Amstrad Fidelity CKX100, computerphonic PLAYRIGHT mode, Computer Aided Harmony [CAH]digital keyboard with lovely pop organ sounds, great demos & MIDI

This keyboard is a historical oddity, because although it was released in 1988 (box copyright date), its sound and case style, box design and especially the long and lovely demos resemble much more the early digital sound estheticism from around 1984. The main voice is made from very clean digital dual waveform timbres those imitate FM and resemble very much the great Yamaha MK-100 from 1983, and like the latter it has even a trio mode ("auto harmony") for cheesy pop organ chords.

An unusual special feature of this instrument is the "playright mode", which switches the right keyboard section to pentatonic tone scales, those vary with the actual chord (selected with the left hand) to force the player to play what establishment considers harmonious (a bit like an Omnichord). The cheesy single finger accompaniments are quite over- orchestrated and thus less versatile. The percussion is made from simple low resolution samples and resembles much Bontempi GT 759. There is also a MIDI- out jack and a simple sequencer with data storage on audio cassettes. The preset sound and rhythm selection is a bit awkward, since you have to press buttons multiple times to step through them, and the demos appear between the rhythms. Initially the PCB stank badly of epoxy.

The Amstrad Fidelity CKX100 seems to be quite rare. I bough mine from eBay; my manual, box and warranty card is in Italian, but also an English version exists. Kees Hink e-mailed me, that he bought his specimen around 1990 in Netherlands. Someone else told me that this keyboard was released earlier under different brand names.

I never saw any other keyboards by Amstrad yet. The company Amstrad was known in 1980th for cheap and often poor sounding stereo sets and the famous Amstrad CPC 464/ 664/ 6128 home computer series. In Germany their HiFi and computer stuff was sold under the Schneider brand.

main features:

SERIAL NO. cT 34 9 0 45519
Amstrad Fidelity, model: CKX100, made in Hong Kong

modifications:

notes:

The Amstrad Fidelity CKX100 is a great authentic piece of mids 1980th sound and design estheticism, that was already a bit anachronistic when it came out in 1988. Especially the demos sound like cheesy digital synth muzak from around 1984, and the control panel looks like a mixture of Yamaha DX7 and late 1980th ghettoblasters; e.g. the background of the volume sliders moves with the slider, and the type labels include in a stylish 1970th font the writing "computerphonic". Also the rastered pictures on the box and manual look rather like from beginning of 1980th. Unfortunately the internal program is stored on EPROMs, those are infamous to loose their data over time by bitrot. (But I hope that these late 1980th ICs are more reliable than the first generation EPROMs, those are well hated in early arcade videogames and pinball machines for spontaneous dying.) The preset rhythms are awkwardly selected in a row- column manner; while the row is selected by 7 individual group buttons, the columns can be only stepped through with a single "rhythm select" button; the 1st column even contains the 7 demo musics, those are selected and started like rhythms. When I bought this instrument, the "tempo up" button didn't work by a faulty keyboard matrix diode.

The main voice sound is made from 2 layered static digital waveforms with independent simple envelopes. All preset sounds are quite bright, but (unlike e.g. My Music Center) don't sound rough by DAC frequency aliasing noise or the like, but have a pure and clean high quality synth appeal. The sound style resembles mostly the Yamaha MK-100 sound generator, that was designed to imitate simple FM timbre sweeps by crossfading between 2 waveforms. Like with genuine FM instruments, the timbres are time- dynamically playable, i.e. the timbre and especially the volume of the release phase changes with the key press duration. But the CKX100 timbres have also many similarities with Casio's early consonant- vowel synthesis instruments (especially Casio MT-30), although the latter had only squarewave muffled by low- pass filters, while the CKX100 uses dull waveforms instead of filters and thus has not problems with quieter sounding high notes. But it still can do the characteristic buzzy "enng!" sound in the bass range in some timbres. The "elec piano" is very bright with a slightly scratchy, percussive attack phase. The "synth 1" resembles a grainy, brassy guitar (or banjo?) sound that fades thinner. The "guitar" begins duller and has a too soft attack phase. The "flute" sounds quite dull with some zipper noise in the attack phase; it resembles more a wooden pipe organ rank. The "bell" resembles a vibraphone without vibrato and with a slightly knocking attack phase. "brass" is a grainy synth- tuba with a scratchy dose of zipper noise in the attack phase, which sounds like archaic Casio consonant- vowel stuff or a C64 SID sound. "synth 2" is a bright and thin timbre that goes "enng!" during attack, holds the note like an organ and has a short sustain. Also the "harpso" sounds thin and bright and has a too slow attack phase for a picked string. The "organ" is a plain, but nicely made Hammond organ timbre with mild percussive attack phase. The "string" sounds more like a saxophone and is likely made from a sawtooth wave; its 3rd and 4th keyboard octave play too quiet, which seems to be no filter problem but a badly programmed stereo panning (or a faulty right loudspeaker?). The sustain button adds to all preset sounds a quite long sustain (3 to 4 seconds), while the vibrato button adds a 7Hz vibrato. (The preset sounds contain no own vibrato when off.) These buttons also affect held notes, while the preset sound buttons only change the timbre of later pressed keys.

The percussion of the rhythms is made from thin and unspectacular sounding low resolution samples of acoustic drum kit stuff and some synth toms. Unfortunately the rhythm plays too quiet in relation to the accompaniment, which makes it badly audible. The accompaniments sound quite thin and cheesy, but most are severely over- orchestrated and thus not really versatile. Generally their timbre reminds much to FM soundcard game music on old PCs. The styles are mainly mellow pop, disco and jazz stuff and have little to do with their names; e.g. "hip-hop" is rather a fusion pattern, while "heavy metal" is more a harmonious pop pattern. Some patterns remind to the game "Sonic the Hedgehog" on Sega Megadrive/ Genesis. There are also a few country patterns. By the lack of a fingered chord mode, only a few different establishment chords can be played, but at least it responds nicely fast to break up the monotony. The "fill-in" button mutes the accompaniment during the fill-in pattern. Annoying is that selecting a preset rhythm always switches the tempo back default value.

With rhythm off,  the single finger chord mode plays chords in a fixed brass timbre, which sounds fairly bright and a bit thin. Due to the slow attack phase of the brass sound, the chord is muted for a short time while changing the chord key, which can be used as a sound effect to chop the chord pad. In single finger chord mode the "auto harmony" button turns the main voice into a trio while a chord is played in the left keyboard section. The additional 2 voices can be also trilled (while holding the main voice note) by trilling the chord key. But during rhythm chords (of the accompaniment) are held automatically (known as "chord memory" on other instruments) and thus also the trio sound can not be trilled here.

An unusual special feature of this instrument is the "playright mode", which switches the right keyboard section to pentatonic tone scales, those vary with the actual chord (selected with the left hand) to force the player to play what establishment academics considers harmonious. This concept resembles much the strange "GLING" stuff on the Philips PMC100 portable sequencer or the behaviour of a Suzuki Omnichord.
 

hardware details

The Amstrad Fidelity CKX100 has relatively complex multi-chip hardware built around the CPU "Hitachi HD63B03XP" with wavetable sound IC "ST M114S". The operating system is in a 32KB eprom 27C256, the waveforms and percussion samples in an additional 16KB eprom 27C128 (I dumped these for preservation). The 8KB sram is likely used for sequencer.
The CPU "Hitachi HD63B03XP" is a romless generic 8-bit microcontroller with 192 byte internal ram that can run at up to 2MHz. Technically it is a variant of Motorola MC6803. The solder damaged writing on the crystal looks like 8 MHz, which is apparently divided by 4 as 2 MHz CPU clock and by 2 as 4 MHz sound ic clock.

I haven't analyzed this hardware further yet, but Scott Nordlund and Traktor helped me to figure out some basics. I was told by e-mail that the sequencer data protocol is described on the audio cassette that came with it, but I haven't checked this either.

Scott Nordlund e-mailed  me the following info about the main voice sound chip: (edited by me)

About the Amstrad CKX100, I was very interested to see that the soundchip is an M114S. I have a very rare and obscure Italian synthesizer, the Keytek CTS-2000  (Keytek previously was known as Siel and made shitty string ensembles and self-destructing synths), that uses two of these (along with CEM 3389 VCF/VCA chips). I wrote about it here if you're interested in more info: http://www.geocities.com/diffused_light/stuff/cts2000.html

The M114S chip is made specifically for crossfading waveforms (like a more advanced version of Casio's Consonant-Vowel synthesis) but evidently it didn't get a lot of use. Actually, because of the page I made, someone emailed me asking for information on that chip, because it was used in some obscure arcade game and he was trying to find out how it worked so it could be emulated for MAME. He said he had hit a total dead end trying to find out about it and that I was the only person who had even replied to his request (though unfortunately I didn't have any information and had similarly failed to turn up anything). I told him to drop me a line if he ever found anything but I haven't heard back.
...
This synth actually used a very unique method of sound generation. Rather than store full samples, there were instead single-cycle waveforms sampled from the attack, decay, and sustain portions of whatever instruments it was attempting to recreate. Each oscillator (2 oscillators per voice, 8 voices) is assigned 3 waveforms (probably 8 bit), and simple envelopes crossfade between them. What's interesting about this is that the waveforms can be arbitrarily used in any order. In theory this is more sophisticated than what Roland did in the D50 (short sampled attack transients and single-cycle waveforms for the sustain portion), but unfortunately the sound is rather poor and the user interface is really terrible, it's not surprising that no one bought them. Anyway I was just interested to see that the same chip was used in something else, maybe the CKX100 uses similar crossfading.

> Sounds like a very rudimentary wavetable sound synthesis (ask a search engine about "Waldorf Wave" or "Waldorf Microwave" to see what I mean).

Yeah, it is kinda, or like the Prophet VS, but none of those synths can completely replicate the functionality. Possibly it was inspired by the Synclavier's resynthesis functions: I read that the Synclavier can perform an FFT of a sample, then the user can pick several points on the resulting spectral display, and the sound will be played back on the FM voices by interpolating between these points (called "timbre windows" I think). Clearly this is what the designers had in mind, to cheaply resynthesize instrument sounds instead of storing entire samples.

Conceptually this is great, especially to crossfade together vibraphone waveforms with vocal waveforms, etc. Anyway the sound of the CTS-2000 means that it's not very useful. Even the raw saw/square waveforms sound somehow really bad, I'm not sure if it's the waveforms themselves that sound bad, or the M114S chips, or some poorly-designed analog circuitry, but it's just unpleasant to listen to.

I.e. because the CPU is generic (romless) and all parts are known now (MC6803, M114, rom image), it would be possible to emulate the whole CKX100 on a MAME/MESS core.

pinout M114A, M114S

The Digital Sound Generator M114 by SGS-Thomson is an NMOS wavetable synthesis sound IC controlled by an external CPU. The version "M114A" (48 pin DIL) supports 32 banks of each 8 KB external wavetable rom (256 KB), while the smaller "M114S" (40 pin DIL) supports only 2 banks of it (16 KB). The IC has 16 polyphony channels those can be routed through 4 analogue output pins. The internal DAC supports 12 bit in absolute or delta coding (which needs an external integrator circuit) with additional 10 bit attenuator (volume control). As a wavetable synthesizer this IC can crossfade among any 2 waveforms within the same bank (each 8 table lengths and reading modes form 58 distinct combinations) and handle simple automated pitch and volume ramp functions. However more complex envelope shapes or chorus and vibrato need to be controlled in realtime by the CPU (which can cause sound glitches by waveform distortion when registers are written with wrong timing). The wavetable rom has to be connected directly to this sound IC, not the CPU, which needs only 6 bus lines to control it.

This pinout is based on pages of the 1988 MOS databook by SGS-Thomson, which also includes programming info. A variant M114AF (seen in datasheet version of May 1991) is like M114A, but supports 6 instead of 4 MHz clock frequency. Some pin names were abbreviated by me, because those in the book were ridiculously long (partly over a dozen characters) and completely unsuited for schematics.
 
M114A
pin		 M114S
pin		 name purpose
1 1 GND ground 0V
2   EA1 rom bank select
3 2 A0 rom address
4 3 A1 rom address
5 4 A2 rom address
6 5 A3 rom address
7 6 A4 rom address
8 7 A5 rom address
9   EA3 rom bank select
10   EA2 rom bank select
11 8 A6 rom address
12 9 A7 rom address
13 10 A8 rom address
14 11 A9 rom address
15 12 A10 rom address
16 13 A11 rom address
17 14 A12 rom address
18 15 VDD supply voltage +5V
19 16 CLOCK clock in (4 MHz)
20 17 RESET reset
21 1 AGND analogue ground 0V
22 18 AVDD analogue supply voltage
23 19 Vref dac reference voltage +2.5V out
24   TAB1/2 rom table select
M114A
pin		 M114S
pin		 name purpose
25 20 +12V OUT stepup voltage out (capacitor >100nF to GND)
26 21 AOUT0 analogue audio out
27   AOUTC analogue audio mixed out
28 22 AOUT1 analogue audio out
29 23 AOUT2 analogue audio out
30 24 AOUT3 analogue audio out
31 25 /E rom enable
32 26 D0 rom data in
33 27 D1 rom data in
34 28 D2 rom data in
35 29 D3 rom data in
36 30 D4 rom data in
37 31 D5 rom data in
38 32 D6 rom data in
39 33 D7 rom data in
40 34 BUS5 bus (from cpu)
41 35 BUS4 bus (from cpu)
42 36 BUS3 bus (from cpu)
43 37 BUS2 bus (from cpu)
44 38 BUS1 bus (from cpu)
45 39 BUS0 bus (from cpu)
46 40 STROBE bus strobe
47   EA0 rom bank select
48   TEST test pin

The pin order hints that M114A and M114S both contain an identical silicon die and only differ in their package. According to the datasheet, the 4 analogue audio outs are split as timeslices from a single internal DAC. The AUTC pin (named COMMON NODE in datasheet) taps the DAC before that splitter and thus contains all mixed channels. The +12V OUT pin comes from an internal 5V/14V DC-DC stepup converter and needs an external filter cap (minimum 100nF) to GND; installing here an additional zener diode (9 to 12V) is claimed to strongly improve DAC and attennuator operation.

The CPU must send to the sound IC per active polyphony channel a programming sequence of 48 bit, that is continuously executed by the sound IC until it receives another bit set for that channel. These consist of:

  1. 8 address bits for the 1st table (ext. ROM)
  2. 8 address bits for the 2nd table (ext. ROM)
  3. 8 frequency bits (4-notes and 4-twelfth of note & +/- 1 or 2/1000)
  4. 6 attenuation or amplitude address bits
  5. 4 interpolation bits
  6. 4 channel address bits
  7. 6 reading mode & table length bits (ext. ROM)
  8. 2 bits for choice between 4 outputs
  9. 1 bit for frequency octave change
  10. 1 bit for disable of gradual envelope
Each wavetable (waveform loop sample) can be between 16 and 2048 bytes long (all lengths are powers of 2). The available frequencies depend on the table length (the longer, the slower their loop rate aka frequency due to unchanged memory scan speed) and also various bit meanings change by it. The frequency change behaviour can be switched between asynchronous mode (frequency changes immediately when a new value is written - good for vibrato) or asynchronous mode (frequency changes only at table end - recommended for pitchbend to avoid clicks). Of course there are plenty of other details how the chip is controlled, but this should give you an idea which functions are handled by the sound IC itself and which need to be handled in software by the host CPU.

This instrument has 7 nicely arranged demos, those play quite long and then repeat in a loop. According to the manual they are:

  1. In the mood
  2. Bread and butter
  3. Variazione su brani di Mozart
  4. Blue suede shoes
  5. I can't give you anything
  6. Pub singalong [Medley: "Oh When the Saints Go Marching In" followed by various classic dixieland country stuff]
  7. Money money money
The demos are selected like rhythms, basically like a 0th preset rhythm column that is active by default. The 4th demo "Blue Suede Shoes" is a rock 'n' roll that contains a guitar solo with a lot a bizarre gritty pitchbend effects (rather glissando than portamento), those remind to Rob Hubbard's famous e-guitar imitations in C64 SID musics. During demos you can also switch the main voice preset sound, which is nothing uncommon but sounds particularly good here. Very unusual is that only during demos (and nowhere else) the chord section of the keyboard switches to a drum kit mode with 6 sound samples those have the icons {whistle, hand clap, dog, cowbell, bell} and repeat in 3 different pitches. The "hand clap" sounds quite dull. The "dog" resembles more a cuica or someone saying "huh" while the cowbell resembles a plop noise (like pulling a cork out of a bottle) and the "bell" is the genuine cowbell. As a play training feature you can also improvise to the accompaniment of the demos, which automatically mutes their main voice and enables the "playright mode".

There is also a sequencer that can save data on audio cassettes. But I haven't fully examined it yet since my specimen came only with a Spanish language manual. The sequencer is controlled through the rhythm and accompaniment buttons, using the "fill/ ending" button for shift. It seems to have no edit mode but at least records the polyphonic main voice with accompaniment and preset sound & rhythm changes. But apparently it can not be used without rhythm. Apparently also the 7th demo can be replaced by the sequencer contents (or data loaded from cassette?), but I haven't figured out this yet.
 

 removal of these screws voids warranty...    
WarrantyVoid
back to tablehooters collection
  
 
back