Mass Storage Technology
Is the hard disk about to crash? Kendall Wrightson looks at the latest developments in mass storage technology, and discovers an optical future.
There have been many developments since the last time we looked at mass storage technology back in January 1990: new mechanisms have finally made it to the market, while older devices have halved in price and doubled in capacity almost every six months. Unfortunately, host demands have more than kept pace with capacity. For example, 32MB is no longer an outrageous amount of memory for a professional sampler. As for tapeless recorders, the latest 'budget' systems have jumped from 2 to 4-track simultaneous recording capability. Up in the professional echelons, 20-bit resolution is rapidly emerging as the new mastering standard, demanding at least 25% more storage than 16-bit systems.
In conventional recording, tape is used as both the recording and backup media. However, tapeless recorders still rely on magnetic tape — DAT, Video 8, tape streamers etc. — for backup since no mass storage device has yet been produced that combines, speed, interchangeability and capacity in just the right proportions.
Interchangeable and with capacities of up to 1.4MB (formatted), the floppy disk remains a viable main and backup medium for MIDI sequence, synth patch, desk automation and EDL (Edit Decision List) data. In these applications, data is manipulated in RAM and periodically saved to disk, so disk access time is not of fundamental importance. However, in tapeless recording applications, access time — or more correctly data transfer rate — is vital, as digitised audio is written to disk in real time. Floppy disks cannot offer anything like the speed required for tapeless recording, and even as a backup medium for samplers, floppies are next to useless. For example, it takes over a minute to save a modest 2MB Akai S1000 bank onto two 1.4MB floppies. With a 32MB machine, the backup process lasts 16 minutes and requires 23 floppy disks.
Tapeless recorders also make huge demands on capacity. The rule of thumb is 10MB per stereo minute for a system offering 16-bit resolution and a sample rate of 44.1kHz. With their capacity of 1.4M, standard 3.5" 2FHD floppies are clearly out of the question. Higher capacity floppies will be with us soon, but are unlikely to deliver in the data transfer rate department.
Thanks to the dual attractions of large capacity — 20MB to 1.2Gigabytes (= 1200MB) — and fast data transfer rates, fixed hard disks remain the most popular tapeless recording medium. However, their non-interchangeable nature means they must be erased when full, so an additional backup medium is essential. Hard dives are fragile and susceptible to static electricity and stray magnetism. They may also fail catastrophically should the read/write heads suddenly decide to plough into the rotating disk platters. Thankfully, this doesn't happen too often. On the bright side, hard drive prices have plummeted over the past two years, and the average cost per megabyte currently stands at around £3.
In a removable hard drive, the rotating disk platter is separated from the reading mechanism and encased in a tough plastic cartridge. Thus, when full, a new cartridge can be inserted into the reading mechanism ready for the next project. Such interchangeability means that, like floppies, hard disk cartridges can be used as a backup medium. Additionally, individual cartridges can be formatted by several host systems. For example, a removable hard drive normally hooked up to a tapeless recorder can be temporarily connected to a sampler or computer for back up purposes.
Some removable hard drive mechanisms are more expensive than fixed hard drives offering similar capacities. However, the cartridges are far cheaper than the reading mechanisms themselves, so the initial high cost of a complete drive can be recovered fairly swiftly, as you buy more cartridges, in applications requiring large amounts of data storage. Discounting the cost of the drive itself, a £45, 41.5MB SyQuest removable cartridge gives a cost per megabyte of around £1.
The SyQuest cartridge remains the most popular removable format for sampling applications, proving popular with 2MB and 8MB sampler owners desperate to find a way out of floppy hell. As a tapeless recording medium, SyQuest 44MB drives offer around four stereo minutes per cartridge, which is tiny, though perhaps sufficient for editing/mastering individual songs or commercials.
Like all hard disk mechanisms, SyQuest cartridges are fragile, and unlikely to survive a drop from desk to floor. This is also true of SyQuest's new 88MB (eight stereo minutes) cartridges which are physically identical to their 44MB cousins — six inches square, half an inch thick and weighing about 500 grams. Unfortunately, the 88MB drives are not downwardly compatible, ie. an 88MB drive cannot read a 44MB cartridge.
Ricoh, Iomega (Bernoulli) and GCC are the latest players to produce removable (and mutually incompatible) hard drive mechanisms. All three drives are physically similar in size to the SyQuest, but lighter. Each claims to be more reliable than SyQuest too: the new Iomega 90MB drives feature an improved head design, freeing them from the dreaded head crash scenario. GCC claim that their 50MB cartridges — badged as UltraDrives in this country — will function normally after being thrown to the ground, a fact demonstrated at the drive's UK launch.
Standard fixed hard drives utilise 5.25" or 3.5" disk platters, but in the past 18 months, new 2.5" hard disk mechanisms have emerged, providing a massive shot in the arm for the rapidly expanding portable computer market. Pocket-sized hard drives are handy for samplers too, though expensive when compared with both fixed and removable formats: £795 for 80MB is typical. There are two further disadvantages: portable hard drive capacities are limited to 80MB (so far) and, like all fixed and removable hard drives, they must be handled with kid gloves and kept clear of loudspeakers and computer monitors.
Clearly, the ideal tapeless recording medium should be physically strong and combine high capacity with fast data transfer and interchangeability, and eliminate the need for a separate backup medium. The closest technology has come to delivering this ideal (so far), is the erasable optical drive.
Like a conventional tape recorder, floppy and hard disks operate on the principle of magnetic polarisation. However, with a hard disk, the magnetically sensitive medium spins rather than spools, and the recording heads are not fixed but step across the disk surface under instructions issued from a computer. Magneto-optical drives utilise a combination of magnetic and optical technology. The surface of a magneto-optical disk has a uniformly magnetised active layer in which each magnetic domain is aligned perpendicular to the plane of the disk.
Unlike conventional magnetic materials, each magnetic domain on the disk's active layer can only be polarised above a certain temperature, called the Curie Point. Therefore data can only be written — using an electromagnet — by first heating the active layer with a high powered laser beam.
The first hurdle scientists faced when developing erasable optical technology was to find a medium that could withstand being repeatedly melted by a laser beam without distorting permanently. A polycarbonate disk coated with metallic film turned out to offer all the properties required.
Challenge number two was to develop a laser that could deliver a beam powerful — yet thin — enough to heat the disk to the required temperature within the fraction of a second during which the target spot passes under the laser.
The rate at which the laser beam can heat a target spot is determined not only by the rotational velocity of the disk and the width of that spot, but also by the wavelength of the laser light. The closer towards the blue end of the spectrum is the laser, the faster it can be modulated (turned it on and off) by the data at a given power level in order to heat the disk.
Currently, the speed at which most magneto-optical drives write to disk is slower than the speed at which they can read from disk. This is because the magneto-optical recording process requires that old data be erased before new data is written. To make matters worse, the newly written data is then verified, taking up yet more valuable time. At first glance, it may appear that the magneto-optical drive's sluggish access time is the only reason why some magneto-optical drives can play back stereo 16-bit/44.1kHz digital audio, but cannot record it.
"Removable hard drive mechanisms are more expensive than fixed hard drives offering similar capacities, but the initial high cost of a complete drive can be recovered fairly swiftly as you buy more cartridges."
However, the average seek time of a magneto-optical or hard disk — the average amount of time it takes for the read/write head to move into position — is not the only factor to be considered in determining whether disk X will work with system Y. For example, Digidesign's Sound Tools and Pro Tools (which both run on any Apple Macintosh II), will not record at 44.1 kHz to most magneto-optical drives. Yet a standard cheap and cheerful Atari 1040 or Mega ST will record to magneto-optical drives at 44.1kHz using the Atari version of Sound Tools. Why? The answer is that the Atari achieves a higher sustained data transfer rate than the Macs.
Data transfer rate (measured in kilobytes or megabytes per second) is the rate at which data is received/transmitted to and from the host and drive. Data leaving the host's DSP is temporarily stored in a RAM cache (an area of fast memory) before being passed to the peripheral port: (SCSI in the case of the Mac).
Theoretically, SCSI can support data transfer rates of up to 4MB per second, but in reality, this is rarely achieved, and certainly not on the Apple Macintosh. The Atari's peripheral port uses a protocol called DMA (Direct Memory Access) and achieves a faster data transfer rate than most Macs! On the other hand, the DMA format offers less data verification, and is therefore potentially less reliable.
The data transfer principle operates within hard and optical drives too. For example, data may arrive at a drive's SCSI port at up to 900 kilobytes per second, while the average rate at which data is written to disk may be around 300 kilobytes per second (due to the drive's seek time). The larger the cache, the longer the drive can sustain a high data transfer rate.
For drives with poor seek times, a massive cache is not really an answer since it would mean that there would be a delay in all operations as the cache was emptied. In tapeless recording, for example, it would mean that drive activity would continue for several seconds after issuing a command. For optimum operation, the sustained data transfer rates for both host and drive must be matched.
Over the past few years, it has become standard practice to quote the average seek (or access) time as the most important specification when determining a drive's performance. Indeed, drives with low seek times are described as "fast". However, from the above explanation, I hope it's clear that a drive's sustained data transfer rate is by far the more important specification. For computers, identifying relevant specifications is even more difficult, since the sustained data transfer rate will depend on many variables including the speed at which the DSP operates, the size of the cache, the SCSI and the CPU clock speed.
There are currently four main optical formats, the largest sellers being the magneto-optical drives (made by Sony, Ricoh and Panasonic amongst others) that utilise cartridges incorporating 5.25" double sided disks capable of storing 300MB to 600MB per side, unformatted. The cartridges measure about 8 x 6 x 0.5" with a write protect tab and a metal access door just like a floppy. After formatting, 600MB cartridges offer approximately 25 minutes per side of 16-bit stereo audio at 44.1kHz. The cheapest drives retail at around £2,300, but cartridges are sold for around £100, offering an amazingly low cost per megabyte of only 20p.
Over the past 18 months Ricoh and MaxOptix have both introduced new magneto-optical drives featuring a technique called Zoned Constant Angular Velocity (ZCAV). The Ricoh disks offer 300MB per side (600MB total), while the MaxOptix offers 500MB per side, so the disks offer 25 and 45 stereo minutes per side respectively.
The ZCAV technique delivers higher performance than the Sony/Ricoh drives, so the former can be used to record direct to disk with Sound Tools or Pro Tools on a Mac IIcx and above. However, there is a premium to pay for the extra performance (and the extra capacity in the case of the MaxOptix): the latter costs around £5,500, and the Ricoh comes in at £3,168.
Like the MaxOptix drive, Panasonic's recently introduced 'phase change' mechanism offers 1GB disks — 500MB per side. The phase change technique is said to enable a faster seek time, since it does away with the erase-before-write problem associated with the magneto optical devices. In a phase change drive, the disk is heated by either a strong or weak laser beam, representing a binary 1 and 0. Due to different cooling rates, the spot heated with the stronger beam cools to an amorphous state, while the spot heated with the weak beam cools to a crystalline state. A weak beam is used to read the disk, giving either a strong or weak reflection depending on whether the target spot is in an amorphous or crystalline state.
Last year IBM released a new magneto-optical drive featuring single sided 3.5" 128MB media. The cartridges it uses can be referred to as 'disks' since physically they are exactly the same length and width as a standard 3.5" floppy, only a little thicker.
Like the various 600MB and 1GB 5.25" opticals, the 128MB drive will work with some tapeless systems, but not as yet Sound Tools or Pro Tools on the Mac. However, it is anticipated that by next year the problem will have been remedied. If this is the case, the 128MB format should take off in a big way, since 12 minutes of stereo is very useful (though not brilliant). For samplers, 128MB is ideal, packing 16 8MB banks onto a 3.5" disk. 128MB optical disks have a lot to offer in the backup department too, thanks to their excellent size/capacity ratio.
Apart from capacity and physical size, perhaps the most important difference between the new 128MB optical format and all those mentioned above is that the former conforms to standards controlled by the American National Standards Institute (ANSI) for both hardware and formatting. This means that all 128MB 3.5" disks should work in all 128MB 3.5" drives, regardless of manufacturer.
Sony recently released a 128MB drive, an important indication of how seriously the new format is being taken by the Japanese even though it was developed by IBM. To remind the world of this fact, IBM should, by time you read this, have released a new upgraded 128MB drive that will offer SCSI-2 compatibility, an improved seek time, and the ability to read O-ROM media.
O-ROM disks are 122MB read only disks designed to be used in a similar way to CD-ROMS, ie. to provide large quantities of read only data. With only a fifth the capacity of the well-established CD-ROM format, it remains to be seen how much success O-ROM will find.
"The huge capacities that optical drives can provide should satisfy the future requirements of tapeless audio, graphics, and the new wave of multimedia and desktop video applications."
Just when you hoped that there couldn't possibly be another format to consider, along comes Second Wave with their Floptical drive, a hybrid of standard floppy drive and optical technology as the name so subtly suggests. The Floptical's most notable feature is its ability to read both Second Wave's new 20MB 3.5" disks and existing 1.44MB 3.5" floppies. From the host's point of view, the new 20MB disks are treated exactly like floppies.
For example, when used with a Mac, they can be ejected from the Mac desktop by dragging them into the trash, unlike both removable hard and optical disks. Speed-wise, the Floptical is slower than current optical drives, though standard floppy performance is greatly accelerated.
As to the future of mass storage, it seems likely that fixed and removable hard drives will find it increasingly tough to compete with both large and small capacity optical formats. There are many reasons for this. Optical disks are far more secure than hard drives since, unlike the latter, they are physically robust — you can treat them pretty much like floppy disks. However, unlike floppies, opticals are not susceptible to stray magnetic fields, nor do they suffer the hard disk problem of head crashes, since the optical read-write head is mounted far from the surface of the disk. The poor seek time suffered by some opticals is a drawback, but it will certainly have been overcome by next year.
Finally, optical's data capacity will soon overtake that of the hard drive. Optical drives holding 1GB are already with us (albeit in 500MB per side form). However, if Moore's Law holds true ("memory halves in price and doubles in capacity every year"), by the end of of '93, a 2GB removable 5.25" optical format should be available at half the current price. The same applies to the 128MB standard — reliable sources confidently predict that future 3.5" drives will offer 256MB and ultimately 512MB capacities, thanks to IBM's huge R&D investment.
The increase in capacity will be due to advances in both laser and optical media technologies, and developments in data compression. Hard and optical drives already employ RNL encoding, giving approximately 1.5 times the 'basic' capacity and a corresponding increase in data transfer rate. RNL works by looking ahead to see if subsequent bytes have the same value. If, for example, the next 300 bytes all have the same hex value (not unusual for audio data), those 300 bytes will be encoded as two bytes: the hex value, and a byte representing the number of times the value is repeated. In the future, greater compression ratios should be possible as cheaper, more powerful processors are developed.
The huge capacities that optical drives can provide should satisfy the future requirements of tapeless audio, graphics, and the new wave of multimedia and desktop video applications. In the professional video arena, tapeless non-linear video editors and computer graphics editors are having a huge impact. On a more domestic level, Apple are about to release QuickTime, an extension to the Mac operating system that makes it possible to replay video 'clips' (complete with a soundtrack) in a window on any 68030 Based Mac. (Microsoft have similar plans for 386/486 PC-compatible platforms). Though greatly compressed, video clips will still require hefty chunks of storage.
Back in the land of tapeless audio, the current limit of 4-track simultaneous record/playback per optical disk should improve, though not particularly quickly. Ideally, its best to keep the number of carriers per project as low as possible, but this is far less important than the advantages that tapeless recording provides.
In archival applications, optical's performance is, like DAT, unproven. Currently, IBM give a 7-year warranty on their 128MB drives. A more secure solution is provided by Write Once media, since they cannot be overwritten accidentally. However, for audio applications, CD-R (recordable CD), with its long recording time and rapidly falling cost may well fulfil this function. However, the format will need to evolve a 20-bit recording mode to keep pace with current trends in mastering.
If you press play on an analogue magnetic tape recorder and the reels don't budge, the problem is obvious and the solution liable to be relatively simple. However, if you press a tapeless recorder's play button and nothing happens, there are a myriad number of reasons why the thing won't work.
Perhaps the two most important are software bugs and drive incompatibilities. The latter is less of a problem for turnkey systems where, hopefully, all the component testing has been done by the manufacturers. However, for PC-based systems component incompatibility has been a major headache, since no manufacturer or dealer can possibly test every combination of drive and computer.
Hopefully, if the ANSI standardisation of IBM's 128MB optical drive is repeated with future higher capacity drives required for pro mastering and all multitrack work, the drive incompatibility problem may abate. The next step will be to devise standards for the digital audio files stored on the disks themselves.
Iomega (Contact Details).
AM Micro (Second Wave ViperDrive Floptical) (Contact Details).
Data Peripherals (Ricah) (Contact Details).
Panasonic UK (Contact Details).
DAC (MaxOptix 1GB) (Contact Details).
Feature by Kendall Wrightson
Previous article in this issue:
Next article in this issue:
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!