E-mu Systems Emulator
The rest of the '80s looks set to be the age of the musical biopsy, courtesy of machines dedicated to sampling your favourite miaow / bark / scream / cackle / moo (delete where appropriate) and letting you play with such more or less musical inputs until the synthetic cows come home. Moving to the analogies department, it's a bit reminiscent of the "Carry on Screaming" film where a severed finger (a digital sample...) was regenerated into a fully-fledged hairy monster after a judicious jolt from the National Grid. And whilst instruments like the E-mu Emulator aren't exactly capable of making something out of nothing, they are capable of making a lot out of a little, whether it be an entire string orchestra from a single violin, or a dawn chorus from a single, solitary tweet.
The principle of digital sampling is fairly straightforward: firstly, a sound is 'caught' by converting a certain length of sound into a stream of data via a fast A/D converter; secondly, this is stored in a suitable chunk of RAM and loop points are set up so that a particular segment of the 'captured' sound is repeated when a key is held down; then, thirdly, the data is 'released' out of memory and passed through a D/A converter at rates dictated by pitches entered from a keyboard. As far as putting this theory into practice is concerned, the Fairlight CMI beat everyone else to the goal of high quality sampling, and the reputation that it now enjoys is thoroughly justified. The Emulator, on the other hand, is much cheaper (£4,750 for the 8-voice version, as opposed to £17,500 for the Fairlight), much easier to use, and produces sounds of stunning quality.
E-mu Systems don't have quite the fame (and fortune) of Sequential Circuits, but they have been making their name felt (and heard) for the past 10 years in the States. E-mu's founder, Dave Rossum, moved from working on the biochemistry of ribosomes to designing and manufacturing modular synthesisers with the idea of making the best instrument that money could buy. His modular systems certainly are superb and their use in many studios attests to this. They're also very expensive; the Audity, a computer-controlled polyphonic synthesiser, with dual floppy disks, 16-channel polyphonic keyboard/sequencer with its own disk for sequence storage, and 16 analogue voice cards, costs the princely sum of $69,200!
Dave Rossum was also responsible for much of the development of the Solid State Music series of synthesiser chips and some of the design work behind the Prophet 5 and Oberheim synthesisers. The Emulator is, therefore, something of a radical departure for E-mu Systems and represents their first wholly digital 'synthesiser' (or, more strictly, a 'reproducer'). Modular systems are obviously tricky breadwinners, and the attention that the sampling capabilities of the Fairlight was getting convinced Dave Rossum that the Emulator should be something that the electronic music community would take to.
The design of the Emulator is nothing if not rigorous, and the sleek, grey, roadie-resistant, steel case shows this — even down to the nice touch of a recess in the top for storing floppy disks. The keyboard is a standard, 4-octave unit and, on the left hand side, there's some reassuring Moog-type modulation/pitch bend wheels. In the same area, there's also a 5¼" disk drive that's angled for easy disk insertion, and the controls for the polyphonic sequencer. The rest of the controls are well laid out with plenty of LEDs for status indication and so on. Inside, the same quality of construction is continued (though we can't show it) with a main processor board (high speed Z80, 128K of RAM, firmware ROM, and various parallel/serial interface chips) linked by ribbon cable to the two 4-voice cards, disk drive, keyboard, and front panel controls. Sampling of sounds is from a mike or line input with low-pass filtering and a programmable threshold to start the procedure. A sample rate of 30kHz is used, but the bandwidth is limited to 10kHz — well below the Nyquist limit of 15kHz for this sample rate.
Of the 128K of RAM, something approaching 120K is reserved for sounds, so that, with the normal 2-octave upper/2-octave lower keyboard split, approximately 60K is allocated to each sound, i.e., enough memory for 2 seconds duration. The Emulator uses companding 8-bit A/D and D/A conversion, giving a signal to noise ratio of around 72dB (compared with approximately 52dB for straight 8-bit conversion). There's also that troublesome thing called quantisation noise (a product of the roundoff error encountered with only 256 amplitude steps in 8-bit conversion) to bear in mind, but E-mu's quoted performance of 72dB signal to quantisation error suggests that they have this licked. Rather than using a fixed low-pass anti-aliasing filter on the output (at 10kHz, for instance), E-mu have elected to put a tracking VCF on the outputs. Thus, whatever the pitch that's being outputted, the relationship between bandwidth and anti-aliasing filter cutoff frequency can be constantly adjusted — a very slick idea.
In use, the Emulator is phenomenally quick at loading and saving large chunks of sound/sequence data from and to disks. E-mu use a super-duper Magnetic Peripherals disk drive which makes my Apple disk drive green with envy at the idea of shifting 60K of data in just a few seconds, but, then, that's the great benefit of DMA (direct memory access). It's not for nothing that DMA is called the I/O superhighway! E-mu also have another trick up their sleeves when it comes to outputting sounds at a required pitch: they use high frequency VCOs to direct the data flow rate in the output channels and therefore to determine the pitch of the DMA'd sound data.
So, on the technical side, the Emulator is a very clever piece of design, but, before I get too carried away with what goes on under the bonnet, let's return to its musical attributes.
The first thing that happens when the Emulator is switched on is that all the panel LEDs light up and the disk drive motor starts whirring into action. This is the machine's way of saying that it's waiting to be fed with a diskfull of data. Each Emulator disk comes charged with a couple of 60K sounds (or multi-sample subdivisions of this amount), a 900-note sequencer file, and some software. The software serves to update the routines contained in the firmware ROM. This way, any software improvements are as simple as feeding the disk drive with a new disk. The only trouble with this approach, rather than updating firmware, is that every disk in your catalogue of sounds would also have to be updated (using the Software Replicate function), and that could be quite a task if you collect sounds like a squirrel collects nuts. Both sounds on a disk are loaded into memory immediately after the system software, and the keyboard becomes active just a few seconds after the booting-up procedure.
GET LOWER and GET UPPER load the two halves of the keyboard with sounds from a different disk. A nice feature of the Emulator is that both halves of the keyboard remain active during the load procedure, though the results of playing on the half of the keyboard into which the load is happening depends on the relative lengths of incoming and outgoing sounds. So, if they're both short sounds, it should be possible to play the new sound after only a second or two of the load cycle. Other functions in this area of the Emulator include SWAP which swaps around the upper and lower sounds, and SAVE, which stores both sounds currently in memory on to disk together with anything in the sequencer. The disk operations of the Emulator really seem impeccable; indeed, E-mu mention that they were able to load sounds 56,000 times in succession from a single disk without any problems!
Most of the ever-propagating library of sounds supplied by E-mu are in what they call a 'short release' mode (exceptions are sound effects and percussion sounds, which are preset in 'natural release' mode). With short release, releasing a key always results in the immediate termination of the sound, regardless of whether or not the recorded sample has come to its end in being accessed from memory. The alternative of natural release means that pressing a key results in the complete sample being played, regardless of when the key is actually released. Release time options of either 0 or 2 seconds may seem a bit extreme, but various other features make this choice more flexible. Firstly, a release foot switch can be used to change the release mode from natural to short, or vice versa, i.e., something like a digital sustain pedal. Secondly, the natural release can be tailored by truncating the actual sample length to something in between the minimum and maximum. This takes us to the sustain/filter section of the unit. A couple of sliders are assigned to various functions according to the status of the TRUNCATE/FILTER button. When engaged, this button sets the upper slider as a means of adjusting the truncation or end point of a sound. Similarly, the lower slider then determines the cutoff point of the output filter. With this button's LED off, the sliders then serve to determine the start point and length of the sustain loop. In practice, all this is reasonably common-sense, and it's really just like getting a tape loop into the exact synchronisation that you're after. However, the Emulator goes one step better than this because it also adds a 3-stage envelope to the sampled sound, and, by manipulation of sliders and so on, the attack, sustain loop, and decay portions can be set. Generally speaking, short sustain loops, with a small number of waveform changes, work best with relatively unmodulated sounds, but instruments with a lot of modulation, like breathy flutes, strings, or voices, need rather longer waveform loops. Of course, these sustain controls can also be used to create new timbres in addition to just recreating the originals as authentically as possible.
The satisfactory editing of a sound is probably the most time-consuming (but very satisfying) side of using the Emulator, and there's no way that any micro can actually improve the decision-making that goes on in your brain between hearing a sound and accepting that it's just what you're looking for! In fact, looking for sounds could become something of a preoccupation for Emulator owners — forget the etchings, come up and hear my emulates...
With sounds firmly ensconced in memory, the Emulator also adds a generous helping of real-time performance controls: firstly, a doubling mode, selected by a foot switch, allows one to play notes on the lower keyboard section and have these notes automatically doubled by whatever sound is programmed into the upper section; secondly, there's the vibrato and pitch bend wheels, with the options of assigning vibrato control to either or both of the keyboard sections and using another foot pedal in place of the modulation wheel for vibrato depth. Other features include dynamic keyboard allocation with channels assigned on a last-key-pressed priority basis (great for auto-antiphonal effects), a tuning control for both halves of the keyboard, and a choice of either separate left and right outputs or a mono mix.
Unfortunately, some sounds with pronounced resonances tend to work less well with the Emulator's pitch shifting technique than others. This is particularly true of the saxophone and the male voice, where the shifted resonance frequency automatically resulting from any attempt to change pitch tends to demolish the identifying characteristics of the original sound, or, as E-mu put it, "the effect is not that of a person singing different notes, but of bigger and smaller people". To overcome this, E-mu have introduced a multi-sampling technique that uses 8 samples over the 4-octave range of the keyboard, so that, on playing, no single sample will be shifted more than a few semitones in either direction, thereby preserving the correct resonant frequencies. In addition, by recording different sounds into each half octave section, it is quite feasible to have, for example, eight entirely different percussion sounds or sound effects available on the keyboard at one time. To do the multi-sampling yourself, another piece of software is needed (costing an extra £250). With this Multi-Sample program, each time you make a recording, the Emulator uses all the currently available memory. Truncating that sample then frees memory for use by the next half octave section, and so it continues until all eight samples are loaded or you're out of memory. Furthermore, each half octave section can have its own loop and filter settings and be individually tuned to match the other sections. So, if one's aim in life is to make the Emulator sound like a cathedral organ, Steinway grand, or whatever, then multi-sampling is the ideal way of achieving a result which is virtually indistinguishable from the real thing.
A recent addition to the formidable armoury of the Emulator is the polyphonic sequencer, and this can also be retrofitted to earlier versions of the instrument. The Sequencer section is situated to the right of the modulation/pitch bend wheels and comprises just six buttons. In fact, this simple facade is deceptive, for the E-mu sequencer is a pretty powerful beast. A total of 900 notes can be stored in the combination of two sequence 'bins', though there's also an 'extra' bin designed to be rather like an emergency fuel tank, as it makes sure that one doesn't inadvertently store something over a sequence that was essential to one's musical existence, the old sequence simply being shunted into the extra sequence bin out of harm's way. By judicious manipulation of the STORE, RECALL, and STOP buttons, all the usual behaviour patterns of polyphonic sequencers can be emulated, but more complicated stuff is also possible, including adding notes to the end or beginning of a sequence, truncation, and overdubbing. With the latter feature, the sequencer both stores and recalls sequences at the same time. Even this is rendered fairly painless because the extra sequence bin is always used for storing the original sequence as it was before an overdub was added — E-mu's anti-goof facility! Sequences can be overdubbed as many times as one likes, even with all 8 sounds from a multi-sample, but you're obviously restricted to the number of simultaneously sounding notes on the Emulator and the sequencer memory note limit.
It's hard to pick holes in such a well-designed instrument — it must be the keyboard of '82. And for those wishing to strike a balance between emulation and financial emulsification, it is possible to start off with the 2-voice (£2,450) or 4-voice (£3,650) versions and then upgrade at a later date.
Finally, I can't resist quoting this wonderfully tongue-in-cheek(?) comment at the end of the Emulator manual: "The Emulator can be a very powerful tool for the creative artist. If enough people take advantage of its capabilities, we are going to make one hell of a lot of money." I wouldn't be at all surprised!
The Emulator is distributed in the U.K. by Syco Systems (UK) Ltd, (Contact Details).
Review by David Ellis
Previous article in this issue:
Next article in this issue: