Yamaha have been sitting on the sidelines of sampler developments, quietly taking in the action. Have they used the time wisely and managed to incorporate the best features of other machines in their first sampler?

THE YAMAHA TX16W is one of the new generation of "elegant" samplers that has succeeded the "affordable" line of samplers heralded by Akai's S612 back in 1985. Since then machines like Akai's S700, Ensoniq's Mirage and Roland's S50 (and its relatives) have made the S612 seem quite primitive. Until now, Yamaha have been notable for their absence from the list of affordable sampler manufacturers, though their recent acquisition of Sequential has understandably given rise to speculation about the future. In the meantime, here's Yamaha's only sampler...

Structural Overview

THE TX16W IS a 16-voice sampler with stereo and eight individual outs. It's capable of mono and stereo sampling, and has a set of 16 digital filters that can be applied dynamically to the samples (along with digital "amplifiers"). Sample rates are 16.7kHz, 33.3kHz, and 50kHz (stereo sampling is only available at 33.3kHz), with 1.5Mbytes of 12-bit sampling memory (expandable up to 6Mbytes with optional EMM1 expansion boards). Sampling times are 43 seconds, 21 seconds (mono), 10 seconds (stereo) and 5.2 seconds respectively in a standard machine.

The operating system and samples are stored on a standard 3½" double-sided disk and the MSX-DOS format allows a good deal of flexibility in terms of loading individual sounds and sets of parameters, but has a leisurely loading time of 1¾ minutes (typically 40 seconds for other machines).

The hardware appears to be all digital with a constant clock playback system. The good side of this is that all processing remains in the digital domain, allowing fancier filtering and output mixing, however any amplitude enveloping is also done digitally. This means that quieter sounds use fewer bits to describe them - take something down 60dB in a 12-bit system, and you're only using two bits, resulting in lots of quantisation noise. Having a constant playback clock means that some distortion also occurs whenever a sound is transposed - if you want to play a sample every 19msec when in fact you sampled one every 20msec, you've got to do some fudging. Every constant clock sampler does this, some better than others, but it does mean the bandwidth stays constant across the keyboard (including low notes), preventing clock noise creeping in.

The TX16W has 64 Waves, which are samples with a set of loop points, and 32 Filters in modules containing one of the 16 different filter types and a set of associated parameters (envelope, velocity response, LFO and so on). Next up the chain are 64 Timbres which each include any one Wave, any one Filter and a number of performance parameters (tuning, keyboard response, pitch and amplitude envelopes, another LFO and so forth). Note that a wave and filter may be used by more than one Timbre (helping stretch memory a bit). Above that is a Voice. It contains up to 32 Timbres laid end to end (no overlapping, but individual timbres may fade in and out at their key limits). A Timbre may appear in more than one Voice, and may even appear more than once in the same voice.

Top of the chain are 32 Performances which contain up to 16 Voices. Each Voice in a Performance gets a MIDI channel (more than one voice can have the same, for layering), an assignment of how many and which playback voices it gets (along with output assignments), transpositions and so on. As you've probably guessed, a Voice can appear in more than one Performance, and may appear in the same one several times. A little complex, but no real problem to work with and admirably flexible.

Editing these various functions is divided into several modes - Performance Select, System Setup, Performance Edit, Voice Edit, Filter Edit, Wave Edit, Sample and Utility. Each has its own dedicated, labelled button under the generous 2X40 character LCD. To the right is a keypad including the common numbers plus a sign key, cursor controls, an Enter key, and typical Yamaha On/Yes/Plus and Off/No/Minus keys. You select the mode by hitting the related button, which brings up a menu of one to 13 Jobs. Entering the number of the Job takes you to the selected subfunction (which can take up to a couple of seconds). The Mode keys always take you back to the top of the menu.

Whenever a new Performance is called up, its various parameters are loaded into edit buffers. You can tweak and mutilate a sound to your heart's content, and still recall its original form. This is a great protection device; the slight cost of this is that you have to copy the edit buffer into the real location (any real location - not just the one from whence it came) if you want to keep it. A current oversight is that you can't compare the edit buffer with the original but you can copy it and switch Voices, Timbres, Filters and Waves in and out fairly easily.

All in all, this is a highly structured system that requires a little thought and button pushing to get around, but as structured systems go, I found this faster than any other I've yet encountered.

Modes and Jobs

- Performance Select: Selects one of the 32 Performances. The display does the courtesy of showing you the Instruments in the chosen Performance and which output they're assigned to (left, right, both or individual). These assignments cannot be edited in this Job, however and the same is true of several other Jobs - the display shows you associated information but will only let you edit it in its own mode and Job.

- System Setup: Seven jobs are available in this Mode - Master Volume, Master Tune, MIDI Switch, Control Number Assign, Program Change, Device Number and Protect. Master Volume controls the level and balance of the stereo outputs. Master Tune allows the whole machine to be tuned sharp or flat a half-tone.

MIDI Switch covers who the TX16W listens to over MIDI. Program changes may be ignored, listened to on all channels, or listened to on one specific MIDI channel. Continuous controllers may each be set to be ignored, be received per Instrument on that Instrument's channel or one global channel may be specified to affect all Instruments regardless. This will allow each string of a guitar controller to affect one Instrument or all, but it would have been preferable for them to act normally and have a global channel (for individual string bends and global whammy bar). The unit can also be set to respond to all notes, or just the even or odd notes. This is a substitute for the MIDI overflow mode that appears on some instruments, in that two TX16Ws could be run in parallel to give 32-voice polyphony.

Under Control Number Assign, any incoming MIDI controller number may be turned off or assigned to behave like the mod wheel, breath controller, foot controller, sustain switch, volume or inc/dec switches. Likewise any of MIDI's 128 program changes can be assigned to call up any of the TX16W's 32 Performances. Device Number sets which device it looks like over system exclusive, and Protect allows you to block copying edit buffers into permanent locations.

- Performance Edit: Here you're presented with seven possible Jobs - Voice Assign, Receive Channel, Output, Volume, Detune, LFO, Note Shift, External Trigger and Name (the latter allows a 20-character name to be given to this Performance). All edits here are buffered, and most deal with the individual Voices that make up the current Performance.

The TX16W has an interesting way of assigning its Voices to playback voices; 16 are available for playback, and up to 16 Voices may be fitted into a Performance. You determine how many playback voices a Voice gets, and precisely which playback voice that is. This becomes more important in the two Jobs below. I have mixed feelings about this system - you have a good grasp of what's going on (bass gets voice one, strings get two to five...), but it's not a dynamic allocation system - the string Voice can't get all 16 playback voices one instant, and the bass all 16 the next. Voices and playback voices are interlinked, in that every playback voice assigned means one less Voice allotted.

The Receive Channel/Altemate Assign Job is where you get to select which MIDI channel drives each Voice. If a Voice has multiple playback voices, the TX16W doesn't let you assign multiple MIDI channels for it (cuts down mistakes, but prevents one Voice in an instrument that may want all 16 playback voices from being driven by more than one MIDI channel - but I admit that's rare. If adjacent Voices are assigned the same MIDI channel, they may be stacked or set to rotate through the assigned playback voices.

In the Output Assign Job, you select which audio output a Voice is assigned to - left, right, both or individual - but restrictions on the individual outs. Only the first eight Voices may have an individual out, the Voice must be placed in the position of the output it wants and if one of the first eight Voices goes to an individual output, all of the first eight must go to them. A Voice may not appear in the stereo mix and an individual output, nor may one Voice share an output with, or cut off, another Voice - severe restrictions for drum programming.

Each Voice may have its own volume and detune settings. The detune range is +/-7 and is very subtle; the manual recommends using it for detuning stacked Voices (ones on the same MIDI channel). Each Voice may also be individually transposed in the MIDI Note Shift Job +/-24 semitones.

The Performance LFO may be a sine, triangle, ramp, sawtooth or square wave. Speed is adjustable, and the LFO may be synchronised to restart when a new note is struck. In this Job, a fixed modulation amount may be applied to the amplitude and/or pitch, with a programmable delay. MIDI control of this LFO amount can be set in the Voice Edit Mode. All controls in this Job have useful ranges.

External Trigger allows you to substitute an audio signal (such as drum hit on tape) for a MIDI note plugged into the Ext Trig jack on the front panel (the manual claims a footswitch will also work, but mine didn't). Threshold may be set, along with a gate duration (how long the note is held on, and how long until the next trigger is recognised). I found the TX16W responded pretty quickly, but the gate function was a bit touchy - I'd have preferred a finer selection of the shorter values.

- Voice Edit: Thirteen jobs here: Slot, Wave, Filter, Pitch, Veloc, AEG, PEG, LFO, AMS, PMS, Veloc Bias, Pitch Bend and Name (the latter allowing each Timbre to have a 10-character name). All except Slot deal with the individual Timbres that make up the Voice.

As mentioned earlier, there may be up to 32 Timbres per Voice. These are fitted into Slots. Each Slot has a Timbre, low and high key (in music, not MIDI, terminology), and the number of keys (up to nine) that it fades in and out from at its limits. No positional crossfades may be set up here; they are created by layering two Voices with Timbres that have overlapping fade areas. Cumbersome, but this is a strict machine.

Wave and Filter Assign let you pick the Wave and Filter used in the currently selected Timbre. It's a little odd to "Slot" an empty Timbre first and then fill it, but you can also change the Waves and Filters in existing Timbres here. The Pitch Job lets you select the key that the Wave sounds at its root note, and has a fine tune control.

You can design your own Velocity Curve, including two break points and levels along with slopes from the first point down to minimum velocity and the second point to the maximum velocity (see Figure 1). The manual suggests drawing your desired curve beforehand (helpful with envelopes too), and will display the curve at a touch of the Enter key. This curve can be turned off, and the overall volume of the Timbre may also be set in this Job.

AEG and PEG are Amplitude and Pitch envelope generators, which are just like the four level/four rate EGs on Yamaha's DXs. I found the AEG rates and levels a little unbalanced (hopefully they'll be fixed in a future update); on the other hand, the eight-octave range of the PEG was handled very smoothly.

Each Timbre has its own LFO aside from the Performance LFO. It uses a triangle wave with adjustable speed, pitch amount and amplitude amount. The ranges are very small and the effects very slight but are just the ticket for adding "natural" amounts of vibrato and tremolo. AMS and PMS are the Amplitude and Pitch Modulation Sensitivity. How the Performance LFO reacts to mod wheel, foot controller, aftertouch and breath controller may be set for each Timbre. Velocity Bias Sense gives you the same control over the velocity at which the note was originally struck. Finally, the Pitch Bend sensitivity may be set from nothing to 12 semitones, and be smooth or stepped. High marks to Yamaha for the modulation section on the TX16W.

- Filter Edit: Here reside the digital filters. There are seven jobs in this Mode - Table, EG, LFO, Scaling, LFO Sense, Bias Sense and the ubiquitous Name (10 characters, just like Timbre).

There are 16 filter shapes to choose from - low and high pass filters with resonance, narrow and wide bandpasses, two of each normal, low and high pass filters (having different frequency ranges), a shelving filter that can go from low pass to high pass, one that acts as a notch or bandpass, straight notch, combination notch and bandpass with a lot of ripple (called a "peak" filter) plus two low and high pass filters with variable slopes and a good deal of ripple. (According to Yamaha more types will become available in the future - loaded from disk along with the O/S.) In the back of the manual are detailed graphs of all 16 currently available.

Each filter has frequency and depth (or slope, in the case of the last four filters) parameters. Depths are between 20dB-50dB ranges (or up to 4dB/kHz, for slopes); frequency variance covers 2-3kHz (one to five octaves, depending on range) for all but the resonate filters (they have an absolute range of 2-12kHz). You determine which one of these parameters remains fixed (and give it a 0-9 setting). The other may be set over a range of 0-99, and can be varied by a Filter EG (like the PEG in Voice Edit), its own LFO (just like the Performance LFO in Performance Edit), LFO and Bias Sensitivity (as in Voice Edit), and keyboard scaling (with a selectable breakpoint and end points for tracking the keyboard). The idea of having so many filter shapes (or Tables) with so much control is a very exciting concept, and one I hope other electronic instruments (regardless of sound source) copy.

It is with much sadness that I break the news that they don't work all that well. The depths of effect and frequency ranges tend to be small and sound more like gentle equalisation than the in-your-face effect of the analogue filters in synthesisers. Good for subtle colourations, but nothing drastic. Now the worst part - I found a third to half the filters clicked or choked if dynamically varied. That means moving a mod wheel or applying an envelope often has disastrous effects.

- Wave Edit: There are seven Jobs here - Load to Buffer, Trim, Loop, Loop Cross Fade, Reverse, Mix and Name (eight characters in this case). Waves must be loaded from a location into a buffer to be edited; the best of this is if you don't like what you've done, you can reload the undamaged original. Waves can be loaded in pairs for dealing with stereo samples.

Trim is pretty straightforward; you can move the start and end points by blocks (a block is 64 samples). Afterwards, you can normalise the sample to its full dynamic range. Loop points may be set by blocks and individual samples. There is an AutoSearch function that looks for loop points at zero crossings with a similar slope as its partner. Only unidirectional loops are allowed. When done (as with Trim), the TX16W "collects garbage" (reclaims unused sample memory), which it seems to do automatically whenever a wave is altered.

Loops may also be crossfaded, with the length of the crossfade selectable in blocks. Since crossfade looping alters the actual sample, it's nice to be able to "undo" the attempt with the Reload function.

The Reverse Job has a few tricks up its sleeve. Not only will it reverse a sample, it will reverse portions of a sample and even mirror parts of the wave back on itself. Mix allows two samples to be appended, mixed, crossfaded and processed in general. A two-screen Job, the level, fade points, fade rates and amount of overlap can all be set (see Figure 2). Both of these are Jobs where you'll want to draw out what you're doing ahead of time on paper.

- Sample: Three simple Jobs here - Frequency, Level, and Record. Length may be set in the Frequency Job as blocks or milliseconds (nice to have both), and the available memory may be checked here.

You can select one of six different ways to start sampling: by a threshold set in the Level Job (which turns the LCD into a VU meter with peak detector); when the Yes key is pressed; by stomping on a footswitch; by a signal hitting the Ext Trig jack; by a combination of auto and footswitch or by a combination of the external trigger and footswitch.

The VU meter in the Level Job works fairly well. It seemed to miss a few peaks, and when I ran an identical signal into both inputs in stereo mode, sometimes the left and right meters read differently. A small bug.

The Record Job is, of course, where you sample. After sampling is completed, you may audition the sample at different pitches by hitting the numeric keys - another thoughtful touch. Sampling a sound again may be initiated immediately from this screen.

How does it sound? I sampled an Alesis HR16 into the TX16W and my trusty Prophet 2002, and thought the Yamaha sounded more accurate than the Prophet. The TX16W does have some problems, though. Transposition causes a little distortion and occasional low-level sidebands. Quantisation noise is worse on untreated samples than the 2002, and because the Yamaha is an all-digital system, there is simply no way to get rid of it. This quantisation noise comes in whenever the sound is reduced in level - by envelope, filter, LFO, or by fading. Also, the image of stereo samples shifted slightly with each triggering. To be a true stereo sampler, the paired voices must always be started in exact sync. The Yamaha is close, but doesn't quite pull it off.

- Utility: The final seven Jobs are fairly self-explanatory - Store (for saving the contents of the edit buffers), Disk Load, Disk Save, Format, Init (clears out Performances, Voices, Timbres, Waves, and Filters, individual or in groups), Disk Copy, and MIDI Dump (the TX16W supports the MIDI Sample Dump Standard).

Exhaustive End

BOTTOM LINE TIME - what do I think of the TX16W? Mixed emotions - as far as structure and features go it's a thoroughly thought out machine with just a few rough spots (positional and velocity crossfades having to be performed by layering two Voices). The filters are a great idea - they just don't work right yet. And though sound quality is good, excessive quantisation noise lurks around the corner every time you have to change the sound's level. With slightly better sound quality, I wouldn't hesitate to buy this machine. As it is, I'll wait for a 16-bit version with filters that work...

Prices £1999 basic 1.5Meg: EMM1 memory upgrades, £449 per 1.5Mbytes; full 6Meg memory £3000.

More from Yamaha, (Contact Details).