Persistent serial memory can take the strain out of monitoring circuit design in repetitive event systems such as motor controllers, writes Grant Hulse, senior director of marketing at Cypress Semiconductor.
Our world is filled with machines that perform repetitive tasks. For example, factory floor robots move in a series of small steps that are logged until a movement is complete.
Once completed, the repetitive motion and step logging begins again. Smart power meters take power readings, log those readings, and then every few minutes upload their reading to a local consolidator, reset, and begin again.
Portable medical instruments, factory line equipment, brushless motors, all take a series of measurements, react to those measurements, then begin anew.
In-line parametric loggers, error-loggers, engine monitors, battery chargers, HVAC controllers – the list of industrial products that depend on repetitive event management is both broad and diverse.
Persistent memory, or non-volatility, is, in many of these applications, a ‘must-have’ requirement in a repetitive event. Having a robotic arm ‘lose its place’ because of a power glitch in the middle of a complicated process is unacceptable.
Lost readings can require timely recaptures, sample escapes, material damage, and unsatisfied (or sometimes endangered) operators and users. Though repetitive event memories are important, their dollar value in terms of memory market size has remained small.
Dedicated memory IC’s to fill this need are uncommon. A repetitive event memory tends to be quite small (16kbit to 2Mbit), has a serial interface (low cost, low switching noise), and needs to provide both random access and sequential READs and WRITEs at reasonable speeds, such as required by serial interfaces like SPI and I2C.
The closest fit historically has been the serial EEPROM, a very inexpensive memory that matches this application on cost, density, and READ and WRITE speeds, but that has two repetitive event handicaps: endurance, usually limited to a maximum of 100K STOREs, and chip/block erase times that can exceed one second. Designers spend a significant number of hours working to get around these complications in order to benefit from the low cost that serial EEPROM can provide.
A new type of non-volatile memory, the nvSRAM, was designed with repetitive event systems in mind.
The parts operate as SRAM, with fast read, fast write, random access, sequential burst mode, and none of the long block and IC erase times seen in serial EEPROM or serial flash technologies.
The non-volatility is completely invisible to the user: the IC detects power supply disturbancess and moves all data to the non-volatile (NV) storage automatically where it remains unchanged with the power off. When power is restored, the data is automatically moved back into the SRAM and system operation can continue right where it left off.
The product’s NV endurance is specified at greater than one million STORE events. As these STORE events are used only on power faults, the endurance is effectively limitless for embedded and industrial applications. The SRAM operates normally, with unlimited READ and WRITE capabilities.
In addition, at these small memory densities, the added cell circuitry needed to create an nvSRAM results in a very small and reasonable cost premium to serial EEPROM or serial flash, considering the improvements it brings to repetitive event applications.
A smart power meter block diagram is shown in Figure 1. As power from the grid enters the home, the meter takes continuous power factor measurements in short intervals and records a series of these usage readings in the repetitive event memory.
Since the resident is allowed to make power consumption decisions and move power hungry tasks to off-peak hours based on changing usage rates, it is mandated that the power company show that it is correctly tracking this consumption and matching it to the varying rates.
Power meters are clustered by neighborhoods, and data from each meter is uploaded to the power company via power line communication or wireless mesh to a local consolidator every few minutes.
After each send, the repetitive memory in the power meter begins to capture its next set of readings.
The constant writing into the repetitive event memory second after second, hour after hour will eventually consume the endurance limits of EEPROM or flash cells.
The nvSRAM, on the other hand, performs the function as an SRAM. No endurance cycles are consumed. Only on each brownout or blackout is a single endurance cycle used.
HVAC controllers take environment measurements (temperature, humidity, duct pressure, flow) and track time schedules, settings, user trends, alarms, and communication packets with the system master and other controllers.
Continually changing inputs are combined with history and system settings to manage dampers, valves, fans, pumps, chillers, etc.
Data persistence through power disrupt events is required. Industrial HVAC systems are expected to have long product lives and low maintenance and management costs once installed.
Again, the infinite endurance and coding simplicity of an nvSRAM make it well suited for this application. Code size and firmware simplicity are added benefits of using nvSRAM technology.
There is no need to keep track of write counts, to plan for erase cycles, nor to manage complex write protect schemes: nvSRAM even includes a top-to-bottom wrap around feature on write and read bursts.
Repetitive event systems will only continue to expand their many roles. Targeting these applications with an elegant memory solution will aid that expansion and give designers more time to work on the tougher problems in their applications.
Cypress Semiconductor
