Bubbles and CCD memories-Solid state mass storage
by J. EGIL JULIUSSEN
Texas Instruments Incorporated
Dallas, Texas
INTRODUCTION
very important for small systems. As chips they can also be
packaged directly with the CPU on PC boards.
The characteristics of today's major CCD and bubble
memory components are shown in Table II. The 92K bit
MBM chip has an average access time of 4 milliseconds and
has a transfer rate of 50K bits per second. The dual-in-line
package is larger than most ICs. The MBM package includes
two permanent magnets giving non-volatility and two coil
windings. The two coils produce a rotating magnetic field
which moves the bubbles around the shift registers.
MBM support chips are already available. A minimum
bubble memory system which is interfaced to a microprocessor can be made with 7 chips. The coil drivers produce
the rotating magnetic fields. The sense amplifier detects the
bubble information and is very similar to a core memory
sense amplifier. The function timer and driver produce the
necessary timing signals to control the operation of bubble
memory chips. The controller chip is a LSI device which
controls the memory system operations, buffers data and
handles interfacing with a microprocessor.
The 64K bit CCD chips have a superior access time over
bubble memories. The TI9 and Fairchild devices have a 410
microsecond average access time and can transfer data at 1
to 5 Mbits per second rate. The Intel device has more shift
registers with shorter length that gives better access time,
but at a lower transfer rate. During operation the CCD chips
use less power than bubble memory chips. The power advantage is reversed to MBMs in a standby mode.
None of the CCD products have any support chips. It
should be noted that CCDs do not need an external sense
amplifier. This function and the refresh function have been
integrated onto the chip (one per shift register).
An assessment of MBM and CCD merits are shown in
Table III. The main advantages of bubble memories are the
nonvolatility and the large number of bits per chip. In effect
MBM is the only solid state technology which is nonvolatile.
The major advantages of CCDs are the low access time and
the high transfer rate.
Bubble memories have an advantage over CCD from
fewer manufacturing steps and fewer masking steps. This is
also reflected in a potential higher MBM packaging density.
However, the experience from MOS manufacturing is very
applicable to CCD production and probably outweighs the
In the last year significant technical advances have taken
place in the development of magnetic bubble memories
(MBM) and charge coupled devices (CCD). The first bubble
memory chip has been introduced with bigger and better
chips under development. Although 16K bit CCD chips have
been available for two years, it is the recently announced
64K bit chips which will make an impact on computer system designs. The first commercial products based on MBM
and CCD chips are now starting to appear.
With this progress it is worthwhile to look at the status of
CCD and bubble memories, their applications and future
potential.
MBM AND CCD STATUS
Both CCD and bubble memories are integrated circuit
analogs to rotating electromechanical memories such as
disks and tapes. CCDs store information in rows of capacitors. Bubbles and electromechanical memories store information as rows of magnets. The access methods are different. Electromechanical memories move the media while the
magnetized regions remain stationary. In CCD and bubble
memories the opposite takes place: the media remains stationary while the information, charge or magnets, move
within the media. This avoids all mechanical motion and its
associated drawbacks. The result is two solid state mass
storage technologies-or electronic disks. For further information on the technical aspects of MBM and CCDs see
References 1-10.
The features of MBM and CCDs are shown in Table I.
Both technologies are serial access memories and are organized as shift registers. Bubble memories are nonvolatile
as the information is retained without power. The electrical
analogy of the magnetic bubble, the CCD, does not keep its
charge or information as the power is turned off. CCDs are
volatile and must be periodically refreshed to retain the
information.
Both technologies have the modularity advantages arising
from their chip packaging. The resulting low entry price is
1067
From the collection of the Computer History Museum (www.computerhistory.org)
1068
National Computer Conference, 1978
TABLE I.-MBM and CCD Features
r1BM
•
•
•
•
•
•
•
•
•
CCD
SERIAL ACCESS MEMORY
SHIFT REGISTER ORGANIZATION
BLOCK ADDRESSABLE
NONVOLATILE STORAGE
READ-MODIFY-WRITE
'r10DULAR STORAGE CAPACITIES
LOH ENTRY PRICE
BIT PRICE INDEPENDENT
OF STORAGE SIZE
FEW MANUFACTURING STEPS
MANUFACTURE SIMILAR TO IC PRODUCTION
STOP/START OPERATION
•
LOWERS EFFECTIVE ACCESS TIME
•
LOWERS POWER DISSIPATION
• VARIABLE TRANSFER RATE
•
MINIMAL DATA BUFFERING
•
•
MBM advantage today. As bubble memory producers gain
experience, it will be interesting to see if the potential MBM
manufacturing advantages can be realized.
In packaging CCDs have an advantage due to the added
complexity of the coils and magnets for a MBM chip.
In interfacing complexity MBM currently has an advantage from the availability of support chips. The low transfer
rate and these support chips are very advantageous for microprocessor based systems. The high performance requirements of mainframe and supermini computers give CCDs a
distinct interfacing advantage for these applications.
The bottom line is price. In the next few years, CCD will
probably have an advantage due to MOS manufacturing
experience and CCD availability from three or more manufacturers. With increased bubble memory manufacturing experience, MBM should close this gap and have an excellent
chance of gaining an advantage over CCDs.
Table IV shows the announced MBM and CCD products
the author is aware of. There are undoubtedly more products
under development.
The three first MBM products use TI's 92K bit bubble
memory chip. The TI 7631765 portable bubble memory ter-
•
•
•
•
•
•
••
•
SERIAL ACCESS MEMORY
SHIFT REGISTER ORGANIZATION
BLOCK ADDRESSABLE
VOLATILE STORAGE
READ-MODIFY-WRITE
MODULAR STORAGE CAPACITIES
LOH ENTRY PRICE
BIT PRICE INDEPENDENT
OF STORAGE SIZE
TTL COMPATIBLE I/O
MOS COMPATIBLE MANUFACTURING
STORAGE REFRESH REQUIRED
•
•
minals are part of TI's Silent 70Q@* series ofterminals. Both
terminals have a minimum of 20K bytes of bubble memory
with optional expansion to 80K bytes. 11
Ql corporation has a microcomputer system with 80K
bytes of bubble memory and also has a plasma display. Data
Systems MBM system is a floppy disk replacement which
is PDP-8 and PDP-II plug compatible. Storage capacities of
86-519 Kbytes are available.
AT&T's 13A announcement system uses 68K bit bubble
memory chips developed by Bell Laboratories and manufactured by Western Electric. The system stores pre-recorded messages which are repeatedly played back as telephone system messages. The system is being field tested by
AT&T. 12
The three CCD systems use Intel's 16K bit CCD chips.
Intel's CCD board is an OEM product that can be used to
develop end user systems. Technical Analysis Corporation
has developed a fixed head disk replacement for their Nova
based small business computer. 13 Alpha Data also has a
* ®Registered trademark.
From the collection of the Computer History Museum (www.computerhistory.org)
Bubbles and CCD Memories
1069
TABLE II.-MBM and CCD Chip Characteristics
TEXAS INSTRUMENTS
Tr-1S 306 Lt
TIB 0103
CCD CHIP
MBM CHIP
64 KEIT
410 SEC
5 ~1B/SEC
64 KBIT
410 SEC
5 MB/SEC
64 KBIT
130 SEC
2. 5 r~B/SEC
0.7 H
NONE
o TO 70° C
14-PIN DIP
0.26 H
25 MI1
o TO 70° C
16-PIN DIP
o.3 L! H
66 Mv!
o TO 55° C
16-PIN DIP
0.33 W
41 ~1H
-10 TO 85° C
18-PIN DIP
PRODUCTION
PRODUCTIOn
PRODUCT ANNOUNCED
NONE
NONE
NONE
(1"
PRODUCTION STATUS
SUPPORT CHIPS
AVAI LABLE
INTEL
2464
CCD CHIP
92 KBIT
4 MSEC
50 KB/SEC
STORAGE CAPACITY
AVG ACCESS TIME
MAX TRANSFER RATE
POVIER
STANDBY pmJER
OPERATING TE~1P
PACKAGING
FAIRCHILD
F 464
CCD CHIP
xLI")
PRODUCTI ON
CONTROLLER CHIP
COIL DRIVER
SEnSE AMPLIFIER
FUNCTION DRIVER
FWJCT ION TI MER
fixed head disk replacement which is compatible with their
other fixed head disk products.
There are many products in the development stage which
use the newer 64K bit CCDs, but none are announced yet.
To summarize the status of CCD and bubble memories a
list of the companies with known R&D efforts are shown in
Table V. As could be expected the major semiconductor
manufacturers appear on the CCD list. Until recently only
TI of the semiconductors producers had bubble memory
effort. In the last year both Intel and National have started
bubble memory efforts.
MEMORY SYSTEM TRENDS
It has been the goal for computer systems to have a single
memory technology that has both the lowest cost and the
highest performance. This goal has not been realized and
the memory hierarchy from fast semiconductor RAMs to
low cost disks and tapes must be used. The well-known
memory hierarchy gap in access time and cost has provided
a market pull for new memory technologies. In the early
seventies this was the major driving force to develop CCD
and bubble memories.
Additional market pull forces are now equally important.
The proliferation of microprocessor applications are demanding small, non-volatile mass memory systems. Severe
environments and the high maintenance costs need improved reliability. The popularity of virtual memory operating systems need some real and high performance mass
memory.
The trend towards using fixed heads in combination with
moving head disks is one solution that gives higher performance mass storage. However, there is still a large percentage
of computer systems that are disk-bound.
MBM and CCDs are not removable storage technologies
and this is a drawback vis-a-vis disks and tapes. There is a
trend towards using non-removable disks. This lowers the
disk drive cost and also improves reliability. The trend is
also advantageous for both MBM and CCDs as both technologies can replace the function of non-removable disks.
At the same time, the use of distributed processing systems
From the collection of the Computer History Museum (www.computerhistory.org)
1070
National Computer Conference, 1978
TABLE III.-MBM Versus CCD
MBM
MAJOR ADVANTAGES
MANUFACTURING
PACKING DENSITY
CCD
NONVOLATILE
BITS/CHIP
ACCESS TIME
TRANSFER RATE
FEW STEPS
FEW MASK LEVELS
EXPERIENCE FROM MOS
MANUFACTURING
ADVANTAGE DUE TO MASK LEVELS
PACKAGE
COILS &MAGNETS ADD COMPLEXITY
INTERFACING
SIMPLE DIP
ADVANTAGE FOR MICROPROCESSORS ADVANTAGE FOR MAINFRAMES
PRICE
ADVANTAGE BASED ON
PACKING DENSITY
also lowers the need for removability. The data communication links in many cases are a substitute for removability.
To indicate how MBM and CCDs compare with other
storage technologies, the system prices for end users are
shown as a function of storage capacities in Figure 1. For
ADVANTAGE FROM MOS
MANUFACTURING EXPERIENCE
small storage capacities MBM and CCDs have cost and
physical size advantages. As soon as a few megabits are
needed, the disks have a price advantage, but MBM and
CCDs retain their performance edge. It is expected that
MBM and CCDs, in the future, will be price competitive at
higher and higher storage capacities.
PURCHASE
PRICE ($)
100,000~------~-
MBM AND CCD APPLICATIONS
FIXED HEAD
MOVIilG HEAD
DISKS
10,000
LOOO
100 L-------~------~lO------~10~0------~l,OOO
STORAGE CAPACITY (MEGABITS)
Figure 1-1978 memory system prices including interfacing
To gain insight in the use of CCD and bubble memories'
Table VI shows the various storage peripheral product categories for mainframe, mini and microcomputers. MBM and
CCDs are not fast enough to be used as main memory, and
are too costly to be used for removable mass storage. As
fast auxiliary memory (F AM) both MBM and CCDs have
excellent credentials. This is the product category which
fills the memory hierarchy gap. In the non-removable mass
storage category, MBM and CCDsare very viable, but will
see strong competition from disks.
Due to their packaging flexibility, MBM and CCDs add·
an extra dimension to mass storage applications. They cim .
be packaged directly with the processor and become an
integral mass storage device. No extra box is needed as is
From the collection of the Computer History Museum (www.computerhistory.org)
Bubbles and CCD Memories
1071
TABLE IV.-Announced MBM and CCD Products
~1B~1
•
TEXAS INSTRUMENTS
TI 763/765 PORTABLE BUBBLE
f1Er10RY TER~1 INAL
20 TO 80 KBYTES MBM
(CD
•
•
•
•
QI CORPORATION
• QI/LITE MICROCOMPUTER
• 80 KBYTES MBr·1
•
• DATA DSDSYSTEMS640 FLOPPY DISK REPLACEf1ENT •
• 86 TO 519 KBYTES MBM
• PDP-II &PDP-8 PLUG COMPATIBLE
•
• AT&T13A ANNOUNCEMENT SYSTEM FOR
• RECORDED SPEECH
KBYTES OF MBM FOR 24 SECONDS
• OF68 SPEECH
DATA
usually the case with disks and tapes. Instead the MBM or
CCD can be put on the same or different PC board as the
processor. The TI 7631765 portable bubble memory terminal
is an excellent example of this advantage.
The most likely CCD and bubble memory applications are
shown in Table VII. The main bubble memory applications
are in small systems. The bubble memory characteristics fit
microperipherals and integral mass storage applications very
well. In the miniperipheral category, MBM add-in or addon electronic disks are viable opportunities. Add-in refers
to MBM packaged with the CPU, while add-on means a
separate box. In essence MBM is a main candidate for
distributed mass storage to go along with distributed processing.
The main CCD applications are in closing the memory
hierarchy gap. These opportunities are mostly in the large
computer market place. In other words, CCD will be the
technology which makes virtual memory systems become
real.
With MBM and CCDs having similar features, will they
, compete head-on for the same applications? To answer this
INTEL
•
IN-65 CCD BOARDS
• 128K TO 1 MBYTES CCD
TECHNICAL ANALYSIS CORP.
•
FIXED HEAD DISK REPLACEMENT
• 256K TO 2 MBYTES CCD
• DG NOVA PLUG CDr1PATIBLE
ALPHA DATA
•
CCDISC - FIXED HEAD DISK REPLACEMENT
• 128K TO 1 MBYTES CCD
question, Table VIII shows the probable MBM and CCD
usage.
The high performance characteristics will favor CCDs in
mainframe computer applications. CCDs for fixed head disk
replacement have already seen usage. Virtual memory and
moving head disk cache memory based on CCDs are probably in development.
The nonvolatility and support circuit availability of bubble
memories will favor MBMs as integral mass storage for
microprocessor based systems. The announced MBM products fall in this category, and more are in the development
stage.
The effect is that MBM and CCDs will have little direct
application competition.
In summary, the usage of MBM and CCDs will evolve as
follows: In 1977 many system designers gained familiarity
with the two technologies. In 1978 CCD and bubble memories are designed into many products. In 1979 numerous
products with MBM and CCDs will be in production. In the
1980's both ,MBM and CCDs will have become important
storage technologies.
From the collection of the Computer History Museum (www.computerhistory.org)
TABLE V.-Companies with MBM or CCD Effort
CCD
MBM
•
US Cor1PAN IES
TEXAS INSTRUMENTS
BELL LABORATORIES
•
•
• ROCKWELL INTERNATIONAL
• HEWLETT-PACKARD
• UNIVAC
• INTEL
• NATIONAL
•
• FOREIGN COMPANIES
• FUJITSU
• HITACHI
NIPPON ELECTRIC
• PHILIPS
• PLESSEY
•
IBr~
•
US COMPANIES
TEXAS INSTRUMENTS
BELL LABORATORIES
IBM
FAIRCHILD
INTEL
NATIONAL
MOSTEI<
MOTOROLA
HUGHES
RCA
•
•
••
•
•
•
•
•
•
• FOREIGN COMPANIES
• BELL
• PHILIPS
• PLESSEY
NORn~ERN
•
•
•
•
SIEMENS
TOSHIBA
NIPPON ELECl
MITSUBISHI
TABLE VI.-Storage Peripherals
FAST AUXI LIARY
MEMORY
MAIN MEMORY
PERIPHERALS
CORE·
MOS RAM
BIPOLAR RAM
MINIPERIPHERALS
CORE
MOS RA~1
MICROPERI PHERALS
MOS RAM
ROM
PROM
EPROf1
NON REMOVABLE
MASS STORAGE
MULTI PLATTER
FIXED HEAD DISK
DISK
EBAM
EBAM
CCD
CCD
MBr~
CCD
MULTIPLATTER
DISK
AUTm1J-\TED TAPE
SYSTEM
Mi\GNETI C
TAPE
SINGLE PLATTER DISK
FLOPPY DISK
MAGNETIC TAPE
TAPE CARTRIDGE
FLOPPY DISK
~1BM
SINGLE PLATFIXED HEtl.D DISK
TER DISK
CCD
MBM
MBM
CCD
MB~l
REMOVABLE MASS STORAGE
OFF-LINE
ON-LINE
MBt·1
CCD
CASSETTE
MINIFLOPPY
r~INIFLOPPY
MINICASSETTE
1072
From the collection of the Computer History Museum (www.computerhistory.org)
Bubbles and CCD Memories
1073
TABLE VII.-MBM and CCD Applications
MB~1
•
•
•
MICROPERIPHERALS
• MI CROCor1PUTER MASS STORAGE
•
FLOPPY DISK REPLACEMENT
INTEGRAL MASS STOPAGE
•
PROGRAMMABLE TERMINALS
•
PORTABLE TER~lItJALS
~IORD PROCESSING TERMINALS
BANKING TERMINALS
POINT-OF-SALE TERMINALS
DATA COLLECTION TERfHNALS
PROGRAMMABLE CALCULATORS
MEASUREMENT &TEST EQUIPMENT
MI NIPERI PHERALS
•
ELECTRONIC DISK
FIXED HEAD DISK REPLACEMENT
• OPERATING SYSTE~1 STORAGE
MAINFRAME PERIPHERALS
•
FAST AUXILIARY MEMORY
•
DISK CACHE
•
•
•
•
•
•
•
•
•
•
•
•
WHAT'S AHEAD IN CCD AND MBMs
CCD and bubble memories will follow the rapid increase
in storage capacities and with the resulting decrease in cost
per bit which is the characteristic of semiconductor technologies. Every two to three years the storage capacity of
semiconductor chips quadruple and similar price per bit
reductions take place. Both MBM and CCD chips will improve in this manner. In the beginning new storage technologies improve even faster and CCD and MBMs are likely
to do so.
An overview of solid state memory trends is shown in
Figure 2.14 It shows that a 256K bit bubble chip can be
expected in a year with a 256K bit CCD chip following a
year later. By 1980 a 1 Mbit bubble memory chip is forecasted with the 1 Mbit CCD chip following a year later.
In terms of performance, CCD chips will still have an
edge in access time and transfer rate over bubble memories.
Both technologies will have higher transfer rates than today.
The access time will be a trade-off parameter. To make
larger chips will require longer shift registers. This would
result in longer access time, unless a faster shift rate is used.
CCD
MAINFRAME PERIPHERALS
•
VI RTUAL f1EMORY
• DISK CACHE
•
FIXED HEAD DISK REPLACEMENT
MINIPERIPHERALS
•
ELECTRONIC DISK
•
OPERATING SYSTE~1 STORAGE
• VI RTUAL r1EMORY
•
FIXED HEAD DISK REPLACEMENT
MICROPERIPHERALS
•
~1I CROCO~1PUTER t'1ASS STORAGE
•
FLOPPY DISK REPLACEMENT
BUFFER STORAGE
•
CRT REFRESH
HIGH SPEED BUFFERING
•
It appears that for both MBM and CCD's the increased shift
rate will just keep up with the longer shift registers. The
result is that the access time of MBM and CCDs will be
about the same as they are now.
But how will MBM and CCDs stack up against disks? The
moving head disk technology has had impressive advances
in the last 15 years, and will probably keep on improving
0
BIPOLAR RAM
®
~lOS RA~1
MAGNET! C BUBBLES
@
ROM
EPROM
@
~
@
@
256K
®
CCD
@
~
§
®
@
@
@
1977
@
@
256K
1978
1979
@
1980
Figure 2-Solid~state memory technology trends
From the collection of the Computer History Museum (www.computerhistory.org)
1981
1074
National Computer Conference, 1978
TABLE VIII.-MBM and CCD Usage
MBr·1
MAINFRAME COMPUTERS
CCD
FHD REPLACEMENT IS A FHD REPLACEMENT IS
POTENTIAL APPLICATION
AN IMMEDIATE
APPLICATION
COMMENT
CCD S ~1AJOR
NEAR-TERM APPLICATION
I
ELECTRONIC DISKS AWAIT VIRTUAL MEMORY AND
MBM PRICE REDUCTION
MHD CACHE HILL
EMERGE
MI NICO~1PUTERS
SMALL ADD- HJ ELECTRON IC SMALL ADD-IN FAMS
DISK WILL EMERGE
WILL SOON BE
AVAILABLE
APPLICPJION FOR
BOTH MBM AND CCD
ELECTRONIC DISKS
COMPETITIVE WITH
CARTRIDGE DISKS MAY
APPEAR BEFORE 1980
MICROSYSTEMS
INTEGRAL MASS STORAGE INTEGRAL ~1ASS STORAGE
FOR MPU-BASED
FOR SOME
SYSTEMS IS AN IMMEDIATE APPLICATIONS
APPLICATION
for another five to ten years. The disk advances have been
realized by packing an increasing number of bits in the same
area. The result is that large multi-platter disks that stored
2M bytes in the early 1960's now store 300M bytes, and cost
about the same. Simultaneously lower priced units have
been developed. The cartridge or single platter disk appeared in the late 1960's and now stores about 100M bits.
The floppy disk was introduced in the early 1970's and just
2 years ago the minifloppy disk was introduced.
Impressive as these improvements are, the CCD and bubble memories will become increasingly competitive with
disks. The disk technology is vulnerable because its bit price
is very dependent on the storage capacity. The bit price of
a minifloppy is about 50 times higher than the bit price of
large disks. By 1980-81 a single MBM or CCD chip will
store more than today's minifloppy disk. The single unit
OEM price of the minifloppy is about $350 today and may
decrease another $100 in the next couple of years. This will
be no match for the learning curve pricing of MBM and
CCDs. An analogy of what is likely to happen is the microprocessor price which has decreased from several hundred
dollars to less than ten in just three years.
In summary, it appears that CCD and bubble memory
S ~'AJOR
NEAR- TER~1
~1BM I
APPLICATION
systems will be price competitive with floppy and cartridge
disks by the early 1980's.
SUMMARY
For a new storage technology to succeed it must meet several requirements. The technology must have advantageous
features and characteristics including competitive price/performance. Long term technology improvement potentials
must exist. Significant R&D investment in the technology
by many companies is crucial. There must, of course, be
available markets. Equally important is an initial market
entry niche.
Both CCD and bubble memories meet these requirements
and should therefore succeed in becoming major storage
technologies in the 1980's.
REFERENCES
1. Bobeck, A. R., et aI., "Magnetic Bubbles-An Emerging New Memory
Technology," Proceedings of the IEEE, August 1975.
From the collection of the Computer History Museum (www.computerhistory.org)
Bubbles and CCD Memories
2. Juliussen, J. E., "Magnetic Bubble Systems Approach Practical Use,"
Computer Design, October 1976.
3. Lee, D. M., "Bubble Memory for Microprocessor Mass Storage," IEEE
Compcon, Spring 1977, February 1977.
4. Juliussen, J. E., "Bubble Memory as Small Mass Storage," Electro 77,
New York, April 1977.
5. Bobeck, A. H., "The Development of Bubble Memory Devices," Electro
77, New York, April 1977.
6. Myers, Ware, "Current Developments in Magnetic Bubble Technology,"
IEEE Computer, August 1977.
7. Carnes, J. E., et aI., "Charge-coupled Devices for Computer Memories,"
Proc. National Computer Conference, 1974.
8. Bhandarkar, D. P., "Digital CCD Memory Trade-offs: A System Viewpoint," Microelectronics, Vol. 7, No.2.
1075
9. Barton, J. B., et aI., "A Cost Effective 64K CCD Memory," Electro 77,
New York, April 1977.
10. Bhandarkar, D. P., "Dynamic MOS Memories: Serial or Random Access," IEEE Compcon Spring 1978, February 1978.
11. Flannigan, J. S., "Bubble Memory Terminal, An Added Dimension to
Data Entry," IEEE Compcon Fall 1977, September 1977.
12. Williams, J. E., "Magnetic Bubble Memory in Telephone Systems,"
Electro 77, New York, April 1977.
13. Jamieson, J. M., "The CCD Memory Gains a Foothold," Mini-Micro
Systems, April 1977.
14. Juliussen, J. E. and W. J. Watson, "Problems of the 80's: Computer
System Organization," The Oregon Report on Computing, Portland,
March 1978.
From the collection of the Computer History Museum (www.computerhistory.org)
From the collection of the Computer History Museum (www.computerhistory.org)