roland tr-606   

 

clocks and timings

 

 

 

 

 

 

 

 

 

 

 

 

 

There are three different, totally independent clocks running inside a TR-606. First, there's the CPU clock, specced to run at about 455 KHz (not very impressive in today's terms). Then there's the clock signal driving knobs and buttons scanning, running at 555 Hz (this process is actually 'interleaved' so each knob and button actually gets polled at 1/4 of the scanner clock rate, ie. about 140 times per second). Finally, there's the pulse clock driven by the TEMPO knob or, in case SYNC INPUT is chosen, external DINSYNC clock signal. A weird design decision made by Roland engineers was not to couple the clock pulse and CPU directly, via an input line. Instead, and probably due to lack of CPU input lines they opted to scan the clock pulses by the same process that is used to poll the knobs.

On my machine, the pulse clock rate goes from 16 Hz (SLOW) to 122 Hz (FAST). So, with fastest tempo, length of a clock pulse is about 8 ms. CPU scans the clock only during negative half-cycle, so it sees it for about 4 ms. Scanner process, clocked at 555 Hz, has a chance to see a clock pulse for 2..3 times, with an average delay of half the scanner clock pulse (0.9 ms, varying from 0 to 1.8 ms). With slowest speed, length of a clock pulse is about 63 ms and the CPU can sense the clock for about 17 times during negative half-cycle.  

clock pulse vs scan pulse (fast speed)

clock pulse vs scan pulse (slow speed)

So the question is: does this affect timing? How closely does the gate signal follow the clock pulse which is not that important with internal clock, but could be pretty meaningful with eternal clock while syncing with master device.

CPU scans the clock signal and divides this by 6, ie. one out of every six pulses makes it to the gate output (port I2). As Roland's gear used SYNC24 (24 pulses per quarter note), the CPU sends 4 gate pulses per each quarter note enabling the sequencer resolution of 16th notes, hence the 16 buttons per bar. Scope reveals that the gate signal is delayed compared to the clock signal, the delay is around 2..3 ms and is not constant ('jitter'). Compared to the typical 10+ ms latency in a modern DAW (PC, MIDI, VSTi etc.) this is still pretty tight, but...

clock pulse vs gate signal

To be continued...