The best just keeps getting better. Evolution is a curious thing. According to the theory of natural selection, traits that provide better chances for

The best just keeps getting better.Evolution is a curious thing. According to the theory of natural selection, traits that provide better chances for survival pass from one generation to the next. This can be observed everywhere in the living world, and it's also evident in the realm of music technology. In some cases, improvements in subsequent generations of products are even passed backward to their precursors.

Nowhere is this reverse evolution more apparent than in Kurzweil's V.A.S.T. synthesizers. Talk about survival of the fit; these synths are the best all-purpose electronic instruments on the market. (They are also among the most expensive.) One of the primary reasons for these synths' success is that Kurzweil designs them to be upgradeable and continues to release improvements for older models while incorporating those changes as standard features in newer ones.

The latest instrument in the line - the K2600 - represents a significant evolutionary step beyond its predecessor, the K2500. Many of its new features are also available as options for the older model, although the K2600 has a few tricks that are beyond the K2500's capabilities.

I've reviewed several of these synths in previous issues of EM - the K2000 (March 1992), K2000RS rack module (May 1993), and K2500 (May 1996) - so I won't repeat myself here. Instead, I'll concentrate on the K2600's new features and point out which are available for the K2500 as software or hardware upgrades.

I looked at the 76-key version with sampling, which is officially known as the K2600S. It's also available without sampling as the K2600 ($6,256). The 88-key version without sampling is the K2600X ($6,820); with sampling, it's the K2600XS ($7,700). The new instrument is also available as a rack-mount unit with or without sampling: the K2600R ($5,175) and K2600RS ($5,950). There's even a soup-to-nuts 88-key model with all the options included: the K2600AES ($20,000). As in previous generations, you can always add sampling to an instrument that lacks it ($900). Unless otherwise noted, I'll refer to the K2600 generically; all features and options are available in all varieties of the instrument (with the obvious exception of a keyboard on the rack-mount versions).

EVOLVED HARDWAREThe K2600's front-panel appearance is much the same as the K2500's, except for its deep purple color and lighter background behind the controls, which is endemic to all new Kurzweil instruments (see Fig. 1). According to the company, the 88-key weighted keyboard action of the K2600X has improved from the one in the K2500X and PC88; Kurzweil worked with Fatar to refine the design, which Kurzweil's representative described as more rugged. The 76-note, semiweighted keyboard is unchanged, which is fine with me; I really like the feel of this action.

The circuit boards were redesigned from scratch, which lets the K2600 include twice as much Flash ROM as the K2500 (from 2 MB to 4 MB). This is critical to accommodate the more advanced software features. The new circuit boards also exhibit a higher signal-to-noise ratio at the outputs, which is always a good thing.

Because of the new circuit boards, the back-panel layout is completely different (see Fig. 2). In addition, the 1/4-inch outputs are now balanced, which contributes to quieter operation when using balanced connections. However, that means the outputs can no longer serve as inserts as they do in the K2500, which might be a drawback if you wish to use them in that way. (There is another way to get multichannel audio into the K2600, which I'll discuss shortly.)

Another benefit of the new circuit boards is the ability to accommodate four optional ROM blocks instead of two in the K2500, bringing the maximum sample ROM to 44 MB in the K2600 (up from 28 MB in the K2500). Two ROM blocks are available: Orchestral and Contemporary ($450 each). By the time you read this, a third ROM block will be available: the Triple-Strike Piano that first appeared in the PC2 (see the review in the July 2000 issue of EM). The new piano ROM block adds 8 MB of soft and medium strikes to the hard strikes in the 4 MB Stereo Piano ROM that is standard in the K2600 and was optional in the K2500. Kurzweil is working on a fourth ROM block, which was under wraps at this writing. To install any ROM blocks, you also need the optional daughterboard ($290).

Kurzweil's KDFX effects processor is standard equipment in the K2600; gone are the K2500's lackluster DigiTech effects. (Hallelujah!) Fortunately, you can install KDFX in the K2500 as an option ($795). Among the many major improvements represented by KDFX are four stereo buses, plus a master stereo bus (as opposed to a single stereo bus; see Fig. 3), up to 18 modulators (as opposed to two, plus the wet/dry mix), many more effects and editable parameters, 20 kHz bandwidth (as opposed to 15 kHz), and all-digital operation (as opposed to converting to analog for processing).

Available effects include 15 reverbs, 7 delays, 11 chorus/flange/phaser algorithms, 9 distortion algorithms, 5 tonewheel organ effects, 14 special effects (various filters, shapers, etc.), 24 multi-effects, and 21 studio/mixdown effects (compressors, expanders, gates, and EQs along with tremolo, enhancers, autopanners, stereo image manipulators, and SRS simulated 3D). Perhaps most important, KDFX consists of custom VLSI DSP chips that can be upgraded with new software capabilities at any time.

All of this is wonderful, but how does the instrument sound? In a word: glorious. I played some basic Programs that use the Stereo Piano ROM on my DigiTech-equipped K2500 with similar Programs on the K2600, sending both through some of the available effects. The KDFX reverbs sound much sweeter and smoother than the DigiTech, with full-bodied tone and lush tails. The delays are as clean as a whistle, the chorus and flange effects are luxuriant, and the other effects are equally superb. All in all, this is one of the most significant improvements over the original K2500.

The stock Program RAM (PRAM) has increased from 256 KB to 486 KB, and the maximum PRAM has increased from 1.2 MB to 1.5 MB. (The K2600 PRAM expansion costs $450.) As in earlier models, PRAM is battery-backed and holds Programs, Setups, and other objects (except samples), even when the main power is turned off. Speaking of samples, all K2600 models ship with 64 MB of sample RAM in the U.S. market, but elsewhere, only sampler models ship with sample RAM - and only 4 MB at that. By comparison, the K2500 sampler models shipped with no more than 4 MB. Both units can increase to 128 MB using standard SIMMs.

One of the most important hardware options for the K2600 is called DIO-26 ($700). Derived from the Kurzweil Digital Stream (KDS) option for the K2500, DIO-26 adds eight channels of 20-bit digital audio I/O. (KDS in the K2500 provided only eight channels of digital audio output.) You can reduce the output to 16-bit resolution with various dithering algorithms.

The audio data is encoded in KDS format, which means you need a DMTi breakout box ($1,309) to interface with AES/EBU and S/PDIF. Because of physical space limitations, a stock DMTi provides only four input channels in addition to eight output channels. You can also get optional 8-channel I/O cards for ADAT ($300) and TDIF ($150).

Together, these options make the K2600 a powerful digital audio/MIDI production environment with eight channels of digital mixing, four effects buses, and a stereo master-effects bus in addition to the internal sequencer and sounds. All the K2600 needs to complete the picture is true hard disk recording, although it does provide something close, as you'll see in a moment.

The DIO-26 option provides the only way to keep the audio entirely in the digital domain all the way to the output of the K2600. (The K2500 requires KDS or KDFX for this; oddly, KDFX in the K2600 does not provide direct digital output as it does in the K2500.) Without the DIO-26, the internal audio converts to analog and back again for the sampler's digital output. This was done on the K2500 to accommodate the DigiTech effects - which could process only analog signals - and Kurzweil wanted to use the same sampling hardware for the K2600. The DIO-26 by itself provides two channels of AES/EBU or S/PDIF out; with the DMTi breakout box, you get eight channels out.

The newest software for the K2600 and K2500 allows partitions of up to 2 GB on hard disks with up to 8 GB in total capacity. A stock K2600 doesn't ship with an internal hard disk, but you can add any internal SCSI drive with a Kurzweil mounting kit ($90 for the keyboard, $50 for the rack-module). Of course, you can also connect an external storage device to the SCSI port on the back.

Unlike the K2500, some of the hardware options are user-installable in the K2600, which is a real blessing. These include the ROM blocks, sample RAM, and PRAM. You still need a qualified service technician to install the DIO-26 and an internal hard disk in the K2600, but a K2500 requires a service technician for all hardware options.

LIVE AUDIOThe K2600 has been out for some time now, but I waited until version 2.0 of its software was available to review it. This version includes a number of very cool features I wanted to touch on, and it was worth the wait. Unless otherwise noted, these features are also available in the latest K2500 software (version 2.99 without KDFX; version 4.48 with KDFX).

Live mode is an exciting new feature that lets you process real-time audio streams from the two sampling inputs through the synth's V.A.S.T. and KDFX architectures. These audio streams are treated just like any other samples, which means they can be pitch-shifted and even played in reverse in addition to any other V.A.S.T. or KDFX manipulation.

With the DIO-26 and DMTi options, you can also process up to eight channels of real-time audio, but these streams bypass the Pitch block in the V.A.S.T. algorithms, so you can't pitch-shift or reverse them. Still, there's an awful lot you can do to them. This powerful feature provides almost unlimited sound-manipulation possibilities.

I used a microphone connected to one of the sampling inputs and had great fun sending my voice through various KDFX chains and V.A.S.T. algorithms. The only problem I encountered was feedback from my studio monitors. If you use Live mode with a microphone, you must be careful of feedback; I used headphones to avoid the problem.

A related feature is the Vocoder, which requires the sampling option and uses the V.A.S.T. architecture rather than KDFX. Like a traditional vocoder, this feature applies one signal's spectral profile - called the master (typically a vocal) - to another signal, called the slave (typically a synth sound). In the K2600, the master signal passes through 20, 22, or 24 parallel bandpass filters, each tuned to different bands in the audio spectrum. Each filter's output level is sent to a set of envelope followers used to control the outputs of another bank of parallel bandpass filters through which the slave signal is sent. This imposes the master signal's spectral profile on the slave signal, creating the "talking-synth" effect.

To use the K2600 Vocoder, you must bring the signals into the sampler's analog inputs: master signal to the left input and slave signal to the right input. To use a vocal signal as the master and an external synth as the slave, simply connect a mic to the left XLR input and the synth to the right XLR input. (These are low-impedance, line-level inputs with no mic preamp or phantom power. However, you can add 21 dB of gain, so dynamic mics work fine.) In addition, you need to use a matching transformer or DI box between the synth and the input.

Alternatively, you can use a 1/4-inch TRS insert cable plugged into the high-impedance, line-level stereo input and connect a mic to the left side and the synth to the right side. However, you need to use a high-impedance mic or a preamp to make this work properly. You can't use one XLR input and one side of the high-impedance input together, because they are mutually exclusive.

The Vocoder Setups include two Drum Programs that use Live-Input Keymaps with 20, 22, or 24 Layers of bandpass filters in the algorithms. Each bandpass filter takes one voice of polyphony, which means you can't use the K2600's internal sounds if you select the 24-band Vocoder. (Do the math: 24 bands for the master + 24 bands for the slave = 48 voices of polyphony with nothing left to play an internal sound.) Therefore, you must use the 22- or 20-band Vocoder if you want the K2600 to provide one or both input signals.In addition, physically connect the B outputs to the sampling inputs to use the internal sounds. (Fortunately, these outputs are balanced, so all you need is a balanced cable with 1/4-inch TRS on one end and XLR on the other end.) If you select the 24-band Vocoder, the Setup's third Program sends MIDI to an external synth.This is all somewhat cumbersome, but the results are quite good. I really enjoy the sound of a vocoder, and this one doesn't disappoint. The K2600 provides a lot of real-time control over the vocoding process, including envelope-follower speed, filter bandwidth, and center frequency. The preprogrammed values for those parameters work fairly well, and altering them (especially envelope-follower speed) can make some pretty ugly sounds. Nevertheless, I had lots of fun with this function.

Another very cool audio feature, called RAM Tracks, lets you record audio into the sequencer, much like a digital-audio sequencer on a computer. In this case, however, the audio is recorded into sample RAM rather than straight to hard disk. Of course, this places fairly tight limits on the length of audio you can record, especially if you have lots of custom samples loaded; I hope Kurzweil is working on a software update that provides true hard disk recording.

Meanwhile, this is the next best thing. After enabling an empty sequencer track to record, play audio into any available input (sampling or DIO-26) while the sequence is playing, and the instrument automatically assigns a MIDI note to the phrase. When you play the sequence, the audio plays along with any MIDI tracks.

To test this, I recorded some basic tracks in the sequencer as MIDI data that played the K2600's internal voices. Then I recorded a trombone solo through the sampling input into the sequence, which played right along with the MIDI tracks on playback. Clearly, that is another powerful and useful production tool, although it would be much more useful if the audio recorded directly to hard disk.

MODEL SAMPLESOne of the K2600's time-saving features is the ability to create automatic Keymaps, which is great for importing multiple samples from CD sample collections. If you dump a bunch of samples into the instrument, it automatically assigns them to root notes and assembles them into a Keymap and a preview Program. This feature doesn't loop or normalize, but it does just about everything else to get samples ready for playing.

I tried this feature with a CD of samples I recorded for Hans Zimmer as he composed the score to the movie Gladiator. The samples consisted of me playing sustained notes on various Swiss Alphorns, seashell trumpets, and Jewish shofars (ritual ram's-horn trumpets). I started with the seashell trumpets and dumped them into the K2600, where they were neatly assembled into a Keymap and preview Program. I also created Keymaps for the Alphorn and shofar samples. This dramatically reduced the time it would have taken to assemble those Keymaps by hand.

The only unfortunate thing about the process is that you must start each sample manually; the instrument won't automatically start each one separately. This wouldn't be too difficult to implement; a new sample could start when the input signal exceeded a certain threshold, and the initial transient below the threshold could be recovered from an input buffer. Kurzweil is considering this feature for a future software upgrade.

You might think that 1,000 ID numbers for samples would be plenty, but that's what they said about 640 KB of RAM in the original IBM PC. In fact, those ID numbers get used up in a hurry, especially if you have some ROM blocks installed in addition to using your own samples. To ease this situation, you can join dual-mono samples and use one ID number to address both of them. Also, you can combine up to 255 samples into a single multisample with one ID number, which really helps conserve those precious numbers. Of course, you can't edit the individual samples when they're combined like that, but you can always split them into separate samples if necessary.

KB3 mode is not really new; it's been available in K2500 software since version 2.52, and it's also part of the PC2. This mode bypasses most of the V.A.S.T. architecture (except the pitch and amplitude blocks) and models the behavior of a tonewheel organ, such as the Hammond B-3. The model includes percussion, key click, chorus, vibrato, tube-amp distortion, and rotary speaker simulation with programmable speed control that ramps up and down in a realistic manner. The eight front-panel sliders and the mod wheel act like drawbars with KB3 Programs; pulling them down increases the level of the corresponding partial.

The upper partials are generated by sample playback, while the lower partials are DSP-generated sine, sawtooth, or square waves. Unlike purely sampled organs, KB3 does not duplicate partials when different notes share common harmonics (say, C4 and G4), which prevents beating between shared partials and more closely emulates the behavior of a real B-3. As a result, the factory KB3 organ Programs sound incredibly realistic; I'd put them up against any sampled or modeled organ out there.

However, that realism comes at a price: polyphony. A KB3 Program that uses 79 tonewheels reserves 40 voices of polyphony, leaving eight voices available for other sounds, such as bass and drums. The KB3 voices remain reserved and unavailable for anything else as long as any KB3 Program is selected for any channel, so make sure there is no such Program on any channel if you're not using it. Fortunately, you can specify the number of tonewheels to use in a Program; if you need more polyphony, reduce the number of tonewheels, which you might not need if you're not using the top few drawbars anyway.

EXCLUSIVELY YOURSThere are a couple of features that simply require more Flash ROM space than the K2500 has, which means they are available only on the K2600. Most important among these features is Triple Modular Processing, or Triple mode for short. This powerful new feature represents the first major expansion of Kurzweil's venerable V.A.S.T. architecture since it was introduced in the original K2000.

To review, V.A.S.T. is based on 31 algorithms consisting of several user-selectable DSP blocks that are chained together in a serial fashion and controlled by a variety of modulation sources. A Program is composed of one or more Layers, each of which includes a set of samples in a Keymap sent through an independent V.A.S.T. algorithm that controls pitch, amplitude, and timbre. Normal Programs have up to three Layers, whereas Drum Programs have up to 32 Layers, and all Layers' outputs are routed through the rest of the signal path separately.

Triple mode provides new V.A.S.T. algorithms that can be combined in groups of three Layers. Unlike normal Layers, however, these three algorithms are connected in a serial manner (see Fig. 4). There are 30 new algorithms available for Layer 1, 38 for Layer 2, and 26 for Layer 3 of a Triple Layer. The algorithms for Layer 2 include a block that combines the output of Layer 1 with the sample assigned to Layer 2, and the algorithms in Layer 3 further process Layer 2's output. You can even combine a Triple Layer with normal Layers or other Triple Layers in a Drum Program, although be aware of polyphony limits. A Triple Layer takes the same polyphony as a normal three-Layer Program.

Triple mode offers almost-infinite synthesis possibilities, including the ability to modulate one sample with another in various ways and chain several independent filters together, which is impossible with a standard V.A.S.T. program. In particular, Triple mode offers the potential for much more timbral variation and control. Playing a single note of some Triple Programs and moving the eight sliders results in a swirling kaleidoscope of sound with shifting timbres and - in some cases - looplike beats that you can't duplicate with standard V.A.S.T. Programs.

Another feature unique to the K2600 is its ability to support 16 simultaneous Drum Channels, which means you can have a Drum Program on all 16 MIDI channels. As a result, the entire concept of Drum Channels is rendered moot in the K2600. The K2500 is limited to eight Drum Channels (2 through 8, plus one user-specified channel), so some of the K2600's Setups will not work on the older instrument because it uses channels other than these. Typically, those are the Interactive Groove Setups that depend on groove patterns, which are also not available in the K2500.

I HEAR VOICESThe K2600 uses exactly the same base waveform ROM as the K2500. However, the latest instrument includes all newly developed Programs and Setups, and plenty of them - it ships with more than 200 Programs in ROM and 200 Programs in RAM just to get you started and demonstrate what this beast can do. In fact, Program ID numbers are becoming more and more precious, so some new Programs actually include two related sounds that you can crossfade or switch between with the A slider.

The first 100 Programs include a good cross-section of sounds, including pianos, V.A.S.T. organs, basses, drum kits, guitars, synth leads, strings, brasses, and choirs, making it easy to audition the instrument. There are quite a few synth Programs in the other banks, including a good number of techno/rave-type sounds (some of which provide repeating loops on their own), as well as a variety of Programs reminiscent of Moog, Prophet, and other classic-synth sounds.

Overall, the new Programs are first-rate; the Kurzweil sound designers have done themselves proud with many exceptional entries. As a brass and woodwind player, I've yet to hear a truly convincing sampled emulation of these instruments, but the K2600 has many Programs that use those samples as a starting point and mangle them in interesting and useful ways. There are many great string, guitar, bass, and vocal sounds, and the drums are uniformly excellent. Because the K2600 uses the same waveform ROM, all standard Programs are fully backward compatible with the K2500.

In the K2600, most of the preset Programs use all eight sliders for subtle and dramatic changes to the sound, unlike the K2500, which typically used only the first slider to control presets. The general paradigm is that sliders A through D affect V.A.S.T. parameters (primarily filter and envelope) while E through H affect KDFX. This is a welcome addition to the presets, demonstrating how much you can do with real-time control.

The new Setups are equally adept at demonstrating the K2600's capabilities. Many are designed to show off the ability to synchronize the sequencer and arpeggiator functions. For example, you can do a sort of wave sequencing, in which the sequencer sends Program Change messages to the sound played by the arpeggiator. This produces a effect reminiscent of the Korg Wavestation.

A number of Setups start the sequencer and/or arpeggiator when you play a note, and some latch the playback to continue after you release it. You can't activate the latch from a footpedal, but Kurzweil is considering that for a future software update. You can also program the arpeggiator to change keys only on measure boundaries or immediately when you play a different note. In addition to a wide variety of melodic patterns, the arpeggiator plays all held notes in a pulsing, repeating chord synchronized to the internal clock, which is another interesting effect.

CHILDHOOD'S ENDThe K2600 far exceeds its progenitor, as all good progeny should, while offering its predecessor the fruits of its evolution. The only significant throwback is its relatively skimpy polyphony, which hasn't been expanded since the K2500. Its 48 voices are quickly gobbled up, especially by multiple Triple Programs, KB3 Programs, and complex Setups. Increasing the polyphony would raise the price - which is already stratospheric - but most other high-end synths (and many not-so-high-end instruments) are up to at least 64 voices or more these days.

Despite this one shortcoming, the K2600's power, depth, and flexibility are unsurpassed by any other instrument on the market today, and I suspect it will remain the pinnacle of keyboard workstations - both in terms of performance and price - for years to come. If you can afford to outfit an expedition, you won't find a more advanced species anywhere.