(Original German Page)
(Translated Figure Captions)


Overview of the characteristics of the MiniDisc
System data of the MiniDisc
The disks
The technology
Buffer for skip protection
Random access
Signal format
MD pickup
ATRAC data compression
Magnetooptic recording
Magnetic field modulation

Since 1970 Sony is developing technologies for optical disk memories with the goal to use the advantages of operation without mechanical contact between carrier and write/read head with high memory density, long life span, fast random access and low memory costs for recording and playback. The application steps were video disk, CD and CD ROM. In 1988 the series was completed by the rewritable magnetooptic disk for data storage, the MiniDisc (MD). The possibilities which the MiniDisc will open in the area of audiovisual media can hardly be over-looked today.

The base of the digital storage medium is a disk-shaped carrier with substantially smaller dimensions than the CD, however with the same advantage of the fast, random access to any place of the recording, as the users are used to from the CD. With playback and recording the utilization of the modern mechano optical or the magnetooptic technique and the effective audio data reduction leads to extremely compact dimensions of disk and device. The contactless, wear-free mode of operation in connection with the digital audio technique ensures a high sound quality, independent of the number of playbacks or recordings. During playback the effects of shocks or vibrations on the device are electronically cancelled and so the high quality of the digital technique with easy, comfortable and unproblematic handling, which is indispensable for a trend-setting, portable system, is achieved.

Overview of the characteristics of the MiniDisc

System data of the MiniDisc

Playing time in min max. 74
Cartridge dimensions in mm 72 x 68 x 5
Disk data
Diameter in mm 64
Thickness in mm 1.2
Diameter centering hole in mm 11
Diameter startup area in mm 29
Diameter beginning of modulation in mm 32
Track gradient in µm 1.6
Recording or scanning speed in m/s 1.2...1.4
Audio data
Channels 2 (stereo/mono)
Frequency range in Hz 5...20,000
Dynamics in dB 105
Wow and flutter unmeasurable
Signal format
Sampling rate in kHz 44.1
Source quantization in bits 16
Compression system ATRAC
Modulation system EFM
Error correction system CIRC
Optical parameters
Wavelength of the laser light in Nm 780
Diameter of the laser mark in µm 0.9
Laser output during recording in mW 5 (max.)
Recording system magnetic field modulation

The disks

The two variants differ in special functional details. The dimensions are identical, cf fig. 3. The disk is centered in the central hole like the CD. An additionally built in disk of magnetic material supports the mounting and stabilization of the disk in the device. Thus pressure temples from "above", widely used with CD players, can be omitted, not least due to the small diameter and the smaller mass of MD in comparison to CD.
The variant used for the distribution of prerecorded music by record companies corresponds in structure and production technology to the CD. Erasing or overwriting this recording is impossible. The signals are mechanically casted in a pit structure in the surface of the disk during the production process exactly the same as with CD and plated for the optical reading. Thus low-priced production in large quantities with injection technology is possible. The allocation of the disk surface is illustrated in fig. 4. The table of contents TOC is inserted into the startup area during manufacture. Track arrangement, pit, disk thickness and material correspond to the CD, cf fig. 5. The disk is constantly accommodated in a cartridge and so protected from damage. For the scanning, access is necessary only from one side. The slot used for that is closed when outside of the device.
The recordable MiniDisc uses the magnetooptic principle and makes use of the magnetic field modulation. The disks can be rewritten practically unlimitedly. This technology first presented by Sony in 1989 operates with a focused laser beam and a closely limited magnetic field, which affects the recording place from the rear side of the disk. Only this special arrangement is able to make a new recording. That ensures the durability of a recording.

The body of the disk molded from polycarbonate carries the groove structure with the time code (cf random access). The layer structure shown in the fig. 6 provides for the magnetooptical function. The magnetically effective layer from rare-earth metals (terbium ferrite cobalt) is embedded in auxiliary layers, which provide for high performance with low power consumption and ensure reflection of the laser light. This structure has been proven for some years with MO memory in the computer technology. The active layer on the one hand permits the writing (during a new recording) of more than one million cycles without quality degradation and on the other hand ensures a good long-term stability. The MO recording requires access to both sides of the disk, therefore this cartridge has opposing openings, which are again locked by shutters outside of the device. Because of that the labelling area is quite small. Bild 6

Bild 3: Abmessungen der MD
Bild 4
Bild 5

The technology

The circuit diagram of a MiniDisc recorder (fig. 7) shows a large correspondence with the structure of CD players. The adjustment to the optical channel of the data to be recorded is done by the modulation system EFM (Eight to Fourteen Modulation). It corresponds to CD just like the error correction system CIRC (Cross Interleave Reed Solomon code). Additional modules are the data compression, the anti-skip buffer and the magnetic recording head. Digital inputs and outputs operate as usual with 16 bits at 44.1 kHz sampling rate. The system data are to be inferred from the table.

Bild 7: Blockschaltbild

Buffer for skip protection

With the practical use of portable optical memory the skipping of the pickup due to vibrations or shocks affecting the playback device turned out to be very annoying. By adding a semiconductor memory to the signal path it was possible to reduce the effects of such disturbances drastically (Shock Resistant Memory). The data compression to approximately 1/5th of the original data by the ATRAC module built into the system is aiding this function. The reading of the information stored on the disk in sectors - each sector has its own address - takes place with a signal stream of 1.4 Mbit/s. However, due to the data reduction only a continuous signal stream of approximately 0.3 Mbit/s is necessary. This means that the reading is usually done in bursts. This data is fed into a buffer, fig. 8. A semiconductor buffer with a capacity of 1 Mbit can store the signals for about 3 s, with 4 Mbit, 10 s can be buffered. Fig. 9 shows this in another representation and gives a comparison to CD, with which the signal stream is constantly 1.4 Mbit/s. If the optical pickup loses its correct position as a consequence of a sudden external shock, this has no effect on the signal output from the memory, however the buffer is drained thereby. When the interference has seized the pickup returns to the last correct position known by the address and fills the memory with the max. signal stream of 1,4 Mbit/s without interruption, so that in most cases after 1 s the buffer is already full again and reading in intervals is again taken up (fig. 10).

Bild 8
Bild 9 Bild 10

Random access

For the MiniDisc system the direct access is of special importance, because it not only enables highly convenient operation, but is also functionally necessary in connection with the skip protection described before. Playback-only MDs of industrial production carry a continuous time code and a table of contents (TOC) like the CD. This ensures fast and random access. The"blank" Disks intended for recording (Recordable MiniDisc) contain grooves for the guidance of the recording/playback laser molded during manufacture of the carrier (Pre Groove). Additional deflections (some tenth micrometers) are superimposed over the spiral formed by the groove, fig. 11. They form a time code with a resolution of 13.3 ms in coded form. Thus direct access independent of the actual recording is enabled. The particular system is called ADIP (Address In Pre-Groove). For comfortable handling of the recorded disk a user-definable field for the UTOC (User Table Of Contents) based on the timecode is included in the starting area of the track. As shown in fig. 12, modifications in the table of contents parallel to modifications of the stored information are achievable in a simple manner by editing.
Especially the usage options that random access with short access times offers are possible only with a contactlessly scanned, disklike carrier.

Bild 11
Bild 12

Signal format

The MD uses a modern and flexible signal format, which is very similar to the standard CD ROM, mode 2. The data produced by the data compression system ATRAC is error-protected with the help of the data security system CfRC and adapted to the special requirements of the optical channel by the modulation system EFM. Both steps have worked satisfactorily with the CD.
In the format CD ROM, mode 2, sectors are formed from 98 sequential frames, cf fig. 13. Each sector has sync signals in the header and a complete address. The equivalent play time amounts to 13.3 ms, with 2332 byte of altogether 2352 byte contents being available for data. Interleave in the CIRC-Coder of the CD is 14.5 ms (corresponding to 108 frames) and is thus longer than a CD-sector. With the creation of the clusters of an MD four further sectors are included additional to the 32 data sectors to adjust the length. One sector is used for subcode data (2332 byte) and the other three, called link sectors, are used for CIRC on the recordable MD. In the interest of sufficient interleave one parity block is recorded before and another one after the actual data. The RAM which serves as shock resistant memory during playback is used for this nesting during recording. The recording of a playback-only MD is done with a continuous signal flow already containing the CIRC DATA, with the four additional sectors available for subcode data.
The ATRAC Coder compresses the audio data to 1/5 of their original volume. It supplies so-called sound groups with 424 byte, 212 from the left and right channel, for the subsequent processing. Altogether eleven of such sound groups are recorded in two subsequent sectors, cf fig. 14. Each sector covers thereby 5 x 424 + 1 x 212 = 2332 data bytes. 32 data sectors and 4 additional sectors form a cluster. Such a cluster forms the smallest recording unit of the recordable MiniDisc.

Bild 13
Bild 14

MD pickup

The different types of signal storage of the two MD disk types requires a special optical pickup customized for the different requirements. A normal optical pickup for CDs fails when reading a MO recording. With the help of a polarizing beam splitter the detection of the different polarization directions when scanning a MO recording can be accomplished. According to fig. 15 a laser beam (about 0.5 mW) focused on the surface (land) of a CD reflects like light with the nonexistence of a pit, while in the pit the quantity of light is significantly reduced (approximately to 25 %). This difference is used for the signal acquisition, whereby the actual information is stored in the transition between land and pit (EFM modulation, transition = 1; no transition = 0). The signals which the two optical detectors produce during the scanning of a recording are added, cf fig. 16. During the scanning of a magnetooptic disk, a recordable MD, the phenomenon called Kerr Effect is utilized, by which a polarized light beam is twisted into its polarization direction under the influence of different magnetizing devices at the focal point (some degrees), cf fig. 17. As consequence the relation of the quantities of light reaching the photodetectors is shifted as a function of the direction of rotation. Bild 16

Bild 15
Bild 17

ATRAC data compression

The MiniDisc can store only about 1/5 of the data of a CD due to the smaller dimensions. By the application of the data compression system ATRAC developed escpecially for high playback quality the recording of 74 min playing time is possible, with a quality which practically does not differ audibly from CD playback.
ATRAC is based on the utilization of scientifically secured psychoacoustic basics of the human hearing and transmits only those audio signals which are actually required by the ear for the correct perception of the respective sound signal [2].

Low amplitude resolution leads to quantization noise. But if one ensures that this quantization noise is inaudible, then the playback quality corresponds to that of the CD. Therefore a primary task with ATRAC is to minimize the audibility of this noise by hiding the quantization noise in frequency ranges in which high signal levels occur. The maximum of ear sensitivity is situated in the frequency area by 4 kHz, with the ear being partly substantially more insensitive to other frequencies. A tone, which is just perceived with max. sensitivity, is inaudible with the same intensity, but other frequency. Basically two tones of same intensity, but different frequency are unequally loud perceived. A quiet sound can become inaudible with the presence of a loud one. This effect is defined as masking and is the more pronounced, the closer the tones are in their frequencies and the higher their intensity difference is.
Within a temporally limited block ATRAC analyzes the music signal and determines the sensitivity of each frequency range. Sensitive areas are recorded very exactly with small quantization noise. Areas with less sensitivity are recorded less exactly, associated quantization noise remains inaudible, cf fig. 18.

Bild 18

The frequency and time partitioning applied by ATRAC is shown in fig. 19. The unequal width of the frequency bands is remarkable. This allocation is based on a further psychoacoustic effect, the frequency groups (Critical Bands), which were found in the human hearing. The width of these groups increases with rising frequency, it amounts to e.g. with 100 Hz W=160Hz; with 1000 Hz W=160 Hz and with 10,000 Hz W = 2,500 Hz. The groups of frequencies are thus, like shown in fig. 18, in the lower frequency range substantially closer together than with higher frequencies. The transfer of this allocation into the ATRAC system helps to achieve a high accuracy even with small transfer capacity.
Music signals constantly change, and the ear adapts its sensitivity to the rate of these modifications. In lively passages e.g. ear sensitivity changes rapidly, in moderate sections slowly. Therefore ATRAC constantly analyzes the input signal in short time periods and adapts signal processing to the ear behavior. In lively passages time slots of 1.45 or 2.9 ms are formed, in moderate ones up to 11.6 ms. Longer time slots enable the application of narrow frequency bands and result in high frequency resolution with high reproducible sound quality (fig. 20). This signal-dependent flexibility is a key to a high effectiveness of the data compression with simultaneous minimization of quantization noise. This unequal allocation of frequency and time is implemented with ATRAC by the combination of filters and transformation processes, cf fig. 20. The input signal is divided into three bands: low 0...5.5 kHz, medium 5.5...11 kHz, high 11...22 kHz, and further processed with a modified discrete cosine transformation (MDCT). Before that it is determined whether the signal changes rapidly or slowly, and the time slots are selected accordingly.
When the signal is divided into spectral regions, the MDCT values are assigned according to the 52 unequal groups of frequencies. In these groups the bit rate reduction takes place in agreement with the masking and sensitivity conditions of each group. A special algorithm is used to avoid unnecessarily high bit values. Thus the data word length is kept small, at the same time however audible modifications of the music are avoided.
With the re-conversion of the signals in the decoder first the MDCT frequency values are transferred to time values by inverse MDCT function. Finally the three bands are combined in order to receive a normal digital 16-bit audio signal. The real data stream amounts to 256 Kbit/s for the stereo channel.
The complexity and the high level of these technological solution become clear in the fact that the entire ATRAC signal processing had been implemented in only one LSI already with the introduction of the system.
Bild 20

Magnetooptic recording

MO technologies have already been in use for some time. Specific demands for the application of the MiniDisc existed regarding the practically unlimited number of recording processes on a carrier and particularly the implementation of light and compact devices with low power requirement. Magnetooptic procedures record opto-thermally magnetically. Magnetic, amorphous thin layers of rare-earth metals (Godolinium, terbium, dysprosium) or alloys with these materials, evaporated on plastic carriers, serve as recording media. After passing through the transparent, optically effective carrier layer the focused laser beam locally warms up the magnetic layer to the curie temperature of the respective material. An exterior magnetic field matching the modulation to be recorded creates a domain with this direction of magnetization at this place. After the magnetic field is switched off or the warmed up place of the layer has cooled to the surrounding temperature, this is preserved, fig. 21. With the application of the MO technology with computers, the laser is controlled by the signals to be recorded, the magnetic field is raised continuously. These recordings are erasable by creating a continuous, uniform magnetization, either in a separate run or by attaching a second laser and magnet - similar to magnetic recording - in front of the recording head.

Bild 21

Magnetic field modulation

For the development of the MD, with which the signals are stored in the same format as with the CD, three targets were given: The possibility of overwriting, memory density and track velocity like CD, address formatting already during the disk production.
The address formatting is done by adding deflections to the guiding groove, providing an absolute address with 13.3 ms resolution over the entire disk surface (s. fig. 11).

The overwriting is a basic requirement for the continuous recording of audio signals in real time on an already recorded disk. The application of the laser modulation, like it is used in optical memory of computers, is not suitable for the MD, because here erasing and recording take place separately. Therefore for the recordable MiniDisc the magnetic field modulation was developed.

With the magnetic field modulation the laser is constantly switched on, and the magnetic field is modulated for recording (fig. 22). This allows overwriting existing recordings without a separate erasing pass. With the MD, this technique is based on a highly stable layer of Terbium Ferrite Cobalt, which allows a magnetization modification with comparatively low field strengths of 6,4 kA/m (80 Oe), while so far a value approx. 3 times as high was necessary. The complex requirements of the recording carrier are achieved among other things by the imbedding of the magnetic layer into a multi-layer system (fig. 6). The low field strength makes possible a small magnetic head with low power requirement and practically immediate change of the direction of flow (approximately in 100 ns) when reversing the direction of magnetization.

Bild 22
The signal pits of a conventional CD are created by an argon laser with 460 Nm wavelength and focusing through a lens with NA = 0.9. That results in a diameter of the light spot of 0.4 µm. For the MD system only a diode laser with 780 Nm was available. Focused via a lens with NA = 0.45 it gives a light spot diameter of 0.9 µm. Thereby achieving the memory density of the CD seemed to be impossible.

In experiments with CD recording by means of conventional laser modulation (rate of 1.2 m/s) up to 200 faulty data blocks per second were found, a value just within the limits of the CD standard. With the application of the magnetic field modulation the number got down to 20 per second, fig. 23. Magnetic field modulation is thus not only suitable for overwriting, it also results in recordings with very few errors. The differences concerning the error behavior between the laser and the magnetic field modulation find their explanation in the pit shapes actually left in the track. With the magnetic field modulation the diode laser constantly produces an output of approximately 4.5 mW, and during focusing on the magnetic layer this achieves the curie temperature (about 180 °C). After leaving the light spot the temperature drops. When repeating this process with the presence of a magnetic field with two different orientation directions, dependent on the orientation either 0 or 1 is recorded. In this way the pits shown in fig. 24 are created. If the magnetic field can be switched sufficiently fast, then it is possible to create areas with a length of 0.3 pm also with a laser with 780 Nm wavelength, focused through a lens with NA = 0.45. Thus the demand for a memory density matching the CD is met! Characteristic for the magnetic field modulation is the high symmetry of the pit structure based on polarity switching (from + to -). In contrast to it the laser modulation results in very asymmetrical structures.

Bild 23
Here the magnetic field can be oriented only in one direction, with the allocation a 1 = laser lighting and 0 with the laser switched off (non-recorded area), cf fig. 24. With laser modulation, e.g. the second half of a pit is always thicker, because the temperature rises. By controlling the laser output a balance would be possible. On the other hand a different laser output influences the beginning or terminator point of a pit. Thus distortions and higher asymmetry are created. Simultaneously time errors (jitter) occur. That is particularly critical, as because of the use of the EFM the length of the pits and pauses form the basis of the data transmission.
The magnetic field modulation allows for fluctuations of the laser performance of up to 20 %. Because with magnetic field modulation the laser beam only provides for the heating of the layer during the recording, a tilt between disk and scanning unit is not critical either. Despite the described advantages of the magnetic field modulation the application in computer technology fails because of the high linear speed of more than 10 m/s common to this application and the high frequencies (some 10 MHz) with which the magnetic field would have to be repolarized. In contrast to that, with the MD the necessary frequency of 720 kHz can easily be implemented with suitable head constructions:
Magnetic field modulation allows only the one-sided data recording on the disk, however for consumer applications this is no disadvantage, because a double-sided version would be twice the cost.
Bild 24

No requirement on completeness and correctness of the specification
All rights reserved.

All rights reserve audio map - Internet Publishing.
In the case of for problems and errors ask email to our
Web master