Linn LinnDrum (1982)
Behind the brushed-aluminum panel: a Z-80, a stack of PROMs, and a fourteen-inch rack of analog voice boards that made every producer in 1982 want one.
Overview
The LinnDrum is a programmable digital drum machine built by Linn Electronics. Its service documentation reads like a tour of early-eighties hybrid design - a Z-80 microprocessor running a bank of analog voice circuits. Each drum sound is stored as digital samples burned into PROMs and converted back to audio through multiplying DACs. The system clock is a 4 MHz crystal-controlled digital feedback oscillator, divided down to 2 MHz by half of a 74LS74 to guarantee a clean 50% duty cycle before it drives pin 6 of the Z-80. Somebody cared what the output sounded like.
The voice architecture is the interesting part. Three separate drum sound generator circuits - the SNR/SSTK/Click/Beep board, the TOM/CGA board, and the MUX drums board - share a common method. They decode strobe and data lines from the CPU, clock data out of the proms, and convert that data to an analog waveform. The MUX board handles Bass, Tamb, Cbsa, HiHat, Ride, Crash, Claps, and Cowbell. Each voice strobe consists of two low-going pulses on a voice strobe line, each carrying its four-bit data word. The first pulse clocks a low into the DO latch, resetting the drum circuit. The second clocks a high on DO along with data on the other three lines to select the drum sound and its volume. The counters then sequentially address all 4,096 words in the PROM, feeding each data word to the D/A converter. When the count completes, the latches reset. It's brute-force sample playback, the same principle that would eventually migrate into a single chip, but here spread across a board full of 74-series logic.
The analog path after the DAC is where the LinnDrum earns its reputation. The bass signal runs through a voltage-controlled filter identical to the one used on the TOM/CGA board. The hi-hat gets its own voltage-controlled amplifier - a CEM3360 - with a decay envelope shaped by an LM556 pulse. When a HAT strobe occurs, the LM556 outputs a 5 ms high-going pulse that charges a capacitor. That capacitor then discharges through one of two paths depending on whether the hat is open or closed, and the closed-hat path runs through the front-panel HiHat decay control. The VCA turns that shaped pulse into the amplitude envelope for the hat output. The decay isn't a sample, it's an analog envelope applied to a sampled source, and that's why LinnDrum hi-hats have a character that later sample-playback machines struggled to replicate.
Tuning on the snare and toms is handled by voltage-controlled oscillators whose frequency is set by front-panel pots and, where present, an external CV input. The snare tuning pot on the MXR board supplies a voltage level that sums with any control voltage input before hitting the diode-protected CV input of a 74LS627 VCO. The five tuning pots for the toms and congas feed a 4051 multiplexer, which selects one level based on the data on its A, B, and C inputs and passes it to the same type of VCO. Those VCO outputs clock the counters that walk through the PROM addresses. So the pitch of the drum is set by how fast the sample is read, not by resampling or interpolation. It's a crude method by later standards, and it's why tuned LinnDrum toms have a particular grain that pitch-shifted digital samples don't quite match.
The trigger input circuit shows how the machine was meant to be used in a studio. The trigger inputs put out a digital pulse in response to rapid changes in the input signal, but a trigger is only generated if the envelope of the signal is increasing. Sharp transients trigger, and sustained tones only trigger when they begin. The input signal runs through a trimpot adjustable from 1 to 10 gain, then through a full-wave rectifier built from an op-amp and diodes, a filter stage, and finally an open-loop differentiator-comparator. The output resistors make the ±15 volt output of that third stage compatible with TTL logic levels. The circuit is built to be fed from a tape track or a drum pad without misfiring, which matters when you're syncing the machine to a multitrack.
Sync and cassette inputs are zero-crossing detectors with about 1.4 volts of noise immunity provided by four diodes. A log amplifier controls output transistors that translate the incoming signal to TTL levels. The tempo clock itself is an RC-controlled oscillator - an LM556 - with a front-panel tempo pot that varies its frequency from about 60 Hz to 300 Hz. That signal is divided in half and placed on the data bus, and when the machine is playing or recording, the CPU uses it to determine the length of a 1/192 note. So the tempo resolution is 192 pulses per quarter note. That is fine enough for the swing and shuffle functions that Mix (1982, p.42) noted could be applied after recording to give a producer an idea of what percentage of swing a tune might want. The same review described the sounds themselves as simply incredible, with depth, good attack, fine low end, and good long decay, and singled out the cymbals as particularly strong.
The CPU section is a Z-80 with 16K of address space, split between ROM and I/O in the bottom 8K and RAM in the upper 8K. Address line A13 selects between the two decoders, gated with MREQ. When A12 through A7 are high, the system switches from ROM to keyboard and I/O. A6 distinguishes between keyboard and I/O but is ignored when decoding the ROM. The RAM chips and write signal are gated with a power-down signal to prevent memory access when the TTL supply isn't at sufficient voltage. That signal is essentially the power-stable signal gated by MREQ, which keeps a memory access in progress from being interrupted when power drops. There's also a power-stable circuit on the DRM board that sets up a 1.23 volt reference and delays the signal fed to a comparator by about half a second, the time it takes to charge a capacitor, before indicating stable AC power to the CPU board. That half-second delay is why the LinnDrum doesn't glitch on power-up.
The displays are driven by 74LS47 decoders through 150-ohm resistor packs, with data fed from latches selected by a 74LS138 device decoder. The LED indicators are driven directly from latches through their own 150-ohm resistor packs. One quirk: the output that drives the "pattern" LED also feeds an inverter that drives the "song" LED whenever the "pattern" LED is off, so the two indicators are mutually exclusive by design. The cassette output is driven by a latch output, with two 220-ohm resistors dividing the voltage to approximately 2.5V for cassette recorder line inputs. That matters if you're dumping patterns to tape, because feeding a cassette deck's line input with a hotter signal will clip.
The power supply is a switching design. The transformer output is rectified by four diodes and filtered by a 27,000 µF capacitor on the drum board, producing a front-end voltage of approximately 12V. That feeds through a 7-pin Molex connector to the 5VSR board, where a switching regulator alternately connects and disconnects the front end to an inductor, with the duty cycle controlled to regulate the output voltage. A current-limit circuit pulls down the voltage reference line when current through a sense resistor is too high. According to the service documentation, that feature is used with National regulators only. The output voltage can be varied by adjusting a trimpot on the 5VSR board.
In Linn's product line, the LinnDrum took the LM-1's concept and made it more accessible. Recording-Engineer (April 1982, p.82) reported that Linn Electronics introduced an improved model at a substantially reduced price compared to the LM-1 Drum Computer, with studio-quality sounds. Linn's own advertising in Mix (1984, p.135) billed it as "the programmable digital drum machine with studio quality sounds" and claimed other drum machines might try to compare their sound to the LinnDrum but there was no comparison. That's marketing copy, not an independent verdict. By 1985, Linn's advertising in Mix (1985, p.89) was positioning the instrument as "the most sophisticated compositional tool ever created," emphasizing that it let users control the process the same way they operated a multi-track tape machine, with punch-in capability.
For collectors and users today, the LinnDrum's serviceability cuts both ways. The machine is built from discrete logic and standard analog parts - 74LS series, 4051 multiplexers, 4558 and TL082 op-amps, CEM3360 VCAs, LM556 timers - so many failures are repairable with a schematic and a soldering iron. The PROMs are the long-term concern. If a voice PROM fails, the drum sound it contains is gone unless you have a known-good dump. The Z-80 and its support logic are sturdy, and the power supply, while a switching design, is documented well enough in the service manual that a competent tech can rebuild it. The tuning pots and front-panel controls are mechanical and will eventually need cleaning or replacement, but they're not exotic parts.
The LinnDrum's place in the used market is complicated by the fact that it's both a collectible and a working tool. Studios that want the sound - the long, deep kick, the snare with its particular crack, the hi-hat with its analog decay - keep them running. The pattern and song memory, the swing and shuffle functions, the trigger and sync inputs, and the cassette interface all still work as designed, assuming the machine is maintained. What you can't do is load new sounds. The PROMs define the voice set, and while there are aftermarket solutions, the stock machine is what it is.
The machine's documentation - the service manual with its board-level descriptions of the DRM, CPU-I/O, and 5VSR boards - is unusually thorough for a drum machine of this era. That suggests Linn expected these to be maintained professionally rather than discarded when they failed, and it's part of why so many are still running.