Hướng dẫn cuối cùng: Bộ nhớ không biến động nhúng (eNVM)
Embedded non-volatile memories (eNVM) are memory blocks that are directly integrated semiconductor fabrication on the same die. Since eNVM memories are non-volatile, the data is retained even when the power is off. In addition, eNVMs are reprogrammable and erasable multiple times. Compared to external non-volatile memory technologies, eNVMs have lower power consumption and quick access time as they are on-chip. eNVM is used to store the program code, setting values, cryptographic key, in-field updates, and adjustments of the circuits. This technology includes flash memory, EEPROM, and FRAM. Figure 1 shows the broad categories of embedded system memories.
Figure 1 Semiconductor memory technologies
The direct chip integration of eNVM provides an advantage for fast access time and minimizes chip footprint. The non-volatile memories hold data even when the power is turned-off which is essential to store critical configurations of devices. They store sensitive information such as cryptographic keys and digital certificates. They allow in-system programmability and update storage data without hardware replacement. The advantages of eNVM include lower power consumption in read and write operations and more reliability. The other valuable benefit of eNVM is that it enables chip designs to be customized for variable applications.
Key Characteristics of eNVM
Most of the characteristics of eNVM are related to its on-chip integration, non-volatility data storage, and reprogramming ability. Its fabrication is on the same silicon die as the other logic chip making it more compact in space. The data stored in eNVM isn’t lost due to power interruption. eNVM can be erased multiple times, and customization of the chip. Its performance is better in access time and bandwidth than other memory technologies. In general, eNVM technologies withstand memory operation over time and faster accessibility with low power consumption.
Applications of eNVM in Modern Electronics
In modern electronics, eNVMs are utilized in multiple applications due to their versatility. All on-chip integration advantages of eNVMs are summarized in Table 1.
|
No. |
Application areas |
Descriptions |
|
|
|
1 |
Automotive |
Storing calibrations of engine control units |
|
|
|
2 |
IOT |
Over-the-air update and store devices configuration |
|
|
|
3 |
Consumer Electronics |
Allow device customizations in TV, smartphone …. |
|
|
|
4 |
Medical and wearable |
Data retention in senor and preserving patient record |
|
|
|
5 |
Industrial automations |
Storing encryption keys in PLC, robot controller… |
|
|
|
6 |
Appliances |
Save application-specific setting |
|
|
Overview of Different eNVM Technologies
Flash memory, EEPROM, FRAM, ReRAM, MRAM, and FeRAM are the technologies of eNVM. All technologies have advantages over the other. The users may choose a specific eNVM technology based on the required performance, production cost, write cycle, and endurances. For high-density and low-cost applications flash memory technologies are employed widely. EEPROM has the potential to erase and write at the byte level. A newly ferroelectric RAM comes with byte-level access and faster write speed than flash memories. ReRAM uses resistive switching for quick read/write speed. MRAM employs magnetic junctions to retain data rather than electronic charge for the sater access type. The eNVM memories are valuable for the modern electronics industry since they retain information even when the power is lost.
This article highlights flash memory, ferroelectric RAM, resistive RAM, magneto-resistive RAM, Electrically Erasable Programmable ROM (EEPROM), and other types of eNVMs technologies in detail.
One-Time Programmable (OTP) Memory
It is a single-time programmed eNVM technology with repeated readable content. Once the required content is stored in OTP memory it never be erased or reprogrammed. This nature of OTP is essential for storing unchanged and immutable data like encryption keys, device ID, default calibrations, and settings. It can be employed for RFID tags, device configuration, and hardware security applications.
Mechanism of Operation
OTP memory technology utilizes an anti-fuse system by employing high-voltage pulses. The non-conductive surfaces can be programmed to be conductive using anti-fuse technology, which makes a high-resistance state to a low-resistance state by permanently breaking down the dielectric strength of the oxide layer during manufacturing. The data programmed via the anti-fuse method in OTP is stored permanently since the breakdown occurs once, and the structure never returns back to its original high resistance state. The programmed digital data (1s and 0s) can managed by physical alteration of the antifuse structure. Anti-fuse technique remains dominant for OTP memory applications today even though other technologies are under investigation.
Advantages and disadvantages
The drawbacks and benefits of OTP memory are mentioned below. Even though OTP memories are low-cost, secure, and reliable, their limited functionality and write cycle are mentioned as drawbacks.
Advantages
OTP employs an anti-fuse structure on regular CMOS processing making them low-cost. Non-volatility and permanent storage properties are essential for securing credential information. They are reliable, robust, and consistent in performance across the cell.
Disadvantages
OTP has limited functionality as they are not fit for updates and rewriting applications. They support only a single write cycle during fabrication. They are not fast during writing as well as high voltage is required for the anti-fuse process.
Typical applications
The non-changeable and cheaper nature of OTP memory makes fit with several embedded applications. Most of the applications of OTP memory lie on device identification and authentication by saving undisputable device IDs, serial numbers, and other information related to devices. Applications that use cryptographic keys also use OTP memory to store security keys that never change in the device’s lifetime. In addition, OTP memories are employed to retain calibration, device configuration, and hardware metering in the industry.
Figure 2 Application areas of OTP [1]
https://link.springer.com/article/10.1007/s00339-023-06383-w
Steve S. Chung, “The discovery of a third breakdown: phenomenon, characterization and applications,” Applied Physics A , vol. 129, no. 2, 2023.
Electrically Erasable Programmable Read-Only Memory (EEPROM)
EEPROM is a type of non-volatile memory that enables byte-by-byte erasing and rewriting electrically without circuit removal. EEPROM is preferred for applications that require frequent updates and rewrites since it can change the stored variables, configurations, constants, and in-system information without memory chip replacement. They stored data such as onboard chip identification numbers, device settings, and occasional updates. EEPROM endures a better erase/rewrite cycle with the cost of slower speed, higher power consumption, and lower density relative to flash memory.
Figure 3 Basic structure of EEPRPM
https://www.youtube.com/watch?v=FIQKdb67g0A
Different Types of EEPROM
EEPROM memory technologies are floating-gate transistors, metal-nitride-oxide-silicon structures, and polymer traps. Each type of EEPROM has been chosen for specific applications based on its advantages.
Floating Gate EEPROM: Uses floating gate transistors, a conductive layer isolated by an insulating layer to store charges. Information is stored by trapping electrons on the gate by varying the threshold voltage for programming. Writing can be accomplished by collecting electrons on the floating gate and removing them to erase data.
SONOS EEPROM: Silicon-oxide-nitride-oxide-silicon structure replaces the floating gate for storing better stable charge. It offers low power requirements and better scalability than a floating gate with the cost of low storage capacity.
TANOS EEPROM: a type of MONSO (metal-oxide-nitride-silicon-oxide) using tantalum oxide added for better performance in terms of programming speed.
Ferroelectric EEPROM (FeEEPROM): a ferroelectric material like PZT is added to allow the cell-gate insulator to become non-volatile, low power, fast read/write operations.
Magnetic EEPROM (EEPROM): Adding tunneling magneto-resistor for non-volatile data storage, which is a chance for high density relative to chargeable EEPROM technology.
Programming and Erasing Mechanisms
EEPROM can be easily programmed or erased without high voltage requirements as compared to old memory technologies like EPROM. However, over-program or erasing will affect its endurance over time so the programmer must take care of them.
Programming mechanisms: each byte of information of EEPROM memory is programmed by collecting electron charges on the charge trapping layer (floating gate or nitride layer).
Erasing mechanisms: In order to erase the stored data from EEPROM, electrically remove charges from the trapping layer by applying reverse voltage.
Advantages and Disadvantages
The special advantage of EEPROM memory is related to its re-programmability, which allows multiple time erase and program at byte-level besides its non-volatile benefits. In addition, they have better endurance and use low voltage for programming/erasing. However, EEPROM memories have a higher cost, finite lifetime, slower write speed, and increased complexity than other specific memory technologies like OTP and Flash.
Typical applications
EEPROM’s ability to rewrite, erase, and non-volatility enables to use them in many embedded applications that require some updates. EEPROM is employed in microcontroller programming to store and update firmware, calibrations, and setting information. They are also used in industry for programming workflow sequences, configuration, and calibration of automation systems like robots.
EEPROM may be utilized in many additional technological sectors like automotive electronics, networking devices, medical instruments, and consumer electronics by storing vital and end-user preferences.
Flash Memory
Flash memory is another non-volatile memory type that allows writing and erasing data in fixed block sizes. Flash memory has limited erase/write cycles, less flexibility, and endurance than EEPROM. They have NOR and NAND memory types. A NAND flash allows a high density and NOR flash has faster read speed. Flash memories are commonly used in storage devices like USB drives, SSDs, and memory cards.
Figure 4 Examples of flash memory
Types of Flash Memory
The NOR and NAND flashes are the two major types of flash memory. Flash memory technologies are under study to enhance performance, density, speed, and endurance.
NOR Flash: The memory cells are arranged in the form of a random access type such as RAM and they are apt for direct code execution. They are fast in reading/writing and are used in multiple embedded and CPU applications.
NAND Flash: Its configuration is the same as EEPROM except its read/write occurs using block size. NAND flash have higher density characteristics at lower cost and good sequential speed, they are used in Flash Disk, SSDs, and memory cards. However, they are a little slower at random speed.
Programming and Erasing Mechanisms
The write and erase mechanisms are done by charging and discharging the floating gate using the Flower-Nordheim tunneling technique. The mechanism is employed to write /program on a flash memory. It injects electrons via the gate oxide, and electrons are trapped on a floating gate, which is electrically isolated by oxides for non-volatile storage. Electron presence alters the threshold values of transistor voltage, which aligns digital information states with either 1s or 0s. On the other hand, for erasing a reverse voltage was applied between the control gate and substrate using Flower-Nordheim tunneling. This process makes electrons and varies the cell’s voltage threshold to represent the opposite state. This technique of write/erase requires high voltage and makes certain wear and tear on the insulating layer through time limits the write/erase cycle of the flash memory.
Advantage and Disadvantage
The main advantage of NAND flash memory is it can achieve higher storage capacity in small sizes due to its compact cell design. NOR-type memory, they are very fast in random reading and accessing a block of data. In addition, flash memories provide a lower cost per –gigabyte. On the other hand, flash memories have limited write/erase cycles, and complex controller requirements in the write-erase process are considered to be a drawback.
Typical Applications
Flash memories are utilized in multiple applications since they are fast read speed (NOR) and (NAND) high-density storage. Thus, they are used in mass storage devices (USB Disk, memory card, SSD), BIOS, and firmware. In addition, they are employed in microcontroller, system-on-chips, and system-on-module embedded systems. They are also used in consumer electronics applications including smartphones, game consoles, and digital cameras to store and easily access photos, videos, and other documents. In industrial applications, flash memory is employed in applications like instruments, automation systems, and robotics.
MTP (Multi-Time Programmable) Memory
In some applications, it is important to manage a limited and pre-defined write/erase cycle during production to store and modify some process configurations and then become read-only memory in the device’s operational life. These types of applications use a non-volatile MTP memory. MTP memories allow multiple data updates during testing and configuration, unlike OTP. However, they become read-only after a few cycles. Even though MTP is costlier than OTP, it provides additional flexibility during device production. The limited number of write-erase of memory cells is managed by the floating gate technology.
MTP versus OTP and EEPROM
As presented earlier, OTP is a one-time programmable non-volatile memory and uses anti-fuse technology at the production stage. However, MTP is another rewritable/erasable memory technology that gives better flexibility at the production stage for a limited time to modify device configurations and become read-only after 3-10 times. EEPROM memories have unlimited rewrite/erase cycles, high cost, and better flexibility. On the other hand, MTP uses the same technology as EEPROM and flash memory to write and erase data only for a few cycles. Thus, MTP balances between EEPROM and OTP, and its flexibility is limited during fabrication.
Advantage and Disadvantages
MTP memory’s special advantage over other types of non-volatile memories is its limited rewritable/ erasable characteristics, which are essential to storing and modifying calibrations, testing, and configuration information during the product fabrication process. MTP maintains read-only operation and non-alterability as OTP after production. MTP is also more flexible than OTP and cheaper than EEPROM.
The disadvantages of MTP are more expensive than OTP, limited rewrite than EEPROM making them unapt for applications requiring in-system updates. In addition, they require careful control to limit erase-write to avoid early wear out and slower performance than OTP.
Typical Applications
The most common application areas of MTP are storing factory trim and setting of a device, calibration and configuration retaining, and firmware programming. Moreover, they are also employed for storing device IDs and boot codes for many embedded systems.
ReRAM (Resistive Random-Access Memory)
ReRAM is a novel type of non-volatile memory that uses resistive switchable material technology sandwich between electrodes to retain data. Its operation is guided by varying the resistance level of metal oxide using the electrostatic field to represent the digital information (1s and 0s). Compared to SRAM and DRAM memory technologies, ReRAM memories have faster read/write speed. Based on the methods and materials ReRAM uses for the transition of resistance, ReRAM technology may be categorized as conductive bridge, Oxide-Based, Electrochemical Metallization, Polymer, and phase change RAM. ReRAM memory technology is still an active area of research across resistance state variations.
Figure 5 Crossbar architecture ReRAM
Mechanism of Operation
ReRAM operation is achieved by changing the resistance state to set and reset using a switch via a meticulous reshuffling of oxygen openings, ion migration, and physical phase change in the oxide layer. Resistance switching can be done by altering the conductive filament in the memory cell using electrical pulses. Initially, the material has a high-resistance value which means at reset (erasing) state, and when electrical pulse signals are connected from two electrodes to induce a resistance change means set (programming) state. Different types of ReRAM have their respective operation during the set and reset state of resistance. For example, the applied electrical pulse in conductive material and electrochemical metallization type creates metal ion migration from one electrode into the switching medium, which lowers the resistance state. The high or low resistance states represent the binary (1s and 0s) and reading the resistance state can be done by applying a smaller voltage and measuring the current flow.
Advantages and disadvantages
ReRAM memories have a promised benefit over the other non-volatile memories such as high read/write speed usable for fast computing applications. In addition, they have better write cycles and high endurance which makes them opt for heavy workload applications. ReRAM consumes low power and has a high flash density beyond NAND flash. However, they require a tight resistance state control to minimize bit error rates, environmental effects like temperature may impact the long-term reliability of data, and its production requires a high volume and refinement process.
Current State of development and future prospects
Even though ReRAM development grants a substantial future promise in speed, endurance, and density, it also faces significant challenges in consistency, commercialization, and retention. A tighter control of resistance state and temperature sensitivity requires improvement. In the future, ReRAM technology will become a universal memory solution. If the consistency and fabrication contests are overwhelmed, it will be used in many embedded systems and consumer applications.
eFuse (Electronic Fuse)
Electronics fuse (eFuse) memories apply an electronic fuse structure for one-time programmable services, unlike OTP which uses a conventional anti-fuse structure. The memory cell of eFuse comprises a polysilicon resistor that will break permanently by electro-migration techniques. One-time programming can be done by applying a high current through a physically break polysilicon fuse that alters the cell’s resistance state, which is a non-reversible open circuit and makes the eFuse an OTP.
Figure 6 eFuse programming mechanism
https://www.pufsecurity.com/technology/otp/
Mechanism of Operation
The structure of the eFuse memory cell consists of conductive polysilicon and two electrodes. When a high current is applied via the two electrode terminals, a metal ion starts to migrate. The migration also causes a permanent collection into the center of the fuse using an electro-thermal effect. These effects create an irreversible alteration of the conductivity of polysilicon elements to a high resistance state. This process is said to be a permanent programming operation of eFuse.
Applications
The eFuse memories are commonly used for one-time programmable applications such as storing device IDs and digital signatures. Like OTP they are also used to store device configurations, programming application-specific data, cryptographic information, and production metadata storage in many embedded system applications.
FeRAM (Ferroelectric RAM)
FeRAM uses a combination of capacitor structure and ferroelectric layer to improve the speed of read/write of the memory technology. Information storage happens through the polarized state of a thin film of ferroelectric material between capacitor plates. By applying a reverse electric field, the data is easily erased. They have better endurance, faster read/write speed, and low power consumption over flash memory. However, they require expensive materials, making them costly and difficult to scale down.
Figure 7 Basic structure of ReRAM
Advantages and disadvantages
The main advantages of FeRAM over other non-volatile memory technologies are: that they have unlimited read and write cycles with better endurance and fast read and write speed. In addition, they consume low power, and they have a better density structure with a capacitor-based mature fabrication process. However, they are susceptible to data errors, and thickness control of the ferroelectric film requires additional manufacturing challenges. Moreover, their writing speed is slower than SRAM, and having a higher write voltage than DRAM is considered a disadvantage.
Some Niche Applications
FeRAM memories are used in applications that critically require low power consumption and fast access control types like medical devices, industrial controllers, and automotive electronics. They are opting for real-time data logging in multiple smart meters are industrial sensors, and they are also good for RFID. Moreover, FeRAMs are adopted in wearable devices and space electronics due to their unlimited read/write property.
MRAM (Magneto-Resistive RAM)
MRAM is a non-volatile memory type that uses magnetism as a data storage mechanism. It uses magnetic tunnel injection and consists of a layer of two ferromagnetic and thin insulators between them. The digital information 1s and 0s can be stored based on the direction of magnetic spin (north/south pole) rather than electron injection/ removal as electrical techniques. When the current passes through a wire near the magnetic tunnel junction, it creates a parallel/antiparallel magnetic field on a free layer of ferromagnetic to write data. Reading can be possible by detecting the MRAM level of resistance. It depends on the magnetization orientation. They have unlimited endurance, allow high-density 3D stacking, and fast read and write operation. However, they consume more power than DRAM, and their densities are lower than NAND.
Figure 8 Simplified structure of an MRAM cell
Advantages and Disadvantages
The main advantages of MRAM over other non-volatile memories are: they have fast read/write operation, MRAM has unlimited endurance since there is no wearing in the read/write cycle. Their cross-point architecture allows high 3D staking density. On the other hand, writing on MRAM memories requires a high current to create a switching field. The challenges are small variations in cell resistance level will affect reading accuracy and increase complexity in manufacturing.
Applications
They are used in many embedded system applications like microcontrollers in industries and medical devices. MRAMs are applied in the harsh environment tolerant they are used in automotive electronics as a memory for infotainment, drive-assistance, and powertrain control. Moreover, they are applicable for spacecraft and avionics, servers and storage, and military devices.
The following table summarizes the relative comparative analysis of eNVM memory technologies based on their key characteristics. More importantly, eNVM’s speed, endurance, density, cost, and power consumption properties across several technologies are discussed relative to each other. ReRAM and NAND flash are the fastest techs in reading/writing and MRAM has the highest endurance with very low power consumption among the technologies.
|
eNMV Technology |
Speed (Read/write) |
Endurance |
Density |
Power consumption |
Cost |
|
OTP |
Fast |
One-time |
low |
Low |
Low |
|
EEPROM |
Modest |
Modest |
Modest |
Moderate |
Modest |
|
NOR Flash |
Fast |
Modest |
Modest |
Moderate |
Modest |
|
NAND Flash |
Very Fast |
High |
High |
Moderate |
Low |
|
MTP |
Modest |
Low |
Modest |
Moderate |
Low |
|
eFuse |
Fast |
one-time |
low |
low |
low |
|
ReRAM |
Very fast |
High |
High |
low |
high |
|
FeRAM |
Fast |
High |
Modest |
Low |
high |
|
MRAM |
Fast |
Very High |
high |
Very low |
Very high |
Trade-offs Between Different Technologies
For one-time writing applications like device ID and authentication, utilizing low cost technologies like OTP and eFuse are preferred. Applications requires low power consumption and better endurance with frequent data update like IoT devices; FeRAM and MRAM are better than others. However, MRAM is a bit expensive not suitable for cost sensitive IoT devices. For Portable applications such as mobile applications, battery consumption, density and cost are highly sensitive in memory selection. Thus, NAND flash and ReRAM are ideal for these applications except that ReRAM is slightly costly that NAND flash. For automotive applications where data integrity is critical, EEPROM is preferred and MRAM for automatic drive-assistance system with high cost. For data centers, NAND flash is a cost effective solution and ReRAM technology for high performance computing are preferred with high price.
To select appropriate eNVM technologies for specific applications, the performance, cost endurance, and power consumption must be considered. The reading and writing speed of eNVM is one of the best performance measurement techniques. It measures the memory access time. Fast access time memory like MRAM and ReRAM are preferred for HPC and low latency applications. The other criterion is endurance, which measures the number of write/erase cycles before the memory fails. High endurance is required for frequent write/erase applications. If writing and erasing are not desired after the first programming, OTP and eFuse memories will be better. The density of the memory is also another selection method, which measures the amount of data stored in a given physical area. NAND flash, ReRAM, and MRAM memories have very high-density eNVM. In addition, power consumption and cost of eNVM technologies are other important factors in selecting memory for specific applications. The level of data security to protect against unauthorized access and to store permanently in the operation environment.
Case Studies Illustrating the Selection Process
This article will highlight the automotive electronic control unit (AECU), High-performance computing Server cache, and wearable health monitoring devices, as case studies. The case studies are summarized in the following table for specific applications requirement and recommended eNVM technologies.
|
Case studies |
Specific applications |
Requirements |
1ts Preferred eNVM |
2nd Preferred eNVM |
|
1 |
Automotive electronics control unit |
Tolerance to harsh environment, high endurance and low power speed |
MRAM Unlimited endurance and tolerance to harsh environment |
FeRAM (if power is sensitive) |
|
2 |
High performance computing server cache |
Extreme fast read/write operation and high endurance |
MRAM Fast access time, unlimited cycle
|
ReRAM (if cost is sensitive) |
|
3 |
Wearable Health Monitoring devices |
High density, low power, and low cost |
FeRAM Low power consumption and better endurance |
ReRAM (if high density is sensitive) |
- V. Future Trends in eNVM
The future of eNVM technologies will pass through several advancements in terms of their performance. The speed, endurance, and density of the eNVM will be close to volatile DRAM memory systems. Advancements in materials, memory architecture, and fabrications are all will continue to reduce the cost of the MRAM and ReRAM technologies. The future of eNVM will also peruse in improving integrations and adoption of hybrid memory systems.
Future Outlook for eNVM Technologies
The outlook of eNVM is promising according to multiple facts such as the demand for eNVM with excellent performance will rapidly expand and continue in the market. In order to achieve the market, demand several technological advancements will be examined in material science, device architecture, and fabrication process. These techniques will reduce the cost of eNVM and improve system-level performances. Thus, their niche application will exponentially grow in various fields.
