PART 1: If you have to decide what to delete on your PC before installing a new program, it's time to consider a bigger disk drive. In the first of this two‑part series, Martin Walker explains why size isn't all you have to consider when making your choice... This is the first article in a two‑part series.
If you are thinking of replacing your current PC hard drive, or indeed adding an extra drive, two factors will immediately spring to mind — speed and size. Before we get on to the specifications which determine overall performance, let us firstly address the size issue.
Size Matters
Most modern multimedia PCs will come with a 1Gb hard disk (with 850Mb being the absolute minimum currently supplied). Even if you are careful in deleting unwanted applications, demos, and memory‑hogging games from your PC, you are still likely to fill at least half of this gigabyte with the Windows 95 operating system, normal everyday applications and data. If you only intend to fly in occasional tracks to accompany MIDI data with an all‑in‑one MIDI + Audio sequencer, this should be fine. However, if you are thinking of hard disk recording, then at 5Mb per minute for a single mono track of 44.1kHz digital audio, the remaining 500Mb will only allow 12.5 minutes of 4‑track recording. Even for basic mastering of a CD album in stereo, you will need at least 600Mb for the final result, and to leave any space at all for editing, 1Gb of space is more sensible.
If you want to run the equivalent of an 8‑track system (such as ADAT), each mono track recorded at 44.1kHz will occupy about 300Mb for an album lasting an hour. So, for an hour of 8‑track recording (where all eight tracks are on the go simultaneously) you would need a drive of 2.4Gb capacity. In more typical situations, when the eight tracks are not all full, you would probably get away with 2Gb. However, skimping on hard disk size is very much a false economy, especially with drive prices falling all the time. The last thing you want to be thinking about when adding the final track to your latest masterpiece is whether you dare delete a few unused takes that you think you won't need again, just to create enough room for the final touches.
Many people's impressions of hard disk recording on a PC are somewhat blurred, due to the wide range of choices available, from simple systems using a low‑cost soundcard (which will provide two to four tracks), to high‑end alternatives that offer multiples of eight tracks with dedicated hardware in external cases. As a rough rule of thumb, if you intend to use any more than four tracks, you would be well advised to look at one of the intelligent soundcards, such as the Turtle Beach Tahiti, which provides enough processing on the soundcard itself to remove a significant portion of the workload from the main computer processor when recording and playing back large amounts of audio data. If you need more than a single stereo output to allow separate outboard EQ and processing for each audio track, whilst it is possible to set up two or more soundcards simultaneously in the same PC, this route can be fraught with problems (see my feature on this in last month's SOS). At this level of performance, you should seriously consider a purpose‑designed system, such as Digidesign's Session 8 or Audiomedia III.
Drive Performance
When choosing the hard disk drive itself, two main factors need to be taken into account — sustained transfer rate and access time. Transfer rate is a measure of how fast the drive can read and write data, but commonly this is quoted in 'burst' mode. Many typical computer applications have a lot of disk activity in a short space of time followed by a lot of computation (for example, graphics packages and word processing). By incorporating memory caches in the disk assembly, short‑term performance can be greatly increased when many small files need to be read or written. One typical use of hard drives by musicians is for hard disk recording, where many reads and writes are carried out using much larger total amounts of data over long continuous periods of time. In this case, onboard caches are less useful, and the manufacturer's 'Sustained Transfer Rate' will give a much better idea of overall performance. For typical applications, a figure of 1Mb/second is a good target to aim for, but the faster the better! 'Access Time' is the time taken by the drive to locate where a particular piece of data is stored. With hard disk recording, the data for up to eight tracks needs to be accessed before your software mixes everything down to a single stereo pair for output through a soundcard, which means that the drive read heads will be shooting off all over the place — so it is important to look for a low access time. Again 'Average Access Time' is the figure to look for, and although 18ms is good enough, if you want to scale the heights of multitrack audio performance, 10ms is more appropriate.
AV Drives For Musicians
When choosing a hard drive, one other thing needs to be taken into account in addition to sustained transfer rate and access time — whether or not to go for a drive specially tailored for AV (Audio Visual) applications. All drives have to pause periodically to carry out what are normally referred to as 'housekeeping chores'. These involve error recovery and thermal recalibration as the unit changes in temperature. Most drives perform this recalibration about every five minutes for the first 30 minutes after switching on, and once every 25 minutes or so after that. Normally, this will occur during gaps in the data stream. In continuous AV usage, these gaps do not occur, so periodically a short hiccup of anything up to about a second can occur, which can cause momentary halts to recording and playback. This may seriously affect the sustained transfer rate, so various disk drive manufacturers (particularly Micropolis) have designed special AV drives that have had their housekeeping intelligently optimised for continuous recording and playback, with pauses of typically no more than 50ms at any time.
In the real world, most people make do with standard drives and get away with it most of the time by using a decently‑sized memory buffer that can cope during the drive pause, but with the price of drives falling all the time, it is sensible to consider an AV type. For high‑end hard disk audio (eight tracks and above, with multiple outputs) most dealers will fit AV drives in new systems, as for the small extra outlay it is not worth the risk of occasional glitches. In general, for semi‑pro use, a non‑AV drive will probably be OK — after all, even if you do suffer a glitch, the next time you run the software it will be probably be fine, and you will forget all about it. However, for professional use, choosing a non‑AV drive for a saving of under £50 is simply not worth it.
Drive Interface Types
Today, there are only two drive types to consider — IDE and SCSI. Most PCs come with IDE (Integrated Drive Electronics) drives as standard. These incorporate the electronics needed to control the drive as an integral part of the assembly, so that the drive can normally be plugged directly into modern motherboards. Up to about 1994, there were limitations for IDE drives that caused problems for AV users. The maximum drive size was 504Mb for a variety of reasons, and it was also difficult to have more than two drives in a single PC system. These limitations have been overcome with the more recent EIDE standard (Extended Integrated Drive Electronics), which allows up to four simultaneous drives, and a maximum capacity of 7.88Gb. Nearly all modern motherboards will incorporate EIDE capability as standard, still referring to it confusingly as IDE, but it is worth checking which version you have if you are trying to add a large drive to an older PC.
When choosing the hard disk drive itself, two main factors need to be taken into account — sustained transfer rate and access time.
SCSI (Small Computer Systems Interface) has traditionally been associated with high‑end applications. It has the distinct advantage of allowing up to seven devices to be attached, with none of the size restrictions of the old IDE specification. With the profusion of devices now sprouting from the average PC, such as CD‑ROM and CD‑R drives, backup tape drives, scanners, and removable drives, even the four devices allowed in the EIDE specification may seem restrictive. But the biggest advantage of the SCSI buss is that it allows devices to be plugged in externally as well as internally, and this has resulted in many musicians buying small external removable drives such as the Iomega ZIP, with its 100Mb capacity. This is so portable that you can carry your drive and several data cartridges with you, connecting to any other PC easily and reliably. The disadvantage of SCSI is that it means the purchase of a separate controller card, which sits in one of the PC's valuable slots. SCSI drives are also always a bit more expensive than an equivalent EIDE model.
Comparing performance of the two types can be tricky — many people mistakenly consider that SCSI drives are significantly faster than EIDE types. However, since many manufacturers have identical drive units in both versions, it turns out that the additional overhead of going through the SCSI buss will result in the EIDE version being slightly faster. It is the SCSI buss that allows transfer rates at up to 40Mb/sec, not the drive! In high‑end systems, SCSI devices have the distinct advantage, since they have embedded disk controllers that can function independently of the main processor, and simultaneously with each other. Although IDE drives have similar controllers on board, they cannot operate simultaneously, and so multiple drives in an IDE system will work one at a time.
Interface Performance
All hard disk controllers and SCSI adaptors require the use of one or more of the following PC resources: ROM addresses, IRQs, DMAs, and I/O port addresses. The individual settings for specific devices are many and varied, but the performance implications for the different types of data transfer are worth explaining in more detail, since they will directly affect the ultimate performance of the drive. DMA (Direct Memory Access) is a means of transferring data directly to and from memory without tying up the main computer processor. Although DMA transfers are normally highly efficient, machines with ISA slots have slow DMA controllers on the motherboard, and so most drive interfaces now use a technique called Programmed Input/Output (PIO), which sends bytes through the I/O ports. This type of transfer is normally faster than DMA, especially in its more recent forms which support block‑mode PIO, allowing the transfer of multiple blocks of data with only one interrupt to the computer processor. Any adaptor that needs no DMA setting will normally be a PIO type.
There are five Modes of PIO supported by the EIDE specification, numbered 0‑4, and the highest supported by a drive will normally be printed on its packaging. Most high‑performance modern drives support PIO Mode 4, which offers a maximum transfer rate of 16.6Mb/sec (note that this is the theoretical maximum — the actual transfer rate is determined by the drive specification). This fastest mode must, however, use a PCI connection, and on many computers, only one of the two drive connectors are connected to the PCI buss, the other being an ISA version which will only allow up to Mode 2 operation, giving a maximum transfer rate of 8.3 Mb/sec. If you intend to use several EIDE drives with your PC, bear this possible limitation in mind. Many SCSI drives also use PIO as the method of data transfer, but higher‑performance SCSI devices use buss master DMA, which effectively takes control of the DMA buss and overrides the normal DMA process. Transfers using this technique can be much faster than PIO transfer.
Using Multiple Ide Drives
Setting up a single IDE drive is extremely simple; you normally 'plug and go'. If you intend to use two drives, one for general use and the other specifically for hard disk recording, several complications may ensue. Jumpers (small switches) on each drive may need to be set, so that one is designated the master (containing system files needed to boot up the computer) while the other is the slave. Sometimes, the drive on a slave setting will be delayed for several seconds at startup to allow the main drive to get going first, thus reducing the initial loading on the computer power supply. In the past, certain compatibility problems made older IDE drives temperamental about working with other drives, especially if each was from a different manufacturer — in the worst cases, the computer would fail to boot up at all. Most recent drives will have none of these problems, and installation will consist simply of connecting the appropriate leads and setting one jumper to either master or slave.
Partitioning
To prevent your computer's operating system storing a new file on top of an old one, a hard drive maintains a table of which file is currently using which area on the disk. This is known as the File Allocation Table, or FAT. Modern versions of DOS allow up to 65,536 (or 64K) file entries, and each can occupy one or more Allocation Units (also known as clusters) of hard disk sectors. The size of each cluster is determined by the size of the drive itself. With a typical 1Gb modern drive, a maximum of 64K possible files results in a minimum file size of 16K (1Gb divided by 64K). This will mean that in the worst case, a 1‑byte file will still occupy a single cluster of 16K on the hard disk.
If you look at typical file sizes on your drive, you will probably find hundreds that are less than 16K in size.
All of this means wasted drive space, and this is partly why compressed drive software (see the 'Neat and Tidy' box) can be so effective. The problem has got worse as hard drive sizes have increased, since with each doubling in drive size, the cluster size also doubles. In one classic case, someone upgrading from a 500Mb drive to a 1Gb version found that after copying his files across, they took up 700Mb, simply because so many suddenly occupied double their previous space on the disk. A new version of the FAT (FAT32) will appear in a future upgrade to Windows, which will apparently tackle the problem more intelligently, with an unlimited number of file entries, and 4K clusters with any drive up to about 8Gb.
One more thing is definitely worth bearing in mind — in the analytical atmosphere of the recording studio, there is nothing worse than a whining hard drive (I know, because I have one!).
In the meantime, there are two ways to get back some of the wasted space. Firstly, by 'zipping' up complete directories of rarely‑needed files, not only will the files themselves be compressed, but the disk space will be used more efficiently, since tiny files will use far less space. You can quickly 'unzip' these directories if they are needed again, but have more disk space in the meantime. Say you have a 1Gb drive (16K clusters), and in one directory you have 64 files that are 256 bytes or less. Normally, each would occupy a single 16K cluster, so that the space taken up by all the files would total 1Mb. By compressing the contents of the directory, the total file size will still be under 16K, which would only occupy a single 16K cluster with space to spare; a saving of 1008K!
A second solution is to divide up the drive into two or more separate partitions by reformatting the disk, and then each partition will have a smaller cluster size. Each partition is treated by the operating system as a separate drive with its own individual letter (for example, your drive C: could be partitioned into two logical drives, C: and D:). Although having separate partitions is a bit less convenient than using one large drive, it does have another benefit for musicians. It is far easier to use a large separate partition for hard disk recording, since file sizes here are normally high anyway, and also defragmenting is easier and faster to do, since there will be far less files. Backing up data is also easier to keep track of, since it is separate from the operating system files, day to day programs and their data.
To Boldly Go...
It is difficult to recommend specific drive models, as manufacturers are constantly releasing new models at ever‑decreasing prices, and with ever‑increasing specifications. If you intend to use a hard disk with one specific package, look through the manual, help files, any web site maintained by the software manufacturer, and other product information such as catalogues and brochures, to see what recommendations they make for hard disks. There is no sense in re‑inventing the wheel when manufacturers may already have a list of suitable drives for use with their product, and if you aim at equalling or bettering the specification provided by the software package you intend to use, you won't go far wrong.
One more thing is definitely worth bearing in mind — in the analytical atmosphere of the recording studio, there is nothing worse than a whining hard drive (I know, because I have one!) Although you can place the computer in a padded case (as long as there is still adequate airflow for cooling purposes), it is difficult to move it far from the monitor and keyboard (but see our 'Silent Running' feature elsewhere in this issue for some suggestions!). If you are buying a drive from a retail outlet, ask to hear a machine with the same drive already installed to obtain some idea of its noise output. You may get some funny looks, but you will be the one living with the beast for the next several years.
In next month's concluding part, I will look at the implications of 'going SCSI', its advantages, the extra cost involved, and how to choose the best SCSI controller card for your application. I will also try to make some sense of the mass of often conflicting information about SCSI types that can be bewildering to newcomer and old‑timer alike.
Neat And Tidy — Defragmentation And Compression
Although this has been mentioned on many occasions in SOS, it is well worth repeating. During everyday use of any hard drive, many files will be created and then deleted. The operating system will save files in the first empty space it finds on the hard disk, and if the file size exceeds this space, it will find another empty space and continue saving. With drives that have been in use for some months, a single file may end up split into many separate sections scattered throughout the disk, and the larger the file, the more likely it is to end up in this state. Writing a file under these conditions will obviously take a longer time, due to the additional overhead of jumping between the individual sections, whilst reading the file will have the same performance penalty. When you defragment the hard disk, using for instance the disk defragmenter utility supplied with Windows 95, the entire disk is simply scanned and everything shuffled so that each file ends up in one neat packet. In hard disk recording systems with four tracks or more, defragmenting can produce a considerable reduction in the time it takes to replay tracks, so regular defragmentation is a must if you intend to get the most out of your system.
Many people use DriveSpace, the disk compression software supplied with Windows. This simply acts like a separate drive with a capacity larger than its actual size, and works by compressing all files in a similar way to programs like PKZIP. With the large increases in hard disk sizes coupled with price drops, this technique is probably not needed quite as much as it was several years ago, but increases in drive capacity at no extra cost will always seem desirable. The disadvantage is that compression and decompression take a finite time, so your data transfer rate will drop slightly. Despite its popularity, some people are still suspicious of trusting their data to manipulation of this sort, and prefer instead to zip up particular files as and when needed, rather than commit themselves to anything more general. Whatever your personal preference, there is a performance penalty, since compressing files before saving and expanding them again when reloading will certainly take a finite amount of time. For backup purposes using removable drives, this is not a problem, but for high‑performance audio, never use a compressed drive unless you are prepared to accept a reduced transfer rate.
Formatting
There are three major steps to completing a format on any hard disk. Low‑level formatting is normally carried out by the manufacturer before you buy a drive, and involves the physical laying‑down of tracks and sectors on to the disk. At the same time, any defects (bad sectors) are marked so that the computer will not try to use these areas. Contrary to popular opinion, no disk drive is perfect, and each will have its share of defects. To some extent, a drive with more marked defects will have undergone more stringent quality control. Most users are unaware of these, since most IDE and SCSI drives employ a technique called 'sector sparing'. Although a drive is nominally a certain size, say 1Gb, in fact there will be slightly more capacity on the disk. This means that any defective sector can be replaced with a spare one from the additional space, so that the user seems to have a defect‑free drive which reports no bad sectors.
A low‑level format is not usually necessary, unless you wish to totally remove any trace of previous data before selling a drive, recover from data corruption after a nasty crash, or remove all traces of a bad virus infection. Once you have carried out a low‑level reformat, if you wish to partition your drive into several separate areas (see the section of this article on partitioning for more details) you should use the FDISK program supplied with Windows and DOS. This will write special information to the boot sector of the drive, so that DOS and Windows can then treat the drive as several smaller ones, each with its own letter (C:, D:, E: and so on). Once any partitioning has been done, the Format command can be used to carry out the high‑level format which writes the FAT and directory structure for each partition, so that files can be managed. This is what most people normally understand as 'formatting a disk'. If you are using SCSI drives, the manufacturer will normally provide specific format utilities, which must be used for low‑level formatting, but for high‑level formatting, the normal Windows/DOS command can be used instead.
Problem Areas — Using Scandisk
Many people are happy to leave their disk drive to get on with things, and would never think of using a repair utility unless they received an error message declaring a missing or unreadable file. In fact, regularly using a utility like ScanDisk (provided with all versions of DOS 6 and Windows 95) can help maintain your drive, and spot any errors at an early stage before they get out of hand. Some authorities even suggest placing ScanDisk into the StartUp directory so that it runs every time you boot up your machine. I prefer to run it about once a week, and certainly every time at reboot after any program has crashed. It is not only corrupted files that can cause problems — occasionally an entry in the FAT (see the section on partitioning elsewhere in this article) can get cross‑linked — ie. two files end up with the same directory entry. ScanDisk can rectify most problems like these, giving you the option of deleting the offending data, or more sensibly converting it into a file that can be read by any text editor, so that you can see the contents to determine whether there is anything of use to salvage. Lost file fragments (any data without a corresponding directory entry) can also be recovered. Although ScanDisk is effective and free, many professionals would recommend a commercial program like Norton Utilities for extensive faultfinding and major problems.