How MicroSD cards power embedded systems and applications

MicroSD cards are widely used with single-board computers such as the Raspberry Pi, NVIDIA Jetson Nano (the 2 and 4GB variants), and Odroid. In addition to single-board computers, various embedded devices rely on MicroSD cards for storage, data logging, firmware updates, and booting operating systems.

MicroSD cards are cost-effective, offering the lowest per-gigabyte cost among storage options. Due to their small size, they fit into compact motherboards and other space-constrained embedded devices. These cards can be easily inserted and removed, enabling quick swapping for backups, data transfer, and updates. They’re also highly durable, withstanding accidental drops, vibrations, and extreme temperatures. Their low power consumption makes them well-suited for embedded systems and mobile devices.

Many MicroSD cards also include security features such as encryption, secure erase, and write protection. Despite their small form factor, they can store vast amounts of data, reaching the terabyte range. For these reasons, most single-board computers and many other embedded devices, such as display and camera modules, include a MicroSD card slot.

MicroSD cards are used for various applications in embedded devices, with each use case requiring different specifications. This article explores MicroSD cards’ key features and specifications and how they align with different embedded applications.

Storage in embedded devices

Embedded devices use multiple types of storage, including MicroSD cards, eMMC, EEPROM, NVRAM, and SPI Flash. MicroSD cards are commonly used as bootable storage for installing and running custom operating systems on single-board computers. While many embedded motherboards run their operating systems on eMMC or flash memory, they often use MicroSD cards to expand onboard storage. These cards are also used for data logging, firmware updates, and storing configuration files or user preferences.

Embedded MultiMediaCard (eMMC) is an integrated version of an SD card. It consists of flash memory and a controller soldered directly onto the motherboard. Compared to MicroSD cards, eMMC offers faster read/write speeds and greater reliability since it eliminates connection issues.

SPI Flash is a non-volatile memory that communicates with the controller via an SPI interface. Depending on the underlying memory technology, it can be NOR or NAND flash. NAND flash stores data in blocks and offers ample storage capacities at a lower cost, making it ideal for data logging and multimedia storage in embedded applications. In contrast, NOR flash provides fast read speeds and higher data retention, making it suitable for bootloaders, firmware updates, and application code storage.

Non-Volatile Random Access Memory (NVRAM) combines the fast access of RAM with the non-volatility of ROM. It’s highly reliable and allows quick data access and modification, making it useful for storing critical data such as Real-Time Clock (RTC) data or system configurations.

Originally, NVRAM was implemented as Battery-Backed Static Random Access Memory (BBSRAM), which retained data as long as power was supplied. Some NVRAM solutions combined SRAM with EEPROM or flash memory, using SRAM for fast access while maintaining a backup copy in EEPROM or flash. Modern NVRAM primarily uses Ferroelectric RAM (FRAM), a non-volatile memory built with ferroelectric materials for data storage.

Electrically Erasable Programmable Read-Only Memory (EEPROM) is another type of non-volatile memory that allows selective data erasure and reprogramming. Unlike flash memory, EEPROM permits updates to specific data sections without erasing the entire chip. It is commonly used for storing configuration files, calibration data, and other small but critical pieces of information in embedded systems.

Choosing storage

Different embedded applications require different storage technologies. MicroSD cards are often used for operating system installation, data logging, and multimedia storage. eMMC is also used for OS and application storage. NAND flash is primarily utilized for mass storage and data logging, while NOR flash is preferred for bootloaders and firmware updates.

EEPROM, typically with I2C or SPI interfaces, is used for storing configuration files and critical data. NVRAM is valuable for maintaining real-time data and other essential system information.

Selecting the appropriate storage solution depends on speed, endurance, reliability, and capacity, ensuring that embedded systems operate efficiently and meet application-specific requirements.

MicroSD cards are categorized into four main capacity classes:

MicroSD (Secure Digital Standard Capacity) – The original SD card standard, offering up to 2GB of storage.
MicroSDHC (High Capacity) – An extension of the SD standard, providing storage between 2GB and 32GB.
MicroSDXC (Extended Capacity) – Designed for larger storage needs, ranging from 32GB to 2TB.
MicroSDUC (Ultra Capacity) – The highest capacity class, supporting storage from 2TB to 128TB.

While the MicroSD class is now largely obsolete in embedded systems, MicroSDHC cards are still used for basic storage applications such as data logging, firmware updates, and storing small amounts of critical data. MicroSDXC is the most commonly used class in embedded applications, offering ample capacity for running operating systems, applications, multimedia files, and large datasets. MicroSDUC cards are relatively new and are primarily used for high-resolution video storage and managing massive datasets on edge devices.

MicroSD speed classes

Speed is a crucial factor when booting an operating system, running applications, and performing read/write operations. MicroSD cards are classified into different speed classes based on their minimum write speeds.

Speed class ratings

These older speed classifications are denoted by a number following the letter “C” (e.g., C10, C6, C4, C2). While largely outdated, they may still be found in older systems:

C2: Minimum write speed of 2MB/sec.
C4: Minimum write speed of 4MB/sec.
C6: Minimum write speed of 6MB/sec.
C10: Minimum write speed of 10MB/sec.

UHS speed classes

Ultra High Speed (UHS) classifications are denoted by a number following the letter “U” (e.g., U1, U3). These are the most suitable classes for modern embedded devices:

U1: Minimum write speed of 10MB/sec. Suitable for booting operating systems and running applications.
U3: Minimum write speed of 30MB/sec. Ideal for embedded applications requiring high-speed performance, such as 4K video recording, gaming, and fast data logging.

Video speed classes

Video speed classes are designed for high-resolution video recording. These classifications are denoted by a number following the letter “V” (e.g., V6, V10, V30, V60, V90). Each class indicates the minimum write speed required for specific video recording applications:

V6: Minimum write speed of 6MB/sec. Suitable for standard-definition (SD) video recording.
V10: Minimum write speed of 10MB/sec. Suitable for high-definition (HD) video recording.
V30: Minimum write speed of 30MB/sec. Suitable for ultra-HD and 4K video recording.
V60: Minimum write speed of 60MB/sec. Suitable for 8K video recording.
V90: Minimum write speed of 90MB/sec. Suitable for high-frame-rate video recording and gaming.

SD Express speed classes

SD Express cards use the PCIe interface to achieve extremely high data transfer speeds, making them ideal for demanding applications such as real-time data processing, high-resolution video recording, and burst photography.

These speed classes are denoted by a number following the letter “E” (e.g., E150, E300, E450, E600). Each class represents the minimum write speed in megabytes per second (MB/sec):

E150: Minimum write speed of 150MB/sec.
E300: Minimum write speed of 300MB/sec.
E450: Minimum write speed of 450MB/sec.
E600: Minimum write speed of 600MB/sec.

MicroSD card APC

While the previously discussed speed classes focus on sequential write speeds, which are essential for video recording, MicroSD cards also have Application Performance Classes (APC) designed to measure random read/write performance. This is a crucial factor for running operating systems, applications, and frequent data access tasks.

The APC rating indicates how efficiently the card handles small, random data transactions rather than continuous data streams. These classes are denoted by a number following the letter “A:”

A1 class: This classification ensures a minimum random read speed of 1500 IOPS (Input/Output Operations Per Second) and a random write speed of at least 500 IOPS, with a minimum sequential write speed of 10MB/sec. It is best suited for running basic embedded applications that do not require intensive data access patterns.
A2 class: This class offers significantly improved performance, featuring a minimum random read speed of 4000 IOPS and a random write speed of at least 2000 IOPS, while maintaining a minimum sequential write speed of 10MB/sec. It is designed for real-time embedded applications with demanding access patterns that require high-speed data handling.

The A1 class is suitable for data logging, while the A2 class is better suited for running operating systems and real-time embedded applications.

MicroSD cards in embedded devices

MicroSD cards are widely used in embedded systems for applications such as operating systems, storing firmware updates, configuration data, user preferences, sensor and network data logging, video recording, real-time data acquisition, machine learning, and industrial control.

Selecting the ideal MicroSD card

Each embedded application has specific storage, speed, and endurance requirements. Selecting the right MicroSD card depends on storage capacity, write speed, application performance, and additional features like security and encryption support.

Booting an operating system: One of the most common uses of MicroSD cards in embedded systems is booting an operating system. The card should have sufficient capacity to store and run the OS. Typically, 8GB to 32GB cards are adequate for lightweight operating systems, while full-version OS installations may require at least 32GB. The card should have a minimum C10, U1, or V10 rating. While A1 is generally sufficient, A2 is preferred for real-time operations or higher data rates.

Running moderate applications: Moderate applications require a balance between speed and random access. The card should have enough capacity to accommodate both the OS and application data for running such applications. Typically, 32GB to 128GB cards are adequate. The card must be U3/V30 rated with an A1 API rating.

Running demanding applications: For more demanding applications requiring higher random access performance, the card should provide sufficient capacity for the OS, applications, and generated data. A minimum of 64GB, up to 256GB, is recommended. The card must be U3/V30 or higher, with an A2 API rating.

Simple data storage: The card should be selected based on the dataset size and should have at least a U1 speed rating for basic storage needs. An A1 API rating is beneficial.

Data logging: For data logging, the key considerations are data size and logging frequency. Generally, U1/V10 or U3/V30 cards are well-suited. An A1 API rating is beneficial, and endurance is crucial for long-term data logging applications.

High-speed data acquisition: Data rates and storage capacity are the primary concerns for high-speed data acquisition. The card must have excellent sequential write speeds. SD Express cards rated E150 or higher with an A2 API rating are recommended.

Video recording: The card’s storage capacity should be chosen based on recording duration and bitrate for video recording. HD video recording typically requires 32GB to 128GB, with a U3/V30 speed class. Sustained write speed support is crucial, but the API rating is less significant.

4K/High bitrate video recording: For high-quality video recording, storage capacity should be selected based on recording duration and bitrate. Typically, 64GB to 256GB cards are required. A V60/V90 or SD Express card is recommended. The API rating is not a significant factor in standard video recording.

Industrial control: Industrial applications require high durability and resistance to extreme temperatures. Industrial-grade cards should have an extended temperature range (-40°C to 80°C), high endurance, data retention, and a controlled Bill of Materials (BOM). Storage capacity between 16GB and 64GB is usually sufficient. The card must be U3/V30 rated or higher and may require an A2 API rating.

AI inference: High storage capacity and read speeds are essential for AI inference and edge training. A minimum of 128GB is recommended, with a U3/V30 or higher rating and an A2 API rating.

By selecting the appropriate MicroSD card specifications, embedded system designers can ensure optimal performance, reliability, and longevity across different applications.