Until relatively recently some of the more upmarket digital sampling keyboard manufacturers had somewhat disdained the MIDI specification as not being up to the standard of their product. However, as it has proceeded to sweep tolerably successfully across the industry, all but a few are now including it in their instruments, and even Synclavier have a MIDI card in the pipeline although it isn't currently available.

The term 'synthesiser' is often used generically to mean simply an electronic musical keyboard instrument which isn't an organ: wrong, but then nobody likes a pedant.

For the sake of argument we can consider a synthesiser to be a device that can be programmed to create synthetic versions of real sounds, plus a variety of non-specific voicings — usually the product of an aberrated musical sensibility and wild and abandoned knob twiddling.

Pure sampling keyboards might be considered less as synthesisers, and more as 'modifiers' or 'extenders', because what they do is take recordings of real sounds and modify or extend them. In spite of this pleasing logic, however, the term 'synth' remains in general usage, and in fact many sampling instruments now include a standard synth anyway.

You will probably have heard of such names as Fairlight, Emulator, Kurzweil and Synclavier, but it is unlikely that you will have one of your own sitting at home. In fact, it's probable that the most expensive Synclavier package would actually set you back more than the cost of your house AND the furniture in it. Until very recently, sampling keyboards were the luxury of the wealthy, but with the introduction of the Ensoniq Mirage at the last Frankfurt exhibition, and now the Prophet 2000, the £2000 barrier has been broken, and even the moderately well-off can get in on the act.

So why the huge amount of money just for a few old recordings? One of the problems is that these digital samples (recordings) have to be stored in some form of internal solid-state memory; ie RAM or ROM. These memory chips, which can be considered as the digital solid-state equivalent of magnetic tape, take up a large amount of power and, although they are getting steadily cheaper, cost a great deal of money. The quality of the aforementioned samples is usually directly proportional to the amount of memory used to record them. To use the jargon — it is better to utilise a 16-bit system than an 8-bit system, and to record a wide bandwidth to allow all the details of the sound to be captured faithfully. Both of these provisions result in the need for larger amounts of memory, and as memory costs money, the price goes up.

For professional musical applications it obviously isn't acceptable to produce poor quality samples, but what is debatable is how long each sample should last. A transient sound, such as a bass drum, will obviously require a recording of very short duration. In contrast, sustained notes continue to resonate usefully for a good 30 seconds, and would therefore require a great deal of memory to capture in full. Even if 30 seconds of ROM/RAM per note was practical (and it isn't), what about reproducing an organ which can produce infinite sustain? No, no, I'm afraid some cheating is most urgently called for.

A primary source of cheatery is as follows: suppose that we record one second of an acoustic piano playing middle C and then on playback, when we come to the end of the recording, we return instantly to a point nearer the beginning and start playing it back again. In the relevant terminology, 'playing back' is called 'reading', and so the trick is to continually re-read the same section of memory over and over for as long as we require the note to sound. This is known as 'looping'. There is generally a bit of hiccup or 'glitch' at the point of 'flyback', but this can be largely mitigated by computer assisted editing, trying to ensure that the phase and amplitude of the beginning and end of the loop are comparable. If a sample is read out faster than it was written in (recorded) its pitch rises, and conversely a slower read speed produces a commensurate drop in pitch. Thus reproducing an instrument's complete register would theoretically appear to be a simple matter of taking a sample at any single given pitch and then reading it out at different speeds for different pitches. This, however, is not the case...

When looping an instrument like a piano, the volume of the note will have to be continuously decreased to approximate the instrument's natural envelope, and in fact the tonal quality of the note also changes with time. Not only is it difficult to anticipate the envelopes of a single note played on, for example, the piano, but worse than that, the pattern is different for each note on the keyboard and indeed for each different combination of notes. Thus it is not possible to sample a real piano just the once, at one frequency, and then digitally pitch change it to produce the notes for the rest of the keyboard. It just doesn't sound natural, and the further away from the original note you get, in terms of pitch, the worse the problem is.

Costs make it impractical to take a separate sample for every note, and so the general tendency is to mitigate the effect of pitch changing by taking as many samples as possible (ie as many as can be afforded) using pitch change only to fill in the gaps. It's rather a brute force approach, but it is quite effective. This technique is called 'multisampling'.

That isn't the whole problem, however. We still haven't considered the difficulties associated with 'touch response'. With most instruments, and especially the piano, the characteristics of both the dynamic and tonal envelopes are also contingent upon the force with which the note is struck: everyone knows a piano is mellow when quiet and bright when loud. Once again, this can to a degree be approximated by the use of filters and amplifiers to modify the sound in respect of key velocity and time but in reality this too is different for each note on the keyboard, and so with the exception of the Kurzweil 250 (see entry), no instrument can really lay claim to accurately replicating the true time related dynamic and tonal characteristics of an acoustic instrument such as the piano, and the fidelity of each voice is usually contingent upon the sophistication of the modifying system used.

Some instruments going under the banner of 'sampling synths' could be more accurately described as 'sample playing synths' since they don't offer the facility to actually record your own samples, but rather use real sampled sounds as the basis for a number of internal presets. This is becoming less common, and in fact the only instrument that might fit into that category is the Technics SX-PV10, details of which can be found in the 'Electronic Piano' section.

The coming of MIDI has meant that the same technology found in sampling keyboards can now be made available at more affordable prices in the form of modular or keyboardless samplers to be controlled from a separate MIDI keyboard — ie your existing synth. There are already a number of these around and they will be increasingly common in 1986.

E-MU II

The Emulator (I) was the first instrument to bring the price of sampling synths down to a reasonable level, and many famous names continue to swear by its updated version, the E-II. You have a total of 17 seconds internal RAM to be used as desired, whether that be in a single 17 second sample or a three second, 5-way multisample, or a number of different sounds, etc. Offline storage is via twin built-in floppy disks or a Winchester hard disk is available as a very expensive option. It has an unweighted plastic velocity sensitive keyboard with 8-note polyphony and standard VCFs, VCAs, EGs and LFOs to modify a sampled sound, statically or via the keyboard's velocity sensing. It's an eight-bit system with a sampling rate of 30k giving an audio bandwidth of 12k.

Earlier this year a new software package was offered free of charge to existing E-II owners which provided a means of using a variable amount of the existing sample memory to store sequences. It's a kind of trade-off whereby you lose approximately one second of sample time for every 5000 notes of a sequence. The same update provided MIDI and SMPTE time code facilities in addition to the existing RS-423 computer interface.

An optional software package, the Macintosh Sound Lab, allows the use of a Macintosh computer providing assistance with sampling and editing such as wave form displays and frequency analysis plots.

ENSONIQ MIRAGE

The Ensoniq represents nothing short of a revolution. A real sampling keyboard for under £1,695 that also includes its own 8-note polyphonic, 16 oscillator synthesiser and a polyphonic real-time sequencer. It has a sampling rate of 30kHz which can be used to give a single two second sample with a 15kHz bandwidth, a four second sample with 8kHz bandwidth or an eight second sample with 4kHz bandwidth. The VCFs, VCAs and LFOs associated with the synthesiser can also be used to modify the sampled sounds and there is an automatic or manual looping mode. An optional software package allows interfacing of an Apple II micro computer which graphically simplifies edit and looping procedures. The sequencer records key velocity, pitch bend and sustain pedal information as well as standard note details, and thus the standard 333-event memory is rather limiting. There is, however, an optional memory extension cartridge (£69.50) providing capacity for an extra 1000 events — still hardly gargantuan, but at least positively useful.

Sounds are stored on 3.5" micro disks, up to three complete keyboard samples per disk, which can include up to 16 splits/sounds, and layering. Although the disks are of a standard type, it is not possible to format them for use yourself, and ready formatted blank disks will set you back £14.95 apiece. An ever growing library of extra factory sounds currently consists of three disks including between them brass, string, drum, percussion and piano (electric and acoustic) voicings. Each of these is yours for a very reasonable £27.95. It isn't up to the E-II, but why should it be expected to be? For the price it is quite excellent.

FAIRLIGHT CMI SERIES III

The price of the Fairlight continues to keep it in the world of fantasy, or at least occasional hire, for most musicians. Purchasing this instrument can really only be justified if you are continually involved in ultra-high quality, complex multitrack production work in which case it is more or less a complete production tool. With this in mind you are given some extremely sophisticated facilities as standard. Apart from the spec given below these include a 16-track fully multi-timbral (up to 16 voices) sequencer; twin floppy disk drives; a 60 Megabyte hard disk; a high resolution VDU with light pen facility; a QWERTY keyboard for non-performance compositions with on-screen musical notation, playback and hard-copy print out; additive synthesis, and 32 library disks of factory sampled sounds — oh, and a full function word processor facility.

Although the wooden, weighted keyboard is only velocity sensitive, the system will respond to after-touch and key-release information via MIDI.

The MIDI facility includes software for interfacing with the Synthaxe guitar-style controller and can also work with SMPTE/EBU time code using 24, 25 or 30 fps rate, ie standard film or TV in Europe or America.

An incredible sound creation/sampling/processing tool with a price to match.

GREENGATE DS:3

An English sampling system based around the Apple II(e) personal computer that is eminently affordable — assuming you already have the Apple plus monitor and disk drives. What you actually get for your money is a small keyboard, plug-in PCBs for the Apple II(e) and some disk-based software which includes a powerful real-time sequencer capable of very high resolution (7500 events per minute) and several minutes of music — Greengate aren't quite sure how many events that is. It's a four-voice system with four separate outputs, a maximum sample time of two-and-a-half seconds, waveform editing on screen and the keyboard can be split as many times as you like or sounds can be layered. Via a MIDI retrofit card velocity and pressure information can be applied to the DS3's VCA's and VCF's though its own keyboard is actually static.

KURZWEIL 250

The now almost legendary Kurzweil 250 is presented to the world in a single sleek looking black package measuring 56"(L) x 27"(W) x 9"(H), and weighing a hefty 951bs; plus 221bs for the power supply.

Included, as standard is a 3500 note, 12-track, real-time sequencer (internally expandable to 9000 notes), with extensive editability: in addition to the normal note correction facilities, you can change the time signature, key, voices and speed of an existing recording and you can store your sequences off-line on disk.

In its basic form the instrument comes with 30 preset sampled voices available at the touch of a button or two. Rather than straight forward looping (explained in the introduction to this section) designer Ray Kurzweil has applied his experience as a past master of artificial intelligence to the problems of musical synthesis and has come up with a system he calls 'timbre extension', which is apparently some form of sophisticated looping using a computer-derived 'sound model' of the instrument to give unprecedented fidelity.

In the case of the famous acoustic grand piano sample, a recording of each note on the piano keyboard played at a number of different dynamic levels is fed into Kurzweil's 'Unspeakably Top Secret' 68000-based computer analysis system. The notes are recorded directly, via a microphone, on to a Sony F1 digital tape recorder to preserve, as far as possible, absolute fidelity. The computer is then able to draw the aforementioned 'sound model' of the instrument.

Referring to this model by means of 'pattern recognition' (one of Mr Kurzweil's main areas of expertise) the computer creates a unique and accurate characteristic for each note on the keyboard, including a fully dynamic touch response. This basic technique is obviously not limited to replication of the acoustic piano, and the various shades of brass, strings, woodwind and percussion are equally impressive.

The 'MacAttach' software package available to work in conjunction with an Apple Macintosh personal computer (which is also based on the 6800 microprocessor), offers a 'Sound Modelling' package that allows you to sample your own sounds up to 100 seconds at 5kHz down to 20 seconds at 25kHz. These are fully editable and loopable, with keyboard split and layering and optional velocity sensitivity, although the results are obviously a little less impressive than the factory-created samples. For the future we are promised such software packages as the 'Kurzweil Sound Laboratory', allowing additive synthesis and the 'Composition/Notation' package, providing an instant VDU display of what you're playing, and a subsequent printout of the same.

For the top professional keyboard player who doesn't want the hassle of wrestling with external computers, VDUs and wave tables etc, the basic 250 has to be the perfect package.

PPG DIGITAL MUSIC SYNTHESISER

The complete system comprises four components:

1. The Wave 2.3
The Wave 2.3 can be used in isolation as a MIDI compatible, eight-voice digital synth using standard analogue-type wave forms such as sawtooth, square and sine etc, plus a massive wave table of 1800 complex wave forms that can be called up and edited to create the desired result. A built-in real time sequencer allows up to eight different voices to be used in a single sequence. The new MIDI software now works in the Omni-off Poly mode and allows pitch bend information to be transmitted. The keyboard split cannot be transmitted over separate MIDI channels as it can on the PRK-FD.

2. The EVU (Expansion Voice Unit)
This is basically a 19" rack mounting modular (keyboardless) 2.3, also with MIDI facilities, offering a further eight dual-oscillator voices and the facility to store sampled sounds. Sequences written on the 2.3 can be down loaded into the EVU allowing it to drive itself.

3. The PRK-FD
This is a MIDI compatible controller with a 72-note velocity sensitive, wooden weighted-action keyboard, designed to use sounds from the 2.3 or the EVU. The new version now has its own floppy disk drive allowing sampled sounds, sequences and MIDI parameters to be down loaded so that it can be used in isolation and/or as a mother keyboard for a MIDI system. The PRK-FD will work in the Omni-Off Poly mode and can send the two sides of its split keyboard on separate MIDI channels.

4. The Wave Term B
Optional central control unit of complete system storing and distributing data around the system and also providing VDU displays to aid editing and sound composition.

360 SYSTEMS

The 360 Systems Digital Keyboard is effectively a digital version of the Mellotron — without the hassles, man. The old Mello was cool. Yes indeed it was a hot little instrument, but it had internal problems. It was sensitive, delicate, temperamental and sometimes just couldn't seem to get its heads (playback) together: much like the long haired hippy types who played it, in point of fact.

The main difference between the old and the new is in the recording medium. Rather than tape, the 360 uses digital memory chips, commonly known as ROMs (Read Only Memory), offering a 16k bandwidth. An original recording of an instrument is made on tape in a normal, though high quality, recording studio, and this is used as the master from which the ROMs are copied.

Inside the 360 there is room for up to 5 printed circuit boards (PCBs) each of which can hold a maximum of 48 ROMs. The number of chips required for each sound is determined by the length of the sample and the number of multi-samples taken across the keyboard.

The decision makers at 360 have no truck with the looping of samples, saying as they do, that it is detrimental to overall musical quality. Consequently, at the end of the prescribed sample period, there finds itself a bleak and pointed silence.

One of its greatest assets, and perhaps limitations too, is its operational simplicity. However, unlike other (cheaper) sampling keyboards the 360 doesn't allow you to record your own sounds into memory, but it does offer compactness, relative ruggedness, general portability and an excellent range of well recorded sounds with 8-note polyphony. Another drawback is that there is no touch sensitivity, although the new software now allows touch information to be transmitted through the 360 when it's being used in a larger MIDI system.

It is possible to use two voices at once in either the 'Stack' mode, where any key pressed will produce both sounds at once, or in the 'Split' mode (with assignable split point), wherein one side will play one voice and the other will play the other voice to allow us to try it out.

In the light of the Ensoniq Mirage and the Prophet 2000, the big drawback with this instrument is its price. The basic machine is £3,680, and you then have to amass your software, and if you're interested in including the strings and the acoustic grand piano (and who isn't), you can probably add another few grand to that. Prices range from an electric piano at £66.70 to the violins or the grand piano, either of which will witness the disappearance of a further £833.75.