Batteries Gates Cyclon Battery Application Manual OCR

Professional Program Session Record

Battery Backup for the
Life of the Product

CATALYST RESEARCH CORPORATION

1421 CLARKVIEW ROAD • BALTIMORE. MARYLAND 21 209-9987 • U.S. A
TELEPHONE (301) 296-7000 • TELEX. 87-768

Session 17

Battery Backup For the Life of the Product

Session Organizer
Edward S. Hatch, Jr.
Catalyst Research Corp.
Baltimore, MD

Session Chairman
Dr. Alan A. Schneider
Catalyst Research Corp.
Baltimore, MD

17/Q

17/1

17/2

17/3

Battery Backup For the Life of the Product

Dr. Alan A. Schneider
Catalyst Research Corp.
Baltimore, MD

Long Life Primary and Secondary Batteries

Thomas Irwin
RAYOVAC Corp.
Madison, WI

Ultra-Low Power CMOS Memory for Battery Backup
Applications

Dr. Isamu Kuru, Msayoshi Nakane, and Kaoru Tokushige
Toshiba America, Inc.
Irvine, CA

Applications of Lithium-Iodine Batteries to CMOS
Memory and Microprocessors

John L. Stegman Jr.
Catalyst Research Corp.
Baltimore, MD

Abstracting is permitted with credit to the source.
Libraries are permitted to photocopy articles in this
volume beyond the limits of U.S. copyright law if for
private, noncommercial use of patrons. For other copy-
ing, reprint or republication permission, write to Elec-

tronic Conventions, Inc., 999 North Sepulveda
Boulevard, El Segundo, California 90245. All rights
reserved. Southcon 1982 Professional Program Papers.
Copyright© Electronic Conventions, Inc.

BATTERY BACKUP FOR THE LIFE OF THE PRODUCT -
AN OVERVIEW

Alan A. Schneider, Ph.D.
Catalyst Research Corporation
1421 Clarkview Road, Baltimore, Maryland

21209

INTRODUCTION

As more computing power is being
built into devices such as terminals,
cash registers, data loggers, etc.,
more attention must be directed to how
these devices will function during and
after main power interruption. A de-
sirable, and almost necessary feature
is that such devices power up in a con-
dition that allows them to continue
their function without reinitialization
by the operator. This means that de-
vices like system time/date clocks
should continue to run and that vola-
tile data should be retained.

POSSIBLE NONBATTERY SOLUTIONS

There are many methods by which
data retention might be accomplished.
Core memories have been used, but they
are large, require more power and are
expensive. The same can be said for
other conventional storage media such
as hard disk and floppy disk. Bubble
memories are an alternative but infor-
mation retrieval is slow because access
is sequential and expense is also a
problem.

2

EAROM's and E PROM's are a popular
alternative but even these have their
drawbacks — access times are measured
in tens of milliseconds, the number of
reliable writes is finite, and there is
a sizeable software burden associated
with storage of data after detection of
incipient power failure and another to
restore function after power up.

Shadow RAM's offer a method of
avoiding some of the software burden by
transferring with a single command the
contents of a RAM to an E 2 PROM contain-
ed in the same package. The burden of
detecting incipient power failure still
rests with the user.

BATTERY BACKUP AS A SOLUTION

Perhaps the most straightforward
method of assuring data retention is to
supply continuous power to the CMOS RAM

chips which are already in use on the
board. Continuous battery power is
also the only way by which devices
like clocks can continue to function
during power down.

Several companies provide batter-
ies for RAM backup applications and
at least one provides a battery-RAM-
switch module which senses impending
power failure and provides a signal
to the processor that the RAM will soon
be on battery backup in a write-pro-
tected mode.

CHOOSING THE RIGHT BATTERY

Not all batteries are suited for
CMOS RAM backup. Some are short lived,
some leak, some have a questionable
safety record. The ideal battery for
CMOS RAM backup should exhibit these
characteristics :

(1) Printed circiut board mountable

(2) Wave solderable

(3) Provide data retention for the
life of the product

(4) Retain data over a wide temper-
ature range

(5) Not require recharging

(6) Not require replacement

(7) Be hermetically sealed to
prevent leakage

(8) Have a proven record of
reliability

(9) Exhibit a high degree of safety

To this point, two generic battery
types have been used as power backup
systems: Nickel-cadmium rechargeables
and several lithium systems. Recharge-
ables have several disadvantages: Some-
what narrow operating temperature range;
replacement required after about four
years of service; recharging electronics
is required; leakage problems can occur;
and backup time is limited to a few
months in the absence of recharge.
They are an excellent choice, however,
when high currents or low cost are im-
portant factors.

1

17/0

LITHIUM BATTERIES

Lithium batteries in general can
provide very long service life, ability
to operate over a wide temperature
range, and high energy' density (watt-
hours/cm^), factors which are essential
for backup and which conventional bat-
teries do not provide. Almost all are
hermetically sealed and some are PC
board mountable using wave soldering
techniques. However, lithium batteries
bring with them their own set of con-
cerns for designers:

(1) Some are limited in the maxi-
mum current they can provide

(2) Safety is a concern with some
systems

(3) Some systems have not been
available for more than a few
years so long-term reliability
may be questionable.

"Lithium battery " is a generic
term. It is important that the de-
signer recognize the differences among
the lithium battery systems and choose
the one best suited for his application.
It is not wise, for example, to choose
a lithium battery only on the basis
that it has the highest capacity rat-
ing and fits in a particular slot on
a PC board. That same cell, in a few
years, may leak or lose much of its
capacity to internal discharge pro-
cesses .

TYPES OF LITHIUM BATTERIES

There are three basic types of
lithium cells which have been offered
for memory backup applications:

(1) Lithium-inorganic electrolyte

(2) Lithium-organic electrolyte

(3) Lithium-solid electrolyte

All provide a distinct advantage over
conventional cells in terms of watt-
hours per unit volume and all have the
promise of long life.

Lithium-Inorganic Electrolyte Batteries

Only one version of this type cell
has been seriously promoted for use in
backup applications — the lithium-
thionyl chloride (Li-SOCl-) cell. It
is manufactured by Tadiran, GTE, Altus
and others.

In addition to the lithium anode,
the cell uses thionyl chloride (S0C1 2 )
the electrolyte solvent and a high-sur-
face-area carbon as the cathode current
collector. SOCI2 serves not only as
the electrolyte solvent but also as the
active cathode species. Because one
component serves two roles, the cell is
very energetic (Ca. 1.1 watt-hours /cm3 ) .
It functions well over a wide temper-
ature range and has an attractive
single cell voltage (>3.5 volts). It
is capable of delivering moderate to
high current (1-1000 ma) depending on
cell design.

The system has also shown several
disadvantages over the past few years.
Self-discharge, the rate at which the
cell consumes its own capacity intern-
ally, can be very high and unpredict-
able. Several manufacturers claim to
have solved this problem, but multiyear
testing is necessary to be certain.

Perhaps the biggest question with
this system is safety. Very large bat-
teries, designed for the military, have
shown a few serious problems. Certain-
ly the smaller versions of this cell
with limited current capability are
much safer. Moreover, several manu-
facturers claim to have solved the
safety problems by mechanical means or
by the use of additives. Nevertheless,
some questions about this system still
exist in the minds of battery profess-
ionals, and the electronics system de-
signer would do well to be certain that
for his application, all the safety re-
quirements are met. Again, time and
the testing of very large numbers of
cells will tell the story.

Lithium-Organic Electrolyte Batteries

At least three types of lithium-
organic electrolyte cells have been
offered for memory backup:

(1) Lithium-manganese dioxide
(Li-Mn0 2 )

(2) Lithium-carbon fluoride
(Li-CF X )

(3) Lithium-sulfur dioxide
(Li-S0 2 )

The Li-Mn0 2 system is offered by
Sanyo, Ray-O-Vac , Mallory and others.
It is a moderate rate system (0.1-10ma)

17/0

2

with a voltage near 2.8 volts. In addi-
tion to the lithium anode, the cell uses
a pellet containing MnC>2 as the cathode
and an electrolyte composed of an in-
organic salt dissolved in an organic
solvent. Its energy density is moder-
ate (Ca. 0.6 watt-hours /cm3 ) . It can
operate over a fairly wide temperature
range with low self-discharge. It has
thus far shown itself to be a safe
system and has promise to be a long-
life (5+ years) system.

Its principal disadvantages are
lower cell voltage, lower energy den-
sity, and lower rate than Li-SOCl2-
To many designers, these are outweighed
by its advantages.

The Li-CF X system is offered by
Panasonic, Eagle-Picher and others. It
is similar in most aspects to the
Li-Mn02 systems except for a slightly
higher current capability.

The Li-SC>2 system is different
from the other lithium-organic electro-
lyte in that it uses a soluble gas
rather than a solid pellet as the cath-
ode. It can build up pressure as it is
discharged or heated and some of the
larger cells use safety vents to pre-
vent case rupture. A few safety-re-
lated incidents have been recorded with
some of the larger cells, but certainly
the smaller cells should cause little
or no problem. Its energy density is
about the same as other Li-organic
systems and it operates over a rather
wide temperature range. Its current
capability is somewhat higher than
Li-MnC>2 or Li-CF X .

Lithium-Solid Electrolyte Batteries

Lithium-solid electrolyte systems
are provided by Catalyst Research,
Mallory and others. Unlike other lith-
ium batteries, the electrolyte is solid,
a distinct advantage for long life.

The solid electrolyte system which
is most used is the lithium-iodine
system. Its electrolyte is solid lith-
ium iodide which accumulates as the
cell is discharged. Unlike other cells,
this cell uses no separator to prevent
internal shorting of anode and cathode.
The electrolyte itself is the separator,
and should it form a crack, it "heals"

itself. This is another important
factor in the performance of this long-
life system. The cathode is a pellet
of iodine mixed with a small amount of
organic to make it conductive.

This system is the cell of choice
of the pacemaker industry, powering
more than 90% of the worlds pacemakers.
It boasts a real 10-year data base en-
compassing more than a million cells.

Its energy density is quite high
(Ca. 0.9 watt-hours /cm^) and its abil-
ity to deliver current is somewhat low
(0.1 - 2 ma), but still high enough for
data retention for many CMOS RAMS.
Open circuit voltage is 2.8 volts.

There have been no safety-related
incidents in its 10-year history.

INTERFACING BATTERY AND ELECTRONICS

Once the choice of battery has
been made, the interfacing to the pro-
cessor or other electronics may seem
trivial. In fact, this is far from
true. The designer must worry about
the characteristics of the switch,
logic level incompatibilities created
by two power supplies, and drawing too
much current from the battery during
switch transitions. These and other
topics will be covered in a later
paper .

SUMMARY

Battery backup seems to be the
ideal solution for data retention and
powering devices like real-time clocks.
When used with CMOS RAM's, battery back
up offers the speed and cost advantages
of RAM storage with the additional ad-
vantages of non-volatility.

Lithium batteries offer real ad-
vantages over conventional and re-
chargeable batteries, but care "must be
exercised to find the right lithium
battery. Factors such as safety, en-
ergy density, and proven reliability
must be considered.

With the proper choice of battery
and interfacing electronics, it is pos-
sible today to design a system with a
backup battery that can last the life
of the product, a system in which the
battery is treated like any other non-
replaceable component and soldered
directly to the PC board.

3

17/0

LONG LIFE PRIMARY AND SECONDARY BATTERIES

Thomas Irwin
Application Engineer
RAYOVAC Corporation
101 East Washington Avenue
Madison, WI 53703

INTRODUCTION

Since the development of the elec-
trochemical storage battery almost a
century ago, a growing number of appli-
cations have employed batteries both as
their only power source and as a back
up supply during commercial power out-
ages. In the last 100 years, a variety
of battery chemistries have evolved both
in primary and secondary batteries.

Today we are entering a microcom-
puter age where the demand for standby
power to preserve volatile memories and
to insure uninterrupted clock operation
is increasing rapidly. Applications for
standby power are in computers or process
control equipment using real time clocks,
computer printers where the customer
initializes his printing format in
volatile memories, portable data ter-
minals used in inventory control, tele-
communications, portable test equipment,
and a growing list of other applications.

Miniature batteries are becoming an
important tool to the microcomputer
system designer. When complimented with
power line sensing and write inhibit
semiconductor devices, it enables the
designer to use standard NMOS and CMOS
RAMs without risking program or data
loss. The designer can then continue
using high speed low cost memories with-
out sacrificing write speeds or having
to resort to EPROMs or high priced low
speed multi-voltage EEPROMs. Nor, is
the number of write operations limited
to some number X due to wear-out
phenomena .

The purpose of the battery is to
protect the microcomputer, its volatile
memory, its clock circuitry, or any com-
bination of these from AC line distur-
bances. To design a battery backup
system, an engineer needs an under-
standing of power line disturbances.
The types of abnormal power line con-
ditions to be reckoned with are 1)
Transients: a short term voltage or
current disturbance having a duration
between 1 nanosecond and 5 milliseconds.
2) Voltage Sags: a drop in voltage

lasting more than 5 milliseconds but
less than 5 seconds. The voltage value
must be below the lower specification
limit for nominal voltage. 3) Power
Interruption: a complete loss of voltage
for any period greater than 5 milli-
seconds, but less than 500 milliseconds.
4) Power Outage: a complete loss of
voltage for any period exceeding 500
milliseconds. 5) Brownout: a deliberate
reduction in nominal voltage supplied by
the utility company. The value may fall
at or below the lower limit specified
for nominal voltage.

To quantify the problem, IBM did a
study-'- of AC power sources for typical
data processing equipment. The study
involved 49 locations and 125,000 hours
of monitoring. The study was made from
July 1969 to July 1972. The information
is broken down into duration of distur-
bance and percent of voltage lost.

During the 125,000 hours of mon-
itoring 1,790 voltage sages, power
interruptions, and power outages
occurred; that is, one disturbance every
69 hours of operation.

Assuming the power supply can op-
erate down to 80% of nominal line
voltage, Figure 1 shows line disturbances
that will cause memory problems. In
125,000 hours of operation there would
be 1,040 disturbances, or one problem
every 120 hours of operation. With the
addition of a battery backup system,
errors due to line disturbance can be
controlled .

Additional considerations to be
aware of are the electronic cash
registers which may be shut off at night
or on weekends and without battery backup
could encounter loss of sales totals
and/or inventory sold. Computerized
office equipment involved in office
moves and rearrangements can be subjected
to power outages of several minutes,
hours, or days. Without battery backup
initialized data, accumulated data, and
clock functions can be lost.

17/1

1

DURATION

PERCENT OF

LINE VOLTAGE

SAG

IN

CYCLES

20-60

60-70

70-80

80-90

TOTAL

.5 to

1

3

16

673

692

1

2

7

11

35

54

2

3

1

3

13

33

50

3

4

3

8

49

60

4

5

1

9

26

46

82

5

6

5

13

14

39

71

6

7

1

12

6

23

42

7

8

1

2

6

7

16

8

9

3

2

3

9

17

9

10

3

5

4

12

10

15

1

6

9

17

33

15

30

1

4

13

18

36

30

120

1

4

8

43

56

120

900

97

97

over

900

1

465

466

TOTALS

16

71

139

1588

1784

FIGURE 1 VOLTAGE SAGS FOR 125,000 HOURS

One additional problem is worth
mentioning. Some computer products are
manufactured and shipped with pre-
programmed information. Batteries must
be used for memory retention in the
interim between shipment and instal-
lation by the end user.

The purpose of this paper is two-
fold: 1) to review the history of
primary and secondary batteries used for
memory backup applications; and 2) to
present the most recent developments in
lithium batteries as they relate to the
microcomputer/microelectronics
environment .

Before I get into the different
battery systems, I'd like to establish
some general characteristics of an
"Ideal Power Cell" and use it as the
yardstick to which all batteries are
compared and measured. The "Ideal Power
Cell" would have:

1. Maximum Energy Density - offer a
maximum amount of energy stored in
a minimum amount of space.

2. Unlimited Shelf Life - no self-
discharge irregardless of storage
period.

3. Rechargeable - fully rechargeable at
any charging rate.

4. Light Weight - combine maximum energy
density with a minimum amount of
weight in the materials used.

5. Safe - the ideal cell would be com-
pletely safe - no leakage or gassing.
The contents would be harmless to
people and the environment.

6. Good Performance - deliver the
required current and voltage under
a wide range of temperature and
load conditions.

7. Cost - the materials used to make

the cell would be in plentiful supply,
multisourced , and manufacturing costs
would be minimal.

SECONDARY BATTERIES

Nickel-Cadmium Batteries

Nickel-cadmium rechargeable bat-
teries are secondary batteries. They
were first developed in the 1940 's.
Figure 2 is an overview of several bat-
tery systems and their respective energy
densities. Although the nickel-cadmium
battery has a low energy density, the
fact that it can be recharged compensates
for this characteristic

Advantages of nickel-cadmium bat-
teries are: they offer excellent rate
capabilities, have a stable discharge
voltage, perform very well at low tem-
peratures (-20°C and below) , and have an
operational life of 3 to 5 years, or 500
to 1000 charges, whichever comes first.

Disadvantages are: it has a low
energy density, an individual cell
voltage of 1.2 volts requiring a stack
of several cells to form a usable battery
for memory backup/clock applications, a
high self-discharge rate of 1% per day
at room temperature and several percent
per day at temperatures above 4 0°C, and
if subjected to many shallow discharge/
charge cycles can develop a "memory
effect" where it can fail to deliver its
full capacity when called upon to do so.

Figure 3 compares the nickel-cadmium
to the ideal power cell.

Most applications employ batteries
consisting of a 3 or more cell stack.
Typical applications are for memory
backup/clocks, calculators (LED), test
equipment, toys, etc.

Sealed Lead Acid

Sealed lead acid rechargeable bat-
teries are members of the secondary bat-
tery family. Conventional lead acid
batteries with the liquid electrolyte
have been available for over seventy
years. Sealed lead acid batteries have
been available for about 7 years.
Figure 2 compares its energy density to

2

17/1

other battery systems. Although sealed
lead acid batteries have a low energy
density, the fact that they can be re-
charged compensates for this
characteristic .

PRACTICAL ENERGY DENSITY FOR CELL SYSTEMS

ALKALINE

MID-LIFE

ENERGY

LITHIUM

MID-LIFE

ENERGY

SYSTEMS

VOLTAGE
V

DENSITY
WH/1

SYSTEMS

VOLTAGE

DENSITY
WH/I

AIR/Zn

1.3

970

S0Cl 2 /Li

3.5

760

AgO/Zn

1.55

650

AgjCrO^/Li

3.0

670

HgO/Zn

1.35

550

CF x /Li

2.8

570

Ag20/Zn

1.55

460

MNO2/L1

2.9

520

MNC7/ZN

1.3

230

SO2/L1

2.8

150

CARBON/Zn

1.3

110

SOLID STATE

2.8-2.0

500-530

N1CAD

1.2

75

SEALED

2.0

70

LEAD ACID

CELL SIZE: 11.6 x 1.2 mm
DRAIN: 5 MICROAMPERES

FIGURE 2

Advantages of sealed lead acid
batteries are: they have a nominal cell
voltage of 2 volts, can deliver high
currents, have a flat discharge, can
operate over a temperature range of
-40°C to +60°C, operational life of 3 to
5 years or 100 to 250 charges, and an
ability to tolerate cell reversal with-
out damage.

Disadvantages are: its low energy
density, a self-discharge rate of 6%
to 8% per month at room temperature,
limited number of recharges, the smallest
size available is comparable to a "D"
size, and it is physically heavy.

Figure 3 compares the Sealed Lead
Acid battery to the ideal power cell.

Sealed Lead Acid batteries are
available in sizes from 2.5 AH to over
25 AH and in 2 volt increments from 2
volts to 48 volts. Typical uses are
standby power for powering large blocks
of memory at a 2 volt data retention
level, uninterruptable power supplies
for small to medium size computers,
alarms, emergency lighting, and as

cyclic power for tools, instruments,
engine starting, televisions, and
video tape recorders.

Mercury Batteries

Mercury primary batteries were
first mass produced during World War II.
Figure 2 shows mercury's energy density
relative to other battery systems. As
you can see, mercury is very high in the
table and at the time of development was
the highest energy density battery
available. This characteristic also
facilitated development of miniature
button cell batteries having capacities
in the hundreds of milliampere-hours .

Advantages of mercury batteries are
their high energy density, flat dis-
charge characteristic, good rate capa-
bility at currents from microamperes to
hundreds of milliamperes , and an oper-
ating temperature range of -2 0°C to
60°C.

Disadvantages of the battery are:
a cell voltage of 1.35 volts, a shelf
life at room temperature of 1-2 years,
problems of mercury migration during
discharge which can internally short out
the cell, the environmental concern of
mercury disposal, and relatively high
cost.

In some of the first computerized
cash registers, 1 AH mercury batteries
were used for memory backup. However,
it was found that the batteries did not
stand up well under the daily temper-
ature cycling. Too often the cells
would leak or go dead after one-half of
their capacity had been expended.
Coupled with the 1-2 year shelf life,
the mercury battery bowed out of the
memory backup market.

Figure 3 compares the mercury bat-
tery to the ideal power cell.

Present day applications for the
mercury battery are in hearing aids,
cameras, instruments, and smoke detec-
tors, which benefit from its high energy
density characteristics.

Alkaline Batteries

The alkaline manganese, or simply
alkaline battery as it is commonly
called, was developed about 25 years
ago. While most of the alkaline bat-
teries available today are primary

17/1

3

COMPARISON TO IDEAL POWER CELL
CHARACTERISTICS

CHARACTERISTIC

IDEAL POWER SEALED

CELL LEAD ACID NIC AD MERCURY ALKALINE SILVER LITHIUM

MAX ENERGY DENSITY High
SHELF LIFE Excellent

Low

Low

RECHARGEABLE
LIGHT WEIGHT
SAFE

GOOD PERFORMANCE

COST

Yes
Yes
Yes
Yes
Low

Low/Med Low
Yes

Yes

No

Good

Med

No
Good
Yes
Med

High
Low
No
No

High

Med
Good

No

No
Good

Yes

Med

High
Good
No

Med

High

High
No
Yes

Good*
Yes

High

♦Depends upon particular system

FIGURE 3

batteries, there is a rechargeable
version. Figure 2 is an overview of
several energy density chemistries. As
you can see, Alkaline Manganese is an
improvement over the zinc carbon and
nickel-cadmium, but has only 40% of the
energy density of mercury.

Alkaline batteries are available
in sizes from 40 MAH button cells to
multiampere hours "D" cells and battery
voltages from lh volts to 9 volts.

Advantages of the alkaline battery
are its high rate efficiency, improved
shelf life of 3-5 years, good to excel-
lent low temperature performance, cell
voltage of 1.55 to 1.6 volts, and low
cost.

Disadvantages are its low to
moderate energy density. Alkaline
button cell batteries have insufficient
energy to meet memory backup power re-
quirements for several years of oper-
ation. Elevated temperatures accel-
erate the self-discharge rate, limiting
its overall service life.

Figure 3 compares the alkaline
manganese battery to the ideal power
cell.

Present day applications for this
battery are LCD calculators, cameras
(exposure control system and strobes) ,
and toys.

Silver Oxide

Silver oxide batteries were
commercially developed about 10-15 years
ago. There are two similar chemistries
used in commerical silver oxide button
cell batteries, monovalent and divalent
silver oxide.

Figure 2 lists the energy density
of both silver oxide chemistries. Mono-
valent silver oxide has the formula
Ag 2 and falls between alkaline and
mercury. Divalent has the formula AgO
and offers the highest energy density
outside of the Zinc-Air system. (The
mechanics of activating/deactivating a
Zinc-Air battery prevents its use in
battery backup applications.) The high
energy densities lend themselves to
miniature button cells. Almost all
commercially available silver oxide
batteries are button cells. The devel-
opment of larger cylindrical size silver
oxide batteries is possible but is
restricted by the price of silver.

Characteristics of the monovalent
and divalent cells are their high energy
densities - about 2-2^ times that of
alkaline, an operating voltage of 1.55
volts, light weight, good low and high
rate current capacities, a shelf life
at room temperature of 3 to 5 years in
high rate cells, and 4 to 6 years in
low rate cells, good low temperature
operation, and low leakage rates.

17/1

Disadvantages of both types are the
increased self-discharge rates which
occur when cells are subjected to high
temperatures for extended periods of
time. The fluctuating market price for
silver in the past couple of years has
increased silver oxide button cell
prices significantly. Currently, silver
is below $10/Troy Oz . and it is again a
viable cost product.

Differences between mono and
divalent cells occur in two areas.
First, divalent has a higher energy
density, about 30% more, than mono-
valent. In the same size cell, divalent
can offer significantly more energy.
Second, divalent chemistry produces a
higher cell voltage of 1.8 volts com^
pared to 1.55 volts for monovalent. In
commercial divalent silver oxide cells,
special techniques are used to reduce
the voltage to 1.55 volts without
sacrificing the higher energy density.

In terms of low temperature oper-
ation, a divalent silver oxide watch
cell measuring 11.56 mm diameter by 5.36
mm high and rated at 1.5V, 220 MAH, has
delivered 113 microamperes at -70°F at
a closed circuit voltage of 1.44 volts.

Silver oxide button cell batteries
range in size from 2.10 mm high by 6.80
mm diameter to 5.36 mm high by 11.56 mm
diameter and capacities of 17 MAH to
220 MAH.

Figure 3 compares both silver oxide
battery systems to the ideal power cell.

Lithium Batteries

The most exciting new battery
system entering the marketplace today
is lithium. Actually, it should be
referred to as lithium systems since
there are in excess of 10 different
systems that use lithium.

Figure 4 is a partial listing of
companies and the lithium technologies
they are or have investigated.

The primary incentive for the
development of lithium batteries is
their potential for very high energy
densities which in small batteries can
realistically be expected to be 450 to
760 Watt-Hours/Liter delivered. Another
incentive is the higher cell voltage of
2.6 to 3.6 volts. Lithium batteries
also exhibit shelf lives of 5 to 20

PARTIAL LISTING OF COMPANIES INVOLVED IN LITHIUM
CELL DEVELOPMENT/MANUFACTURE

COMPANY LITHIUM SYSTEMS

RAYOVAC CORP. Mn02, CuS, Cu 2 S, S0Cl 2 , Ac^CrO^ FeS, CF x ,
SOLID STATE

TOSHIBA RAYOVAC Mn02

SANYO Mn0 2

MATSUSHITA CF X , CuO, Mn02

UNION CARBIDE Mn0 2 , FeS 2 , S0Cl 2

MALLORY Mn0 2 , SOj

HITACHI-MAXELL FeS

GENERAL ELECTRIC Mn0 2

HONEYWELL S0Cl_ 2 , V 2 5

GTE S0Cl 2

EAGLE-PITCHER CF X

VARTA Mn02, Bi 2 3

BEREC MnC^

ALTUS S0Cl 2

ULTRA ENERGY S0 2

CATALYST RESEARCH SOLID STATE

WILSON GREATBACH Br 2 COMPLEX, SOLID STATE, Ag COMPLEX

SAFT CuO, Bi 2 Pb 2 5 , Mn0 2 > S0Cl 2 , AG 2 CR0t|

PCI S0 2

TAD I RAN S0Cl 2

FIGURE <t

years depending upon the particular
lithium system. They can also operate
at low temperatures to which conventional
•batteries would be inoperative.

Lithium is the lightest metallic
element known to man and is very re-
active. It is a far more reactive
material than zinc and even surpasses
potassium and sodium. The extreme
reactivity of lithium metal, which
theoretically makes it such an attrac-
tive anode material, also makes it very
difficult to work with. In the presence
of water vapor or water it can ignite
spontaneously. Hence, assembly must
take place in dry rooms and the seals on
the cells must be of a very high quality.

In terms of lithium chemistries
which are at a production or prototype
level, the major ones are solid state,
sulfur dioxide (S0 2 ) , carbon monofluoride
(CF X ) , manganese dioxide (MnC^), and
thionyl chloride (SOCI2). Figure 2
compares the energy densities for these
lithium systems. Figure 5 is a graph
of their discharge characteristics.

17/1

5

ENERGY SYSTEMS

Lithium/Manganese Dioxide

CELL SIZE: 11.6mm O.D. x 1.2mm HT.
DRAIN: 5 Hicroamp

10 SO

120 160 200 210
CAPACITY (mAh/cell)
Characteristic Discharge Voltage of Energy Systems
FIGURE 5

Solid State

Lithium solid state batteries are
as their name implies, composed of a
solid anode, solid electrolyte, and
solid cathode. Combined with hermetic
sealing, they may achieve shelf lives of
10 to 20 years or longer.

Solid state batteries were ini-
tially used in cardiac pacemakers where
high reliability, low drains, and long
operational life were extremely impor-
tant. As power requirements for CMOS
memories were reduced to the low micro-
ampere range, a new application for
solid state batteries developed.

Solid state batteries offer a high
energy density, approximately twice that
of the alkaline battery, a long shelf
life of 10 to 20 years, a wide temper-
ature operating range of -55°C to +125°C,
no leakage, a cell voltage of 2.8 volts,
and a flat discharge characteristic.

In button cell sizes the principal
limitation is a typical current drain of
20-25 microamperes, 50 microamperes
maximum. They also have a moderate to
high cost.

As the standby power requirements
for CMOS RAMs continue to fall into the
low and sub-microampere range, more
applications will develop for solid
state lithium batteries.

Lithium manganese dioxide cells
consist of a lithium anode, separator,
manganese dioxide cathode, and an
organic electrolyte. They have a
nominal cell voltage of 3 volts and
deliver approximately 2.9 volts under
load. Its energy density is better than
twice that of an alkaline battery. They
usually employ crimp seals but can use
a hermetic seal if the application so
requires. The shelf life is estimated
at 5 to 10 years. Lithium manganese
dioxide cells are available in button
cell and cylindrical sizes ranging in
capacity from 30 MAH to 1000 MAH.

Advantages are its high energy
density, 3 volt cell voltage, good rate
capability, low self-discharge rate, has
a temperature operating range of -2 0°C
to +50°C, and a sloping discharge which
lends itself to end of battery life
sensors.

Lithium/Carbon Monofluoride

Lithium carbon monofluoride cells
consist of a lithium anode, separator,
carbon monofluoride cathode, and an
organic electrolyte. The nominal cell
voltage is 3 volts and 2.8 volts under
load. It has an energy density slightly
higher than that of lithium manganese
dioxide. Lithium carbon monofluoride
cells are available in sizes from button
cells to "C" size cylindrical cells.
Crimp seals are used in both the button
and cylindrical cells. Shelf life is
estimated at 5 to 10 years. Capacities
range from 40 MAH to 5 AH. The button
cells can deliver continuous currents
up to 250 uA with pulsing up to 10 milli-
amperes. The larger cylindrical cells
can deliver currents from microamperes
to hundreds of milliamperes .

Advantages are its high energy
density, cell voltage of 3 volts, good
rate capability, low self-discharge
rate, and an operating temperature
range of -20° to +60°C.

Lithium Thionyl Chloride

Lithium thionyl chloride cells are
a high voltage, high energy density
lithium system. Figure 2 shows lithium
thionyl chloride at the top of the list.

6

17/1

If lithium systems were classified
as low, medium, or high rate, thionyl
chloride would fall in the high rate
category. It has a nominal voltage of
3.6 volts and is commercially available
in sizes from .85 AH to 10.8 AH.

They are available both in low
profile prismatic and conventional
cylindrical shapes. The basic con-
struction consists of a lithium anode, a
separator, a carbon cathode, and a
thionyl chloride electrolyte/depolarizer.

Advantages are: a very high energy
density, a high cell voltage, excellent
rate capability, with hermetic seals - a
shelf life of up to 10 years, a very low
self-discharge rate, and a temperature
operating range of -55°C to +70°C.

Disadvantages of the system are its
susceptibility to a voltage delay problem.
If initially unused for some period of
time, a passive layer develops on the
anode. If the cell is subjected to a
light current drain, the voltage drop
will be very small to non-existant . It
is only on a heavy drain that the cell
voltage drop is pronounced. The delay
time during heavy drains can be several
seconds in length. Under both light and
heavy drains the layer is dissipated.

Lithium Sulfur Dioxide

Lithium sulfur dioxide cells have
been in production for several years,
primarily for military markets. They
are available in sizes of .43 AH up to
30 AH. The nominal cell voltage is 3
volts and 2.8 volts under load. The
basic cell is spiral wound consisting of
a lithium anode, a separator, a carbon
cathode, and a sulfur dioxide rich
electrolyte. The cells are usually
hermetically sealed. The sulfur dioxide
cells are operated at a positive internal
pressure which can reach 100-200 PSI
(pounds per square inch) at elevated
temperatures. The operating temperature
range is -65°F to +165°F. At temper-
atures between 2 30°F to 25 0°F and
pressures of 450 to 500 PSI, a safety
vent is activated rendering the cell
inoperative. The sulfur dioxide system
can sustain currents from microamperes
to amperes in its "D" cell size. Shelf
life is estimated to be about 10 years.

Lithium sulfur dioxide cells are
also susceptible to the same type
voltage delay seen with lithium thionyl
chloride. Here too, good progress is
being made to eliminate the problem.

17/1 7

Lithium as a composite battery
system has the advantage of a high
energy density, a cell voltage of 2.6
to 3.6 volts, a wide temperature oper-
ating range, greatly improved shelf
life, low leakage rates, light weight,
good rate capabilities and long term
reliability.

Lithium is not a high current drain
battery system. On an equivalent size
basis to an alkaline battery, it cannot
supply the same high currents to a load.
Lithium is high priced primarily because
it is not yet in mass production. There
is a lack of universal availability and
there are government restrictions on
shipment of lithium cells having over
one-half gram of lithium.

Today's applications for lithium
button cells are strongest in the watch
and calculator markets. However, a
growing number of applications for >
button cells as small as the 4 MAH to
as large as multiampere hour thionyl
chloride cells, are appearing in indus-
trial controls, test equipment, telephones ,
memory backup, and computer clock/timing
functions .

Figure 3 compares the lithium systems
with the ideal power cell.

CONCLUSION

The microelectronic system designer
now has a broad portfolio of battery
systems with which to meet his backup
power requirements. Secondary batteries
like the Nickel-Cadmium and Sealed Lead
Acid should be considered where the
application could require frequent re-
placement of primary batteries, where
high current drains are required, where
primary batteries may not be replaced,
and if rechargeability makes the device
more desirable to the user. Primary
batteries like the silver oxide and
lithium systems can offer many years of
reliable protection for volatile
memories and clocks, can reduce mainte-
nance, and in some cases may be good
for the life of the equipment, can
eliminate the need of charging circuitry,
can operate over broad temperature ranges ,
and can be directly mounted on the
printed circuit board.

While this paper presents some
design guidelines for selecting and
applying primary and secondary batteries,
it is by no means complete. It is in-
tended as a guide to you, the systems
designer, for determining your battery
requirements .

In the process of making your bat-
tery selection, it is very important to
solicit more detailed information and
assistance from the battery manufac-
turers. Generally, manufacturers have a
group of trained application engineers
to provide technical assistance and
specification guidelines. By working
with the battery manufacturer, you can
minimize risk in product design,
optimize product performance and con-
tinue the successful alliance of bat-
teries with microelectronics.

REFERENCES

1. Memory Support for Data Processing
Equipment; Mr. David Kuykendall,
1979 Wescon Professional Program,
Session #12.

2. Power Sources for Volatile Memories
Joseph Carone, 1980 Wescon Pro-
fessional Program, Session #24.

8

17/1

ULTRA-LOW POWER CMOS MEMORY FOR BATTERY BACK-UP APPLICATION

Dr. Isamu Kuru,
Msayoshi Nakane,
and

Kaoru Tokushige
Toshiba America Inc.
2151 Michelson Drive
Irvine, California 92715

It was becoming widely acknowl-
edged that the Complimentary-Metal
Oxide-Silicon (C-MOS) process provides
many advantages over other processes in
fabricating modern integrated circuits
such as Random-Access-Memories (RAMS)
and logic circuits. The major
advantages of the C-MOS process over
other MOS processes are 1) wider
operating temperature range; 2) wider
operating voltage range; 3) higher
noise margins; and 4) low power
consumption. Until recently, the
major disadvantage to the C-MOS process
has been slow operating speed but, as
we shall show in this paper, this
disadvantage has been overcome by
advancement in both circuit design and
process capability by Toshiba
Corporation which has culminated in the
very popular TC5516/17/18 series of 16K
static C-MOS RAMs currently in
production.

The ability to produce a high-
density RAM with fast access time (200
nSec) and exceptionally low power
consumption (2-3 nA) during stand-by
operations has opened many new
application opportunities where
non-volatile memory retention or low
operating power are key issues. This
article will describe this new family
of ultra low standby power CMOS RAMs
and their use in non-volatile battery
backup applications.

CMOS RAM

The characteristics which make
CMOS RAMs ideal devices for non-
volatile memory storage applications
are their very low power dissipation
and their very low voltage requirement
(2V) during standby operation where
the only function of the device is to
retain the memory information already
stored in the device.

Figure 1 through 3 are presented
to describe how these merits are
achieved in CMOS RAMs.

17/2

Figure 1 describes a typical CMOS
memory cell structure. This basic
circuit is used to store one bit of
digital information and is replicated
many times within the completed RAM
device. This circuit is known as a
"bi-stable flip-flop". The bi-stable
flip-flop has only 2 stable operating
modes: either Trl and Tr4 are both on
(resulting in the storage of a "0"
logic level) , or Tr2 and Tr3 are both
on (resulting in the storage of a "1"
logic level) in the cell. Either of
these two stable conditions will be
maintained as long as VDD is larger
than the sum of the voltages necessary
to turn on one P-channel transistor
(Trl or Tr2) and one N-channel tran-
sistor (Tr3 or Tr4) . These two
voltages are essentially the threshold
voltages of the transistors and
usually are in the range of 1-2.0
volts and it is this characteristic
that gives CMOS RAM the ability to
maintain stored data with VDD down to
the 2V level.

Further investigation of this
circuit indicates that when the bi-
stable flip-flop is in either of its
two stable states, no DC path exists
between VDD and VSS because one pair
of transistors is always off. This
characteristic results in the very
low standby current consumption of
the CMOS RAM during inactive periods
of operation.

Figures 2 and 3 describe this
structure in more detail. Figure 2
shows a cross section of the CMOS
inverters used to implement a memory
cell. Figure 3 is a schematic
representation of the inverter. When
this inverter circuit is in either of
the two stable conditions, it can be
seen that the P-channel transistor
and the N-channel transistor cannot be
turned on simultaneously. This means
that essentially no power is
dissipated in the inverter when the
circuit is in a stable state.

In reality, however, a very small
amount of power is dissipated even in
the stable states because the tran-
sistors are not "ideal" and therefore
a small leakage occurs even in the off
condition. The magnitude of the
leakage current is proportional to the
area of the p-n junction used to form
the transistor, and, in the case of
the TC5516/17/18, the dominant
component arises from the junction
formed by the Pwell and the N substrate
(the Pwell-Nsub junction) . In a normal
CMOS RAM, the total area of the Pwell-
Nsub junction is very close to the
total chip area. Using 5mm X 5mm as
an approximate chip size and 0.1 NA/mm2
as a typical leakage current density at
room temperature, the expected CMOS
RAM leakage current in standby mode
is 2.5 nA.

The leakage current is dominated
by the "Carrier Generation/Recombi-
nation" effect and therefore is
temperature sensitive. The expected
change in G-R related leakage with
temperature predicts that this leakage
current will double with every 10 C
increase in temperature. Thus, if InA
of leakage is observed at 25 C, the
expected leakage of the same device at
85 C would be 64nA.

Low Standby Power CMOS RAM

As was described above, the
standby power dissipated in a CMOS RAM
is determined by the leakage current
across the reverse-biased P-N junction
in the device. In the actual device,
the dominant causes of Generation-
Recombination Centers in the device
are metallica impurities found in the
depletion layer of the P-N junction.
These metallic impurities in the
depletion layer act as additional
Generation-Recombination Centers
(beyond the normal expected due to
crystal imperfections or intersticial
sites) and therefore add significantly
to the background leakage current.

Toshiba has developed proprietary
CMOS process technology which enables
the reduction of these metallic
impurities in the P-N junctions by a
significant amount. Toshiba utilizes
this process in the production of the
currently available TC5516 APL device,
a 16K Static CMOS RAM. Typical
standby current values at room
temperature for this device are 2-3 nA.
Figures 4 and 5 describe the leakage

current characteristics of the TC5516
APL versus voltage and temperature.
Figure 6 is the standby current dis-
tribution data for a typical diffusion
lot. Approximately 9 0% of the
distribution have standby current less
than 10 nA and the center of the
distribution is 2-3 nA.

The above standby characteristics
make the TC5516/17/18 APL devices very
attractive and practical for non-
volatile memory storage applications
where a primary battery such as a
Lithium Cell (rather than a
rechargeable cell like Ni-Cd) is the
main power source.

Table 1 lists the current products
from Toshiba which exhibit these very
low standby power characteristics.

Battery Back-up Applications

Since these devices exhibit such
exceptional standby characteristics,
their major applications are in
battery backup systems. Toshiba has
incorporated several "easy-to-use"
features in the fundamental design of
the parts with these applications in
mind .

The first is compatiblity with
existing N channel MOS RAMs which have
been widely used for years. The
RC5504 APL (4K X 1) and TC5514 APL
(IK X 4) 4K CMOS RAMs are pin-
compatible and functionally equivalent
to the MK4104 (4K X 1) and i2114
(IK X 4) NMOS RAMs respectively. The
memory speed, which in the past has
been one major drawback in the use of
CMOS memory, has been dramatically
improved in recent years and the
timing of the TC5504 APL and TC5514
APL are essentially equivalent to the
corresponding NMOS RAM. This, then,
makes it very convenient for users to
change existing designs from 4K NMOS
RAMs to non-volatile 4K CMOS RAMs
without changing the memory board
itself.

"Easy-to-use" features in battery
backup applications was given full
consideration in the design of the
recently developed and marketed 16K
CMOS RAMs.

The 16K CMOS RAMs are available
in the industry standard 24 pin, 600
mil dual-in-line package and with a
pin-out compatible with the very

2

17/2

popular 2716 16K EPROM. Additionally,
to provide even greated flexibility
for assembly on the memory board, two
different pin-outs were developed.

The TC5517 APL/BPL was developed
to achieve pin-out compatibility with
NMOS 16K RAMs such as the popular
TMM2016P. The TC5516 APL and TC5518
BPL were developed specifically for
battery backup applications where one
of the two available chip-enable
signals can be used to control the data
retention mode for the entire memory
board. For these devices, CE2 is
the signal used to place the device
into the standby mode. This signal
can be commonly connected to all
devices on the board and when simply
pulled to a high logic level, used to
place all memories on the board in the
standby, power-down, data retention
mode .

Another merit of the TC5516/18
devices in battery backup applications
is that the input voltage level
required for all inputs to the memory
during standby operation need only to
be TTL compatible signals. This is
due to the internal design of the
device which gates all of these inputs
with the CE2 signal in a NOR con-
figuration. This fully eliminates
the need for external pull-up
resistors which are normally required
for other types of CMOS RAM with only
one chip-enable signal.

Replacement of EAROM and EEPROM with
Lithium Battery Backup TC5516 APL

As was shown in Figure 6, the
TC5516 APL has extremely low standby
current. Figure 7 is an estimated
data retention lifetime of a typical
TC5516 APL memory system using a long
life Lithium battery for backup.
The worst case standby current in the
TC5516 APL was assumed from Figure 6
to be 50 nA, which is 5 times the
average standby current in the device.
A fully populated 64K byte system,
which would use 3 2 of the above
devices, exhibits a data retention
time of greater than one year. An
8K byte memory system, using 4 of the
above devices, would have an expected
data retention time of greater than
10 years for just a single small
button type Lithium cell. This is
comparable to the predicted data
retention time of both EAROM and
EEPROM. Additionally, when you

compare other characteristics such as
write cycle time, you find that the
CMOS RAM implementation is superior
by several orders of magnitude.

A third alternative over EEPROM
and EAROM is that both of these devices
experience a "wear-out" phenomenon
each time information is written into
the memory. The current predictions
show a maximum number of writer cycles
for each of these devices on the order
of 10 5 cycles. In any common computer
or microprocess application, 100,000
write cycles can be accumulated very
early in the installed life of the
equipment making these non-volatile
types of memories unusable for a
majority of applications. CMOS RAM
does not experience this wear-out
phenomenon and therefore can be
written to an unlimited number of
times without wearing out. A major
and distinct advantage of CMOS RAM
over EAROM and EEPROM.

Basic Circuit for Battery Backup
CMOS Application

Figure 8 shows the basic circuit
used to provide backup power from a
battery to a CMOS RAM. When the
system power supply is at its normal
operating voltage, diode Dl is
conducting (passing current) and diode
D2 is non-conducting, which isolates
the CMOS RAM from the battery. In
this condition, the actual voltage
provided to the CMOS RAM is VS (power
supply voltage) - VFl (forward voltage
drop of diode Dl) . When the power
supply fails and VS begins to fall,
diode D2 will begin to conduct when
VS-VF1 < VB-VF2 is established. The
resulting voltage to the CMOS RAM
will then be VB (battery voltage) -
VF2. Since this value can drop all
the way to 2V using the TC5516 APL,
data retention will be guaranteed for
battery voltages down to approximately
2.7V.

This circuit is the simplest
possible backup mode and operates in
a fully passive mode.

Battery Backup Memory Circuit Using 16K
CMOS RAMs with Two Chip-Enables

For this example circuit, Figure
9, when the system power is at its
normal operating voltage, diode Dl
and transistors TR1 and TR2 are all on
(on conducting) , and the resulting

17/2

3

VDD to the CMOS RAM is VS-VSAT2.
VSAT2 is the saturation voltage of
transistor TR2 and is typically 0.2V
for common transistors. Referring
to the previous example. Figure 8,
the forward voltage drop of diode
Dl, VF1, is typically in the range of
0.6 - 0.8V, which is considerably
higher than VSAT2._ Thus it can be
seen that the circuit shown in Figure
9 will deliver actual voltage to the
CMOS RAM much closer to VS than in
Figure 8. This difference is
occasionally important to ensure the
entire system works within the
guaranteed voltage range. Guaranteed
operational voltage range for systems
using TTL or LSTTL gates if 5V+5%
which CMOS RAMs usually allows
5V+10%. Using a circuit similar to
Figure 9 with VSAT2 0.2V, 5V+5% -
VSAT still satisfies the RAM require-
ment of 5V+10%.

When the system power supply
fails and the voltage starts to drop,
the voltage from the backup battery
is fed through diode D2 as soon as
Dl, TR1 and TR2 turn off and the
relationship VS < VBZ + VBEl is
established. The voltage level at
the memory during backup operation
will be VDD=VB-VFZ. However, it is
necessary, before VS < VBZ + VBEl is
fully established, to make the Data
Retention Control Signal, CDRC, a
logical high and therefore place the
memory system in the data retention
mode. Figure 10 shows the data
retention voltage characteristics
of the TC5516 APL .

It should be noted here that a
low-power shottly gate (LSTTL) should
not be used for the Data Retention
Control Gate. In LSTTL circuits,
the output terminals are connected to
the power supply rail through an
internal resistor which when used as
the Data Retention Control Gate, will
sink the battery current and degrade
the life of the backup battery.
Standard TTL gates and CMOS logic
gates which have no such internal
current path should be used.

In Figure 9, the benefits of
two chip-enables found in TC5516 APL
and TC5518 APL have been fully
utilized with CE2 commonly connected
to all memory devices on the board
through a single control line.

The following are additional

advantages users can expect with
TC5516 APL and TC5518 BPL in battery
applications .

TC5516AP/APL

High speed memory systems can be
implemented by using the fast
access from chip-enable feature.
Access time from CEl for the
TC5516AP/APL is 100 nsec max.

TC5518BP/BPL

Both CEl and CE2 can be used to
control the power down mode so
that for large arrays, an X-Y
decode arrangement can be used
to control the active selection.
This results in only one device
being active at any time and
therefore the average power con-
sumption of the system is reduced
dramatically.

Large Capacity Memory Application

This example, Figure 11, shows
the use of the two chip-enables of
the TC5518BPL to make two-dimensional
memory capacity expansion very simple.

The configuration of the battery
backup circuits and its operating
principles are identical to those
discussed earlier in Figure 9. The
TC5518BPL is most suited for this type
of system because the two chip-enable
pins have identical functions and can
be used interchangeable. Futhermore,
the number of decoded circuits
necessary to decode the multichip
memory system is minimized in this
2 dimensional memory array where 2n
decoders can decode n2 memory chips.

In an actual application where
the memory system is used with a CPU
whose data must be transferred and
stored in memory when the power fails,
it is necessary to detect the power
failure as early as possible so that
the DC voltage in the system is still
sufficiently high to ensure accuracy
during the memory write operations
while VDD is falling. By utilizing
power down voltage detection circuit
which detects off the AC power supply
line as is shown in Figure 12. The
earliest warning of a power failure
can be generated. This signal can
be detected by the CPU resulting in
the storage of critical data which
VDD is still at its maximum, and then

4

17/2

to generate the Data Retention Control
signal which puts the memory system
into the standby mode with battery
backup.

When the memory system is used
only as a ROM replacement and no
writing to the memory occurs, the power
fail signal, when detected, would
simply place the system into the
backup mode .

SUMMARY

TH OUT

Figure 2. Cross Section of CMOS
Inverter

The capacity of high- technology
semiconductor memories, supported by
improvements in finer, more advances,
process technologies, has quadrupled
every 2 years essentially without
increasing the chip size of the devices.
This indicated that the standby current
of CMOS RAM will remain essentially
constant even though the storage
capacity of the devices will continue
to increase. This fact and the basic
benefit of low power dissipation of
CMOS make the CMOS RAM increasingly
more attractive for future applications,
where low power and non-volatility are
key issues.

The low power performance of CMOS
RAMs in the standby mode has improved
to such a stage where even a small
button type Lithium battery can be
used to maintain data as long as 10
years .

New devices from Toshiba, TC556 4
and TC5564, 64K Static CMOS RAMs, which
are soon to be released should con-
tribute heavily toward this direction
in the future.

OUT

Figure 3. Circuit Diagram of CMOS
Inverter

I DDS2 vs. VdD

3.4

3.2

<:
c

e

H

3.0

2.8

Ta = 25'C
CE 2 =V DD-0.5V

'

3 4 5 6

V DD (V)

Figure 4. Standby Current vs. VDD

Data

Data

Figure 1. Circuit Diagram of CMOS
Memory Cell

17/2

5

IDDS2 vs. Ta

500
300

100

50
30

10

<:
c

cn
o
o

M

5
3

1

0.5

0.3

0.1

v D D = 5.5V

CE2 = 5.0V

A —

/

lPv ia ry

II il o ||

a <\j ^

> O O Ft

-■40

40
Ta(C)

80

Figure 5. Standby Current vs. Ta

m in
jr <_»

!3 -

Figure 6. TC5516 APL Standby Current
(IDDS2) Distribution Data

/

/

/

7

MM , , ,

III 1 1 1 1 1

null i i

in 1 1 l i i

(sunoH) J

8IUU UOlJUSlSa BJtQ

Figure 7. Data Retention Time vs.
System Size

System
Power (V s )

Figure 8.

01

vdd

CMOS
RAH

D2

«F2

Li thf urn Battery

777

Basic Circuit for Battery
Backup Application

6

17/2

System
Power ,

K r 1 r-

1 [p\

r

TC40H000

Littiiun
Battery

Figure 12. Battery Backup Circuit with
Power Down Signal

Figure 9. Simple Application Circuit
for Battery Backup

Data Retention Period

Figure 10.

System
Power

(1 merory cycle time
tirre rain.)

Address A, 3 , A, 4

1/2 TC40H139P

Address
All .*12

1/2 TC40H139P

Figure 11. Large Capacity Memory
Application

Package

18 pin
Plastic
and
Cerdip

24 pin
Plastic
and

Ce rd i p
and

Plastic
Flat

Function
Mode

Address latch
and synchronous

asynchronous

asynchronous

Operating

Current

(mA/Wz)

„

Access
Time{ns)

200/300

200/300

200/250

O

O

eg

Organization

TC5504APL | 4,096 x 1

a.

in
O

CO
X

to

Device
Name

TC5516APL

a.

TC55178PL
TC55138PL

v ca

2 2

17/2

7

APPLICATIONS OF LITHIUM IODINE BATTERIES TO CMOS MEMORY AND MICROPROCESSORS

John L. Stegman, Jr.
Electronic Design Engineer
Catalyst Research Corporation

1421 Clarkview Road
Baltimore, Maryland 21209

I NTRODUCTION

Applications of batteries for con-
tinuous long-term memory backup are not
as easy or as simple as one might first
assume. It requires consideration of
not only the battery but also the type
of switch to be used to turn the bat-
tery on and off, and the interface be-
tween the memory (or microprocessor)
and those components not under battery
power .

We have chosen a long-life, solid-
state, lithium-iodine battery which has
been used in the cardiac pacemaker in-
dustry since 1972. Today, this battery
powers more than 90% of the world's
pacemakers, and its reliability is well
documented .

We have also designed and tested
the switching, control and interface
electronics needed for reliable data
retention under battery operation, and
during the transistions to and from the
normal (DC) power supply.

THE BATTERY

The battery is a solid-state, high
energy density, long-life primary cell.
It uses lithium metal as the anode, a
solid pellet of iodine imbedded in an
organic polymer as the cathode, and
solid lithium iodide as the electrolyte
and separator. It cannot suffer from
separator rupture or loss of electro-
lyte, which are common failure mecha-
nisms in most batteries, because the
separator /electrolyte is solid and self-
healing.

The open circuit cell voltage is
2.793 volts at 25 C and at end-of-life
is only 3 millivolts less. The temper-
ature coefficient of open circuit volt-
age is about +1 millivolt per degree
Celsius. Most CMOS memories will re-
tain data at 2.0 volts, leaving almost
0.8 volts for loss in the battery due
to polarization, temperature change and
voltage loss in external circuits.

The cell current available is
limited by the internal impedance of
the battery and the minimum operating
voltage required by the device under
battery power.

The cell impedance varies with
capacity extracted and the cell temper-
ature. For a given capacity extracted,
cell impedance is essentially indepen-
dent of the drain rate at which the
cell had been used. Practically, this
means the impedance is constant and
predictable .

The available cell current, at a
fixed voltage, increases with increas-
ing temperature and decreases with de-
creasing temperature. The apparent
cell impedance also varies with temper-
ature, decreasing with increasing
temperature .

Self-discharge is the electrochem-
ical process in which battery energy
is lost internally without doing any
external work. The extent to which
this process occurs in the lithium-io-
dine system is about 10% of capacity
remaining over 10 years at 25 C. This
is what makes the battery desirable
for long-term use; most batteries will
completely self-discharge in much less
time .

Lithium-iodine cells can accept
reverse currents both continuously and
momentarily, provided the magnitude of
the reverse current is kept below the
limit that causes damage to the cell.
A small reverse current of about 1
microampere will not adversely affect
the battery. Preliminary tests at
Catalyst Research indicate that these
small reverse currents may prolong the
life of the battery. It should not be
assumed that the battery is being re-
charged by these currents; the capacity
of the battery is not significantly
increased. Momentary reverse currents
should be kept below 1 milliampere for
periods of 1 second since excessive re-
verse current can cause permanent cell

1

17/3

damage .

THE POWER SUPPLY SWITCHES

To readily interface the battery to
a CMOS device requires the use of a
switch to select battery or normal sys-
tem power. The switch can be modelled
as two single-pole single-throw switch-
es with a common terminal. This con-
figuration is necessary for make-before-
break operation. The battery switch
must operate under battery power, and
need only pass small (<100 microamperes)
currents. The other switch, for the
normal supply, must pass sufficient cur-
rent to operate the CMOS device in worst
case situations without excessive loss
of voltage. In addition, the normal
supply switch must not present a load to
the battery when the normal supply is
off.

battery's internal impedance; in the
off-state the channel resistance is
almost infinite . Because the bias cur-
rent for a FET is so small it uses no
appreciable amount of the battery en-
ergy. There is some difficulty in
using a FET since the gate-source bias
voltage is at most the battery's open-
circuit voltage, but by proper select-
ion a suitable device may be found.

Si/Ga

s

FET

d

+

9 1

e»

The Battery Switch

The battery switch can be a sili-
con diode, a germanium diode, or a field
effect transistor (Fig. 1A,B). A relay
may be used although its reliability
over the life of the product must be
taken into consideration.

The silicon diode is inexpensive
and easy to- use. The primary disadvant-
age of the silicon diode is the forward
voltage drop. Although this drop may*
be as high as 0.7 volts at moderate
currents, tests we have conducted show
that a type 1N4148 diode at a forward
current of 10 microamperes has a volt-
age drop of only 0.35 volts at 25 C.

The germanium diode is also an
attractive switching device for the bat-
tery. Its forward voltage drop is much
lower than that for a silicon diode.
For a type 1N34A germanium siqnal diode
at 10 microamperes forward current, the
voltaqe drop is less than 100 microvolts
at 25 C. However, because of the maqni-
tude of the reverse bias current in qer-
manium diodes, the battery may see
siqnificant reverse current under normal
supply ooeration. Small reverse cur-
rents present no difficulty provided
these currents do not exceed the capa-
bility of the battery.

The field-effect transistor has
shown to be an almost ideal device for
this application. In its on-state, the
channel resistance is much less than the

Fig. 1(A) Fig. 1(B)

The Normal Supply Switch

The normal supply switch must pass
sufficient current to operate the CMOS
device without voltage loss great
enough to cause logic level conflict.
For CMOS devices this loss cannot ex-
ceed 0.3 volts, typically. The normal
supply switch must also not load the
battery when the CMOS device is under
battery power. Two possible choices
for the normal supply switch are a
diode and a bipolar transistor (Fig.
2A,B) .

The silicon diode is inexpensive
and easy to use but its forward volt-
age drop in this application will be
0.7 volts because of the current re-
quired under normal operation. To
prevent logic-level incompatibility, a
separate supply voltage may be needed
and a bias resistor to hold the diode
drop at 0.7 volts at all times. The
reverse bias current thru a low-leak-
age silicon diode is small enough that
it will not load the battery during
battery operation. A germanium diode
would load the battery, and should not
be used here.

The PNP-bipolar transistor is a
very attractive switch in this appli-
cation. A low emitter-collector satu-
ration voltage and low reverse current
thru the collector-base junction meets
the requirements of the normal supply

17/3

2

switch. Suitable transistors are avail-
able for a wide range of normal supply
currents. Caution must be exercised to
avoid forward biasing the collector-
base junction, and thereby loading the
battery, during battery operation.

Si

to-

Fig. 2(A)

Fig. 2(B)

Transitions in Power Supplies

The steady-state operation of the
switch, either in battery mode or in
normal supply mode, is easily accom-
plished with combinations of solid-
state devices as described. It is
during the transition states, from
battery to normal and normal to battery,
that difficulties arise.

It is obvious that even short dur-
ation losses of power to a CMOS device
will cause errors in the data contained
in the device. Practically, this means
that during power-up, the normal switch
must close before the battery switch
opens; and during power-down the bat-
tery switch must close before the nor-
mal switch opens. If a common signal
is used to activate the changeover in
power supply, the response of the
switches must be guaranteed to meet
this condition.

Lithium-iodine cells can accept
reverse currents both continuously and
momentarily, provided the magnitude of
the reverse current is kept below the
limit that causes damage to the cell.
A continuous small reverse current of
about 1 microampere will not adversely
affect the battery. Preliminary tests
at Catalyst Research indicate that
these small reverse currents may pro-
long the life of the battery. It
should not be assumed that the battery
is beinq recharged by these currents;
the capacity of the battery is not sig-
nificantly increased. Momentary re-
verse currents should be kept below 1
milliampere for periods of 1 second or
permanent cell damage can occur.

THE CONTROL MECHANISM

The control mechanism must deter-
mine power supply status and generate
signals to switch between normal supply
and battery supply. To facilitate the
use of simple interfaces, the change
from normal supply to battery supply
should occur at about the nominal cell
voltage; switching at this voltage im-
plies logic-level compatibility during
changes in power supply. A signal to
indicate the mode of control of the
memory (or microprocessor) must be gen-
erated to determine if access will be
permitted by the system hardware; under
battery power the memory (or micropro-
cessor) must not be accessed because of
excessive demand on the battery.

Power-Up/Down Sequences

In a typical application, a normal
operation sequence consists of : power-
up, operation on normal supply, and
power-down. It is essential that the
data retention system be capable of
handling the shortest length sequence;
the length of this sequence is depend-
ent on particular normal power supply.
Abnormal power supply situations must
also be kept in mind to assure data re-
tention during these times.

The power-up sequence can be ap-
proximated by a smooth rise from zero
volts to the normal supply voltage.
When the normal supply rises to the
nominal cell voltage, the change to
the normal power supply should occur.
When the normal supply rises to the min-
imum DC operating voltage for the system
hardware, access to the control of the

3

17/3

memory (or microprocessor) may be per-
mitted.

A power-down sequence can be ap-
proximated by a smooth drop from the
normal supply voltage to zero volts.
When the supply drops below the minimum
DC operating voltage for the system
hardware, access should be denied to
the memory (or microprocessor) . When
the normal supply voltage falls to the
battery nominal cell voltage, the change
should be made to the battery from the
normal supply.

This technique of two level volt-
age detection will protect the memory
(or microprocessor) hardware from most
types of power supply failure and nor-
mal on/off use. The use of hysteresis
at the critical voltage detectors will
further enhance the hardware protect-
ion, designing fast power supply switch-
es, characteristics of which were de-
scribed before, will enable the hardware
to handle abnormal power supply fluctu-
ations with reliable data retention.

False-Write and Read Protection

A matter of great concern in bat-
tery backup is the prevention of a
"f alse-write" . A false-write is any
attempt: (1) To write into memory data
of questionable validity, or (2) The
premature termination of a normal write
cycle. The first case of false-write
can be effectively prevented by in-
hibiting operation of the external hard-
ware whenever the normal power supply
voltage is below minimum DC operating
voltage, and granting access on a power-
up sequence only after establishing
normal system operation. The second
type of false-write, premature termi-
nation, occurs during a power fail se-
quence. Premature termination is pre-
vented by sensing the impending loss of
power before minimum DC operating volt-
age is lost, and halting system oper-
ation before insufficient voltage is
available to complete the write cycle.

The power required for a normal
read of memory usually exceeds the capa-
bility of the battery. Pead access
must be prevented during battery oper-
ation or loss of data may occur due to
a momentary drop of battery voltage be-
low the minimum data retention voltage.

DIGITAL INTERFACING

The digital interfaces to battery-
backed memory (or microprocessors) can
be divided into three classes: Control
signals, address signals, and data sig-
nals. Control signals present the most
difficult problem in interfacing be-
cause the interface must take over con-
trol of the device during battery op-
eration. Because the battery cannot
supply enough energy for normal oper-
ation, the change in control must pre-
cede the change to the battery supply,
and follow the change to the normal
supply .

Control Signals

Control signals asserted low (neg-
ative) must be pulled high during bat-
tery operation. We have developed and
tested a simple and reliable interface
circuit that performs this function
(Fig. 3A) . A type 74LS03 open-collect-
or, dual-input nand gate has the follow-
ing property: If one of its inputs is
held at ground potential, then its out-
put transistor will remain completely
off (not conducting) for any voltage
at the Vcc terminal greater than zero
volts and less than +5 volts. This
property has been theoretically and
empirically verified for several major
manufacturers (Tx Inst., Natl Semi,
others) . This property has not been
checked for more sophisticated LS-TTL
devices; it is the simplicity of the
74LS03 internal construction that gives
it this property. Therefore, a pull-
up resistor connected to the CMOS
memory (or microprocessor) Vdd terminal
will hold the CMOS input connected to
the 74LS03 output at cmos Vdd during
the rise and fall of the normal power
supply voltage. The signal that is
held at ground potential at the nand
gate input we call "Access Grant", for
obvious reasons. The remaining nand
gate input is the system control input
with inverted polarity; we call this
"Access Request." It is much easier to
generate an access grant signal held at
ground than to attempt to disconnect
the CMOS input from the system control
output with a switch. There is no
need to be concerned with the access
request line; it can fall to ground in
any manner provided its voltage does
not exceed the voltage at the Vcc term-

17/3

4

inal of the 74LS03.

Address Signals

Control signals asserted high
(positive) present a somewhat less
difficult problem, since the CMOS in-
put must be held at ground. Using any
open-output (collector or drain) device
such as gates, comparators, or more
sophisticated function devices, and
connecting the pull-up resistor to a
switch that turns the pull-up into a
pull-down (to ground) will hold the
CMOS input at ground. (Fig. 3B) is an
example of this technique using a 74LS
05 open-collector inverter. Designing
the appropriate switch for the pull-up/
down resistor is an easier task than
attempting to disconnect the CMOS in-
put from the system control input.
(Note: Analog switches usually do not
function correctly during their own
power-up/down, and are unreliable in-
terface devices.)

CMOS

Fig. 3(A)

Fig. 3(B)

The problem of address signal
interfacing is essentially one of
avoiding voltages at the CMOS address
inputs that fall in the CMOS input
"Linear Region." This linear region is
between the maximum input-low voltage
and minimum input-high voltage. Input
voltages in this interval cause both
the N-channel and P-channel transistors
to conduct heavily. This current is
pulled through the power supply inputs;
if the CMOS device is under battery
power then the battery will attempt
to source this current. Not only does
this condition waste battery power, if
it occurs simultaneously on many inputs
it will saturate the battery and cause
the cell voltage to drop below the min-
imum data retention voltage for the
memory (or microprocessor) , and the loss
of data. This problem is solved by
holding the address inputs at ground
potential during battery operation. A
careful choice of address bus-driver/
interface will facilitate this. In
any case, an interface like those for
control signals described above will
also work for address signals.

Data Signals

Data signals are the easiest to
interface. If the data in/out to the
memory (or microprocessor) is of the
three-state bidirectional type then
no interface is required if the data
in/out terminals are in the high imped-
ance state. The particular memory (or
microprocessor) should be checked for
excessive leakage current in the high-
impedance state. Memory (or micropro-
cessor) devices with separate data in-
puts and outputs can be treated like
address inputs for data inputs; and
data outputs should be disconnected
from their loads, either internally or
externally .

SUMMARY

The use of lithium-iodine batter-
ies for long-term, uninterruptable
power supplies for CMOS memory (or
microprocessors) is quite feasible.
Their applications in products can be
separated into problems independent of
each other, and the solutions of the
problems can be integrated to form a
highly reliable system without the use
of expensive or dif f icult-to-f ind de-
vices. We have developed a series of

5

17/3

successful products based on the lith-
ium-iodine battery; in each case the
probability of lost data was reduced
to the probability of failure in the
actual memory (or microprocessor) in
the product.

Note :

The author would like to thank
Alan Schneider for his guidance in
the preparation of the manuscript.

17/3

6