter, are not thread-safe. which means the user should avoid calling them from multiple tasks.
-:ref:`CONFIG_RMT_ISR_IRAM_SAFE` controls whether the default ISR handler can work when cache is disabled, see also :ref:`rmt-iram-safe` for more information.
-:ref:`CONFIG_RMT_ENABLE_DEBUG_LOG` is used to enable the debug log at the cost of increased firmware binary size.
The encoding for the RMT transmission is carried out within the ISR context. Should the RMT encoding process be prolonged (for example, through logging or tracing the procedure) or if it is delayed due to interrupt latency and preemptive interrupts, the hardware transmitter might read from the memory before the encoder has written to it. Consequently, the transmitter would end up sending outdated data. Although it's not possible to instruct the transmitter to pause and wait, this issue can be mitigated by employing a combination of the following strategies:
- Increase the :cpp:member:`rmt_tx_channel_config_t::mem_block_symbols`, in steps of {IDF_TARGET_SOC_RMT_MEM_WORDS_PER_CHANNEL}.
- Place the encoding function in the IRAM with ``IRAM_ATTR`` attribute.
- Enable the :cpp:member:`rmt_tx_channel_config_t::with_dma` if DMA is available.
- Install the RMT driver on a separate CPU core to avoid competing for the same CPU resources with other interrupt heavy peripherals (e.g. WiFi, Bluetooth).
Different ESP chip series might have different numbers of RMT channels. Please refer to [`TRM <{IDF_TARGET_TRM_EN_URL}#rmt>`__] for details. The driver does not forbid you from applying for more RMT channels, but it returns an error when there are no hardware resources available. Please always check the return value when doing `Resource Allocation <#resource-allocation>`__.
The callback function, e.g., :cpp:member:`rmt_tx_event_callbacks_t::on_trans_done`, and the functions invoked by itself should also reside in IRAM, users need to take care of this by themselves.
