Top five non-volatile memory considerations for data logging applications
Data logging is essential across a broad range of IoT and industrial applications. Data loggers, generally small, stand-alone, portable devices, collect bits of data at set intervals and store that data for analysis.
While a data logger can have internal or external memory, the stand-alone aspect implies on-board memory. Memory size varies depending on how long the unit will be unattended, how often data is logged, and other considerations. The memory may be battery-backed static random access memory (SRAM) or NVM (Non Volatile Memory), but because it is important that data be safely stored in case of power loss, NVM provides a key advantage.
Here, Shin (Shimon) Raviv, Director of Product Marketing, Weebit Nano, outlines five important considerations in NVM selection.
1. Power consumption
Some data logging applications must operate from battery, or even ambient energy sources, so data loggers must be designed to be extremely power efficient. Part of this is ensuring the device is kept in deep sleep mode, with zero refresh current, for long periods. It’s also important that the device wakes up as fast as possible and quickly writes the data.
Historically, flash memory is the typical NVM used in data loggers, but flash has drawbacks compared to emerging NVM technologies. Power consumption is one of these drawbacks.
A typical flash device consumes 100x more programming power compared to an emerging memory such as resistive non-volatile memory (ReRAM or RRAM). Flash also requires programming voltage above 10V while ReRAM requires less than 2.5V. This means flash can require a 10x larger battery (or multiple batteries), resulting in a larger or more expensive data logger, or one with a significantly shorter lifetime. Using compression to reduce flash size requires more computing power – so it’s still a higher power solution.
When it comes to other emerging NVMs, FRAM is also used in data logging. But FRAM requires refresh cycles after each data access (read) leading to higher power consumption.
2. Reliability in all environmental conditions
Given the importance of the data they collect, data loggers must be highly reliable regardless of environmental conditions. Environments can have very low or high temperatures, high vibration, electromagnetic interference (EMI), radiation, or other extremes. And since devices may need to operate for long periods of time in remote locations or without human intervention, it is imperative that so long as they have power, they will not fail to log data, and will not lose data, for any reason. This means the system’s memory must be able to carry out a high number of reads/writes and retain data reliably for significant time periods under any conditions.
ReRAM has typically 100x faster write access time than flash
One drawback of flash memory is that it can’t withstand high levels of radiation, so when used in harsh environments, it often requires shielding or other complex design workarounds. Emerging NVM technologies offer alternatives. ReRAM cells are inherently immune to various types of radiation and electromagnetic fields. ReRAM also demonstrates high thermal stability and reliability, with Weebit’s own ReRAM able to retain data for more than 10 years at 125+°C.
3. Write speed
In some applications, the last seconds of data logging are most important. If a system suddenly powers off, this ‘last gust’ information may be key to finding out what happened.
Data loggers for such applications need not only highly reliable NVM, but also fast write speed. With an external flash device, data must be written to external memory using serial interfaces, so flash memory programming time is slower compared to embedded ReRAM. Moreover, programming flash (either external or integrated) requires a page erase cycle before programming a new page, making the write cycle even longer.
With slower programming time, critical data can be lost. To guard against this, such flash-based systems often require large, expensive, and unreliable capacitors to maintain enough energy for the ‘last gust’ write operation.
A fast embedded emerging NVM solution such as ReRAM provides clear advantages, with direct read/write functions and bit addressability, leading to fast cell read/write and enhanced system performance. ReRAM has typically 100x faster write access time than flash.
Data loggers can be a weak point for cyber-attacks. With external NVM, it’s possible to ‘listen’ to and change data via external buses – another reason to consider embedded NVM. Embedding NVM with the system’s microcontroller adds very little cost while saving power, increasing write speed, and improving security.
There are other types of cyberattacks to consider. With floating gate devices like flash, it may be possible to sense or change the device’s internal state using electron beams. Magnetic and optical (laser) attacks are also possible.
Since ReRAM does not use any charged particles, it is more difficult to access with electron beams, and ReRAM is immune to electromagnetic fields. Because the ReRAM bit cell is deeply embedded between two metal layers integrated at the back-end-ofline (BEOL), it is more immune to attacks compared to flash, which is integrated at the front end of line (FEOL).
Cost is of course a prime consideration for small, widespread data logging devices. ReRAM has much lower silicon cost compared to flash and other emerging NVMs like MRAM, FRAM and PCM due to fewer layers and masks needed in production. MRAM comprises more than 10 layers, which can be so thin that they can be difficult to accurately deposit and monitor. MRAM’s magnetic materials make it extremely unfriendly for advanced fab environments, and it requires special tools, cleanroom, and handling, directly affecting cost. Weebit ReRAM requires only two additional masks (versus 10+ added masks for flash). It also uses fab-friendly materials, so it only adds ~5% to wafer cost, compared to flash which adds ~10-20% and MRAM which can add ~30-40%.
FRAM is today a leading emerging memory technology for data logging since it has a very high write endurance. However, there is a price to pay in terms of memory size. A typical FRAM cell is ~2X larger than the Weebit ReRAM cell.
Selecting the right memory is a critical consideration for data loggers, and choosing off-the-shelf flash isn’t necessarily the right choice. With new IoT use cases, growing security threats, an increasingly competitive landscape, and with emerging embedded NVMs like ReRAM, you can now choose a proven NVM that is low power, low cost, high performance, and highly reliable and secure.