Ten billion bits on a disk Kees Bulthuis, Marino G. Carasso, Jaques P.J. Heemskerk, Peter J. Kivits, Wilhelm J. Kleuters, Pieter Zalm - 1979

Compact, high-power laser diode systems should make optical-disk recorders attractive for general office use. Their price/performance ratio will make possible such devices as electronic filing cabinets for electronic mail. For these applications, they have advantages over recently announced optical-disk recorders that employ high-power gas lasers and external modulators. Such recorders of digital information are relatively bulky and expensive and are mainly applicable for mass-storage systems.

But high-power AlGaAs laser diodes, for example, have the dual benefits of small size and inherent direct-current modulation capability. External electro-optic or acousto optic modulators are not needed. In addition the use of pregrooved disks simplifies considerably the system mechanics. A compact optical disk system being developed by Philips--an offshoot of optical video-disk tech-nology--uses high-power AlGaAs laser diodes. The recorder resembles strongly in concept consumer optical video-disk and audio-disk players. It gives random-access times, both in recording and in reading, similar to those obtained with present magnetic disks. Information capacities of over 101 bits per 30-cm disk have been achieved, with a mean access time of 70 ms and recording bit rates of 10 Mb/s. Raw bit error rates between 10- and 10- have been measured.

Laser both records and reads

The AlGaAs laser diode in this recording system is used for both recording and reading. The laser has a wavelength of 820 nm; it is a double-heterostructure type with a 5-um-wide stripe defined by proton bombardment. The laser operates in the fundamental transverse mode with an elliptical Gaussian beam profile. Typical values for the beam divergence are 50° full width, half maximum (FWHM) perpendicular to the plane and 25° FWHM in the plane of the junction. The laser's output power vs. drive current is shown in Fig. 1.

Recording materials have been discussed at length in previous articles. In principle, many materials and phenomena can be used for optical recording. If the material has any property that can be changed by absorption of laser irradiation--and can be observed opticaily-it is of great interest. The sensitive recording material (tellurium-based semimetals) in the Philips recorder is of the relatively simple pit-formation type. With these one-layer systems, the recording power can be sufficiently low to meet the requirements set by the use of a laser diode. With a throughput of 35 percent of the optical system, power levels between 10 and 20 mW in the spot are obtained for pulse lengths of 50 ns.

Pregrooving for simplicity

Having selected a good laser diode and recording material, one can construct a compact optical recorder. To avoid the use of high-precision mechanics in the recorder, Philips developed a pregrooving disk concept. Conventional recording requires accurate and heavy mechanics to control both the rotational speed of the disk and the translation of the optics relative to the disk during recording of information. Such systems are used as optical recorders and also as mastering machines for producing photoresist masters for video disks. By using electromechanical actuators, the optical reading units in players can track on the information formed by a relief pat-tern. The reading spot is kept on the groove within an accuracy of 0.1 pm by a radial servo system.

With a pregrooved recording disk containing a continuous-spiral groove with a pitch of 1.67 pm, the servo system controls the radial tracking of the optical recording head during both reading and recording. The pregrooved disk is made the same way that video disks are: by replication from a master disk produced on an accurate mastering machine. In the recorder only a simple tracing control is needed instead of a precision slideway. The pregroove concept also allows for storing synchronization information on the disk. In this way the radial and rotational accuracy of the mastering machine is transferred to the surface of the recording disk. In practice, the groove is 0.6 um wide and 0.07 km deep. Two such disks are sealed to each other with spacers in between in a disk-protection mechanism tradenamed Philips Air Sandwich (described by Kenney, et al in the February 1979 issue of Spectrum). Data is written in the groove by melting holes in the recording layer.

Extending the pregrooved concept

The pregrooved disk concept can be extended further by insertion of special information in the groove as a modulated relief pattern. Such information can be prerecorded at the same time that the continuous data groove is formed during the mastering process. The information is read on reflection, just as in the case of the video disk. The combination of prerecorded and on-line recorded data offers potentially easy data handling. For example, on-line recorded data in the groove must be accompanied by an address; otherwise it cannot be located.

For many applications, the system must allow data to be recorded anywhere on the useful recording area. One can do this by dividing the tracks into sectors and prerecording a heading address at the beginning of each sector. The heading contains information about the synchronization of the sector, bits, and words, as well as the track and sector number. Figures 2-4 show the grooves and heading addresses. The pregrooved disk concept offers other attractive possibilities. Information can be prerecorded on one part of the disk, while empty grooves remain available for use in the recording layer of the disk. In this way the mass distribution of information is combined with capability for recording customers' private information. The prerecorded information, as well as the information recorded in the layer, can be video, audio or data--or a combination of these.

The mechanics, servo systems, and signal processing of the diode-laser pregrooved disk recorder are similar to the diode-laser video disk player or the compact audio disk player. The optical systems for recording and reading are also similar.

A choice of two optical systems

Two optical systems have been developed for the Philips recorder. One is used for both recording and reading, the other for reading only. The two systems employ different solutions for information tracking on optical disks. The major components of the optical system for the recording head are shown in Fig. S. The throughput of the optical system is 35-40 percent from laser to recording spot.

This percentage is relatively low because a spherical col-limation lens is used, while the radiation pattern of the laser diode is elliptical with an axis ratio of 2. A throughput of 50 to 60 percent could be obtained by cylindrical beam-shaping, but this would require complicated optics. Spherical optics were selected to avoid the complicated cylindrical optics at the expense of collection efficiency.

In addition to concentrating sufficient light into a diffraction limited spot, the recording head has three other functions. It must generate servo signals to correct focusing errors, servo signals to correct tracking errors, and it must be able to retrieve the information that has been recorded (the data signal). These three signals are obtained from one spot. Because of the quarter-wave plate, the polarization of the light reflected from the disk is rotated over 90 degrees. As a result, the returning beam passes through the polarizing beam splitter and is subsequently divided by a semitransparent mirror into two parts. One part is detected by a split photodiode for track error correction. The diode detects the asymmetric diffraction from the shallow pregroove caused by possible tracking errors. The other part of the beam generates the focusing error signal and the RF signal as follows: After reflection, the beam is split up by a wedge into two halves, which are focused onto two split photodiodes. From simple ray tracing, it can be seen that the focusing error signal can be obtained by adding and subtracting the appropriate photocurrents of the diode parts. The RF signal is obtained by summation.

Recording by modulation of drive current

Recording is performed by modulation of the laser drive current. Light pulses with powers of up to 20 m W during 50 ns are concentrated in a spot on the disk. For the tellurium-based layer, powers of nearly 12 mW were used during 50-ns pulses for recording with optimum signal-to-noise ratio. Once the servo signals have been generated, the errors can be corrected by means of actuators. The position of the objective lens, relative to the recording layer, is controlled by a voice coil and is held to within 1 pm. The same linear motors are used for tracing as for the random-access loop (a linear sled or a rotating arm). The light spot can thus be focused onto the disk and guided through the pregroove during recording and reading of data.

The major components of the second optical unit, used for reading only, are shown in Fig. 6. Apart from focusing error generation, the optics are quite similar to the optics of the recording head, except for the F detection. A special RF signal detection method is used employing the optical feedback or scoop effect (self-coupled optical pickup). This effect is the result of a decrease in threshold current that results when the output beam is partly reflected back into the laser cavity. The relative increase in output power is maximum around the threshold current. In practice, a relative output increase of a factor of five was found. The feedback effect is useful for RF signal pickup because the disk surface is essentially a rapid succession of reflecting and nonreflecting areas. The laser output is modulated by the RF signal. The light source also acts as the light detector.

The output variations of the laser are measured at the rear mirror by means of a photodiode. This detection method has all optical components on one axis, thereby making easy the alignment of the components. Apart from reading the rF information, an optical reading head must also generate servo signals for focusing and tracking. The error signal for tracking in this one-spot optical path is obtained by a wobbling method. A piezoelectric transducer causes a small-amplitude (0.1 um) vibration of the laser in a radial direction. Phase-sensitive detection of the corresponding variations in the F signal subsequently yields the error signal for tracking.

The actuators for focusing and tracking are driven with the error signals obtained. The complete optical system is mounted in one tube, which is moved as a whole along its axis for focusing. The tube is tilted in the radial and the tangential directions for tracking and for time correction. All movements are induced by small electromagnetic drivers. The length of such an optical head is typically 40 mm, the diameter 10 mm, and the weight can be as low as 10 g. For all three servo systems, a bandwidth of 2 kHz was obtained, which is more than sufficient for disk reading on a 30-cm-diam disk with 25 г/s.

Fast random access easy to obtain

For many applications, it is important that the access time be as short as possible. Relatively fast random access is easily achieved in diode laser recorders because of the small size of the recording head. The use of a pregrooved disk with a prerecorded heading structure permits fast access during data recording and retrieval (Fig. 7).

The linear electric motor, which has a moving coil, can be driven with peak powers of 15 W. As a result, the time taken to move from the inner radius at 7 cm to the outer radius of 14 cm is 100 ms. The mechanical stability of the arm is also of interest for the bandwidth of the tracking ser-vo. A bandwidth of 1 kHz has been achieved, meaning that the first mechanical resonance of the rotating arm occurs near 4 kHz. The average access time on the optical disk, when rotated at 2.5 г/s, is 250 ms. This system can also be used for 25 г/s, with an average access time of 70 ms to the 6 × 10° bits on one side of the disk. By further optimizing the power in the electric motor and the mass of the actuator, the time taken to move from the inner to the outer radius can be reduced to 10 ms, resulting in an average access time of 25 ms at 25 r/s. For faster access, the rotational speed of the disk must be increased. The access time for data on optical disks can thus, in principle, be made similar to the times achieved on magnetic disks. A second example of an optical recorder with access to 5 × 10° bits is shown in Fig. 8, where the recording head is mounted on a sled driven by an electric motor.

Error free digital recording

In any recording system, the signal-to-noise ratio and the raw bit-error rate are important parameters, and both depend upon the laser power used during recording on the disk. This dependency is shown in Fig. 9, where a pulse length of 100 ns was used. The measured noise is white and originates from the roughness of the recording layer or substrate. Preamplifier noise, shot noise, and the noise due to variations in the hole diameter are small compared with the surface noise, and such noise can be neglected in the system under discussion. With the good signal-to-noise ratio obtained, noise does not cause er-tors in digital signals. Errors are caused by defects in the optical recording layer and the substrate. The density of information in optical recording is so high that imperfections as small as 1 um may result in a loss of information.

One source of defects may be dust on the outer surface of the disk. Since focusing takes place through either the transparent substrate or a coating over it, these dust particles are kept out of the focal plane of the objective lens, preferably by more than 400 um, and so do not affect the recording and retrieval of information. A more serious source of defects may be irregularities in the recording layer or on and near the inner surface of the disk, pregrooved or not. These defects must be kept to a minimum--which calls for careful substrate preparation and homogeneous evaporation of the recording layer in a clean environment. Once the disk has been made, the layers are fully encapsulated and the disks can be handled without special care. Typical values of between 10- and 10- are found for bit-error rates, including missing bits caused by the recording and reading of information. For digital data, a much better value for the bit-error rate is generally required. A bit-error rate of better than 10-1 in data retrieval is essential for many applications.

A block diagram of a system that can achieve the desired bit-error rates is shown in Fig. 10. Before writing a data block, the data is encoded. Redundant bits are added. This makes it possible to correct bit-error bursts when the bursts are not too long. By writing the encoded data interleaved on the disk, the performance of the error-correction scheme is often improved. The effect of a long, noncorrectable error burst is then spread over a whole sec-tor, and the error-correction code can often corsect these interleaved error bursts. The data is buffered, modulated, and recorded on the disk.

The data spectrum is matched by modulation to the special requirements of the record/read process. This procedure improves the bit-error rate of the read information, but the system is not error-free because of the effect of large dropouts. Dropouts are detected by reading the information immediately after it has been written and comparing the detected and demodulated data with the original informa-tion. When the comparison indicates that the information in a sector is unreliable, that sector is invalidated; the information is rewritten in another sector. In this way, all the errors are corrected during the recording process. One important requirement to be met by optical disks is that they permit archiving of written information. The raw bit-error rate was measured during accelerated aging tests.

During several months of a temperature/humidity cyclic treatment no significant change was observed in the raw bit-error rate. This testing indicates that good archivability of plastic sandwiches with tellurium-based materials can be achieved. When information is read after storage, it is, of course, essential that the data be reliable. For this reason, the coding scheme must not only correct errors but also detect situations with errors that cannot be corrected at that mo-ment. Such errors can be caused by malfunctioning of the recorder or by external interferences. At such times, it is possible to reread the information. If this also fails to correct errors, an error signal can be given. All of these measures, together with the tellurium-based recording material, ensure that the system is essentially error-free after long storage.

Limitations on the bit rates

The bit rates achieved at present are limited by the laser diode's lifetime and the sensitivity of the recording material. Lifetimes of 2000 hours at 15 mW are reported for commercially available, single-mode laser diodes, and these can be used for bit rates up to 1 Mb/s. Lifetimes of more than 100 hours have been measured in the laboratory for diodes used for 10 Mb/s recording in tellurium-based materials.

As both the laser technology and the sensitivity of recording materials are improved, laser diode-recording will also become reliable for high bit-rate recording. The current modulation of laser diodes can go up to hundreds of megahertz. The light intensity, however, will become high in short laser pulses and may limit the bit rate.

Erasable recording materials are feasible, but the signal-to-noise ratio is still somewhat low and the archival quality has yet to be proven. The first available systems will have nonerasable materials with good archival quality.