The PCNT (Pulse Counter) module is designed to count the number of rising and/or falling edges of input signals. The {IDF_TARGET_NAME} contains multiple pulse counter units in the module. [1]_ Each unit is in effect an independent counter with multiple channels, where each channel can increment/decrement the counter on a rising/falling edge. Furthermore, each channel can be configured separately.
PCNT channels can react to signals of **edge** type and **level** type, however for simple applications, detecting the edge signal is usually sufficient. PCNT channels can be configured react to both pulse edges (i.e., rising and falling edge), and can be configured to increase, decrease or do nothing to the unit's counter on each edge. The level signal is the so-called **control signal**, which is used to control the counting mode of the edge signals that are attached to the same channel. By combining the usage of both edge and level signals, a PCNT unit can act as a **quadrature decoder**.
-`Resource Allocation <#resource-allocation>`__ - covers how to allocate PCNT units and channels with properly set of configurations. It also covers how to recycle the resources when they finished working.
-`Set Up Channel Actions <#set-up-channel-actions>`__ - covers how to configure the PCNT channel to behave on different signal edges and levels.
-`Watch Points <#watch-points>`__ - describes how to configure PCNT watch points (i.e., tell PCNT unit to trigger an event when the count reaches a certain value).
-`Register Event Callbacks <#register-event-callbacks>`__ - describes how to hook user specific code to the watch point event callback function.
-`Unit IO Control <#unit-io-control>`__ - describes IO control functions of PCNT unit, like enable glitch filter, start and stop unit, get and clear count value.
-`Power Management <#power-management>`__ - describes what functionality will prevent the chip from going into low power mode.
-`IRAM Safe <#iram-safe>`__ - describes tips on how to make the PCNT interrupt and IO control functions work better along with a disabled cache.
-`Thread Safety <#thread-safety>`__ - lists which APIs are guaranteed to be thread safe by the driver.
-`Kconfig options <#kconfig-options>`__ - lists the supported Kconfig options that can be used to make a different effect on driver behavior.
Resource Allocation
^^^^^^^^^^^^^^^^^^^
The PCNT unit and channel are represented by :cpp:type:`pcnt_unit_handle_t` and :cpp:type:`pcnt_channel_handle_t` respectively. All available units and channels are maintained by the driver in a resource pool, so the user doesn't need to know the exact underlying instance ID.
Install PCNT Unit
~~~~~~~~~~~~~~~~~
To install a PCNT unit, there's a configuration structure that needs to be given in advance: :cpp:type:`pcnt_unit_config_t`:
-:cpp:member:`pcnt_unit_config_t::low_limit` and :cpp:member:`pcnt_unit_config_t::high_limit` specify the range for the internal counter. Counter will back to zero when it crosses either limit value.
Unit allocation and initialization is done by calling a function :cpp:func:`pcnt_new_unit` with :cpp:type:`pcnt_unit_config_t` as an input parameter. The function will return a PCNT unit handle only when it runs correctly. Specifically, when there are no more free PCNT units in the pool (i.e. unit resources have been used up), then this function will return :c:macro:`ESP_ERR_NOT_FOUND` error. The total number of available PCNT units is recorded by :c:macro:`SOC_PCNT_UNITS_PER_GROUP` for reference.
If a previously created PCNT unit is no longer needed, it's recommended to recycle the resource by calling :cpp:func:`pcnt_del_unit`. Which in return allows the underlying unit hardware to be used for other purposes. Before deleting a PCNT unit, one should ensure the following prerequisites:
To install a PCNT channel, users must initialize a :cpp:type:`pcnt_chan_config_t` structure in advance, and then call :cpp:func:`pcnt_new_channel`. The configuration fields of the :cpp:type:`pcnt_chan_config_t` structure are described below:
-:cpp:member:`pcnt_chan_config_t::edge_gpio_num` and :cpp:member:`pcnt_chan_config_t::level_gpio_num` specify the GPIO numbers used by **edge** type signal and **level** type signal. Please note, either of them can be assigned to `-1` if it's not actually used, and thus it will become a **virtual IO**. For some simple pulse counting applications where one of the level/edge signals signals is fixed (i.e., never changes), users can reclaim a GPIO by setting the signal as a virtual IO on channel allocation. Setting the level/edge signal as a virtual IO will cause that signal to be internally routed to a fixed High/Low logic level, thus allowing users to save a GPIO for other purposes.
-:cpp:member:`pcnt_chan_config_t::virt_edge_io_level` and :cpp:member:`pcnt_chan_config_t::virt_level_io_level` specify the virtual IO level for **edge** and **level** input signal, to ensure a deterministic state for such control signal. Please note, they are only valid when either :cpp:member:`pcnt_chan_config_t::edge_gpio_num` or :cpp:member:`pcnt_chan_config_t::level_gpio_num` is assigned to `-1`.
-:cpp:member:`pcnt_chan_config_t::invert_edge_input` and :cpp:member:`pcnt_chan_config_t::invert_level_input` are used to decide whether to invert the input signals before they going into PCNT hardware. The invert is done by GPIO matrix instead of PCNT hardware.
-:cpp:member:`pcnt_chan_config_t::io_loop_back` is for debug only, which enables both the GPIO's input and output paths. This can help to simulate the pulse signals by function :cpp:func:`gpio_set_level` on the same GPIO.
Channel allocating and initialization is done by calling a function :cpp:func:`pcnt_new_channel` with the above :cpp:type:`pcnt_chan_config_t` input parameter plus a PCNT unit handle returned from :cpp:func:`pcnt_new_unit`. This function will return a PCNT channel handle if it runs correctly. Specifically, when there are no more free PCNT channel within the unit (i.e. channel resources have been used up), then this function will return :c:macro:`ESP_ERR_NOT_FOUND` error. The total number of available PCNT channels within the unit is recorded by :c:macro:`SOC_PCNT_CHANNELS_PER_UNIT` for reference. Note that, when install a PCNT channel for a specific unit, one should ensure the unit is in the init state, otherwise this function will return :c:macro:`ESP_ERR_INVALID_STATE` error.
If a previously created PCNT channel is no longer needed, it's recommended to recycle the resources by calling :cpp:func:`pcnt_del_channel`. Which in return allows the underlying channel hardware to be used for other purposes.
The PCNT will increase/decrease/hold its internal count value when the input pulse signal toggles. User can set different actions for edge signal and/or level signal.
-:cpp:func:`pcnt_channel_set_edge_action` function is to set specific actions for rising and falling edge of the signal attached to the :cpp:member:`pcnt_chan_config_t::edge_gpio_num`. Supported actions are listed in :cpp:type:`pcnt_channel_edge_action_t`.
-:cpp:func:`pcnt_channel_set_level_action` function is to set specific actions for high and low level of the signal attached to the :cpp:member:`pcnt_chan_config_t::level_gpio_num`. Supported actions are listed in :cpp:type:`pcnt_channel_level_action_t`. This function is not mandatory if the :cpp:member:`pcnt_chan_config_t::level_gpio_num` is set to `-1` when allocating PCNT channel by :cpp:func:`pcnt_new_channel`.
Each PCNT unit can be configured to watch several different values that you're interested in. The value to be watched is also called **Watch Point**. The watch point itself can't exceed the range set in :cpp:type:`pcnt_unit_config_t` by :cpp:member:`pcnt_unit_config_t::low_limit` and :cpp:member:`pcnt_unit_config_t::high_limit`. When the counter reaches either watch point, a watch event will be triggered and notify user by interrupt if any watch event callback has ever registered in :cpp:func:`pcnt_unit_register_event_callbacks`. See `Register Event Callbacks <#register-event-callbacks>`__ for how to register event callbacks.
The watch point can be added and removed by :cpp:func:`pcnt_unit_add_watch_point` and :cpp:func:`pcnt_unit_remove_watch_point`. The commonly used watch points are: **zero cross**, **maximum / minimum count** and other threshold values. The number of available watch point is limited, :cpp:func:`pcnt_unit_add_watch_point` will return error :c:macro:`ESP_ERR_NOT_FOUND` if it can't find any free hardware resource to save the watch point. User can't add the same watch point for multiple times, otherwise it will return error :c:macro:`ESP_ERR_INVALID_STATE`.
When PCNT unit reaches any enabled watch point, specific event will be generated and notify the CPU by interrupt. If you have some function that want to get executed when event happens, you should hook your function to the interrupt service routine by calling :cpp:func:`pcnt_unit_register_event_callbacks`. All supported event callbacks are listed in the :cpp:type:`pcnt_event_callbacks_t`:
-:cpp:member:`pcnt_event_callbacks_t::on_reach` sets a callback function for watch point event. As this function is called within the ISR context, user must ensure that the function doesn't attempt to block (e.g., by making sure that only FreeRTOS APIs with ``ISR`` suffix are called from within the function). The function prototype is declared in :cpp:type:`pcnt_watch_cb_t`.
User can save their own context to :cpp:func:`pcnt_unit_register_event_callbacks` as well, via the parameter ``user_ctx``. This user data will be directly passed to the callback functions.
In the callback function, the driver will fill in the event data of specific event. For example, the watch point event data is declared as :cpp:type:`pcnt_watch_event_data_t`:
-:cpp:member:`pcnt_watch_event_data_t::watch_point_value` saves the watch point value that triggers the event.
-:cpp:member:`pcnt_watch_event_data_t::zero_cross_mode` saves how the PCNT unit crosses the zero point in the latest time. The possible zero cross modes are listed in the :cpp:type:`pcnt_unit_zero_cross_mode_t`. Usually different zero cross mode means different **counting direction** and **counting step size**.
Registering callback function will result in lazy installation of interrupt service, thus this function should only be called before the unit is enabled by :cpp:func:`pcnt_unit_enable`. Otherwise, it can return :c:macro:`ESP_ERR_INVALID_STATE` error.
The PCNT unit features filters to ignore possible short glitches in the signals. The parameters that can be configured for the glitch filter are listed in :cpp:type:`pcnt_glitch_filter_config_t`:
-:cpp:member:`pcnt_glitch_filter_config_t::max_glitch_ns` sets the maximum glitch width, in nano seconds. If a signal pulse's width is smaller than this value, then it will be treated as noise and won't increase/decrease the internal counter.
User can enable the glitch filter for PCNT unit by calling :cpp:func:`pcnt_unit_set_glitch_filter` with the filter configuration provided above. Particularly, user can disable the glitch filter later by calling :cpp:func:`pcnt_unit_set_glitch_filter` with a `NULL` filter configuration.
The glitch filter is clocked from APB. For the counter not to miss any pulses, the maximum glitch width should be longer than one APB_CLK cycle (usually 12.5 ns if APB equals 80MHz). As the APB frequency would be changed after DFS (Dynamic Frequency Scaling) enabled, which means the filter won't work as expect in that case. So the driver will install a PM lock for PCNT unit during the first time user enables the glitch filter. For more information related to power management strategy used in PCNT driver, please see `Power Management <#power-management>`__.
Before doing IO control to the PCNT unit, user needs to enable it first, by calling :cpp:func:`pcnt_unit_enable`. Internally, this function will:
* switch the PCNT driver state from **init** to **enable**.
* enable the interrupt service if it has been lazy installed in :cpp:func:`pcnt_unit_register_event_callbacks`.
* acquire a proper power management lock if it has been lazy installed in :cpp:func:`pcnt_unit_set_glitch_filter`. See also `Power management <#power-management>`__ for more information.
On the contrary, calling :cpp:func:`pcnt_unit_disable` will do the opposite, that is, put the PCNT driver back to the **init** state, disable the interrupts service and release the power management lock.
Calling :cpp:func:`pcnt_unit_start` will make the PCNT unit start to work, increase or decrease counter according to pulse signals. On the contrary, calling :cpp:func:`pcnt_unit_stop` will stop the PCNT unit but retain current count value. Instead, clearing counter can only be done by calling :cpp:func:`pcnt_unit_clear_count`.
Note, :cpp:func:`pcnt_unit_start` and :cpp:func:`pcnt_unit_stop` should be called when the unit has been enabled by :cpp:func:`pcnt_unit_enable`. Otherwise, it will return :c:macro:`ESP_ERR_INVALID_STATE` error.
The returned count value is a **signed** integer, where the sign can be used to reflect the direction. The internal counter will overflow when it reaches high or low limit, but this function doesn't compensate for that loss.
When power management is enabled (i.e. :ref:`CONFIG_PM_ENABLE` is on), the system will adjust the APB frequency before going into light sleep, thus potentially changing the behavior of PCNT glitch filter and leading to valid signal being treated as noise.
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 user enables the glitch filter by :cpp:func:`pcnt_unit_set_glitch_filter`, the driver will guarantee that the power management lock is acquired after the PCNT unit is enabled by :cpp:func:`pcnt_unit_enable`. Likewise, the driver releases the lock after :cpp:func:`pcnt_unit_disable` is called.
By default, the PCNT interrupt will be deferred when the Cache is disabled for reasons like writing/erasing Flash. Thus the alarm interrupt will not get executed in time, which is not expected in a real-time application.
This will allow the interrupt to run while the cache is disabled but will come at the cost of increased IRAM consumption.
There's another Kconfig option :ref:`CONFIG_PCNT_CTRL_FUNC_IN_IRAM` that can put commonly used IO control functions into IRAM as well. So that these functions can also be executable when the cache is disabled. These IO control functions are as follows:
-:cpp:func:`pcnt_unit_start`
-:cpp:func:`pcnt_unit_stop`
-:cpp:func:`pcnt_unit_clear_count`
-:cpp:func:`pcnt_unit_get_count`
Thread Safety
^^^^^^^^^^^^^
The factory function :cpp:func:`pcnt_new_unit` and :cpp:func:`pcnt_new_channel` are guaranteed to be thread safe by the driver, which means, user can call them from different RTOS tasks without protection by extra locks.
The following functions are allowed to run under ISR context, the driver uses a critical section to prevent them being called concurrently in both task and ISR.
-:cpp:func:`pcnt_unit_start`
-:cpp:func:`pcnt_unit_stop`
-:cpp:func:`pcnt_unit_clear_count`
-:cpp:func:`pcnt_unit_get_count`
Other functions that take the :cpp:type:`pcnt_unit_handle_t` and :cpp:type:`pcnt_channel_handle_t` as the first positional parameter, are not treated as thread safe. Which means the user should avoid calling them from multiple tasks.
-:ref:`CONFIG_PCNT_CTRL_FUNC_IN_IRAM` controls where to place the PCNT control functions (IRAM or Flash), see `IRAM Safe <#iram-safe>`__ for more information.
-:ref:`CONFIG_PCNT_ISR_IRAM_SAFE` controls whether the default ISR handler can work when cache is disabled, see `IRAM Safe <#iram-safe>`__ for more information.
-:ref:`CONFIG_PCNT_ENABLE_DEBUG_LOG` is used to enabled the debug log output. Enable this option will increase the firmware binary size.
Application Examples
--------------------
* Decode the quadrature signals from rotary encoder: :example:`peripherals/pcnt/rotary_encoder`.
Different ESP chip series might have different number of PCNT units and channels. Please refer to the [`TRM <{IDF_TARGET_TRM_EN_URL}#pcnt>`__] for details. The driver won't forbid you from applying for more PCNT units and channels, but it will return error when all available hardware resources are used up. Please always check the return value when doing resource allocation (e.g. :cpp:func:`pcnt_new_unit`).
:cpp:member:`pcnt_event_callbacks_t::on_reach` callback and the functions invoked by itself should also be placed in IRAM, users need to take care of them by themselves.
