Predicting Data Retention
Times of NVRAMs
Designers who need to provide non-volatile storage in their applications can choose from several types of memory, the three most popular of which are Flash, EEPROM and battery-backed SRAM (NVRAM or Non-Volatile RAM). The best choice depends on numerous factors such as cost, required write speed, memory size and how often individual bytes need to be written but in applications that demand high write speeds, NVRAM is the most popular solution.
NVRAMs such as ST’s ZEROPOWER® range consist of three parts: the SRAM, a lithium button cell battery and a control circuit that continually monitors the SRAM’s Vcc line and automatically switches to battery back-up if the supply voltage goes outside of tolerance. In the case of real-time clock-RAMs such as ST’s TIMEKEEPER® range, the devices also include a quartz crystal.
Unlike Flash and EEPROM, where the data retention time is determined solely by the properties of the memory cell, the SRAM used in NVRAMs is inherently volatile so the data retention time depends entirely on the lifetime of the battery, which is affected primarily by temperature and the time for which the SRAM draws current from the battery. Because NVRAMs are often used in applications such as industrial control, where high temperatures may be encountered, or remote data loggers, where field servicing is often difficult and expensive, designers often need to be able to predict with confidence how long the battery will last.
A NVRAM will reach the end of its useful life for one of two reasons: either its original charge will be used up providing back-up current to the SRAM (Capacity Consumption) or aging effects will have made the battery inoperative before the stored charge has been consumed (Storage Life). These two effects have very little influence on each other so that they can be considered as two independent but simultaneously present mechanisms and the data retention lifetime of the device is determined by whichever failure mechanism comes into effect first.
Storage Life
The finite storage life is mainly due to slow evaporation of the electrolyte and is therefore primarily a function of temperature. Figure 1 shows the results of tests performed by ST. The two lines, SL1% and SL50%, represent different failure rate distributions for the cell’s storage life. For example, by looking at SL1% it can be seen that at 60°C the battery has a 1% chance of failure after 28 years and a 50% chance of failure at around 50 years. Clearly, storage life is not a significant factor in overall battery life unless temperatures above 60 -70°C are involved.
If the designer knows that the equipment will be used with a more or less constant ambient temperature, the expected storage life can be read directly from Figure 1. In many cases, however, the ambient temperature will vary, often in a broadly predictable way.
If there is a known pattern to the ambient temperature, Figure 1 can still be used to predict storage life by using the following formula:
SL=T/(t1/SL1 + t 2/SL2 + ... + tn/SLn)
where SL is the predicted storage life, Tn /T is the proportion of the time that the device is at ambient temperature TAn, SLn is the storage life at temperature TAn as found from Figure 1, and T is the total time t1 + t2 + ... + tn.
For example, suppose the battery is expected to operate at temperatures up to 90°C for 600 hours per year and at 60°C or less for the remaining 8160 hours per year and we want to predict the worst-case storage life using the SL1% curve in Figure 1. Then SL1 is 1.8 years, SL2 is 28 years, t1 is 600 hours/year, t2 is 8160 hours/year and T is 8760 hours/year. Using the above formula, we obtain a predicted storage life of about 14 years.
Capacity Consumption
When the SRAM’s Vcc supply is within its specified range, the current drawn from the battery is zero; conversely, when Vcc is out of tolerance, the battery supplies all of the current required by the SRAM. The Vcc Duty Cycle, which is simply the proportion of time (expressed as a percentage) that the device draws power from the external supply and therefore draws no current from the battery, is clearly a major factor governing how quickly the battery is discharged.
The working temperature is also an important factor as it largely determines the current (IBAT) drawn from the battery when the device is operating in back-up mode.
For a given Vcc duty cycle (VDC) and back-up current (IBAT), the battery lifetime in years (LTCC) due to capacity consumption can be very easily calculated from the formula:
LTCC = battery capacity/ (8760 x IBAT x (1-VDC/100))
For example, for the M48T35Y TIMEKEEPER device, which integrates a 32Kx8 NVRAM and a Real-Time Clock, the value of battery capacity would be 0.048Ah (48mAh) and the typical battery current (IBAT) at 70°C is 2666nA. For a given Vcc duty cycle (VDC), the battery lifetime at 70°C is simply calculated from the formula:
LTCC = 0.048/(8760 x 2666 x 10 -9 x VDC/100)
which gives a battery lifetime due to capacity consumption of about 4.11 years at 70°C.
Of course, the value of IBAT is considerably lower at lower ambient temperatures and is also lower for ZEROPWER devices than it is for TIMEKEEPER devices as there is no clock circuit to supply. In fact, a device such as the M48Z08 ZEROPOWER NVRAM consumes around 23nA at 250°C.
Inserting these values into the equation gives a battery life due to capacity consumption of over 20 years when VDC=0 there is no Vcc supply and the device is permanently in battery back-up mode.
In conclusion, it is evident that in the vast majority of applications, the data retention lifetime of NVRAMs is usually significantly greater than the specified value of 10 years. For equipment that is expected to operate for longer than the predicted data retention lifetime, NVRAMs also offer an important benefit in that ST’s proprietary SNAPHAT® package allows the battery to be replaced in seconds without losing any data, providing that the Vcc supply is present.
|
|
|
