Computer history: Bubble memories

 

Bubble memory as a successor to core memory

The invention

Bubble memory was invented at Bell Labs in the USA in the early 70s. By the mid-70s, many research organisations were interested in using this type of memory. Experimental devices could store only a few kilobits, but that was already a threat to the magnetic core memory technology used in computers of that time. Notable research was undertaken by Bell Labs, IBM, Crouzet and Plessey in the USA as well as in Japan.

The first bubble memory devices were not sold commercially. The first bubble memory module, the Western Electric type 29A, was used by AT&T in the US telephony system to store automated voice announcements. The 29A package comprised four 68,121-bit chips, making a total of 272,484 bits, which could store twelve seconds of 24 kbps encoded speech. These chips comprised simple shift registers without a major-minor loop structure. Four of these modules were placed on a single circuit board.

Example of a core memory manufactured by TNO
Example of a core memory manufactured by TNO (the early 70s)

The operation of a bubble memory

A magnetic bubble memory device of the cassette type includes at least one magnetic bubble memory chip packaged in a box with upper and lower lids. Encompassing the magnetic bubble memory device are two permanent magnetic plates fixed to the inner surfaces of the upper and lower lids to shield the external influences of strong magnetic fields. Also, preventive measures against the electrostatic destruction of the bubble information and provisions for heat dissipation are taken.
The memory device itself consists of a very small film of an orthoferrite which has a crystal structure and is weakly ferromagnetic.

Photo
Photo of a film of orthoferrite that has a crystalline structure

Note that garnet has a similar magnetic structure. While researching the degree of magnetisation of permalloy magnetically thin films, it was discovered that it was possible to move magnetic information in perpendicular directions in these films. In this way it turned out to be possible to use small magnetised areas to indicate data bits: if a magnetic bubble (bubble) is present at a certain point and within a certain time within a bubble domain, this is considered a digital “one”. The absence of a bubble represents a digital “zero”.
An alternating electric field causes these bubbles to move. Magnetic bubble memories are non-volatile, which means that the magnetic bubbles will not disappear when power is switched off.

In the late 1960s, Bell Laboratories’ employees made several discoveries which led to a working bubble memory:

  1. Two-way control within a single demarcated area in permalloy films.
  2. The uses of orthoferrite.
  3. The discovery of a stable cylindrical region.
  4. The discovery of a method for reading and writing information.
A bubble memory with its control electronics
A bubble memory with its control electronics

 

An exploded view of a bubble chip with the orthogonal magnetic control coils
An exploded view of a bubble chip with the orthogonal control magnetic coils

The bubbles within the film are given directions with rod-shaped magnets. The material is shaped in a way that yields several parallel paths. These magnetic areas (bubbles) in these paths are moved by an external magnetic field. The bubbles can be read or modified at the edges of the film surface.

Read and (re)write actions
Read and (re)write actions

 

Four bubble memory chips with 80 x 8,192 bits each = 655,360 bits (source: Handbook p800 AR241 (see US patent 4459680))
Four bubble memory chips with 80 x 8,192 bits each = 655,360 bits (source: Handbook p800 AR241 (see US patent 4459680))

Use of bubble memories

Starting in 1975, bubble memories were applied at the Physics Laboratory RVO-TNO and replaced earlier core memories. The lab bought some small Fujitsu plastic cassettes with the inscription “FBM43CA bubble memory cassette”. These memory cassettes were packed in transparent plastic boxes. The capacity of this type of cassette was 256 kbit = 32 kbyte.
A warning was placed on each cassette to avoid touching the pins of the connector at the rear as the chips were sensitive to electrostatic charges which could damage the internal electronic circuits.

Cassette holder with the bubble memory chip
Cassette holder with the bubble memory chip

 

Contents of a cassette
Contents of a cassette

 

Bubble geheugenchip met besturingselektronica
Bubble memory chip with control electronics

Fujitsu manufactured cassettes with different memory sizes, such as the FBM-C128GA. That cassette contained an FBM54DB unit with a 1 Mbit memory. The encoding of the product name contained the kbyte size of the memory: 128 denotes 128 kB or 1 Mbits.

Fujitsu 256 kbit module: FBM42DA without lids
Fujitsu 256 kbit module: FBM42DA without lids (source: Andrew Wylie http://www.wylie.org.uk/technology/computer/bubblmem/bubblmem.htm)

Experimental use of bubble memories at the Physics Laboratory

In the late 1970s and early 1980s, there was a need to store more information for later use. For instance, to be able to compare measurement results and also to investigate the content of fast impulses as the duration of an impulse could provide information about its origin. To properly assess this information, impulses had to be sampled. Such sampling provided a lot of information that had to be stored temporarily.
Memory in the form of SRAM (Static Random Access Memory) was very expensive at that time. Bubble memory was a much cheaper alternative.
Bubble memory could also be used in combination with microprocessors allowing the development of portable measuring systems. These could then be used during field tests. The commonly used microprocessor at that time was an 8-bit Intel 8085. The bubble memory was coupled to the microprocessor via a 7220-1 controller. The bubble memory turned out to be usable but was quite slow in combination with the microprocessor.

The major drawback of bubble memory for further application was its sensitivity to electrostatic discharges. Out of three bubble memories purchased, two quickly died. This was insurmountable for portable systems with uncertain voltage supplies. In the end, we had to wait for the semiconductor memories to become cheaper.