Sound Sampling System for Apple II/IIe
Mainframe's sound-sampling add-on for the Apple is now available to the general public. David Ellis puts the system through its paces.
Originally developed simply for the band's own use, Mainframe's Apple-based sampling add-on is now available to the general public at an attractive £250. David Ellis
For once, there's very little needed in the way of an introduction about either the subject or the product: 'Sampling' in some way or another is just about the most popular musical pastime around at the moment, while the band Mainframe have occupied a good few television screens over the past few months, showing how it's done with their self-built Apple II/IIe-based sound sampling system, the Greengate DS3.
Well, 'self-built' is perhaps a slight exaggeration. The truth is that John Molloy and Murray Monro - Mainframe's central core - were fortunate enough to have a tame hardware/software designer in tow by the name of David Green, and the DS3 is really his baby, though supported by the umbrella of Greengate Productions, the company that the entire Mainframe team set up to market the final commercial product. What's extremely unusual, though highly laudable, is to find a rock band taking the bull by the horns, so to speak, and going into the marketplace with a product born out of kitchen table R&D and rather less than megabucks of financing. So, it's hats off to Greengate for getting this far!
However, Mainframe/Greengate Productions do have a slight tendency to the hype syndrome which tends to detract from their more positive virtues, especially when they insist upon saying that they're 'the most computer-literate pop/rock band of the '80s' and claiming that the DS3 system 'rivals the performance of the current 'big' units costing up to one hundred times the DS3's modest £250'. Now, I can't prove or disprove their first claim, though I'm sure that Thomas Dolby acolytes and a good many Fairlight/Synclavier owners would have something to say on that matter, but I hope that what follows should at least put the DS3's capabilities in the right pecking order alongside the competition.
The first page of the DS3 manual informs you that 'that small printed circuit board with its load of 'chips' is a jewel, a masterpiece of modern electronic engineering which could not have existed a few short years ago'. Just in case you hadn't guessed, what it's referring to is the card that plugs into one of the Apple's expansion sockets. And whilst it's true that there's a load (well, 19) of chips on it, I wouldn't exactly describe it as a masterpiece. More specifically, the card does three things:
1 It converts line-level analogue signals fed into a single op-amp input stage into data that gets whisked off to the Apple's memory. Sampling is eight-bit via a well-scrubbed ADC (presumably of Ferranti type) clocked at a rate of 30kHz, but no low-pass filtering, compansion, or pre-emphasis/de-emphasis is involved.
2 It converts data from the Apple's memory into four analogue output channels that go straight from four ZN428E DACs to an 'audio box' for connection to a mixing desk or hi-fi. No low-pass filters or buffers are provided to separate the DACs from any subsequent output load, and the mono output option is achieved by shorting the output of all four DACs, resulting in a total which is less than the sum of the parts.
3 It scans a five-octave passive keyboard (marketed for an extra £200), which plugs into a 16-way IDC connector situated above one of the scrubbed-off 6522 VIA chips. This keyboard is internally identical to the keyboard used by Passport Designs' Soundchaser: however, the difference in connectors means that an owner of both the DS3 and Soundchaser is obliged either to purchase both keyboards or to cobble together an adaptor to use one with the other. The alphaSyntauri keyboard will not work with the DS3 under any circumstances.
Viewing this as an example of 'modern' electronic engineering, I'd make the following comments. First, a low-pass filter prior to the ADC input is important if you want to avoid nasty noises (aliasings) on playback at low pitches - especially if sampling rates are used where the Nyquist limit is less than the highest frequency being sampled.
Second, the combination of compression/expansion and treble pre-emphasis/de-emphasis (used in units costing up to one hundred times the cost of the DS3...) is the industry's standard way of reducing the noise that's inevitable with eight-bit sampling.
Third, DACs don't take kindly to being connected directly to the hi-fi or mixing desks, and get a bit narked if their outputs are shorted together. Come to that, the average impulse current drive amplifier (or whatever it is in non-Star Trek terminology) and tweeter won't like it either - especially if ultrasonic clocks et al are making their unhindered way through the wires.
Fourth, the DS3 keyboard is a dumb keyboard: so are those with the Soundchaser and alphaSyntauri systems. The outlook for the 5000 or so owners of either of these systems is therefore doubly dumb if a company with a new Apple sound synthesis addon fails to take them into account. If Greengate had really got their thinking caps on, they'd have cottoned on to the fact that a superannuated micro like the Apple needs every dangled carrot it can get in order to attract users. An obvious move would have been to make the system compatible with the Soundchaser and alphaSyntauri keyboards (as Decillionix have done with their DX1 sampling system, for instance), but even better would have been a MIDI In facility (at least) on the card so that the system could be used with anything chucking out bytes of MIDI data. Personally, I think it's pretty thick to expect someone to buy a £200 dumb keyboard that's of no use to man or beast once outside the DS3 system.
To be honest, I'd feel more kindly disposed towards this item of hardware if less time had been spent scrubbing-off chips and more on the finer details of interfacing. To my eyes, removing identification numbers on big chips with an impressive leg count looks designed to fool the customer into thinking that the hardware contains some special custom chips rather than just common-or-garden VIAs.
Well, those are the major criticisms off my chest.
Despite all that, the curious thing is that what the DS3 actually produces sounds remarkably good. To be sure, if you sample a noisy signal, it'll sound even noisier when it comes out, but a clean, percussive signal sounds pretty good over quite a wide range of playback pitching. Part of this lies with the fact that the present software limits the recording sampling rate to 30kHz, so aliasings aren't going to be much of a problem unless you sample something that's way over the top in terms of upper harmonics. In addition, the sampling software is sufficiently flexible that if some noise is apparent after you've put a sample into memory, it's usually quite feasible to edit it out - but more on that anon.
Where everything stands or falls is in the software, and not just as far as functions and features are concerned: 'usability' is equally important if you want to keep musicians on your side. So, contrary to ever-popular belief, appearances are all-important, and a messy display and convoluted commands won't help one iota.
The DS3's software actually comes on three disks: one (the system master) holding all the programs that run the DS3, and the other two containing sample and sequence data. Because of the way the system is structured, programs are loaded up to order from the system master disk in order to do a particular task, whether that be sequencing, sampling, editing, or whatever. Which means that the system master disk is forever being accessed for this or that. Which means you can use a single disk drive, but you'll find it exasperating forever swapping system master and data disks according to the wants and whims of the software. Which means you forego the pleasures of a dual disk drive Apple at your peril...
The starting point of all this program-loading action is the DS3's main menu. The first thing to note is the message 'Ramcard DOS loaded'. This means that the Apple's 10K DOS (Disk Operating System) has been shifted onto the language or RAM card in slot 0, freeing 10K of memory for extra sampling space: good news. To perform some sampling, the first menu option is selected. The disk drive then whirrs into action looking for the relevant program, and a fairly short time later, you're ready to put theory into practice.
There are four further options at this juncture: sample, trim, play, and save. The first thing to do is to set the threshold at which you want sampling to commence. Otherwise, you'll just end up sampling your own breathing, the Grandfather clock, or a load of neighbourhood juggernauts. But before you actually take the plunge and press the space bar to start sampling, it's worth getting levels adjusted so that the signal-to-noise ratio is optimised. Fortunately, this is particularly easy with the DS3, as a real-time oscilloscope display shows whatever's happening input-wise, though one thing that's missing is any Y-axis calibration of amplitude, a feature which would help a lot to avoid the dreaded clipping. As the manual says: 'the reason you wish to avoid clipping is that when played back, a clipped waveform produces a distorted sound and a distorted sound is of no use to man or beast'. Quite right, too. If only they'd followed their own advice when constructing the disk of sample samples!
Initially, the sample goes into 32K of memory. To chop this down to more reasonable lengths, there's a trim option which enables you to whittle away at the sound from either end. What I found intriguing about this is that even a fairly long sound like 'One' can be trimmed down to as little as 6K while still retaining ail of the essentials. That's important bearing in mind that there's only a limited amount of memory to play with. The next stage of the game is to play the sampled sound, and two options are available to the user: first, to trigger it from the QWERTY keyboard at the pitch it was sampled: or second, to play it from the DS3 keyboard over a five-octave range. Needless to say, if you've only got the basic £250 DS3 (ie. without the keyboard), that latter pleasure will be denied.
All this works efficiently and it's reasonably friendly. But there's also a major problem. Let's suppose you've got a sound source that's only available pitched at a quarter-tone up from middle C (a struck brake drum on a car, for instance). Well, there's nothing to stop you from sampling this, but because of the pre-programmed 30kHz sampling rate on recording, and the fixed range of sampling rates on playback, playing the A above middle C on the DS3 keyboard will produce a quarter-tone up C rather than an A.
"The software is sufficiently flexible that if some noise is apparent after you've put a sample into memory, it's usually quite feasible to edit it out."
In short, the only way you can get a sound out at pitches correlated to the keys on the keyboard is if all your samples are of things producing As above middle C.
Clearly, you should be able to fine-tune the sampling so that you get out what you want rather than what you had originally. In addition, the 30kHz sampling rate should be variable (as in units costing up to one hundred times the cost of the DS3...), so that if you're after a longer sample time at the expense of decreased bandwidth, that option is available for you to exploit. The point is that whilst a 30kHz sampling rate is just fine if you want to capture the authentic flavour of a drunk crashing into a pile of milk bottles, such exalted heights of fidelity are an overkill on more treble-less objects such as bass drums, for example.
OK, so that's the first stage of taming a sound over. Next on the agenda is saving it to disk before departing from this area of the DS3's activities. In fact, if there's a general lesson to be learnt, it's 'If in doubt, save it'. And that's all because on leaving this program for the wider territory of the main menu, you lose whatever was previously in memory, ie. the sample you'd finally got sounding just right.
The editing side of the simple sampling story comes under the guise of the second option on the main menu. But before you can edit anything, the program needs ammunition, so it's out with the system master disk, in with the disk that's got the sound you want to edit, and then in again with the system master in case something goes wrong. Of course, you avoid all of this pushing and pulling if you've got a second disk drive.
The aim behind the editing is really two-fold: (a) to allow the user to see what makes a particular sample tick, and (b) to provide the option of removing glitches, sibilants, and other audible bugs. All this is achieved by another push/pull operation - courtesy of the Apple's game paddles - which allows you to step through the entire sample, removing and altering bytes as and when you wish.
Again, all this is very functional and to the point, but sadly lacking in a couple of areas. First, finding where the paddles are about to scramble up your sample is quite tricky because of the pin-point size of the moving edit point: this really needs to be a decent 'X' for the old retina to zoom in on. Second, unlike the sampling software discussed above, the editing program restricts playback of the sample to only a single pitch triggered from one of the paddle buttons.
This is rather limiting when you're trying to fathom out what to do to the sample to make it sound better. For instance, there may be some breathy sound in the sample which is bearable at the pre-programmed pitch but sounds ghastly played lower down.
Now, this is where life gets complicated. Sequencers really are the most difficult things to get right - you know, the one man's meat is another man's poison syndrome. The DS3's sequencer has four tracks, each of which can access any of up to ten different sampies in memory (mighty short samples in that case, though). The sample-sharing is important - you're not tied down to the one track, one sample, one output channel way of working (used in units costing up to one hundred times the cost of the DS3...)
That's the good news.
The bad news is that contrary to Greengate's current advertising, this isn't a steptime sequencer. In fact, all the events have to be entered in time with a metronome 'blob' that flashes merrily at you in LORES graphics and annunciates itself via the Apple's internal speaker. This isn't what I'd call the best way of going about putting in an accurate percussion track.
The problem is really two-fold. First, the minimum event timing is simply the metronome rate, which is fine if you like epileptogenicity (oh yeah, know just what you mean -Ed) but bloody hard to follow with any accuracy. Which leads us on to the second problem. Once a track's events have been entered, you'll find on playback that these have been automatically quantised to the nearest metronome tick, so if you didn't keep up with it, you'll find that your sequence is somewhat transmogrified from the original. For the life of me, I can't understand why Greengate haven't done what everyone else does in this sort of situation, and that's to subdivide the quarter-note event with multiple pulses. The Roland standard of 24-pulses-per-quarter-note would have done fine, and the DS3 could then have offered degrees of autocorrection to boot. Oh well.
The nominal way of entering the aforesaid events is to trigger the sounds from the first four rows of the QWERTY keyboard. These are set up by another program (called 'Keyboard Setup'), which configures the first ten keys on each row with different sounds and pitches. Whatever 'kit' has been set up on each row will then get piped down its respective channel to the DS3 card.
So, to recap, each of the four channels can be fed with up to ten sounds at pre-programmed pitches. But if you've got the DS3 keyboard to hand as well, all this gets transformed, because these events can then be pitched across a five-octave range, which is great fun. Anyhow, once events have been entered into file sequencer, the next step is to decide what you're going to do with them. Well, while you're thinking about that momentous question, it's wise to put the sequence out of harm's way in one of the nine sequence stores by keying SHIFT plus 1 to 9. Next time round, you can play back that sequence, overdub with a different sound, and then store the overdub as a separate sequence in another sequence store. This can be repeated again and again, but the major constraint is the number of events that can be handled at once, ie. four channels four simultaneous sounds.
The final stage is to manipulate all the sequences into a nice well-seasoned whole, and that's accomplished by first 'merging' up to four of the stored sequences into a single one, and second, by 'chaining' the composite sequences into a song. So for instance, the merged sequences might look like the following with the sequence chain just running 1 through to 8 in turn:
SHIFT 1 Intro
SHIFT 2 Verse
SHIFT 3 Fill
SHIFT 4 Chorus
SHIFT 5 Break 1
SHIFT 6 Middle 8
SHIFT 7 Break 2
SHIFT 8 Coda
Again, all very easy to understand when it's on paper. And, indeed, it's all very usable in practice - especially when the sync in/out facility (via the Apple's cassette port) is brought into play. So, where's the beef? Well, point one is that all this sequencing involves a lot of key pushing, which is confusing at the best of times. It's a shame, therefore, that the sequencer display merely pre-occupies itself with flashing LORES event blocks at you, rather than telling you what sound is where, which sound is located on the DS3 keyboard, and so on. Point two is that the manual has its own idiosyncratic way with the English language, which doesn't make understanding the DS3 inside-out any easier.
Where Greengate go from here is difficult to see. They're talking about adding looping (so that the sample sustains when a key is held down), pitch bend, and all sorts of other niceties, but the problem they're going to come up against is that all these sorts of realtime interactions with playback on four channels use up a fair amount of processor time. Frankly, I don't see how that's going to be possible given the Apple's rather slow processor. Still, proper step-time sequencing should be included in one of the three free software updates they're promising with the system, so that at least should make event entry a good deal easier than the present battle of minds with the metronome.
The MIDI side of the matter is clearly another problem altogether, and I gather that Greengate are giving a MIDI In retro-fit facility serious consideration.
If there are two clear messages that emerged from playing with the DS3, it's that (a) despite the frankly silly omissions in the hardware, it sounds remarkably good, and (b) even though the overall package is pretty impressive when it comes to facilities and flexibility, the software needs more work on the user-interface side. The truth is that however laudable a home-spun product might be, what counts in the final analysis is sheer professionalism, and whether we're talking about the hardware, software, or manual, the DS3 just doesn't quite live up to that requirement.
Let's hope they put it right in the next version!
The sampling add-on card and software self for an RRP of £250, while the five-octave polyphonic keyboard is a further £200.
Review by David Ellis
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!