J. L. Archer

Magnetic bubble domain devices

A review is given of recent advances which have been made in the area of magnetic bubble domain devices with emphasis being placed on new device components and the problem of device optimization. The subject of bubble domain propagation is introduced by considering the results obtained on the basis of a simplified model which is applied to individual bars and bar pairs to illustrate the basic principles involved in field access drive circuits. Experimental data is presented which qualitatively confirms various aspects of the model. New device component designs are then introduced, such as the multiple-bar keyhole generator and the chevron stretcher detector, in which bubble stretching occurs orthogonal to bubble motion. Preliminary data on these as well as on various other circuit components are presented and discussed in the light of future device applications.

Device component margin evaluation using generalized field interruption technique

A generalized field interruption technique is evaluated for bubble memory chip margin analysis. Using a short bubble stream and measuring the averaged propagation steps for the first error in a given data pattern, field interruption offers a statistically significant measurement of the failure probability for a small segment of a complete bubble memory device. This technique also allows an individual component margin measurement independent of the rest of the device. A practical example is given using a 100K bit serial loop memory chip which is composed of various components and which also contains a weak processing defect. The component margin measurements show that at low driving fields the chip operation is limited by the passive replicator component, while at high driving fields operation is limited by the permalloy defect.

Experimental Confirmation of Field Access Device Modelling

In order to experimentally verify the predictions of a two‐dimensional field access model, two types of experiments have been performed on simple bar patterns. VSM measurements have been used to measure the average magnetization in bar arrays as a function of applied field and field angle relative to the major bar axis, and the results have been found to be in excellent agreement with theory. Effective z‐component fields for several bar patterns have been measured with and without an in‐plane applied field. Comparison with theory is complicated by the presence of an ion‐implanted layer on the surface of the garnet which presumably affects the field of the bubble and, therefore, the outcome of the experiment. Calculations based upon the idealized bubble field give good qualitative and quantitative agreement with experiment, provided an effective spacing is used which takes into account the ion‐implanted layer and consequent reduction in bubble field.

64K fast access chip design

A 65,664 bit, 4 μm, minor loop chip of 128 loops of 513 bits is described. The goals of fast read and write access, and high average data rates are shown to lead to a decision to use a replicate switch of compact design to minimize loop length and data housekeeping. Operation of the device is described, followed by an example of a read and write of a single block of data and examples of read/write of multiple blocks. Use of modulo 128 and modulo 151 block address assignments illustrate means of obtaining either gapless and gapped record delivery.

Experimental techniques for studying the reliability of bubble memory devices

The reliability of a large capacity bubble memory device was studied by measuring the error accumulation in the stored data resulting from the random annihilation of bubbles during the operation of the device. The bubble‐population dependence of the error‐rate plot indicates a strong dependence of the device reliability on the bubble pattern. To compare the long‐term margin degradation the initial error rate at a fixed bubble population, the inverse of the mean step to failure, is used as the comparison parameter rather than a 50% failure criterion. In this manner, the measurement time required to provide good statistics is greatly reduced when using large capacity chips. In addition, chips of this size are practical ones which can be used in real memory systems. The analysis of the chip reliability is complicated by the many different types of components in the chip and the higher probability of material or processing defects. A generalized field interruption technique is also used to characterize localized areas or components within the large chip. Using this technique in conjunction with the technique of measuring the bubble decay rate at different sites of a repetitive bubble pattern, the direct bubble‐bubble interaction in the circuit can be related to the Permalloy geometry. For example, the bent‐H‐type 180° corner with a diagonal bar was found to cause a significant increase in the error rate in the third bit site of the word pattern 01110111 over that of the pattern 01110000. When the device is operated in start–stop mode, the error rate is further complicated by the static coupling between the bubble and the Permalloy. The effect of bias‐field tilting on the device operation reliability can also be studied through a margin degradation‐type measurement for the start–stop mode.

A magnetic bubble domain flight recorder

An all electronic bubble memory system is aimed toward direct replacement of present flight tape recorders which have a high rate of failure due to the tape transport mechanism. A small 60k bit bubble recorder has been built to demonstrate the feasibility of this approach. It consists of six 10k bit chips mounted in three separate packages operating as a FIFO at 150 KHz bubble data rate. In addition to a FIFO the bubble recorder can be organized as a block oriented memory and also can be operated asynchronously to very low frequencies adding dimensions of flexibility and reliability not found in the present tape recorder system. Recorder nonvolatility is achieved by power shutdown protection circuits and a magnetically closed structure which also prevents magnetic contamination of surrounding equipment. A 10-8detection error rate is achieved with a 300μV detector signal at 150 kHz using power strobing and d.c. restore. Projection on a 50M bit prototype recorder is made. A 105bit chip capacity is proposed based on the electronic cost and MTBF considerations. This prototype operating at 150 kHz is capable of meeting or surpassing the capacity, data rate and reliability specifications of most present flight tape recorders.

High Speed Characteristics of a Chevron Stretcher Detector

The characteristics of a chevron stretcher detector consisting of a thin permalloy bar (200–900A thick) placed transverse to a multiple chevron track are presented for frequencies up to 100 KHz. Data has been taken at quasi‐static and high frequencies on the output as a function of bias field, detector current, in‐plane field and the length of bubble stretching. The characteristics of the bubble signal were also observed as a function of the rotating field and the length of bubble stretching. The characteristics of the bubble signal were also observed as a function of the rotating field direction for detector elements at two positions on the chevron propagation track. The data indicates that the position is very important in determining the structure of the bubble signal. The output drops off with increasing bias and drive field and can be found to vary with frequency depending upon the value of the in‐plane drive field. The output is also linear with the length of bubble stretching.

Data unreliability in bubble memory devices caused by spontaneous annihilation

The unreliability of the bubble devices caused by spontaneous bubble annihilation in a large capacity chip is different from that measured in a small test loop because of the direct interaction between neighboring bubbles. This direct interaction can be separated from the indirect interaction by measuring the individual bit decay rates for different word patterns. Measurement on a 16 micron 20 K bit chip shows that the bubble annihilation rate caused by direct interaction can be several orders higher than that caused by indirect bubble interaction through the remanence effect in the permalloy overlay. Because of this strong pattern dependence, the inverse of the initial slope of the error accumulation (mean step to failure, MSTF) should be used as the comparison parameter to evaluate the long term reliability effects in large capacity bubble memory devices.

Compact replicate/transfer switch for field access bubble devices

A compact transfer/replicate switch has been developed which is also capable of bubble annihilaation. The switch utilizes a new 180 degree multi‐chevron corner adapted to a conventional T‐I propagation pattern (Figure 1) and permits two bit spacing of the minor loops on the access track resulting in minimum access time for any given data block. A two level version of the switch has been employed in a 64K bit block access chip design.1 The manipulation of the domains is accomplished in a fashion similar to an all chevron, one level switch previously introduced by other workers.2 The propagation margin of the minor loop at −25°, 25°, and 70°C is 16, 17, and 13 Oe respectively, for a drive field of 45 Oe at 100 kHz. The annihilation function overlaps the propagation margin completely over this temperature range. At −25°, 25°, and 70°C replicate‐in has 12, 13.5, and 10 Oe; and transfer‐out has 8, 14, and 9 Oe. All the switching functions have at least 40 degree margins.

The operating characteristics of a compact bubble transfer replicate switch

The characterization of a bubble transfer/replicate switch was performed over the temperature range of -25°C to 70°C at 100 kHz. The two level switch utilizes a new 180° degree multichevron corner adapted to the conventional T-I propagation pattern and allows two-bit separation of minor loops on a major input/output track. The manipulation of the domain is similar to an all-chevron single level switch previously introduced by other workers. In addition to transfer and replication, this switch can perform a third function, annihilation of bubbles in the minor loops. It was found that the switching pulse phase and width for all of these functions are relatively insensitive to temperatures. On the other hand, the pulse amplitude margins are somewhat sensitive to temperature because of the decrease in the domain wall energy with increasing temperature. Replication and annihilation have adequate pulse amplitude tolerance over the entire temperature range. However, for a given pulse amplitude the transfer margin degrades noticeably at the temperature extremes. The propagation margin of the minor loop at -25°, 25°, and 70°C is 16, 17, and 13 Oe, respectively, for a drive field of 45 Oe. With the optimum pulse settings, the annihilation function overlaps the propagation margin completely over this temperature range, At -25°, 25°, and 70°C replicate out has a margin of 13, 15, and 10 Oe; and transfer out has 8, 14, and 9 Oe. All the switching functions have at least 40° phase margins.