Printer Friendly

Low Power Memory for IoT Wearables and Portable Medical Devices.

Power consumption is critical in today's embedded systems. Emerging low power applications include battery-powered Internet of Things (IoT) sensor nodes, wearables, and medical electronic devices that are power and energy-constrained. With the ultimate goal to reduce overall system power and increase battery life, these applications rely heavily on low power processors and energy-efficient memory solutions. The memory requirement can be fulfilled by several memory technologies available in the market. Each of these technologies have unique advantages. Ferroelectric RAM (FRAM) is finding its way into more applications, because of its instant non-volatility, low operating power, fast write speed, and high endurance.

Non-Volatile Memory in IoT Applications

The Internet of Things is the world of connected devices. It encompasses all things connected to the Internet. Data is collected in a central system, or in the cloud from various "Things," that sense their environment. The collected data is then used to obtain meaningful results and make process decisions. Results or commands can be sent to various "Things" through the interconnected networks.

Key attributes of an IoT device are self-existence and connectivity. When the IoT device is not connected, it should be capable of collecting meaningful data and, if required, make decisions locally. This is called computing at the edge. When connected, it becomes part of a bigger network, working as a team to achieve a larger goal.

Consider a smart city. Sensors installed in the city capture details such as traffic movement, events that are happening, availability of parking spaces, and so on. Parking lot sensors in standalone operation can guide drivers to an available parking space. When multiple parking lots are connected, the available parking spaces within the entire city can be shown. Drivers could even reserve a space online, making parking in the city much easier.

IoT technology can be deployed in places where continuous power can't be supplied. This requirement has major implications on the way IoT nodes (Figure 1) are designed. Many of today's devices are powered by batteries, harvested energy, or a combination of both. In energy-constrained environments, IoT designers must select low power components that can save overall system power. Power-efficient memory plays an important role in reducing overall system power.

With the rapid growth of the IoT market, IoT applications require various types of memories. In sensor-based IoT nodes, nonvolatile memories play an important role. As remote nodes sense their environment, they transmit data to the central collector or upload to the cloud. Using a non-volatile memory in a remote sensor node increases the reliability of the system, giving developers the option to trade off between data size, time, and power. This is especially important in energy harvesting applications.

Consider a remote IoT device harvesting energy from the sun to sense the temperature and humidity of the surrounding environment. The designer must ensure there is enough power to transmit reliably. An important design consideration is the transmission data size and duty cycle to minimize power consumption, thus optimizing use of the available power. In these scenarios, adding a non-volatile memory increases system reliability. Logging data locally gives the system designer multiple tools to manage how data is securely transmitted from the sensor node to the central collector with its available power. FRAM, with instant non-volatile writes and low power operation, enables developers to maximize power efficiency for these applications.

Wearable electronics (Figure 2) is one of the fastest growing IoT applications. Wearables include smartwatches, fitness trackers, sport watches, smart clothing, and smart jewelry. Wearables incorporate a processing unit, smart sensors, memory, and some form of communication to allow access to the data in real-time. Typically, these devices collect important data such as number of steps, pulse rate, sleep time, etc.

Like wearables, portable medical devices are an emerging IoT application. Adoption of connectivity in medical devices is quickly transforming the health care industry by making remote sensors cost-efFective. Consumers can now monitor personal vital signs and transmit data from their home, avoiding cosdy visits to the doctor's office. Portable medical devices measure and store data such as temperature, glucose levels, and blood pressure locally. The data is then transmitted periodically to the doctor's office for analysis. Measured data can be evaluated locally by a microcontroller in real-time, or it can be stored in a non-volatile memory, along with the date and time stamp, for later processing.

With an emphasis on miniature size and low power consumption, the type of memory used becomes one of the key elements in wearables and medical designs. Multiple factors such as write endurance, power consumption, and package size are important considerations in selecting the right memory. Key design factors include:

1. Endurance and density: To store collected data, it's important to use a memory that has adequate density, as well as sufficient endurance capable of supporting constant logging of data.

2. Power consumption: Wearables and portable medical devices typically operate using small batteries. Therefore, it is very important to select components, including memory, with low active and standby current to maximize operating time between battery charges.

3. Package size: Wearables and portable medical devices are size constrained. Selecting small system components plays a major role in reducing the size of these devices. Hence why small form factor, low-pin count memories are gaining popularity in these designs.

4. Start-up time: Memory should have instant start-up time for code-in-place options. Code-in-place allows the device to immediately execute boot code, eliminating the need for a separate memory chip. In this way, having a single, fast, non-volatile memory serves the dual purpose of supporting code storage and real-time data logging.

There are many types of memories that are worth considering for the IoT applications, including Hash, EEPROM, MRAM, and FRAM. Each of these technologies have their own advantages and disadvantages. For example, Flash is available in high densities, but has comparably low endurance, making it ideal for code storage but not data logging. EEPROMs have better endurance than Flash, but not enough for continuous data logging. FRAM combines fast speed writes with instant non-volatility and infinite endurance. However, unlike Flash, FRAMs are not available in high density. Hence, one memory may not fit all requirements. Designers need to consider the needs of their application and make a suitable choice. Table 1 compares the various non-volatile memory technologies available in the market today.

Flash is available in large densities but has low endurance cycles. As an IoT memory, Flash is more suited for code/configuration data storage (i.e., code or data that is updated a limited number of times over the life cycle of the product). Though Flash supports page writes, without byte write, it is not a preferred choice for logging the fast changing sensor data.

EEPROMs are available in small densities with high endurance and byte writability, which makes them better suited than Flash for the data logging requirements of IoT applications. However, EEPROMs consume a certain amount of time, called soak time, after every write. Although this delay does not affect most data logging applications, from an IoT perspective it increases current consumption by requiring the system to be in high current mode for a longer time to complete data writes. FRAM technology provides an alternative to Flash and EEPROM by offering virtually infinite endurance, instant non-volatility, higher write speed, and low power operation.

To demonstrate how FRAMs reduce the overall system power that is essential to IoT systems, an experiment was set up with FRAM and EEPROM. In the experiment, one byte of data was periodically logged into FRAM and EEPROM. This was a typical IoT sensor node application where a few bytes of data were written periodically. The results of the experiment are given below.

Periodically logging a few bytes of data to memory is generally performed by IoT sensor nodes. Hence, fast writes not only reduce the power consumed by memory but also lower overall system power consumption by reducing the system power-on time. The advantage of FRAM over EEPROM increases with frequency, as is illustrated in Figure 4 and Figure 5.

This demonstrates that for certain applications, FRAM's faster write speed delivers both performance and energy advantages that arc desirable for many IoT applications.

Resources

FRAM Technology Brief Energy' Comparison of FRAM and EE-PROM

By Harsha Medu and Girija Chougala, Cypress Semiconductor
Table 1: Comparison of non-volatile memory technologies

                 FRAM                      EEPROM

Density          4 Kbit - 8                1 Kbit - 2
                 Mbit                      Mbit
SPI Speed        108 MHz                   20 MHz
Byte Write       Yes                       No
Write Endurance  1E+14                     1E+6
(Cycles)
Active current   3.2 mA @ 50               2 mA @ 5
                 MHz                       MHz
Standby current  2.15 uA                   0.08 uA
Hibernate        0,10 uA                   N/A
Data Retention   151 years @ 65[degrees]C  100 years @ 55[degrees]C

                 FLASH                    MRAM

Density          512 Kbit - 1             128 Kbit - 4
                 Gbit                     Mbit
SPI Speed        133 MHz                  104 MHz
Byte Write       No                       Yes
Write Endurance  1E+5                     Unlimited
(Cycles)
Active current   40 mA @ 50               42 mA @ 40
                 MHz                      MHz
Standby current  20 uA                    3 mA
Hibernate        N/A                      N/A
Data Retention   20 years @ 85[degrees]C  20 years @ 85[degrees]C

Table 2: Energy Consumption in Different Practical Scenarios

                               64 KB Device               2 MB Device
Function          F-RAM        EEPROM        F-RAM        F-RAM
                  FM25CL64B    AT25640B      vs. EE-ROM   FM-25V20A
                  ([micro]J)   ([micro]J)    (Ratio)      ([micro]J)

Write Full         17.0          3.9 x 103   229.4        900.0
Memory [2]
Write 1 byte        0.5         13.1          29           28.7
every 100 ms
with standby
during idle [3]
Write 1 byte                                                1.8
every 100 ms
with F-RAM put
to sleep during
idle [4]
Read Full          17.7        138.5           7.8        844.0
Memory [5]
Write-Read        154.3        108.8 x 103   705.1          9.6 x 103
1 byte (full
memory) [6]

                     2 MB Device
Function          EEPROM       F-RAM
                  M95M02-A     vs.
                  ([micro]J)   EEROM
                               (Ratio)

Write Full        7.7 x 103      8.6
Memory [2]
Write 1 byte      5.8            0.2
every 100 ms
with standby
during idle [3]
Write 1 byte      5.8            3.2
every 100 ms
with F-RAM put
to sleep during
idle [4]
Read Full         1.7 x 103      2.0
Memory [5]
Write-Read        1.5 x 106    156.3
1 byte (full
memory) [6]

Notes:
2. Energy = Energy for 1-byte write enable command + Energy for writing
1-byte command + Energy for writing 2-byte / 3-byte address + (Energy
for writing 1 byte of data + Total number of bytes)
3. Energy = Energy for 1-byte write enable command + Energy for writing
1-byte command + Energy for writing 2-byte / 3-byte address + Energy
for writing 1 byte of data + Energy consumed during 100 ms of standby
mode
4. Energy = Energy for 1-byte write enable command + Energy for writing
1-byte command + Energy for writing 2-byte / 3-byte address + Energy
for writing 1 byte of data + Energy consumed during 100 ms of Sleep
Mode (For EEPROM standby mode)
5. Energy = Energy for writing 1 -byte command + Energy tor writing
2-byte / 3-byte address + (Energy tor reading 1 byte of date + Total
number of bytes)
6. Energy = Energy for 1-byte write enable command + Energy for writing
1-byte command + Energy for writing 2-byte / 3-byte address + Energy
for writing 1 byte of data + Energy tor writing 1-byte command + Energy
for writing 2-byte / 3-byte address + Energy tor reading 1 byte of data
COPYRIGHT 2018 Advantage Business Media
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2018 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Applying Tech Low Power Memory
Author:Medu, Harsha; Chougala, Girija
Publication:Medical Design Technology
Date:Nov 1, 2018
Words:1864
Previous Article:Mobile Apps Drive Future of Patient Care: Apps enable users to monitor chronic conditions outside medical facilities as part of daily routine.
Next Article:Assessing Wearable Device Adhesives for Optimal Fixation: Understanding adhesive properties ensures maximum use for wearable devices.
Topics:

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters