When Is A Computer? (Part 6)
The science of the drum machine.
What goes 'boom-kas, boom-kas'? Andy Honeybone when his word processor's packed up. Alternatively, it could be the life story of the electric drum machine, from rhythm boxes to real sound digital samplers. Here be the beats and the volts...
Drums were banned by the plantation owners for fear that their power of communication could assist the slaves to revolt. It's hard to equate this same force with the tick-a-boom sounds which have been standard on front room organs for many years.
Electronically, the rhythm box was a great step forward because until then the percussion synthesiser circuitry (usually two transistors and a minimum of components) had been triggered by key depressions on the pedal board and accompaniment manual. By this method a walking bass would generate synchronised ride cymbal strokes, and left hand, offbeat stabs could bring in some tasteful castanets (noticeably missing on the Drumulator and Co.). It really makes you shudder!
The heart of the stone age rhythm box was a ring counter. Although definitely digital, it would require a great deal of optimism on the part of a salesman to qualify description as a computer. A ring counter is the electronic equivalent of 12 fully grown adults standing in a circle and transferring an orange from one to another under their chins. The orange represents an electronic pulse with can trigger a percussion voice, and each person represents a semi-quaver or quaver triplet step depending on whether each complete round produces a 3/4 or 4/4 signature pattern. This may sound confusing because of the sprinkling of musical terms, but each person is just one twelfth of the repeating measure.
By connecting isolating diodes to positions around the ring, pulses from different steps can be added to give a trigger pattern for one voice. The procedure can be repeated to produce triggers for many voices in as many rhythms as required. Switching between rhythms calls for some cumbersome hardware but the overall circuit is very simple.
The sound generating circuitry was at first a valve design but as transistors became available so they found their way into rhythm boxes. The technology may have changed but the sound hadn't; drums remained damped oscillators and cymbals stayed as envelope shaped white noise.
The drum circuits generally consisted of a twin-T filter network strapped between the input and output of an inverting amplifier stage. The gain of this amplifier was adjusted so that the circuit did not quite oscillate but when given a jolt in the form of a trigger input, it would 'ring' for a few cycles at its resonant frequency, before decaying into silence.
By suitably choosing frequency and resonance, the same circuit was pressed into service for bass drum, toms, congas, timbales, claves and, let's not forget, castanets.
It's quite an achievement for one transistor to produce a frequency and to be an envelope generator and VCA, but this low cost simplicity meant that improvements become expensive. Luxuries such as pitch drop during decay were considered far too esoteric at this stage of development.
In the past, cymbals and brush noises have been the most unrealistic of electronic percussion voices. This has been due to their reliance on white noise which is cheaply produced from a 'duff' transistor.
White noise has always been a good standby for complex sound of no fixed pitch, and for 'surf on the beach' noises it can't be beaten. It usually gets thrown in for the rattle of a snare but, as you will have heard, crash cymbal circuits sound like a cistern re-filling. This adds to the evidence that cymbals do have a fixed pitch and a very complex series of overtones. Many of the present day 'metallic' cymbal generators cross-modulate several oscillators to produce a more acceptable sound.
Prompted by the increased demand for domestic organs, several rhythm pattern 'micro-chips' became available in the form of Read Only Memories (ROMs). The rhythm information is stored in these devices in the way shown on the liquid crystal display of the new Dr Rhythm Graphic. The memory is like a grid eight rows wide (one for each instrument) and 512 columns long. It can hold enough info for 16 patterns of 32 steps.
Each square of the grid corresponds to one memory bit and the bit pattern is fixed at manufacture after which no change is possible. Columns are selected by an address in the range 0 to 511 and these must be cycled by a counter for the rhythm sequence to appear at the output of the chip. Present day ROMs can hold 32 times more data than these early designs.
The rhythm selections on these chips were the usual march, waltz, fox-trot, bossa, tango etc and were of little use outside their intended application. An Electro-Harmonix machine which worked on the same pre-programmed principle suffered because the patterns were so 'over the top' and 'clever-clever' that they detracted from whatever else was being laid down, and the ear soon became tired. It seems that a pattern has to be simple to stand up to endless repetition.
The user-programmable drum machine was an effective solution to these problems but no commercial product could be considered until there was some cheap way of retaining the rhythms after turning the power off. The type of memory needed was one which used so little power when it was idle that it would draw no more current from a battery than would be lost by natural leakage.
This low power consumption is a feature of circuits fabricated by the CMOS (Complementary Metal Oxide Semiconductor) process. The output stage of a CMOS circuit element consists of two field effect transistors stacked one on top of the other between the two voltage supply rails. These switches are complementary so that when one is on the other is off.
The output of the circuit comes from the junction between the switches, and the input is fed to both gate connections which control the transistors. For the output to go to logic '1', the top switch turns on and the bottom switch turns off, so connecting the output to the positive rail. To output a logic '0' the reverse occurs, and this time the ground rail is switched.
The power saving comes from one switch always being off (open) and blocking current flow. Some current is drawn during switching but this falls to virtually nothing at other times.
Early CMOS memories were very expensive, had a low storage capacity and were prone to damage from static electricity. These problems have now been corrected and a 2k x 8 (2048 byte) chip costs about a fiver. Programmable synthesisers use the same method for voice storage and instead of a couple of HP7s supplying the juice, a nickel-cadmium rechargeable battery is used which trickle-charges from the main supply when the instrument is being used.
The list of features for these robot drummers began to expand to include programmable fills and variations. This was an indication that the manufacturers had more memory space available than they knew what to do with. For a variation every 4th bar, a signal is derived from the master clock divider chain which causes a jump to a pattern in a different memory area on every 4th pass. For this type of A-B sequencing, a microprocessor is not necessary because discrete logic elements are quite able to cope.
The feature which sorts the men from the boys is the ability to chain patterns to form complete songs. The microprocessor is a great manipulator of data and the upmarket DMX and Drumtraks-type machines all weigh in with around 100 patterns of 100 bars and 100 song capability. The presence of a micro also enables control over pitch, tempo and dynamics in reward for hours spent punching in numbers.
Without wanting to detract from the power of these one grand's worths, there does appear to be rather a lot of 'specmanship' going on. As if the poor British Brochure Browser is not confused enough with Stateside talk such as kick drums, sock cymbals, quarter notes and measures, he/she is now required to know how many events there are in a song (dry ice for the first verse, inflatable doughnuts for the chorus?).
One hundred is the maximum number of selections that can be shown on a two digit display (0-99) and the total available memory has to be juggled between the various sections according to demand. Even taking into account the "up to" disclaimers, this latest generation of machines must cater for most needs (except Flamenco dancers).
Of course the most impressive aspect of the Linn and pals is that the percussion voices are real drum sounds which have been digitally recorded and blown into memory chips. Capturing natural sounds for real time playback is nothing new as Mellotron owners will be quick to point out. Optical disks have also been used for sound storage as in the Optagan of Steve Hackett fame. The difference with the solid-state approach is sound quality and reliability.
Although the sounds are trapped in circuits intended for computer use, the computer side of these big boys need only provide the trigger pulses. Just a simple gated oscillator and ripple counter are needed to wrench the sound samples from the memories and thrust them into a Digital to Analogue Converter (DAC) to produce an audio output.
Without the expense of a computer, a real drum sound machine with factory set rhythms could be produced for well under two hundred green ones. It would be like a no-moving-part, 'Drum Drops' type rhythm accompaniment record with the advantages of tempo control, unlimited repeats and portability. The Simmons digital Clap-Trap uses the same technology and costs just over a ton.
This method of sound production needs a DAC for each noise, but the extra cost results in a very clear sound. If the sound samples are addressable by the micro then it can perform time slot summation and throw the entire mix out through one DAC. This makes the processor work very hard and limits the degree of resolution possible.
Digital to analogue converters are responsible for the sound quality of all digitally based instruments and effects. They come in many different forms but the majority have an output which is a linear function of the input value. Where memory is tight, a companding DAC can be used to increase the dynamic range that can be stored in a given number of bits. This is a converter with an output which changes logarithmically in decibel fractions per step. A drum machine with this type of converter will have a good signal to noise ratio and sound much more pokey.
At the bottom (affordable) end of the market there has been a re-emergence of the ready programmed drum machine with new models by Yamaha and Boss. True that one also has 'live' pads and the other a programmable section, but could it be that we're not such an inventive bunch after all and would happily settle for an interesting metronome? Whatever would 'Big' Sid Catlett have said?
Feature by Andy Honeybone
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!