Yamaha FS1R

FM synthesis was the success story of the mid‑'80s, and synths based on its principles, like Yamaha's DX7, sold by the bucketload — until affordable sample‑based synths arrived at the end of the decade. Now, with their new FS1R, Yamaha have updated the technology for the late '90s.

When Yamaha launched their SY85 in 1992, it seemed like the end of an era. For the first time since the advent of the hugely successful DX Frequency Modulation (FM) synths in the early '80s, here was a Yamaha hi‑tech instrument without FM in any form. Sample‑based synthesis had seemingly taken over the world, and perhaps FM was looking like tired technology at Yamaha HQ. What goes around comes around, however: after all, analogue synthesis has undergone an astonishing renaissance in the '90s, and more recently, retro interest in FM has been bubbling under, with instruments such as the DX100 and TX81Z attaining cultish appeal in techno circles for their FM basses and electric pianos. In current manufacturing terms, though, FM had all but disappeared. Until now, that is...

Apparently out of the blue, Yamaha have released their first new FM instrument for several years, the FS1R. But — and this is an important but — it's FM with a difference. Not only is it an eight‑Operator variant (like 1982's trailblazing, megabuck GS1), as opposed to the six‑ and four‑Operator FM of the DX/TX range, but it also incorporates a new, potentially very powerful technology called Formant Shaping Synthesis. Formants are, in this case, the spectral patterns making up the sounds of human speech, so one of the FS1R's great strengths is the creation of vocal‑like timbres.

If you've seen any of the limited press info for the FS1R it might come as a surprise to find that this is an FM instrument at all, as the marketing approach seems to be to play down the FM angle and play up the FSS angle. There's one thing you can be sure of: there are no samples anywhere on board, nor is there any physical modelling trying to convince you the FS1R is a saxophone or a vintage analogue synth.

From The Outside In

The FS1R doesn't break with the traditional rackmount synth module look. Indeed, anyone who knows Yamaha's MU100R/90R will be in familiar territory as far as appearance is concerned, as the FS1R uses a similar casing and an identical display. One major difference between the FS1R and the MU modules is the presence of four control knobs on the far right of the FS, which have different functions depending on the mode of the synth, and can be used for real‑time sound modification. Other front‑panel features comprise a headphone socket, volume control, a large backlit LCD (with lettering easily visible even from across the room), and a total of 15 small, round editing buttons, arranged in one group of six and one group of nine. Next up are the four knobs, plus two more small buttons which light up red when activated: these select one of two sets of knob control parameters in Play mode. Finally, there's a large, recessed power switch.

The obligatory look at the back panel reveals a power inlet for the built‑in power supply — whoopee! — plus MIDI In, Out and Thru, and two pairs of stereo outputs.

Yamaha FS1R rear connections.

Inside the FS1R is a synth engine capable of 32‑note polyphony and 4‑part multitimbrality. These days, four is a pretty low number of simultaneous voices, but Yamaha justify it, quite reasonably, by saying that the FS1R wasn't designed as an all‑round studio in a box for the creation of entire orchestral arrangements, say, but rather a dedicated synthesizer whose forté is interesting, individual sounds. You'd have an FS1R as an addition to a general‑purpose workstation, not instead of one.

A total of 1408 preset Voices and 384 preset Performances are on board, plus 128 user Voice and 128 user Performance memories. Three separate effects processors are built in — reverb/delay, variation effects including chorus, flange, phase and so on, and 'insertion' effects, including some basic multi‑effects configurations — and there's 3‑band EQ.

The FS1R & FM

If you're familiar with the workings of any other FM synth, that's a head start when it comes to understanding the FS1R. For those who haven't used an FM synth before, we'll take a brief look at how FM relates to this new instrument.

Put simply, FM synthesis involves the modulation of the frequency of one waveform (the carrier) by another (the modulator). This interaction generates a complex range of subsidiary frequencies, which results in timbral change; the greater the modulation, the greater the change in timbre. As commercially implemented by Yamaha, these waveforms were all sine waves generated by oscillators, and each oscillator (in combination with a series of fairly familiar synth parameters, such as an envelope generator, level, keyboard scaling and so on), was dubbed an 'Operator'. Several Operators — six in the case of the massively successful DX7 — were organised into algorithms, which essentially preset the interaction of the carriers and modulators in a voice. The simplest algorithms laid the six Operators side by side, with no modulator/carrier interaction, offering basic additive synthesis facilities. More complicated algorithms organised the Operators in parallel or serial configurations — a modulator could itself be modulated, for the generation of even more complex timbres. Every algorithm also included a feedback loop, whereby one or more modulators modulated themselves. Algorithms were depicted graphically as linked blocks on the top panel of all Yamaha's FM keyboard synths.

Later FM synths offered variants on these ideas; the Operators on these instruments offered a choice of waveforms, rather than the basic sine wave of (for example) the DX7.

This outline applies pretty well to the FS1R's basic voice architecture, though the new synth is understandably much more complex. Up to eight Operators are configured in an algorithm, and there are 88 algorithms on board (the DX7 offered 32). There's one algorithm per FS1R Voice, and up to four Voices are combined in a Performance.

An FS1R Operator is more comprehensive than a DX Operator, since the basic sine wave is joined by seven other waveform options — or 'spectral forms' in FS1R‑speak — including broad‑ and narrow‑band all‑harmonics, broad‑ and narrow‑band odd‑harmonics, resonant broad‑ and narrow‑band, and formant, for a wider range of potential sounds. Original DX synths lacked a filter, whereas the FS1R includes a very good one, and until late in their run, older FM‑equipped synths had little, or nothing, in the way of effects. Again, the FS1R is well endowed in that department.

But it's at the Operator level that the FS1R reveals its complexity and power, especially in the creation of voice‑like timbres. In order to accurately emulate the workings of a human vocal tract, Yamaha have provided two sets of formant Operators, eight 'voiced' and eight 'unvoiced', that work side by side in pairs — so you could say that the FS1R is actually a 16‑Operator synth. The voiced formant Operators, which are basically oscillators, mimic the larynx, producing the basic sounds, or vowels. Speech needs consonants, however, which brings us to the eight unvoiced Operators. Essentially noise generators (which can be used as such when synthesizing percussion or sound effects), the unvoiced Operators allow the FS1R to imitate consonants.

The FS1R manual says convincing speech can be emulated using three to five formants, so there are evidently enough to be going on with in the FS1R! The synth can even be made to talk (or sing), using so‑called Formant Sequences — more on this later.

Modes

The FS1R powers up in Play Mode, from which you can select and audition its Performances. There's no separate Voice mode as such: Voices are accessed and edited from within Performances. In practice this is quite elegant and logical, but it may feel strange at first to those accustomed to how other synths work. It has three edit Modes, Performance, Voice and Effect, plus a Utility Mode, for global housekeeping chores such as display contrast, master tuning, and global MIDI options. To audition, or change, individual Voices, you need to access Part Assign Mode, wherein Voices are assigned to Performance Parts (four per Performance), and basic Part parameters such as MIDI receive channel, volume, pan, effects send and note shift are set.

At The Operator & Voice Level

In Voice Edit mode, parameters are divided into two groups — Operator and Common. The former are available for each Operator, while the latter apply to a Voice as a whole. The upshot is that there are lots of parameters!

• Operator parameters: Each Operator in an algorithm can be modified by a set of synthesis parameters. The exact parameters vary depending on whether you're editing voiced or unvoiced Operators, and on which 'spectral form' is chosen for the voiced Operators. All Operators, whether voiced or unvoiced, have frequency (pitch) parameters, an amplitude envelope generator (with 'hold' parameter for delaying onset of the envelope), frequency EG (which produces a filter EG effect), and various sensitivity parameters that determine how Operator amplitude and formant frequency respond to velocity, amongst other parameters.

All voiced Operators have detuning, key sync and a range of level‑scaling parameters. One use for the latter would be to split groups of Operators into virtual 'keygroups' in a Voice, enabling a single algorithm to apparently produce different sounds at different points on the keyboard. Unvoiced Operators are cut back to one level‑scaling parameter, which alters their volume response either side of middle C. This is a shame, because there would have been scope for creating percussion sets with up to eight voices had level scaling been as comprehensively implemented as for voiced Operators.

Formant Operators are a special case, though they do have access to standard voiced‑Operator parameters. It's possible to think of a formant Operator as a mini synth, combining an oscillator and filter in one unit, and its sound is shaped by parameters which include centre frequency, level, bandwidth and the amusing 'skirt'. This last parameter determines the shape of the flare at the bottom of a formant's bell‑shaped response curve — higher values produce a wider skirt!

Unvoiced Operators don't have quite as many parameters as voiced, but they do have the same bandwidth, resonance and 'skirt' parameters, and can even be made to resonate to self‑oscillation. During the review, this latter feature was used to provide a couple of excellent sub‑oscillators that, whilst almost inaudible when solo'd, added weight and body to the main sound.

• Common parameters: Additional modifications can be made to an algorithm as a whole, using the so‑called Voice Common parameters. Included in this set of parameters are two LFOs, a pitch EG, and a filter. The latter is in addition to the filter‑like parameters available to each formant Operator, and is more like a VCF on an analogue or S&S synth. It's rather well‑specified, with low‑pass (12dB, 18dB or 24dB per octave), high‑pass, band‑pass and band‑reject options, plus cutoff frequency and resonance controls, and even comes equipped with its own EG. Using the filter does have one unfortunate side‑effect, however: bafflingly, it cuts a Voice's polyphony in half.

At The Performance Level

Performance editing parameters are organised, as are Voice parameters, in two groups — Common and Part.

• Performance Common parameters treat all four Parts in a Performance, and include individual output assignment, controller assignment (whereby various control sources, including the front‑panel knobs and various incoming MIDI controllers, can be used to alter parameters in real time), and the choice of Formant Sequences (FSeqs).

FSeqs are what allow the FS1R to 'talk' or 'sing' recognisable words or phrases. This is achieved by sequencing the frequency, fundamental pitch and level for each voiced/unvoiced formant pair. Only one FSeq per Performance is possible, so Voices will never say or sing anything too complicated, although the presets manage phrases such as "I Love You", "Yamaha FS1R" and "Are you ready?". A nice touch is that playback speed of phrases can be increased or decreased, and even sync'd to MIDI clock. And since the sound of formants isn't pitch dependent, speed and pitch can be varied over a wide range without altering the basic timbre of the sounds — try that with your sampler! FSeqs can also loop, with Performance B 013 Shooby Do a prime example: it sings 'shooby do wop', with the 'wop' bit looping if a key is held down. This Performance also illustrates a nifty triggering option, where the FSeq is triggered by the first key press; as long as one of the first keys pressed is held down, the loop will kick in, and you can play tunes within the FSeq without retriggering the phrase.

Unfortunately, FSeqs are a silver cloud with a grey lining: there are 90 preset FSeqs, which provide a wide range of actual phrases, fun gibberish, laughing and, surprisingly, arpeggio and drum patterns, but there are only six memory positions for your own FSeqs. And it gets worse: there's no way of creating your own FSeqs on the FS1R. A software editor — whether from Yamaha or a third party — will offer the only solution, and none is available yet. And if you could save your own FSeqs, doing so would halve the user Voice memory — you'd have space for only 64 of your own Voices instead of 128. This takes some of the shine off what may be one of the most exciting aspects of any new synth of recent years. If there's a way to fix this with a software or hardware update, Yamaha should be urged to try to find it. We hear that there's a development tool which turns WAV or AIFF files into FSeqs, and it would be fabulous if the public could have access to that — imagine, you could sing or say something and have it transformed into an FS1R synth voice!

• Performance Part parameters: Even from within Performance Edit, it's possible to tweak many of an individual Voice's parameters (for example, LFO speed, filter cutoff, resonance and EG depth, EG attack, decay and release times). However, when you change, say, LFO speed for a Voice within Performance Edit mode, all you're actually doing is adding an offset to that Voice's stored LFO speed value. The Voice itself will sound as though it's been edited, but remains unchanged — only the offset value is stored as a Performance edit. This is excellent and means that you can tweak away to your heart's content at Voices in a Performance, secure in the knowledge that you won't have screwed up that Voice or its appearance in any other Performance.

An easy way of tweaking a Performance without venturing too deeply into the OS is to use the four front‑panel knobs to alter a total of eight parameters (four preset and four you can assign yourself) for all four Parts at once. As above, when a Performance edit made with the knobs is saved as a new Performance, the constituent voices remain unscathed.

There are a few parameters which are unique to Performance edit mode, covering pitch settings (detune, note shift), portamento, pitch bend, note assign, polyphony reserve, mono/poly mode, plus high/low note and high/low velocity values, for splits and velocity layers.

FS1R In Use

Programming and editing the FS1R is very different from programming and editing an S&S synth. For a start, there are no banks of raw samples as a basis for programming — instead there are 88 eight‑Operator algorithms. You don't know what any of the algorithms are going to sound like until you play through them, and a few suggestions in the manual for what each might lend itself to would have been nice. The display does attempt to help, but the graphic it uses to display algorithms is so small it's practically invisible unless you're all but on top of the synth.

Of course, when programming a bass‑type sound, say, you'd start by tweaking a bass preset, and realistically many people will start their own sounds from one of the 1400+ factory presets. DX7 aficionados will like the fact that the FS1R has almost complete compatibility with 6‑Operator FM synths: you can send, via MIDI, a patch from a DX7, for example, and the FS1R will convert it to an identical‑sounding patch in the new synth. A lot of the preset Voices actually come from the DX7's library. It's also possible to program sounds from DX7 sound charts, finding an algorithm that has an equivalent layout to the DX algorithm and turning off the unwanted Operators.

The FS1R isn't the friendliest synth to program. Changing parameters from the front panel can be a bit tricky — the buttons are small, the display isn't ideally suited to the synth, and the many parameters are organised in a hierarchy that's often three levels deep. However, there is an alternative to pecking at the tiny buttons, since the four knobs can scroll though Performance Parts or Voice algorithm Operators, navigate menus, select from lists of parameters, and change their values. And there are nice touches: individual Parts in Performances and individual Operators within Voices can be muted or solo'd in order to fine‑tune an edit, a Search button quickly finds themed Voices or Performances within named categories, and the Play button provides a quick audition of whatever you're editing. The display, though obviously borrowed from the MU family, provides some help: the bottom strip shows MIDI channel, category setting and volume level, amongst other useful information.

But any OS shortcomings the FS1R has are mitigated to a huge extent by the sounds it can produce. It really does sound like no other synth — both lush and digital at the same time, without being thin or cold. If you were to compare its character to anything, in some ways it's rather like the super‑expensive digital synths of the '80s, such as the Synclavier and PPG Wave. Loads of presence, loads of depth, a big sound that's bright without being strident. Some of the factory presets knock you backwards on first hearing (the 'Favourite Presets' box gives a personal 'best of' selection), without being gimmicky. The presence of unvoiced formant Operators introducing a noise element even allows drum‑like timbres to be produced from a synth that has no drum samples.

Some might say the Formant Sequence presets are gimmicky, and no doubt they'll be appearing all over chart records within a very short space of time. But they're great fun, reminding of Kraftwerk and Electric Music, in the way that something that's not quite human, but not quite electronic, seems to be speaking and singing. The FSeqs almost give the FS1R a personality, and will become much more than a gimmick when some way for users to create their own is introduced.

Remember, too, that the FS1R's formant Operators are not just confined to voice‑like timbres — they can make almost any type of sound feel more 'real', partly because the formant component of an FS1R sound is completely independent of pitch. Thus sounds using formants will play much more naturally at the extremes of the keyboard than sample‑based sounds.

Conclusion

Genuinely novel synthesis systems are increasingly hard to come by. The last big new thing was physical modelling, and its main task these days seems to be recreating decades‑old analogue technology. The FS1R isn't completely revolutionary — it features a development of 15‑year‑old technology, after all — but it does manage to offer a sound and method quite unlike any other synth on the market. Even for those who don't want to do in‑depth programming the FS1R has lots to offer; there are tons of quality presets, which can be easily layered into new Performances and then tweaked with the front‑panel knobs, all with no real knowledge of FM needed. And though programming enthusiasts will probably find that the FS1R requires a bit of effort to come to grips with, the results should more than repay their time and trouble. Formant Shaping, in particular, is a serious, exciting achievement on the part of this synth's designers and a breakthrough at any price, let alone at £699. Looks like FM's back to stay!