DIN sync in x0x boxes (was RE: [sdiy] Help-Blacet midi-din sync chip problems!)

[Colin Fraser]

> -----Original Message-----
> From: owner-synth-diy at dropmix.xs4all.nl 
> [mailto:owner-synth-diy at dropmix.xs4all.nl] On Behalf Of David D
> Sent: 10 February 2002 8:33
> To: cyborgzero at home.com
> Cc: synth-diy at dropmix.xs4all.nl
> Subject: Re: [sdiy] Help-Blacet midi-din sync chip problems!
> 
> >
> >Problem is, this 50% duty cycles screws
> >*some* 808s and 606s up pretty badly,
> >and they begin to slip.
> 
>   Thats weird cause my 606,808 and CR 8000 all output a nice 50% duty
> pulse and sync to each other perfectly with each others sync output.

Well, I got fed up speculating what is going on inside these boxes, so I
poured myself a double Smirnoff and Coke, warmed up the O-scope, and did
some empirical science...

The CPU in a 303, 606 or 808 has a number of jobs to do - 
1. scan the keys and switches, and refresh the LEDs
2. read the DIN sync input pins to detect the RUN status, and arrival of
a clock pulse
3. playback a sequence at a rate controlled by the incoming sync pulses
(which are switched externally from the CPU, so it doesn't know where
they came from)
4. do the track edit and sequence edit stuff

The 303, 606 and 808 all have an 'interrupt clock' with a 1.8ms period
(1.9ms on the 808) that drives the interrupt line of the CPU.
I've often wondered which tasks are performed during the interrupt
routine, and which during the main program.
The answer to this question will determine what might happen if the
interrupt timer is badly adjusted.
Given the high probability that by now someone has been inside every
303, 606 or 808 ever made, and many will have been looking for an easy
way to make it sound 'better', a lot of the interrupt timers are likely
to be far from their intended setting. 

The trimmer on my 303 will adjust the interrupt period between 0.6ms and
2.5ms.

Looking at the scanning pulses to the keyswitches and the drive pulses
to the LEDs on the scope, while adjusting the interrupt clock, it's
clear that the interrupt routine is responsible for the key scanning and
LED refresh.

Driving the 303 from it's internal 50% duty cycle oscillator, and
adjusting the interrupt timer over it's full range has no effect.
Syncing the 303 from my MIDI to SYNC convertor (with a clock pulse
length of 2ms), and adjusting the interrupt timer shows that as the
period of the timer is increased above 2ms, the reliability of the sync
rapidly falls off.

This shows the interrupt routine is also responsible for polling the DIN
Clock input.
A pulse shorter than the interrupt period can start and finish in
between 2 interrupts, and not be seen by the CPU.
If the main program loop was polling for the clock input, it wouldn't
lose sync if it was being interrupted less than usual.

So, if you're syncing a 303 from a MIDI to DIN sync unit with a fixed
pulse width, the interrupt timer *must* have a period less than the sync
pulse length, or sync will be unreliable.
Given this fact, I'm surprised the Blacet chips work well at all in the
recommended configuration. 1ms just shouldn't be long enough.
I'm also tempted to increase the pulse length I use to 3ms - this only
limits the maximum tempo to somewhere over 400bpm, and means that even a
badly adjusted 303 etc. should be able to see the pulse.

Why a 303/606/808 would fail to sync correctly to a 50% duty cycle sync
pulse, I have no idea.
It's possible that if the interrupt timer is set too high, the main
program loop is starved of CPU time.
This could have strange effects...

I'll need to have a think about how to determine whether the sequence
playback is processed in the main loop, with the interrupt routine
setting some flag to say a sync pulse arrived, or if the sequence
playback is performed by the interrupt routine itself.

Until then, more vodka :-)

Colin f