Emax & Emax II

EMAX SE

SYNTH FUNCTIONS
UPDATE

By Dominic Milano

EGADS, IT'S ANOTHER UPDATE FOR THE Emax. First there was the hard disk option, which we reviewed in Nov.'87. Then there was Optical Media's CD-ROM with over 3,000 Emax samples on it, which we reviewed in March '88. And we reviewed the Emax itself back in Jan. '87. So what's this new $300 update all about? Aside from adding enough initials to its name to tie even the most dexterous of tongues into knots (try saying Emax SE HD ten times fast), the SE update adds a powerful additive synthesis engine called spectrum synthesis, and several types of digital signal processing.

Spectrum Synthesis. Wait a minute. Let's call a spade a spade here. E-mu-speak calls it Spectrum Interpolation Digital Synthesis, but it's time-variant additive synthesis with a funky user interface. The concept is identical to Digidesign's Softsynth (Keyboard Report, Dec. '86)-sine waves of different frequencies and amplitudes are added together to build complex waveforms. The SE gives you control over 24 partials (Softsynth has 32), each of which can be retuned to a whole- or non-whole number ratio of the fundamental. Each partial has its own 24-stage amplitude envelope (Softsynth's have 40 stages), and its own 24-stage pitch envelope (plus or minus a half-step).

The waveforms you synthesize can be thought of as hand-tooled samples. To determine the length of the waveform, you select a sample location and record a null sample (nothing) in that memory slot. Sample length can be any amount of time up to the maximum amount of memory in your Emax.

The length of a spectrum-generated sample is divided into 24 equally spaced time segments called time slices. Each time slice holds a wavetable.The transition between time slices can be either smooth or stepped. The latter gives you a PPG-like effect. As you might imagine, drawing 24 spectra takes time. Lots of time. Happily, there's an interpolation function that will fill in the blanks between two time slices, so you can draw time slices one and 24 and have the machine draw all those in between. Very handy.

As for visual feedback, it's all supplied by the Emax' LCD (ugh). The data entry slider is used to draw waveforms (in overtone bargraph form) and amplitude and pitch contours-shades of the Korg DSS-1 (Keyboard Report, Nov.'86). Contours show up as tiny bars. While you draw the contours, the cursor advances through the display at a fixed rate, which makes it next to impossible to know exactly what partial you're affecting as you move the slider up and down.

Once you've created a spectrum, an edit mode provides exact control over the amplitudes of the overtones. You get the choice of drawing a spectrum either horizontally (across all 24 partials in one time slice) or vertically (one partial across all 24 time slices).

So what good is having synth waveforms in a sampler? For starters, you can layer them with samples (or paste them onto the tail end of a sampled attack, for that matter) and then process the layer with the Emax' VCF, VCA, and envelope generators. Remind you a little the Roland D-50's architecture? It should. The difference is that you can build your own synth waveforms and sample your own attack transients too. And unless you want to layer more than one sample and one waveform, the Emax' dual mode lets you preserve eight-voice polyphony.

Now, before the thought of turning your Emax into a super version of a D-50 for $300 leaves you thinking you died and went to heaven, let us all recite Mother Product Review's third law of reality: All cool things have a price. For the Emax SE, it's dealing with the aforementioned user interface, which requires that you have the patience of a saint. There are two reasons why: First, it takes an ungodly number of keystrokes to do just about anything with the digital synthesis module. Second, you can't listen to the sound you're synthesizing while you're tweaking it, since none of the additive synth functions in the Emax are real-time functions. The length of the wait depends on the number of spectra you've drawn using the data entry slider, the length of the sound

you're synthesizing, and the complexity of the defined parameters (pitch-bends, retuned partial frequencies, etc.). Unless you're working from a known concept such as odd-numbered or even-numbered harmonics, there's no way to be certain what the results of your drawing will sound like until you select the synthesize option and wait. And wait. And wait. And... Well, maybe it's not always that bad. We clocked computing times at anywhere from 30 seconds to three minutes. The latter was for a 1.5-second long spectrum interpolation. We shudder to think what the wait for a full 18-second interpolation might be - can you imagine waiting a half hour to hear the result of your synthesis and then deciding it isn't what you were after? Our advice: Start your experimenting with short sounds. When you find something you like, then go ahead and do a longer version of it.

Transform Multiplication. The first demo we heard of transform multiplication (a process that multiplies two samples together) completely knocked us out. The timbres produced by this process have an eerie and exotic quality. Very experimental sounding. Very cool. Then we inquired about how long the process takes. Depending on the length of the two samples being multiplied, a transform multiplication can take up to 41 minutes! We're talking major number crunching here. There's nothing to your end of the process: Just select two samples, read the warning in the display that tells you this could take as long as 41 minutes (talk about intimidating), and it's time to take a long coffee break.

Since the results of a transform operation can be very unpredictable but are almost guaranteed to be interesting, we found ourselves wanting to experiment with this feature a lot. But who's got 4,100 minutes to spare for a hundred experiments? We wish E-mu had included an automation feature that would let you set up five or six transforms, check to make sure there's enough memory for you, and then process and save them to disk automatically. That would let you turn the Emax loose on the problem while you sleep.

Miscellany. Other new features in the SE include sample rate and pitch conversion functions. Again, this could take as long as 41 minutes, but the average time is closer to two minutes. The pitch conversion function allows you to resynthesize the original sample at a new pitch, which could be handy if you want to splice together two sounds that weren't sampled at the same pitch. Another hip possibility: Create a microtonally detuned version of a sample and mix it with the original to create digital chorusing without using up voices.

The SE has a couple of new MIDI func-
tions: Program change reception can be
disabled, and a special program change code can be set up to allow remote hard disk loading: Also, a new preset stack mode allows up to four voices, each playing a different preset, to be triggered from a single key.For those who don't want to spend hours and hours creating spectra, E-mu includes 95 basic types on the update disks; already synthesized versions of these are available on the disk, so you can hear them without having to go through the synthesis routine yourself. The factory sounds range from traditional waveforms, through Rhodes-like and presampled attack transients, to finished sounds created by combining samples with synth sounds.

Update policy. The SE lists for $300 more than the original Emax. If you bought your Emax before March 1, 1988, the update will only cost you $95 if you get it before July 1, 1988. If this means you, don't wait; $95 is a steal for these features.