The Synth That Samples
A synth that samples, or a sampler with a built-in synthesiser? Either way, the DSS1 takes music-creation a step further, with sampling, waveform synthesis and analogue processing. Paul Wiffen fiddles to his heart's content.
The flood of 12-bit sampling machines at man-in-the-street prices goes on. This month, Korg join the club with a sampler that boasts many traditional synth features, plus waveform creation and additive synthesis. Its potential is enormous.
Not so long ago it was Akai, then it was E-mu, this month it's Korg. Next month it'll probably be Roland. It seems that at the moment, all the synth manufacturers are releasing 12-bit samplers one after another. How is anybody supposed to choose between them?
The spec hunters will tell you to compare memory, sample rate, number of sample locations and so on. Others will point to large libraries of samples available for older machines and recommend them. But both these approaches miss what many consider to be the most important factor: what each machine allows you to do with samples once you've made them.
And this is where the new Korg DSS1 scores over some of its rivals, even though they may have larger memories or more disks currently available.
It features more sound-creation facilities and after-the-event processing than any other 12-bit sampler. These encompass long-established traditional analogue features like sync and oscillator detuning, along with the latest digital techniques - waveform drawing, additive synthesis and digital delay lines (two of them).
But we're getting ahead of ourselves. Let's begin with sound quality. The paper spec gives good cause for optimism. Sample quality is a product of (among other things) sample analysis and sample rate, and the DSS1 uses 12-bit analysis, which puts it on a par with all the samplers released this year, while its maximum sample rate of 48kHz puts all the others in the shade. In fact, there are four possible sample rates: 48, 32, 24 and 16kHz.
The factory disks supplied with the DSS1 are of variable quality. Some samples - like the acoustic guitar and solo strings - are superbly authentic, but others - like the pianos - give away their multi-sampled origins all too easily, with glaringly obvious changes between samples as you play across the keyboard.
On the other hand, the synth sounds the DSS1 comes with are uniformly excellent, ranging from the synth strings and brass (which everybody started to ask for as soon as samplers made accurate acoustic sounds available) to 'pure' synthetic bass and lead sounds.
Accessing a new set of sounds on the DSS1 requires a fair amount of patience. The double-sided disks hold four complete memory setups of the DSS1 which are known as Systems, with each one containing samples, synth sounds, and (initially) blank user locations. But it takes five separate actions to set the loading process going (four switches and a fiddly slider movement) and then the disk drive takes over 50 seconds to load one of these systems. Korg don't try to hide the fact (the display does say "Please wait a minute" while loading), but 50 seconds is a very long time to load 256K of memory. The Prophet 2000, Akai S900 and Emulator II manage to load twice as much memory in 40 seconds or less. I don't know how people using the DSS1 live will cope with the minute or so it will take to change systems, though inevitably there'll still be a couple of other keyboards around to make noises while the DSS1 is otherwise engaged. Let's hope a future update will speed up the Disk Operating System...
Making your own samples on the DSS1 is straightforward enough, and the machine guides you through any decisions you have to make before you can begin sampling. Hitting the Sample button gets the ball rolling by putting you in the correct mode, and the display prompts you to choose your sample rate (which you do by moving the data entry slider). Hitting Enter then moves you on to setting the Total Time for your sample. Depending on the sample rate you've selected, this gives you a choice of times which correspond to half or all of the DSS1's memory.
Now comes the point where you have to decide if you're going to use multi-samples (to make a more accurate representation of the instrument you're sampling over the five octaves of the keyboard) or just one sample. You do this by specifying how many memory divisions you want, and the DSS1 automatically divides the time set between the number of samples you want to make (1,2,4,8 or 16). Then you input the number of the particular sample you want to record. The DSS1 now offers default keyboard positions (C3 for a single sample or various positions up the keyboard for multi-samples), or you can place the imminent sample wherever you like on the keyboard.
Now, if you want to, you can begin sampling straight away by pressing zero on the keypad (all such parameters are listed on the front panel). But it's as well to check the level of the incoming signal first (2 on the keypad gives you a VU-type reading in the display), and alter it if necessary via the data sliders. You might also want to set a trigger threshold, which is represented as a star within the VU-type display, so you can see just when your threshold is being exceeded by the signal. At any point, you can press zero and make a sample.
The whole system is breathtakingly easy to use, but I can't help feeling that, for the experienced sampler (person, not instrument), the procedure could get a little long-winded and restrictive with continuous use.
With only the bare minimum of preparation and familiarisation, we successfully created some impressive samples from sound sources ranging from a dynamic Peter Gabriel CD to 4AD's recently released album of Bulgarian vocal music (which the Cross-Fade Loop parameter looped very smoothly).
Of course, you don't get anything for nothing in this world: the 48kHz rate uses up all 256K of memory in no more than 5.5 seconds. A 32kHz rate gives you 8 seconds, while the 24kHz sample rate gives a generous 11 seconds, and the 16kHz rate (best reserved for special effects) gives 16 seconds of continuous sampling.
In terms of memory, the 256K of the DSS1 compares favourably with that of the unexpanded Prophet 2000, but as far as we've been able to establish, it's not possible to expand the memory to the 512K which looks like being standard soon for samplers in the Korg's price bracket.
Once you're happy with the quality of your sample, you have the choice of naming it and saving it to disk, which you can do either straightaway or after some editing. The first thing Edit Sample mode allows you to do is truncate the start and/or length of your sample. This means you should set the start point first, otherwise the point at which your sample ends will move when you move the start point, as the loop length fixes the end by reference to the number of individual samples after the start.
More experimentally, you can reverse, link, or mix samples (with splice or crossfade) to build up new sounds in a quasi-Musique Concrète approach. The penultimate function of Edit Sample mode (just before the Save function) allows you to view and edit the value of each 'word' of sample data individually. No other sampler on the market allows you to do this without a VDU, and it's an incredible feature - though it would be a brave programmer who'd attempt to make major changes via this method - it could well take all night. That said, it should prove invaluable for removing small blemishes in samples caused by clicks and pops.
Strangely, a more standard function which you'd normally expect to find in the Edit Sample section - looping - is actually to be found in the Multi-Sound area of the machine. But having found it, it is well worth the search, as it offers you three levels of help with your looping. Loops are set up by specifying a start point and a length for each sound within the Multi-Sound (this terminology can get confusing, can't it?). This means that each sample can be looped independently of all the others.
The most basic way of looping is to set the start and loop lengths manually, a method which virtually all samplers on the market offer. These points can be altered in four ways. Data entry slider A (there are two of them) allows you to move through the sample memory in steps of 10K (10,000 samples), while the associated Up and Down switches move in 1K jumps (1000 samples). Data entry slider B gets even finer resolution with every 20 samples, and using the switches next to this slider, you can step through each individual sample.
Assuming you don't instantly find the ideal sample points manually (and it can take a long time, as programmers who have samplers with only this method of looping will testify), the DSS1 offers you the chance to use an Autoloop function. This is all very fine when it works, but unless you've already moved the start and length values to something suitable, the Autoloop often comes up with 'Not Found' in the display. Once you get to the right place with your manual settings, though, you can keep pressing Enter to step through suitable zero crossings until you find the best one.
And if that still doesn't find you a perfect loop, you can bring in the heavy artillery and use Crossfade Looping. This feature is nothing new (it's been available for a while now through Digidesign's Sound Designer, and will be on the Emax), but the DSS1's implementation is unique in that it allows you to hear what the Crossfade Loop will actually sound like, before the sample data is permanently altered. This is invaluable, as it allows you to try numerous different crossfade lengths and sample points before you decide which one gives the best results.
So, you can make a finite sample sound continuously in the same way as a traditional synthesiser oscillator. And by using the Multi-Sound section, you can get a good representation of your original source over the whole five-octave keyboard (or 10½ octaves via MIDI).
Yet the DSS1 offers you two alternative ways to produce digital waveforms in addition to sampling. This makes the machine a proper synthesiser as well as a sampler, and seeing as the digital waveforms you can create from scratch take up very little memory, there should be no worries on that score.
The first method of waveform creation is drawing. This is done by moving one of the data entry sliders to plot the waveform in a slowed-down, real-time operation which represents one cycle. This lasts about eight seconds, and in that time, any movements you make with the slider are recorded and used to plot the waveform (see Diagrams).
This movement of the slider generates a 512-segment waveform, or 512 different levels, all of which can then be viewed and edited in a similar way to the sample data. Such analysis and editing doesn't represent quite such a marathon task, either, as there are only 512 levels to look at, rather than the thousands involved in just a couple of seconds of sampling. And the neat thing about this process is that you can set up a complex waveshape very quickly using the drawing, and then tailor it by means of the View/Edit function.
What waveforms can you come up with via this method? Well, I soon found how to make traditional shapes like sawtooth (slowly raise the level over the whole eight seconds) and square (start with the slider at full level and pull sharply down to the bottom after four, seconds), but shortly afterwards discovered these waveforms are available as starting points for Harmonic Addition. More unconventionally, pushing the slider up and down rapidly gives you a lot of high harmonics; moving it less jerkily gives you a purer waveform.
What eventually transpires from all these hand movements is that, in the static mode you obtain from a single cycle, many radically different waveforms end up sounding-fairly similar. To correct this, you apply a slow filter sweep in the analogue section (which we'll come to in a while), so that the different harmonic structures of each waveshape become audible.
The same turns out to be true of Harmonic Synthesis, the other option in Create Waveform mode. This method allows you to specify a different level for each of 128 harmonics. Again, this is a flexible system that can involve the user in a rather long-winded procedure. Fortunately, Korg offer you a good selection of starting waveforms besides the Blank one (all harmonics at zero level). This means you can start with an old friend like 'sawtooth' or an organ-type sound, or even the 'current' waveform - which can be one you've just drawn, or even one cycle of a sample.
You can't hear the difference that changing the level of a harmonic makes instantly. When you've made any changes you think are appropriate, you press Enter and the DSS1 computes the resulting waveform. You then go through this process as many times as is necessary, until you're happy with the result.
When you've finished creating your waveform - whether by drawing, editing or altering harmonic levels - you can save it to disk with a name, just as if it were a sample.
And now it can be used as a source waveform for the DSS1's synthesiser, just as though it were a sample. On most sampling machines, the 'synthesiser' section is there simply to 'clean up' samples, and is rarely a very versatile beast. But on the DSS1 you have a fully-fledged synth section: each of the eight voices boasts two oscillators plus noise, and a VCA and VCF with independent envelope sections for each - in other words, virtually all the features of Korg's flagship polyphonic synth, the DW8000.
However, the synth section is not multitimbral, so all samples within a Multisound are affected by the same set of synth parameters. This isn't to say you can't create multisamples and split effects through sampling on the DSS1. But as soon as you start synthesising these samples, you hit the monotimbral limitation.
And it's symptomatic of the DSS1's design that to make the best use of the instrument, you'll probably want to use your samples (which of course take up precious memory) in both synthesised and unsynthesised forms.
To be fair, Korg have conceived the DSS1 as an extension of the synthesiser tradition - more specifically, I suppose, their synthesiser tradition, in which multitimbral synths don't figure. Thus you won't find individual audio outs on the instrument's rear panel, or an implementation of MIDI Mode 4 (the 'multitimbral' mode).
Anyway, all the Program Parameters (as the synth section is labelled) are accessible in three ways: by typing the parameter number on the keypad, by moving data entry slider A, or by stepping through associated Up/Down switches. This system is no substitute for traditional knobs and dials, but once you get used to it, you can whizz around between parameters without too many headaches.
Generally speaking, the available parameters are explained quite fully in the display when you call them up (which is just as well, as the preliminary manual we received ignored them completely).
There simply isn't the space here to examine each synth parameter in detail, so I'll just pick out the unusual and unique features for comment, bearing in mind that standard features like filter cutoff and VCA level do their job admirably.
The most remarkable feature of this section appears early on, with the OSC1 and OSC2 Multi-Sound parameters. As I've already implied, these allow you to select different sound sources for each oscillator. These can be drawn from the full range of standard (square, sawtooth and so on), created (drawn and/or harmonically synthesised), or user-sampled sources.
So, you can put a multi-sample on OSC1 and an 'analogue' waveform on OSC2, and then mix them, detune them, sync them, switch between them with velocity, and so on. The system's potential is huge, ranging from standard synth setups with two traditional waveforms, through hybrid arrangements combining synthesis and sampling, to purely sampled setups such as velocity crossfades. Of all the digital sound-creation systems whose possibilities are claimed to be truly infinite, this is the one I can foresee programmers becoming less bored with less quickly. Truly awesome, as they say in California.
One 'analogue' technique that's an old favourite of mine appears early on in the parameter list, but the Editor has asked me not to rave on about it for too long. It's called sync (that's enough - Ed), and it allows you to force OSC2's waveform cycle to restart every time OSC1's does. In combination with pitch-shifting of OSC2 (via the built-in LFO or Auto-Bend), this can be used not only for Jan Hammer-type lead synth sounds (of which there are a couple of beauties on the factory disks), but also for strikingly realistic effects like bowing at the beginning of string samples.
The next unusual parameter the DSS1 offers is a filter cutoff slope switchable between 24dB and 12dB/octave, which allows you to determine how sharply the brighter frequencies are cut off.
The filter also features keyboard-tracking and resonance as well as the expanded ADBSSR (Attack, Decay, Breakpoint, Slope, Sustain and Release) envelope. The latter is a design Korg have been using since the launch of the Poly 800 synth, and it's a fairly versatile one, as it allows a second attack or decay to be inserted before the sustain section.
The DSS1's velocity-sensitivity can be used to affect VCF cutoff and VCA level, the attack, decay and slope of each envelope, plus the amount of autobend. You can also use it to switch or crossfade between the oscillators (especially useful on samples, this). Aftertouch (or pressure-sensitivity, as it's often known) can be routed to bring in OSC MG MOD INT (snappy Korg jargon for vibrato) and filter mod, or to open up the filter or VCA.
Korg's now standard joystick (not my favourite type of performance control, though many musicians swear by them) can control pitch-bend and filter cutoff (left-right movement) and modulation (up for oscillator - another oblique way of referring to vibrato - and down for the filter).
The Key Assign section is excellent, allowing two different types of Poly mode - cyclic and fixed assignment - and a great Unison mode where the number of voices used (2,4,6 or 8) and the detuning between them can be preset.
Finally in the Parameter section, we come to the controls for onboard signal processing. The programmable EQ allows both bass and treble to be cut or boosted over a -4 to +8dB range. But the real beauty of this section lies in the two DDLs. These can be used in series or in parallel, so you can have two different effects from each stereo out, or gang one effect after another. The parameters available allow for a wide range of DDL effects such as delay, chorus, flanging and ADT, and whatever effects you set up are stored as part of the program in question. This means that, live, you can be sure that both your synthesiser and sampled sounds are getting the signal-processing they need.
And so to MIDI, and that 10½-octave MIDI note range. The most notable thing about this is that you can put drum samples and other sounds outside the keyboard range, and trigger them from a sequencer while you play other sounds live. Aftertouch is transmitted as channel pressure, and active sensing ensures you're not left with any embarrassing drones should MIDI connections be broken.
The lack of Mono Mode implementation means the DSS1 isn't quite the MIDI control centre it could have been, but Poly and Omni modes are fully supported. You can also send and receive System Exclusive data, and Digidesign are already hard at work converting the excellent Macintosh Sound Designer program for use with the new Korg.
The DSS1 is one of the most inspiring all-round instruments to appear in quite a while. It goes far beyond the scope of machines which are limited to sampling as their only means of sound-generation, and you'd need to add something like Digidesign's SoftSynth program and the Macintosh it runs on to make those other samplers compete with the Korg.
Its problems lie in its slothful disk drive, hardware-limited memory and lack of MIDI Mono Mode implementation. But for me - and, I suspect, for a lot of other musicians - the DSS1 more than makes up for this by making waveform-creation as easy as it makes sampling, and by taking the synthesising and signal-processing further than any other sampler currently available. Bar none.
Review by Paul Wiffen
Previous article in this issue:
Next article in this issue:
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!