TB Sequencer Replication by Software (was another @#!! TB t hread)

[Fraser, Colin J]

Thanks to Mark for the most accurate description of the 303 sequencer
function, but I'd like to add my take on what might make it sound the
way it does.

The slide on the 303 is enabled at the start of the note that is to be
slid to.
There is no 'premonition'.
I confirmed this using an oscilloscope.
The exponential curve of the slide may explain why some people 'hear'
the 'pre-slide' effect.

However, I'm not sure the circuit is just a simple RC intergrator. 
The d to a convertor in the 303 is constucted from a hex latch and a
R-2R resistor network.
The integrating capacitor that causes the slide is switched by a 4066
analogue switch between the input and the output of the op-amp that
buffers the R-2R network - so if slide is on, the capacitor is charged
via the output impedance of the R-2R network. I haven't done any
analysis of this, IIRC from a textbook this impedance is nominally the
value of R (in this case 100k), but possibly the impedance changes
depending on the binary output value. Anyone know more about this ?

The other aspect of the slide circuit that may affect the sound is that
the same control line used to load the dac latch is used to control the
slide circuit ie. the latch is loaded on a rising edge of the control
line, the line stays high if the slide is to be enabled, but goes low
again if slide is off.
So when every note is loaded into the latch, the slide circuit must be
activated momentarily. I don't know how long the latch pulse is when
slide is off, I'll need to measure it, but it's possible that this has
an effect on the start of every note. 
Certainly, when my friends and I were comparing the 303 sequencer
driving my clone, to an *identical* pattern triggered from Cubase via a
cv convertor with a simple RC slide circuit switched in and out by a
4066 (my design), we all noticed a difference at the very start of each
note. The real 303 sequencer seemed to add some 'punch' to the start of
the note not present with cubase + cv convertor.

The other difference we noticed is that with a short repetetive pattern,
the real 303 sequence had more 'life'. You could hear a constant
movement in the pattern, like a very subtle shuffle. With cubase the
pattern seemed noticably more robotic and precise.
A possible explanation for this might be that the processor in the 303
does not always respond to the clock pulse immediately and this may
cause a timing error. There is an interrupt triggered on the 303 every
couple of milliseconds that presumably initiates switch scanning / LED
updating - possibly a clock pulse arriving duriong the interrupt cycle
can not be responded to until after the interrupt routine ends.
This is just speculation though. I'll need to do some more research.

Colin f