Bias Margin Research Articles

Temperature distributions in a megabit bubble chip

For a large-capacity large-size bubble chip, various operating parameter nonuniformities across the chip as well as garnet material and device processing inhomogeneities could be a source of considerable bias margin loss. An example is chip temperature nonuniformity due to detector heating. In an attempt to estimate the chip margin loss due to local temperature nonuniformities, temperature distributions in a Mbit chip have been mapped as a function of detector current amplitude. Temperatures at several regions in the chip were monitored by measuring the resistance of active device components, such as transfer-in gate and replicate/transfer-out gate conductors, and detectors themselves that are of full-shorted 200-element chevron stretchar type. A simple physical model has been developed which, together with various measured local temperatures, provides complete two-dimensional mapping of temperature distributions in the chip. Nonnegligible temperature differences are shown to exist among minor loops (especially at the replicate/transfer-out gate), and those minor loops located close to the detectors are shown to suffer considerable margin losses even at moderate detector-current amplitudes. Operating margin measurements of the minor loops have confirmed the finding.

290KBit, 15μm cell bubble chip

A block replicate bubble memory chip with a 14×16μm2 bit cell size has been developed using state-of-the art bubble technology. Organized into 282 minor loops and 1025 bits per minor loop, the total capacity of the chip is 289,050 bits. At 25°C and 45 Oe drive the start-stop operating bias margin is 18 Oe up to at least 300 kHz field rate. For operation over the temperature range −25°C to 100°C and up to 300 kHz, all chip drive parameters have at least ±10 percent operating range indicating that no drive parameter is required to be temperature compensated.

256 kbit bubble memory chip fabrication and characterization

A 256 kbit bubble memory chip, which uses asymmetric chevron propagation elements has been designed, fabricated and characterized. Average circuit period is 8 μm and minimum feature size is 1 μm for all chip functions. Chip size is 4.5 mm×5.1 mm. A new stretcher detector and a planar chip structure enable the 256 kbit chip to operate at low rotating field. For a typical device processed on a (YSmLu)3(FeGa)5O12 magnetic garnet film, the overall bias margin for all functions is 15 Oe with a 50 Oe circular rotating field at 200 kHz.

Characterization of 1.5 µm Bubble Devices built on low Magnetization Garnet Films for minimizing Drive Field

Minimization of the drive field by using garnet films with low magnetization has been studied for 1.5 µm diameter bubble devices. Test chips designed on 8 µm period half disks with a 1 µm minimum feature have been built on three sets of garnet films with different magnetizations of typically 350, 400 and 490 Gauss (G) by using planar processing. Results for devices made on garnet films of various magnetization show that the normalized bias margins increase with decreasing magnetization and minimum drive fields for these devices decrease from 30 to 20 Oe. Furthermore, when the drive field was defined as the field necessary to produce normalized bias margins of 20% for propagation and 15% for the stretcher, we obtained a drive field value of 30 to 60 Oe for a magnetization of 350 to 490 G.

Block-replicate chip operation at 250 kHz over the temperature range -25°C to 75°C

A block-replicate chip employing asymmetric half-disc propagating elements, pickax replicator, and swap gates was designed and fabricated. One bit per rotating field cycle data rate was accrued for this design through the use of a data merge and one detector. Packaged chips were operated and characterized at 250 kHz over -25 to +75°C temperature range. A composite bias margin of 7 Oe under worst case loading at any given temperature in the temperature range was achieved.

Gap-tolerant half-disk bubble device margins

A simple model is presented which allows accurate prediction of bias margins of gap-tolerant half-disk propagation tracks for bubble domains. After this is verified by comparison with experimental margin data, an "isomargin" plot is derived to show how the margin varies as a function of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">W</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">G</tex> , where <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">W</tex> is the minimum linewidth and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">G</tex> is the inter-bar gap. The bias margin is shown to decrease along a fairly straight line which goes to zero when <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">W + G</tex> equals the runout diameter, i.e., when <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">W+G \approx 1.5 W_{s}</tex> , where W <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">s</inf> is the bubble stripwidth or average bubble diameter. This agrees with experiment, and means that the minimum resolvable feature for half-disk type patterns must be less than <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.75W_{s}</tex> , and probably will not be much larger than <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.5W_{s}</tex> to <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.6W_{s}</tex> . It is concluded that, if made with perfect Permalloy, T-bars and half-disks should propagate isolated bubbles equally well. The advantages of half-disks over T-bars are 1) the fatal bar-crossing problem of T-bars with multiple bubbles is avoided, 2) the minimum propagation field is lower than for T-bars, and 3) half-disks seem more tolerant of "bad" (e.g., high-coercivity) Permalloy. Also tabulated are the effects on margins of variations in the device parameters of a representative design, as might be encountered in a fabrication process with finite tolerances. A brief discussion of stop-start margins is given in conclusion.

Design and characterization of 3 μm bubble memory chips utilizing Y-Y patterns

Major-minor organized bubble memory chips have been developed utilizing 14 μm period Y-Y propagation circuits. First type is a 75 kbit memory chip consisting of four major-minor units for high performance demands. Designed access time and data rate for 300 kHz drive are 0.3 nsec and 2×600 kbits/sec, respectively. Second type is a single unit major-minor 78 kbit memory chip for medium access time use. Main chip design features are as follows: (1) 14 μm Y shaped Permalloy circuits are used throughout the chip, except for the 18 μm chevron stretcher. (2) Y-Y transfer gates have been tailored in order to adapt to Y patterns. (3) A thin film Permalloy detector is used for the sake of large output (15 mV/2 mA) amd O-π phase detection. The 3 μm bubble chips are fabricated on (YSmTmCa)3(FeGe)5O12 garnet films. Wafers are processed through 3 evaporation steps and 5 masks. Total 20 Oe operating bias margin is obtained at 300 kHz for 50 Oe drive field.

Replicate/transfer bubble switch

A replicate/transfer switch compatible with the gap tolerant structure has been developed. The switch characteristics are superior to that of the existing switches both at the the 8-μm and 16-μm periods. The switch offers the advantages of good bias and phase margins, ease of fabrication, and reduced drive field requirements. In the 8-μm version the device requires a substantially lower drive field than the pickax design by virtue of reduced bubble-bubble interaction in the minor loops.

Operation of a 100-bit, 2.6 µm bubble shift register

Complete circuit operation for a small 100-bit serial shift register memory with a 10.6 μm period fabricated on a garnet wafer using electron beam lithography and single level all-permalloy technology is reported. Overall circuit operation including generation, propagation, and detection was achieved with a 6 Oe bias margin using a 1 kHz rotating field. Although the circuit design and fabrication techniques were not optimized, we believe that this is the first published report of complete circuit operation for a single level device with bubbles as small as 2.6 μm. The performance of two different types of chevron expander detectors and two different generators was evaluated and circuits combining disk generators with herringbone detectors were found to provide the best overall operating margin. At low speeds (∼5 Hz) high-bias failures were caused by failure of the domains to strip out fully in the detector while the low bias limit was determined by the introduction of spurious bubbles into the track near the generator.

Fabrication of 3 µm Bubble 80 kbit Chips

A 3 µm bubble 80 kbit chip was designed by 1/2 scaling from the pattern for 6 µm bubble chip. To obtain a 1 µm minimum feature over a 5 mm square, projection printing was employed. By reducing the reflectance of the surface to be exposed below 33%, standing wave effects with monochromatic illumination could be minimized, and uniform 2 µm width and 1 µm gap pattern could be obtained, notwithstanding the variation in photoresist thickness and the unevenness of substrate. A standard deviation of this distribution in the pattern gap width in the fabricated 80 kbit chips was 0.067 µm when the flatness of the wafer is less than 0.5 µm/chip size. The effects of the variation in the pattern gap on the propagate margins could be ignored, if it fell in the range of 1.03 µm to 1.38 µm. Consequently, the resolution of 1±0.3 µm and uniform bias margin over 80 kbit chip area could be achieved with high yield.

Some temperature characteristics of a bubble memory

A 28-μm period, all chevron bubble memory using 6-μm diameter domains in Y <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1.32</inf> Sm <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">.38</inf> Lu <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">.38</inf> Ca <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">.92</inf> Fe <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">4.08</inf> Ge <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">.92</inf> O <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">12</inf> was operated at 100 kHz from -45° to +100°C with suitably chosen pulse and drive field amplitudes. At high drive currents the minimum all-function bias range among 4 accessible storage loops in the major-track, minor loop organized memory increased from 12G at -45°C to 17G for a broad maximum about room temperature and then decreased to 6G at 100°C. The behavior of bias margins with temperature and drive-current amplitude is shown. The stretch, cut, and annihilation-pulse operating ranges decrease with increasing temperature. Delay tolerances for the annihilation and cut pulses are shown as a function of the temperature.

Growth and mobility measurements of (YLuCa) 3(FeTiGe) 5O 12 films for magnetic bubble application

The liquid phase epitaxial growth conditions and properties of garnet films of nominal composition Y 1.8 Lu 0.2Ca 1.0Fe 4.0Ti 0.3Ge 0.7O 12 are described. Ca+Ge reduces the lattice parameter when substituted into yttrium-lutetium iron garnet. On the other hand, Ca+Ti increases it. Ca+Ti+Ge additions to the garnet make it easy to match the lattice parameter with the GGG substrate and to obtain stable bubbles. This material has low ferrimagnetic resonance damping loss and high mobility bubbles. The bubble mobility, estimated from the damping parameter and other static bubble properties, is 5626 cm/sec-Oe. This value approximately agrees with that measured by translational field gradient techniques. The bubble translation bias margin at 310 kHz and temperature coefficient of the static bubble properties are reported briefly.

The effect of DC in plane field on the operation of field access bubble memory devices

The effect of a small dc in-plane field on the start-stop operation of field access bubble devices has been studied. Experimental results show that the bias margin in this mode is very sensitive to the magnitude of the field and its orientation relative to the start-stop direction of the drive field. In a T-I circuit a complete margin loss was observed for an in-plane field of 3 Oe oriented antiparallel to the start-stop direction. For parallel orientations of the in-plane field the start-stop margin improved and approached that of the continuous propagation margin at an in-plane field of approximately 6 Oe. Dependence of the start-stop margin on the orientation of the start-stop direction relative to the pattern was also observed. Measurements of the bubble collapse field at various points in the pattern show a very strong dependence on the in-plane field and the permalloy geometry. The collapse-field results and magnetostatic energy considerations which take into account local field variations and bubble-bubble interactions provide a basis for understanding the experimentally observed start-stop margins. These results show that a small tilt (2 to 3°) should be introduced in the bias field to overcome normal alignment tolerances and ensure that a favorable in-plane field is always present. This assures reliable start-stop operation.

10 kbit Bubble Memory Chip: Design and Fabrication

A 10 kbit magnetic bubble memory chip was designed and fabricated with a 6 µm bubble material. The chip was designed using a major-minor organization with new elements. The chips were operated in sufficient bias margin at 100 kHz, 40 Oe drive field. From analysis of detector output, it was shown that the serpentine detector could be designed from the collapse field and the mobility of a garnet film. 256 chips were fabricated using dry processing technique and the yield of 37.9% was obtained. From analysis of these results a 3 µm bubble 80 kbit chip was designed and the process yield at a reasonable level is expected.

Design and Evaluation of 64 kbit Magnetic Bubble Memory Chip

A 64 kbit bubble memory chip is designed, fabricated, and tested. The chip contains 128 minor loops, each with 513 bits. The circuit period is 20 µm. The chip size is 6.4 mm by 6.0 mm. The bubble material used is (YSm)3(FeGa)5O12 with a demagnetized strip width of 4.5 µm. The chip is mounted in a memory module and tested by a chip tester at a rotating field frequency of 100 kHz. Complete memory operations are successfully performed using all memory addresses of all minor loops and an overall bias margin of 10% is obtained. Thus, the feasibility of practical 64 kbit memory chip operation is confirmed.

Diagnostic testing of a 10-kbit bubble memory chip

A 10-kbit bubble memory chip has been designed and fabricated. Testing was accomplished using a new diagnostic test system, which can drive the bubble chip at two different speeds with bias fields switched synchronously with the bubble propagation. Bias margins of the fabricated chips were analyzed and it was confirmed that a sufficient bias-margin window could be assured in long-term operation.

Bubble dynamics in T-bar-type circuits

A theoretical model for bubble propagation under T-bar type overlays due to an in-plane rotating field is developed. By making certain assumptions about the nature of the overlay magnetisation, it is shown that bubble driving forces can be readily calculated for a circular bubble of varying diameter moving freely in 2-dimensions. Expressions for the frictional forces acting on a bubble are derived and the resulting equations are solved to yield the bubble centre trajectory, the radius, velocity and phase variations. This analysis is applied to examine bubble propagation along a straight portion and around a corner of a T-bar circuit. Cases of successful propagation of an isolated bubble, as well as of failure to propagate, are given. Bias margins for a straight T-bar track at 100 kHz are obtained theoretically and compared with experimental results for a chip having the same parameters.

New high−speed bubble garnets based on large gyromagnetic ratios (high g)

A new approach to overcoming the problem of dynamic conversion in high−mobility bubble garnets is described based on large gyromagnetic ratios (high g factors). In a film of Eu1.45Y0.45Ca1.1Fe3.9Si0.6Ge0.5O1 2, a g factor greater than 30 has been obtained, which increases the usable domain wall velocity before onset of dynamic conversion by more than an order of magnitude over comparable bubble garnet films with g approximately 2. Using a circuit period of 28.8 μm, propagation data showing no deterioration in bias margins out to 107 steps were obtained on this film at 1 and 2 MHz.

The effects of spacing between garnet film and Permalloy overlay circuit in magnetic bubble devices

An experimental test set for magnetic bubble devices has been constructed in which the spacing between the garnet film and the Permalloy overlay is variable. The experimental uncertainty in spacing is approximately <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">\pm.15\mu</tex> m, and spacings as small as <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">.5\mu</tex> m have been attained. Bias margin data are presented which were taken at 1 Hz on a 20 micron period chevron circuit as a function of spacing. The collapse and strip-out fields begin to be affected when the spacing is comparable to the garnet film thickness, increasing as the spacing decreases. At larger spacing the high-bias failure mode changes from collapse to uncorrelated bubble motion. A theoretical model which accounts for some aspects of the spacing dependence of the strip-out and collapse fields is described. This model approximates the circuit by a continuous Permalloy sheet. At the low spacing required for efficient use of the rotating field, the model indicates that ±10% nonuniformity in a 2 micron spacing over the device area results in a degradation of the bias field operating range by about 12%.

The Static Stability of Half-Bubbles

A variational model is used to calculate the static stability limits and equilibrium properties of “half-bubbles,” magnetic domains residing on one surface of a magnetic bubble material platelet. Stability is achieved through the presence of gradients in the domain wall energy density and/or saturation magnetic moment. The model evidences two distinct types of instability behavior separated by a critical value of the wall energy density gradient. Unlike the standard cylindrical domain, the half-bubble has a minimum stable value of the ratio of domain diameter to height. The half-bubble is shown to possess a number of properties which make it potentially useful for device applications. It is self-biased by its closure wall and in some cases is stable in zero external bias. Bias margins are of the same order as those for the standard cylindrical domain. It stabilizes on only one platelet surface, and its properties are independent of both material thickness and of minor irregularities on the second surface. In addition, its structure may be advantageous in avoiding the undesirable properties of hard bubbles.