Understanding and Using the No-OS and Platform Drivers

Rapidly advancing technology needs software support (firmware drivers and example code) to simplify the design in process. This article describes the use of no-OS (no operating system) drivers and platform drivers for building an application firmware with Analog Devices precision analog-to-digital converters and digital-to-analog converters, which offer a high level of performance in terms of speed, power, size, and resolution.

Embedded firmware examples based on no-OS drivers are provided by ADI to support precision converters. No-OS drivers are responsible for device configuration, data capture from the converter, performing the calibration, etc., while firmware examples based on the no-OS drivers facilitate the transfer of data to a host PC for display, storage, and further processing.

Introduction to No-OS and Platform Drivers

As the name suggests, no-OS drivers are designed for use with generic (or no specific) operating systems. The name also implies that these drivers can be used on BareMetal systems without any OS support. No-OS drivers are designed to provide high-level APIs for digital interface access for the given precision converter. No-OS drivers using these APIs interface with devices to access, configure, read, and write data without knowing about the register addresses (memory map) and their contents.

No-OS drivers make use of the platform driver layer to allow the reuse of the same no-OS drivers across multiple hardware/software platforms, making your firmware highly portable. The use of a platform driver layer insulates the no-OS drivers from knowing about the low-level details of platform specific interfaces such as SPI, I2C, GPIO, etc., which makes the no-OS drivers reusable across multiple platforms without changing them.

Figure 1. Precision converters firmware stack.

Using No-OS Drivers

Figure 2 shows the typical code structure of no-OS drivers.

Figure 2. No-OS driver code structure.
Figure 3. Device configuration enums, structures, and APIs.

The no-OS driver code for precision converters is typically incorporated within two source files written in C programming language: adxxxx.c and adxxxx.h, where xxxx stands in for the device name (for example, AD7606, AD7124, etc.). The device header file (adxxxx.h) contains the public programming interface of device-specific structures, enumerations, register addresses, and bit masks, which are available for public access by including this file into required source files. The device source file (adxxxx.c) contains the implementation of the interface used to initialize and remove the device, read/write the device registers, read data from device, get/set device specific parameters, etc.

Typical no-OS drivers are structured around a common set of functionalities:

  • The declaration of device specific register addresses, bit mask macros, device configuration enumerations, and structures to read/write the device’s specific parameters (for example, oversampling, gain, reference, etc.).
  • Initializing/de-initializing the physical device through the no-OS driver’s device initialize/remove functions and device specific init and driver structures and descriptors.
  • Accessing the device memory map or register details using the device register read/write functions; for example, adxxxx_read_register() or adxxxx_write_register().

No-OS Driver Code Usage

Using device specific addresses, bit masks, and parameter configuration enums and structures:

As stated earlier, the adxxxx.h header file contains the declaration of all device specific enums and structures, which are passed to device-specific functions or APIs to configure or access device parameters. This is illustrated in Figure 3.

The adxxxx_config structure shown in Figure 3 allows users to select the multiplexer channel and set the oversampling rate for that. Both the members of this structure (afe_mux_channel and oversampling) are enumerations that are present in the same header file, which contains the numeric constants of all the possible values for both the fields that the user can select.

The adxxxx_set_adc_config() function defined in the adxxxx.c file takes user-passed configurations/parameters through the configuration structure and further calls the adxxxx_spi_reg_write() function to write the data into the ADXXXX_REG_CONFIG device register through the digital interface (in the previous case, SPI).

Initializing the device using the no-OS driver device structures and initialization function:

Figure 4. Device init and driver structures’ declaration.

Along with the device configuration enumerations and structures, the no-OS drivers provide two additional structures:

  • Device initialization structure.
  • Device driver structure.

The device init structure allows users to define the device specific parameters and configurations in user application code. The init structure contains the members of other device-specific parameter structures and enumerations. Figure 5 shows how the device init structure is defined.

Figure 5. The device init structure definition in the user application.

The device driver structure loads the device initialization parameters through the device init function adxxxx_init(). The device driver structure is allocated at run-time (dynamic) memory from the heap space. The parameters declared within the device driver structure and device init structure are almost identical to each other. The device driver structure is a run-time version of the device init structure.

The typical device initialization function and initialization flow is described in the following steps:

  • Step 1: Create a definition (or instance) of the device init structure in your application (for example, struct adxxxx_init_params) to initialize the user specific device parameters and platform-dependent driver parameters. The parameters are defined during compilation time.

Note: The parameters defined in the init structure vary from device to device.

Equation 1
  • Step 2: Create a pointer instance (variable) of the device driver structure in the application code.

A user application needs to create a single pointer instance of the device driver structure. This instance is passed to all no-OS driver APIs/functions to access device specific parameters. This pointer instance defined in the application code points to a dynamically allocated memory in heap, which is done through a device init function such as adxxxx_init(), which is defined in the no-OS drivers.

Equation 2
  • Step 3: Initialize the device and other platform specific peripherals by calling the device init function.
Equation 3

The adxxxx_init() function defined in the no-OS drivers initializes the device with the user specific parameters passed through adxxx_init_param structure. The pointer instance of the device driver structure and the instance of the device init structure are passed as two arguments to this init function. The user application code can call the adxxxx_init() function multiple times, provided the init call is balanced by a call to the device remove function.

Accessing the memory map (register contents) through device register read/write functions is shown in Figure 6

Figure 6. Accessing the register contents.

Users can access the device register contents (such as the product ID, scratch pad value, OSR, etc.) through no-OS driver device-specific adxxx_read/write() functions.

Most of the time, users don’t use register access functions directly. The device specific functions call through these register access functions such as adxxxx_ spi_reg_read/write(). The use of device configuration and status APIs to access the device memory map is recommended when possible instead of using the direct register access functions, as this ensures the device driver structures are kept in sync with the configuration in the device.

Platform Drivers

Platform drivers are one of the hardware abstraction layers (HAL) that wrap the platform specific APIs. They are called by no-OS device drivers or user application code to provide an independence from the underlying hardware and software platforms. Platform drivers wrap the low-level platform-specific hardware functionality such as SPI/I2C initialization and read/write, GPIO initialization and read/write, UART initialization and receive/transmit, user specific delays, interrupts, etc.

The typical file structure of an SPI platform driver module is shown in Figure 7.

Figure 7. SPI platform driver code structure.

Using Platform Drivers

Platform driver code is typically incorporated within three source files written in C/C++ programming languages.

1) spi.h: This is a platform agnostic file that contains the device structures and enumerations required for the SPI functionality. The C programming interface defined in this header has no platform dependencies.

All the parameters declared within the init and device structures are common to SPI interfaces on any platform.

The void *extra parameter used in device init structure allows users to pass additional (extra) parameters, which could be specific to the platform that is used.

The parameters declared within the SPI driver structure and SPI init structure are almost identical to each other. The SPI driver structure is a run-time version of the SPI init structure.

2) spi.cpp/.c: This file contains the implementation of the functions declared in the spi.h file, which are used to initialize the SPI peripheral and read/write data from it for a specific platform. The term “platform,” in a broader sense, refers to a combination of a hardware microcontroller (target device) and software (for example, RTOS or Mbed-OS). This file is platform dependent and needs to be modified while porting on some other platform.

Figure 9 details the SPI interfaces for the Mbed platform and shows how to initialize SPI and read/write data using these interfaces and device init/driver structures.

Figure 8. SPI init and driver structures.
Figure 9. SPI APIs or functions. Note: the added code for spi_init() and spi_write_and_read() is abbreviated code and details are omitted for clarity.
Figure 10. SPI extra init and driver structures.

3) spi_extra.h: This file contains the additional device structures or enumerations, which are specific to a given platform. This allows user application code to provide the configurations that are not covered in a generic spi.h file. For example, the SPI pins can vary from platform to platform, which therefore can be added as part of these platform specific extra structures.

Porting Platform Drivers

The platform drivers can be ported from one platform (microcontroller) to another by typically creating platform specific .cpp/.c and _extra.h files. Platform drivers reside one layer above the device-specific hardware abstraction layer (HAL) provided by the microcontroller unit vendor. Therefore, porting of platform drivers from one platform to another needs some minimal changes in platform driver code related to calling the functions or APIs present in the HAL provided by its vendor.

The diagram in Figure 12 distinguishes between Mbed-based SPI platform drivers and ADuCM410 SPI platform drivers.

The GitHub source code links for ADI’s no-OS repository and platform drivers are available on the Analog Devices Wiki and GitHub page.

Contributing to No-OS Drivers

ADI no-OS drivers are open source and hosted on GitHub. The drivers not only support precision converters, but also many other Analog Devices products such as accelerometers, transceivers, photoelectronic devices, and more. Anyone familiar with the source code can contribute to these drivers by committing the changes and creating a pull request to review those changes.

There are many example projects that can run in Linux and/or Windows environments. Many example projects are developed with a hardware descriptive language (HDL) to run on FPGAs developed by Xilinx®, Intel®, etc., and that target processors developed by different vendors.

No-OS software drivers in C for systems without an operating system can be accessed at the Analog Devices no-OS GitHub repository.

The Analog Devices Wiki provides examples developed for precision converters using the Mbed and ADuCMxxx platform.

Figure 11. Mbed platform specific SPI init implementation.
Figure 12. Platform driver differences.

Об авторах

Mahesh Phalke

Mahesh Phalke

Mahesh Phalke is a senior software engineer at Analog Devices Bangalore in the Precision Converters Technology Software Group. He graduated from Pune University in 2011 with a bachelor’s degree in electronics engineering.