Internal sound source:

Synthesized drums: (Bass Drum, Snare Drum, Low/Mid/High Tom, Rim Shot, Hand Clap)

Sampled drums: (Open/Closed HiHat, Crash/Ride Cymbal)

Inputs:

MIDI, Tape Load/Sync in (combined), Start/Stop, Din Sync

Outputs:

2xMIDI, Tape Save/Sync out, Trig out coupled to rim shot sound +14V 20ms pulse, L/R Mix, 10x individual out.

Synchronization possibilities:

MIDI, Tape, Din Sync.

Secondary data storage:

Cassette, RAM cartridge (M-64C)

Memory size:

48 patternsx2 banks, 4 tracks of max 896 measuresx2 banks.

Physical size:

486x105x300 mm, 4.5 kg.

Data generation/modification/recording:

Step or real time recording of MIDI events used to play back internal drum sounds, but also to control an external sound source (see below) The unit does not record velocity nor other controller data. Playback is limited to two velocity levels ("Accent"). A row of 16 buttons are used in conjunction with an instrument selector to program the unit, both in step and in real time (in real time the buttons correspond to the various instruments




Bass Drum consists of two parts. One part simulates the clicking attack-phase of the sound, where the other part provides the bottom end, which is responsible for the well known punch of this Bass Drum. Main circuit design is the same as described detailed in the Snare Drum section.

When TRIG (>) is applied to the base of Q1, ENV-3 (part 6) is started by Q2 collector, releasing a pitch modulation of the VCO (part 2) by controlling the Control Voltage Generator (part1). Modulation depth can be adjusted by VR2 (TUNE). There's also a positive pulse at the base of Q11, discharging C14 and resetting VCO at every TRIG period. The triangle waveform of VCO output passes a diode clipper (part 3) to get a sine wave. The amount of this part is determined by VCA Q12 (part 4) whose gain follows ENV-1 (part 5) which is in turn controlled by an ACCENT coming through Q4. ENV-1 is actually an important part of this circuit. It's shape (see small diagram near ENV-1) looks like a compressed Drum sound and that's the reason for the big punch mentioned above.

The second part of this instrument is separated into Pulse and Noise.

Pulse (part 7) is generated by a circuit using a low pass filter (Q8, R5, R6, R14, C22) and a band pass filter (Q9, C2, C3, C4, R7, R8, R15, R16) to form the trigger signal into the right pulse shape. The noise part uses the common noise generator, whose signal (>) is low pass filtered and mixed with the signal coming from pulse generator. The resulting signal passes through a VCA Q6 (part 9) which is controlled by ENV-2 (part 10).

Finally ATTACK and VCO-signal are mixed and amplified at IC11a (part 11) to provide level control.



Snare Drum consists of Drum and Snappy, each further separated into two parts.

Drum voice is composed of VCO-1 (part 10) and VCO-2 (part 11) with associated Control Voltage Generator IC35 (part 8). VCO-1 and VCO-2 have similar circuitry except that charging capacitors C69 and C71 have different capacitance so that they can oscillate at different frequency. VCO-1 runs at lower frequency.

VCO's comprise a hysteresis comparator IC37a (IC38a), inverting buffers configured as voltage-dependant resistor (in IC36) and an integrator consisting of IC37b (IC38b) and C69 (C71) with Q44 (Q45) switcher. In this arrangement VCO-1,2 generate triangle waveforms . When TRIG (>) is applied to the base of Q39 VCO-1 receives a positive pulse from Q40 collector at the following places. There are similar conditions at VCO-2.

a) One input of IC37a via D62. When the pulse is applied, IC37a turns its output to low.

b) The base of Q44 which discharges C69, cancelling VCO-1 output. The combination of a) and c) resets VCO-1 to the starting point at which VCO-2 also starts oscillation, phasing the initial waveforms of both VCO's.

c) The base of Q46 which cuts off VCA Q50, muting unwanted noises in the VCO-1 path.

There is also a positive pulse starting ENV 1-5.

Output of ENV1 (part 9) influences the Control Voltage Generator. The resultant effect is a pitch bend of Snare drum sound for about 20ms. The triangle waveform (integrated square = triangle) of VCO-1(VCO-2)passes through a diode clipper (see parts 12,13) to twist the triangle into a waveform near sine. The amount of drum voice from VCO-1 (VCO-2) is determined by VCA Q50 (VCA Q51, see parts 14,15) whose gain follows ENV3 (Env2, see parts 17,16) which is in turn controlled by an ACCENT (>) coming through Q41 currently gated by the TRIG.

Snappy uses the common noise generator, which is also used for Bass Drum, Hand Clap and Tom's. Noise signal (>) is low pass filtered (IC 40a, see part 1) and then splitted into two parts. The first part is passing through a VCA Q48 (part 5) which is controlled by ENV4 (part 6). The second part has its own VCA Q47 (part3) controlled by ENV5 (part 7), but is high pass filtered (IC 39a, see part 2) before passing through. ACCENT signal is gated through Q41 by the trigger from Q39 collector and is coupled to the base of Q47 VCA as ENV 5. ENV 5 determines the amount of high frequency noise components in the SNAPPY which becomes articulate when noises passing through the high pass filter are combined with the noises from the low pass filter at IC39 (part 4).

Finally Drum and Snappy are mixed and amplified at IC40b (part 18) to provide level-control.



Low, Mid and High Toms have similar circuitry except that VCO's have different main tunings. The following description takes Mid-Tom as representative.

Mid Tom is composed around three VCO's and noise, called Tom noise in this case. Tom noise (>>) is a modified signal from the common noise generator.

VCO's have similar circuitry except that charging capacitors C32,33,34 have different capacitance so that they can oscillate at different frequency. VCO-3 (part 4) runs at highest frequency, VCO-2 (part 5) runs at lowest.

When TRIG (>) is applied to the base of Q25, ENV-4 (part 2) is started by Q24 collector, releasing a pitch modulation of all VCO's (parts 3,4,5) by driving the Control Voltage Generator (part1). Control Voltage can be also adjusted manually by VR13 (TUNE). There's also a positive pulse at the bases of Q30,31,32 , discharging C32,33,34 and resetting VCO's to sync them together. Circuit configurations of VCO's are the same as used in Snare Drum. Refer to this section for more information.

Each VCO output passes through a diode clipper (parts 6, 7 and 8) to get sine waves. The clipper of VCO-1 is also controlled by TRIG (with associated envelope C38 and VCA Q23) to cause a slow change from square (low resistance in collector-emitter path, hard clipping) to sine (softer clipping).

Each modified oscillator signal passes through a VCA (Q26,27,28; see parts 9,10 and 11) with associated envelope (parts 12,13 and 14). ENV-2 (for VCO-2) is adjustable in Decay-Time (VR17). So you hear just one oscillator after some milliseconds, when you adjust longer decay times. VCO-3 signal is also mixed with Tom Noise (>>) before reaching the VCA. This causes a more natural sound in the attack phase of this instrument.

Finally VCO-signals are mixed and amplified at IC23a (part 15) to provide level control.



Handclap uses almost the same circuit as the well known TR-808. It’s based on white noise (> from the common noise generator) which is band-pass filtered by IC26b and associated RC’s (part 1) to get the right sound for simulation. The filtered noise is divided into two parts. The first part passes through a VCA (IC 30, see part 2) which is controlled by a special envelope (ENV-1 see part 3) after TRIG (>) is applied . This envelope is the main feature of this circuit. It simulates clapping of several hands. Basically it consists of four attack-decay envelopes (Opamps a-d in part 3), where finishing the first starts the second an so on.

The second part also passes through a VCA (Q37, see part 4), which is controlled by a simple envelope (part 5). This part produces the atmospheric background sound of the instrument. It sounds like a reverb.

Both parts are mixed and amplified by IC28a (part 6) to provide level control.



The Rimshot is the easiest instrument on this machine. It's based on three bridged T-network-filters (see parts 1,2,3). These kind of bandpass filter is the base for nearly every sound on the TR-808 (except HiHat and Cymbal).

They have different Q- and f-values as you can see on the diagrams above the schematic. According to these you'll get following frequencies : f1=500 Hz, f2=220 Hz, f3=1000Hz. The disadvantage of this circuit is, that frequency depends on every resistor (5% types used) and every capacitor. Every fault sums up by principles of superposition and can cause large differences between the same sound from machine to machine. That's perhaps the reason why Roland used VCO's for the other drumsounds of the 909.

When TRIG (>) is applied, oscillators 1,2,3 get a positive pulse from D89 and start oscillation. The decay time depends on Q-factor which can be calculated by the following equation: Q=(sqrt(R1/R2))/(sqrt(C1/C2)+sqrt(C2/C1)).

The outputs are mixed and clipped by a simple diode-clipper (part 4) D91,92. The clipped signal passes through a VCA (part 5) which gets control voltage from IC48b, Q64 and a simple envelope C119 (see part 7). Last stage is a high pass filter (part 6) which removes unwanted low frequencies to provide a more realistic simulation of the instrument.



HiHat, Ride and Crash cymbals are 6-Bit (!) samples. Circuit configurations and operations of these voices are basically the same. The following description takes HiHat as a representative.

Pressing HiHat button(s) develops a positive pulse (TRIG), resetting Address Counters IC70 and IC71 (part2) to have "0's" on their outputs. These 0's cause IC72a output to swing to H irrespective of a Closed/Open being applied to diode OR's (D196-199, necessary to distinguish between Closed and Open HiHat). Upon receiving this "run" from IC72a, a simple oscillator composed of two NAND-gates (no crystal!!) starts oscillation and outputs about 60kHz (see part 3), which is divided by two and shaped up by IC73 D-Flip-Flop (see part 4), clocking the address counters (pin10 of IC70,71). IC69 (ROM, see part 1) starts clocking out voice data, which are latched into IC68 Hex D-FlipFlop (see part 5) and then converted to analog voltages while passing through RA9 (resistor network, see part 6). This is a very simple configuration for a D/A-Converter. According to the resolution of 6-Bit, the sounds have been compressed before being digitalized in order to have a greater S/N ratio and higher digital resolution. The original envelope is reproduced by a VCA, which is controlled by an envelope.

Open/Closed decay times of this envelope can be regulated individually. This is realised by a little bit tricky circuit (see part 9).

A high Closed/Open (Closed HiHat selected) on Q72 base removes a positive voltage from its collector which in turn allows Q73 to charge decay capacitor C135 through R451 and VR21 (CH Decay). Since this charging path is 1/10 the total resistance of R452 and VR23, the charging rate of C135 depends on VR21 setting only (better say almost). With low Closed/Open (Open HiHat selected), CH charging path is disconnected from the DC supply source at Q73 and OH path becomes conductive. Charging rate of C135 now depends on R452 and VR23 (OH Decay). A disadvantage of this circuit is, that CH Decay is changing slightly when turning OH-Decay (VR23). But this change will not be notified in most cases.

Envelope output passes through an Anti-Log (part 8) to provide a more realistic shape of the Env-signal, which drives the VCA (part 7). Finally HiHat signal is lowpass-filtered (two LPF's in serial, see part 10) to remove unwanted high frequencies after D/A conversion. Level control is provided by an amplifier (part 11)

Some notes on Crash and Ride

These voices also have unique envelopes that are quite different from actual sounds when the data is directly reproduced. The reason is the same as described in HiHat section. According there is no decay-control, restoration of the envelopes are made by the use of ROM addresses as the envelope data. When Crash (Ride) sound data are read successively from ROM with correct addresses, the same addresses are also converted to analog voltage (by a second D/A), anti-log tapered and applied to the VCA. This way of generating the envelope is necessary, because tuning of the instrument is possible. As the signal gets shorter at higher tunings, the decay-time has to change in the same manner.