Digital Sound Effects System
An exciting sound-sampling effects generator
Broadly stated, the DX-1 is a sound sampling add-on for the Apple II or IIe. In fact, it's such an effective little beast that I'm only surprised no one has thought of doing this before on the evergreen Apple. After all, Mountain Computers' multi-channel A/D and D/A card has been around for four years now, and there's no reason why it shouldn't have been pressed long ago into the more-or-less musical service of sampling and regurgitation. Perhaps it takes Fairlights, LinnDrums, and Emulators to alert the add-on marketplace to something as straightforward as the delight of being able to sample your Grandmother's snores...
So, what do you get for the £150 that the hardware and basic software costs? Well, an ADC and a DAC aren't part of the standard Apple, so a card kitted out with the relevant chips has to be plugged into one of the expansion slots on the motherboard. For those who like chip chat, the silicon slices important to the DX-1's story are an AD7574 (ADC), an AD7528 (dual DAC), and a CA3080 (VCA). Basically, what happens is that the line level or mic input (selected with a jumper on the board) is preamplified and directed to the AD7574 for A/D conversion, using the successive approximations technique to provide an 8-bit output that goes onto the Apple's data bus.
The AD7574 is operated in what's called the 'ROM interface mode', which makes the conversion process very straightforward as far as software is concerned. A data READ instruction to the chip's address location results in the RD line going high. This automatically restarts the conversion process, with the BUSY line going high once conversion has been completed. In fact, to vary the sampling rate, all one has to do is strobe the RD line with READ instructions, and let the chip get on with its job. This way, the DX-1's software is able to vary the sampling rate over the range of 0.78 kHz to 23.2 kHz and lodge the sound data in 24K of the Apple's memory.
Now, unlike the Fairlight, which has a separate 16K chunk of memory for each voice, the DX-1 adopts the more flexible approach of the Emulator, where all voices share the same memory. Of course, this is forced on the DX-1 by the architecture of the Apple, but the beauty of this approach is that you can allot more pages of memory to samples that need a larger RAM soundbase than those that are shorter in length or lower in bandwidth. However, unlike the Emulator, which has multiple DACs for as many voices, the DX-1 has to make do with just one, switching from one sample soundbase to another, according to software/user instructions. In fact, there is a second DAC on the board (the other half of the 7528), but this is used for setting the output volume level via a VCA (the CA3080 chip).
The return journey of a sampled sound starts with the initialising of various parameters, including the start address of a particular soundbase plus its length, the playback sample rate, and the playback volume. For instance, the pre-recorded Drum Set provided with the basic software breaks down as follows:
SOUNDBASE 1: Drum Set
REC RATE: 3
REC MODE: 2
PLAY RATE: 5
OK, that's the tough stuff over. Now, we move on to actually playing with the thing. On booting up the system disk, which, for once, isn't copyright-protected (whoops of joy), various defaults such as board slot and volume are assumed by the software, and one of the three 96-page soundbases on the basic disk is loaded into memory. All the sampling, regurgitation, and sequencing programs have their own menus which are approached from a main menu.
The first two selections are really what I'd call fun and quick try-out options. 'Sound Samples' basically does what the menu options suggest it does. For instance, 'Cycle Sounds' is a routine that plays each sound in a soundbase five times, at increasing sampling rates, before moving on to the next sound. By comparison, 'Falling Object' repeatedly plays a sound, but progressively chops off portions of the beginning, creating a sort of 'bouncing ball' effect. 'Roller Coaster' operates in a similar fashion, but by doing an axe job on the end of the sound.
'Preset Rhythms' is a fast way of trying out various rhythmic combinations of the sounds in a soundbase. Unfortunately, most of these are pretty unimaginative, the one exception being 'Rock 2', which is a strong, driving pattern of the 'electro-rock' type. One limiting factor behind using the DX-1 as a drum machine is that you can only get one drum sound out at once (because there's only one DAC receiving sound data), so a continuous hi-hat pulse plus an on-beat bass drum is out. Actually, that's not 100% true, because you could, in theory, make samples of combined drum sounds, and then use these to spice up a pattern with the occasional illusion of a two-handed drummer at work. Of course, if you opt to construct your own custom drum set, you've got to find your drum samples, but, take a hint from me, 'Drum Drops Vol. 3' is a great source of inspiration!
What about the drum sounds themselves? Well, I must admit to being bowled over by what I heard from the DX-1. I tried putting the output through a decent combo and, on most of the sounds, it was hard to believe that there wasn't a drum set or sampling drum machine behind the action. What's more, friends reacted similarly! It's certainly quite a shock hearing such good drums coming from an Apple - especially after being accustomed to the rather woolly and undynamic sounds of the Mountain Computer MusicSystem. Of the eight drum sounds in Soundbase 1, the only two I have any reservations about are the cymbals. The 'crash' certainly crashes very effectively, but the cut-off is on the abrupt side; the same is true of the 'ride', which has plenty of initial metallic zing but cuts off before the shimmering has had a chance to establish itself. But when you consider that the cymbals have been sampled into just 6K, and that most of the sampling drum machines are now using 32K of ROM for such sounds, the DX-1 is still doing remarkably well even with these notoriously difficult sounds.
One thing that Decillionix might consider doing is to use the trick of envelope-shaping (with the on-board VCA) a continuously cycled output from a cymbal soundbase (the technique used by the MCS Percussion Computer, in fact). That way, a faster recording sampling rate could be used to get a crisper top end, and the VCA could be used to shape the sound dynamically as required.
The third option on the main menu takes us away from all this preset stuff to the interactive side of the DX-1 - the real-time record and play facilities. This is where the fun really starts. The left of the picture shows the various rec/play modes, ie. the apportioning of the Apple's memory to individual sounds. We've already come across mode 2 in the case of the drum set sound-base, where the last two samples are allocated three times the memory of the first six. By way of contrast, mode 3 divides the 24K total sound memory equally into eight, and subsequent modes go about their business by allocating more memory to fewer samples.
According to whatever mode you're in, sounds can be played by pressing relevant keys on the Apple's keyboard. So, for instance, with soundbase 1, keys A and S play the snare drum, keys D and F the first tom-tom, and so on. Note that pairs of keys are provided so that you can play rolls. It's little touches like that that really warmed me to the DX-1. Selection of reverse or forward playback through a sound sample is achieved with the cursor keys, the current status being indicated by the arrows at the head of the right display column.
Recording is equally straightforward. First, the REC RATE has to be set bearing in mind the relationships between the members of the sampling menage a trois, ie, rate, quality, and length. For best results, a value of 5 suits percussive sounds, whilst 10 is OK for speech. One thing that's slightly confusing is that the REC RATE entered is actually inversely correlated with the sampling rate. That's also true for the PLAY RATE, so playing back a sound at a slower rate (and, therefore, a lower pitch) entails entering a larger PLAY RATE value. Not surprisingly, all this entails a certain amount of head-scratching.
The FREERUN option on the menu allows continuous input and output via the DX-1 board. The idea behind this is to allow input levels to be set so as to optimise S/N ratio, minimise distortion, and so on. The next step is to set the trigger level, so that sampling commences only once a particular input level has been achieved. The 'R' key can then be used to toggle the system into record stand-by. Pressing the left key of the pairs active within a chosen mode starts the sampling process, and pressing the corresponding right key plays the sound back. You have to watch that you don't monitor at too high a level, as otherwise all you'll end up sampling into the Apple is feedback howl, which is of no earthly use to Man or beast!
Once you've sampled whatever sounds you want in a particular soundbase, the 24K of memory can then be saved to disk. Alternatively, you might want to try out a few rhythms or pitch changes on the material, in which case you can go for the first option on the screen which takes you back to the main menu. Aside from the options we've already covered, such as Sound Samples and Preset Rhythms, there's also the delights of Auto-Sequencing.
Again, this option has its own menu, and one way to start investigating it is to try out the ten demo sequences already on the disk with the sound samples you've just ensnared and dissected into byte-size pieces. Basically, the sequencer allows control of six different parameters for each sequence event:
1 ADD - Address (in pages, 54-149)
2 LEN - Length (in pages, 1-96)
3 DUR - Duration (relative to the time it takes to play one page)
4 PIT - Pitch (1 = high (fast), 254 = low (slow))
5 DIR - Direction (0 = forward, 1-254 = reverse)
6 VOL - Volume (0 = quiet, 254 = loud)
Four groups, each with a maximum of eight events (an example is shown in the lower half of the picture), can be entered using the SEL/LIST GROUP and ENTER STRING options, and these can then be chained together by a further group which sets up repeats of these groups. In fact, it's rather like the 'patterns' and 'tracks' of the Roland TB-303. However, those six parameters make the sequencing of the DX-1 a lot more flexible than most of the competition. In fact, the sky's almost the limit when you're able to leap into a sound sample at any page address, switch from forward to reverse regurgitation, and vary the pitch and volume over such a wide range. All in all, a powerful musical tool.
Mind you, life's not all sweetness and roses - especially when you try producing a rhythmically precise sequence with lots of pitch changing. The problem is that altering the PIT parameter also alters the time that it takes to play through each page. So if you're after something like a run of semiquaver burps going up in pitch, the only way that they'll come out approximately in time with each other is if you also manipulate the DUR and LEN parameters at the same time to balance things out - and that's not easy. Also, there are big problems in getting out playback that's pitched accurately enough for the more demanding musical applications. More on that anon...
A further item of software, called 'Echo', adds real-time audio processing to the capabilities of the DX-1. Unfortunately, the ease of use of the basic DX-1 software is nowhere near as apparent with this addition. This criticism also applies to the manual's explanations of what's actually going on, software-wise. For instance, in describing the looping side of the DX-1's echo business, the manual states, "the outer loop is the outermost echo loop in any given echo selection. The outer loop contains 1 to 146 middle loops. The quantity of middle loops corresponds to the length value. Each middle loop consists of 256 inner loops". OK, but nowhere in the manual is there an attempt to equate these loopy (sorry) ideas with the more meaningful (to micro users, anyway) term of 'pages of memory'. Since the manual for the basic software constantly uses this term to help explain what's going on, it's obtuse in the extreme to confront the user with an entirely different line of reasoning in the Echo manual. In fact, it appears as if the Echo software is an entirely different product, written by an entirely different person.
Anyway, nit-picking apart, the software of 'Echo' sets out to provide a delay line with all manner of control points as to the start, end, and speed of the process. A feedback pot is also provided for connection with the DX-1 so that regeneration can be added. Invoking these control variations is a question of choosing one of the 50 echo routines (a reference card is handily provided as an aide memoire) and pressing various other keys to change speed, volume, and so on. Apart from straight echo, the software also makes plentiful use of random cycling of variables to create unusual (and generally unmusical) effects. The best effect I got was with the 'H' echo routine, where the looping is limited to just 1 or 2 pages of memory. Changing the speed of playback (the sampling rate, in other words) alters the pitch and bingo, you've got a nice metallic-sounding harmoniser.
With high quality machines like the Drumulator and LinnDrum around for comparison, it'd be very easy to tear the DX-1 to shreds on basic hardware considerations, but that'd be silly as the DX-1 is way below their price range, and was probably never intended as anything more than what its makers call it - a 'digital sound effects device'. In fact, it more than makes up for its limitations with ingenious software, and I'm sure it will find much favour with Apple-owning musicians in search of sampling facilities. That said, there are areas needing attention.
The jumper selection between mic and line inputs is a pain, and an extra socket for line input would be a sensible addition. With lower sampling rates, the frequency content of the input signal becomes a fairly critical factor in determining playback quality, and some sort of low-pass filter would be a welcome addition to the input stage. Ideally, it'd be nice to switch the output of the volume DAC to controlling a VCF when recording, so that the input cut-off could be set according to the sampling rate.
The fact that the 8-bit conversion is non-companding (unlike, for instance the Oberheim DX, Emulator, and Drumulator, to name but a few) means that the DX-1 isn't going to be as noiseless as these machines. However, as the DX-1 board is silent unless a sound is being strobed out, the only time you really hear the noise is when low bandwidth sounds are recorded at a low sample rate. Where this is particularly apparent is with the Echo and Sound Processing software, because of the continuous input and output that's going on. One obvious improvement would be to replace the ADC and DAC chips with companding chips, but that's obviously a major hardware re-design job. In fact, the non-companding MXR Drum Computer demonstrates rather well that standard 8-bit conversion can do a pretty convincing job of discontinuous sound regurgitation, and the DX-1 appears to be equally effective.
On the whole, the software is extremely good, but there are a number of improvements that could be made to improve the DX-1's position in the sound sampling stakes. First, I don't see any reason why the software limits itself to the basic 48K Apple: the majority of users now have 16K RAM cards in their machines, and moving DOS onto the RAM card would raise HI MEM by 10K, giving 34K rather than 24K for sound storage. Secondly, the system is crying out for a sync facility to enable it to be interfaced with MC-202s, TB-303S, and the like. Why not use the annunciator I/Os on the games socket for that purpose - like Syntauri, for instance?
Thirdly, there's the rather more significant problem of pitch and tempo resolution. At the moment, not only is it impossible to get out a sound at exactly the pitch sampled, but, more importantly, all the sample rate resolution is concentrated at the slow end of the scale. For instance, if you sample at 15.9 kHz (a REC RATE of 5), then the nearest playback sample rate is 15.6 kHz (a PLAY RATE of 7). Selecting a PLAY RATE of 6 raises that to 17.0 kHz, 5 to 18.5 kHz, and so on up in rather large steps. What this means is that it's a hit and miss affair persuading a DX-1 sound to coincide with equal-temperament tuning. That's unlikely to be a problem with drums (apart from the tuning of toms), but it's clearly disastrous if you're trying to make the sequencer deliver an E-flat arpeggio of woofs!
Fourthly, the on-board VCA could be made to do more than just altering the volume for each soundbase that's trotted out of memory. Why not shape the sample playback with a software-generated envelope? That'd probably give the DX-1 the edge on some sampled drum machines. More importantly, adding an envelope to longer sounds that invariably get cut off in their prime (cymbals, for instance) would enable them to be more naturally decayed than with the usual abrupt transition to zero amplitude when the pages run out.
I gather from Dan Retzinger, the designer of the DX-1, that Decillionix are now conversing with Syntauri and Passport Designs with a view to interfacing the DX-1 with their respective keyboards. I assume the aim is to turn the DX-1 into something like the Mimic (where is that, by the way?), ie. a monophonic, sampling keyboard. To do that, they'd certainly have to sort out the pitch resolution problem, though I don't really see any great problems there, provided they also get their sampling and playback routines more equally balanced than at present. As it stands, then, the real strength of the DX-1 lies in the wonderful things you can do with the auto-sequencer, rather than being able to prod things in real time.
Something that I still can't get over is the degree of realism obtainable from a sound stored in just 8 pages of memory. 2K isn't much when you've got to fit the idiosyncracies of a snare drum, hi-hat, burp, or bark into a sampling system, but it works incredibly well with the DX-1. When I think of the fairly awful quality of the Chamberlain waveform-sequencing approach (through 15 or so pages, mark you), I can't help feeling there's an important lesson to be learnt here. Perhaps it all comes down to the fact that the DX-1 sampling approach captures inharmonic components as well as harmonic, whereas waveform-sequencing is, by definition, limited to those that toe the periodic line. It sure makes you think...
Dynamic sound quality
Ease of use
Flexible sound processing
Friendly and informative manual
Jumper line/mic input selection
Lack of input filter
Lack of sync facilities
Under-usage of VCA
Lack of 16K RAM card option
Lack of mixing soundbases
Noise with Echo software
Echo user interface
DX-1 System (includes basic software) $239
Echo and Sound Processing software $149
4 disks of prerecorded samples (Vols 2-5) $79
Availability: either from Computer Music Studios, (Contact Details), or direct from Decillionix, (Contact Details).