Things You Should Know About Hard Disk Recording

Hard disk recording is what everyone seems to be talking about, with new systems being released, at all levels of the market, almost every month. For the benefit of anyone confused about how hard disk recording works, Martin Russ points out a few things you might not have considered...

The wave of the moment is hard disk recording — the 'multitracker' of the '90s. Instead of recording audio onto magnetic tape, computer memory chips and hard disk storage are used to hold digitised audio signals. There are advantages to doing this: the limitations of physical tapes are transcended — you can re‑record the same segment of audio without waiting for a tape to rewind, and without worrying about wearing out the tape through over‑use or destroying it accidentally. In the digital domain there's no need to ever cut or splice a tape, and so there are no problems with sticky splicing tape, dropouts, shedding...

But there are disadvantages too. Because the technologies used in hard disk recording are so different from conventional audio recording, it is all too easy for the unwary to overlook the traps. So how do you find out about the other side of hard disk recording? Read on.

It's About Time

Magnetic tapes follow a convention that is so obvious and subtle that you may never have even considered it, because it looks like plain common sense. When you buy a tape, it has a length and a width: the length determines how long the tape will play for, whilst the width of the tape limits how many audio tracks of a given quality you can have.

How long the tape will play for can be calculated by knowing the length of the tape and the speed at which it plays — although some tapes are commonly known by their time 'capacity' at a specific speed. Cassettes are an everyday example of this use of time to describe the length of a tape: a C90 lasts 90 minutes. Calling a cassette a C90 is really a way of saying in short that the tape has enough length to provide 90 minutes of recording at the standard cassette tape speed. Double the speed of the tape passing across the record/playback head and a C90 cassette becomes a C45; halve the tape speed and it becomes a C180. This time‑squashing is familiar to users of cassette multitrackers which can run tape faster to improve audio quality.

When we're talking about magnetic tapes, their capacity can be expressed using their recording time and the number of tracks they provide. An 60‑minute half‑ or quarter‑inch open‑reel tape used for 8‑track recording will be able to record eight separate tracks, each lasting one hour. This actually gives a total recording time of eight hours, but it is very unlikely that you would ever hear anyone refer to it as an eight‑hour tape! The reason is that the tracks are all tied together physically — when you play the tape, the tracks all play simultaneously. You can't take the tape and play it continuously for eight hours — at least, not without stopping it and reversing direction, or rewinding!

But for hard disk recording, this 'obvious' relationship between tracks and time does not exist in anything like the same form. The figure which is quoted for the 'record time' may be the total time that can be recorded across all the tracks — which is the equivalent of describing an 8‑track, 60 minute tape as an eight‑hour tape. The reason this figure is sometimes quoted for hard disk recorders is because it can be easily determined from the size of the hard disk and the data rate of the digital audio.

One method of reducing the demands made by a hard disk recording system is to reduce the amount of data which needs to be stored.

For example, suppose we're recording mono, CD‑quality audio — one track of 16‑bit sampled audio, at a sampling rate of 44.1kHz. At the time of writing, the price of 1000 Megabytes (1Gigabyte) of hard disk storage was under £200, and it's possible to calculate that this would provide over 11,000 seconds of digital audio storage, which is just under 189 minutes, or about three hours and 10 minutes. The 'record time' calculation represents the idealised total capacity of the hard disk to store mono digital audio. As the number of tracks increases, so the time goes down to match. Stereo audio requires two tracks, so the capacity of a 1Gb drive is just under 95 minutes. For four tracks, the calculations give a capacity of just 47 minutes.

In order to emphasise this correspondence between time and number of tracks, the term 'track minutes' is being adopted by some manufacturers as a way of describing the capacity of a hard disk recorder. A 'track minute' is exactly what it says: a minute's worth of recording on one track. So a machine which has a capacity of 60 track minutes will be able to record 60 minutes on one track, or 30 minutes on two tracks, or 15 minutes on four tracks, or even 7.5 minutes on eight tracks.

To complicate matters even further, whereas with a stereo tape recorder you can reasonably expect to consistently have an audio signal on both tracks throughout the length of the track, with a multitrack tape recorder this is not necessarily true. On a simple 4‑track recording, two of the tracks might contain the bulk of the audio material, with the two remaining tracks used for additional parts, double‑tracking, and so on.

If the hard disk recorder only allowed the allocation of time to tracks on a strict capacity/number of tracks basis, there would be wasted capacity whenever you didn't have an audio signal on a track. A much better solution is for the hard disk recorder to allow the flexible assignment of time to tracks. In this case, the figure of 15 minutes for four tracks might well be larger: perhaps 20 or 25 minutes depending on how fully the tracks are actually used. The same applies to eight or more tracks, where utilising the unused time within the tracks might allow 10 minutes or more of total record time in a system which only has the capacity for just over seven minutes of fully‑used tracks. This is very different to how an analogue tape recorder works.

The Numbers Game

One of the ways in which cassette multitrack manufacturers manage to produce versatile portable studios at low cost is by reducing the number of tracks which can be recorded simultaneously. Many of the budget 4‑track machines can only record two tracks at once, although all four can be replayed together. Almost all other tape recorders do not suffer from this limitation — the ability to record all the tracks at once is almost taken for granted.

A single mono digital audio track, recorded using 16‑bit samples at a rate of 44,100 per second, requires the hard disk recorder to move 88,200 bytes onto the hard disk every second. Four tracks requires 352,800 bytes per second to be moved, which is approaching the typical continuous throughput of an ordinary SCSI or IDE hard disk interface. So recording eight tracks or more at once may well require some more advanced interfacing — which is fine inside a hard disk recorder box recording to an internal hard disk, but may cause problems for an external hard disk connected via SCSI. This can mean that there is a physical limit to the number of tracks that you can record simultaneously on any given system — due solely to the huge amount of data which needs to be moved around.

One temporary solution to this limitation is to record into the read/write (RAM) memory instead of to a hard disk. Although possible, this requires large amounts of RAM, so it severely restricts the length of time which can be recorded without saving to hard disk. The 4‑track, 16‑bit, 44.1kHz sampling‑rate system outlined above needs just over 1Mb of memory to store every three seconds of audio, so 8Mb will yield 24 seconds of 4‑track audio, 16Mb will give almost a minute, and a typical 'pop' single would probably fit into 64Mb — with the RAM costing around £500. Once stored in RAM, the digital audio could then be saved to hard disk, thus freeing up the RAM for more recording.

Compression

One method of reducing the demands made by a hard disk recording system is to reduce the amount of data which needs to be stored. In a computer context, this is called 'compression', and it is very different from the compression that you might apply to a vocal track, or the compression that happens when you record aggressively onto analogue tape. There are two basic forms of computer‑type compression: lossless and lossy.

• Lossless compression squashes the data in some way so that it can be stored, and then recovers all of the original information again when it is read back. Although this sounds impossible, there are several methods which can achieve it. The underlying aim is to exploit the characteristics of the data. For example, suppose we were noting down the number of cars which crossed a bridge every hour. The first few hours might read like this: 312, 294, 287, 301, 295... and so on. If we take the first figure, and then note down the differences between successive totals, then we get 312, ‑18, ‑7, 14, ‑6. The first number is the same, but all the rest are much smaller. By taking the first number and adding or subtracting the increments, it is possible to restore the original numbers exactly. Audio data has properties which can be exploited by compression schemes — for example, it turns out that the change between successive bytes of audio data is often quite small, so if differences are sent instead of complete sample values, the amount of storage required can be significantly reduced. More sophisticated schemes look for patterns of data, and then code these with shortened 'pointers'. But each process is designed to save all of the original set of data, and subsequently restore it completely. The audio data which is saved to the disk is recovered unchanged, and so will sound exactly the same as previous playbacks before the save to disk.
• Lossy compression attempts to store just the significant parts of the data, and then to recover a close approximation to the original. Many of the techniques for lossy compression are similar, in that they exploit the characteristics of the data in some way, but decisions are made about just how much detail needs to be stored. Using the 'cars crossing a bridge' example, if the information required from the counting of the cars was the peak number of cars, then only one number needs to be stored: 312. Alternatively, if only the total number of cars crossing per hour is required, then we could approximate the hourly counts to the nearest 10 or even 100 cars — '300 cars cross the bridge every hour' may be perfectly adequate. For audio data, some parts of the waveform or the spectrum might well be superfluous to the audible content, and so can be mostly discarded. This is exactly the type of processing that is used by Minidisc and DCC (Digital Compact Cassette) — the most prominent sounds are stored with the most detail, while any background sounds that they 'mask' are not stored with the same depth of detail. When designed carefully, lossy audio compression techniques can provide significant reductions in the amount of data which needs to be stored, whilst remaining almost inaudible to the listener.

With any audio compression system, your approach should be to audition it closely to see if it is acceptable for your music. Unlike the clipping distortion that you may be familiar with in digital systems, compression can introduce much more subtle changes to the sound. If you want to hear these effects, you'll need to make A/B comparisons with a range of material to become familiar with them — and try to concentrate on the background, not the foreground sounds: listen out for a slight coloration.

Get Your Backup

As you know, an analogue tape recorder stores its audio on a removable tape. Hard disk recorders store audio on internal hard disks which normally can't be removed and stored away. So what happens when you've filled the hard disk with music?

Hard disks tend to fill up more quickly than you expect, and they can occassionally fail catastrophically. In order to prevent the accidental loss of painstakingly recorded and edited digital audio data, or just to save it so that you can put something new onto the disk, its contents need to be stored on at least one other alternative storage medium.

Hard disk recorders offer two options for saving digital audio. The simplest and lowest‑cost systems provide access to the digital audio only — so the only way to save the contents is to record them onto DAT using the digital I/O sockets. This is truly 'real‑time' storage, where the time taken to store the digital audio is about the same as half of the total length of all the tracks — assuming that the data is stored using the two tracks of the DAT as two mono audio tracks. Re‑loading data backed up in this way will also take the same amount of time. For example, if a three‑minute single is made up of eight continuous mono tracks, each lasting three minutes, it will take at least 12 minutes to store onto DAT tape, and the same to re‑load.

If the hard disk recorder does provide direct access to the hard disk, this will almost certainly be in the form of a SCSI socket. SCSI (Small Computer System Interface) sockets allow you to copy the entire contents of a hard disk to another storage medium. This could be a data tape, another hard disk, or an optical disk.

The DDS (Digital Data Storage) DAT is one of the most popular data tape formats. It offers storage of at least a couple of Gigabytes on a 90‑minute tape. Tape storage can be slow: a 90‑minute DDS DAT tape will take at least 90 minutes to fill — and probably more because of how files are written to the tape and then verified.

Hard disk drives can be fixed or removable. With a removable drive, you buy the case, the power supply and the interfacing circuitry once, and just change the hard disk. With a fixed drive, you buy the whole lot every time. The capacities and performance of removable drives are increasing all the time, with 1Gb removables costing only just over twice the price of a similar fixed drive — and buying a second 1Gb cartridge is much cheaper than buying a second fixed disk!

Optical drives store data on something very similar to a CD‑R. Until very recently, optical drives have suffered from a lack of standardisation, although the forthcoming DVD (Digital Video/Versatile Disc) may sort this out and should provide just under 5Gb initially, with the promise of 15Gb longer‑term.

Conclusion

Despite appearing to be nothing more than a computer‑based version of a tape recorder, hard disk recording is actually very different, and understanding these differences can help you to avoid expensive mistakes. Because the technology of hard disk recording is new and expensive, a range of alternative approaches are appearing, which are intended for specific applications. Trying to use one of these application‑specific products in another way can be frustrating and occasionally disastrous — but you're now armed with some appreciation of the potential pitfalls and hopefully, you can avoid them!