AW: Re: Sequencer Update

[jhaible]

>Hi Juergen. This sounds interesting. I was looking at the Superseque
article 
>the other day. Then I thought about improvements that could be made to it.
One 
>such thing would be to add a switch for changing the length of each step. 
>Thomas Henry suggest using one row of pots to control the frequency of the
VCO 
>that clocks the sequencer. But that means you will have to twek them
carefully 
>to get a consistent rhythm. Maybe it would be better to add a thumbwheel
that 
>selected among the most used rhytmic values. Like 1/4, 1/8, 1/6, dotted
4ths, 
>8th triplets and so on. With variable length you can get away with fewer
steps. 

I'll have CV inputs for clock speed and PWM (= relative gate length).
Personally,
I am no fan of quantized triplets etc. For some reason, a straight beat of
1/8s
sounds right to me, but quantized triplets sound funny ...
Nevertheless, one can use a quantizer for this function (1V/8ve is also a
good
CV scale for clock speed !). Or maybe even a gate bus and proper scaling of
the FM
input is enough to switch between 1/8s and 1/4s, for example.
But this whole talk about time control stuff brings me to one important
point:
It's crucial that I can remove (skip) selected rows from the vertical clock
to have them
ready for timing stuff of filter control.

>Then 8 steps might be enough. Otherwise I think 12 is an absolute minimum.
With 
>8 equal steps it will be difficult to get triplets.
I>s this something you have already thought of?

I'm quite sure now that I will go 4x8.
It's easy to make 1x12 from that with a "reset" at stage 7.


> 1) will you be implementing a "stage select" pulse input (maybe just a
4051 to
> the load inputs of the 4516)?

Something like the touch keyboard of the Serge ?
I don't think so.
While I admire the usefullness of the Serge for both, sequencer and
programmer
(and a few more) applications, I will go purely sequencer here. I consider
a single
analogue CV input + DAC for the preset function, as some of you suggested,
so I can set the preset stage# by a potentiometer or external CV. But even
then, I see it as a voltage controlled *jump* function rather than manual
preset.
("jump", because I'll use the preset input similarly as the reset input)
But if someone really wants individual keys, a simple DAC will do to create
the required CV ...
All in all, I find the preset input even more important for the *vertical*
stuff,
to select one of four different patterns on the fly.

> I didn't know about this circuit when I built my sequencer, and
> used the "brute force" 16 comparator + 2 encoders method for my 4bit
> flash ADC. 

Chris, do you have the circuit for this one at hand? Have spent a few
thoughts about
economic encoding, but haven't come up with anything convincing yet. Are
there
speacial chips for that? (Four LM324s, or LM339s, for the comparators is no
problem.)

>Using the "jam inputs" on the counter like this is so easy and
>adds some neat functionality to the sequencer, I highly recommend it.
>Controlling it with an ADSR or a sawtooth wave for example. OTOH, you
>(Juergen) can probably do all the same tricks with the interpolating
>scanner, and do them better that way!

The I.S. has a clear advantage on creating smooth rather than stepped
waveforms. But this sequencer project will be pure pattern generation;
audio rate operation not intended. And selecting sequences (rows that
is ...) with ADSRs of LFOs is a great idea. So I guess I need a 2bit
ADC for selecting the rows even more necessarily than the 4-bit ADC
for the steps ...


> 2) If you use the vertical clock (and you should ; >),

I am pretty convinced that I should do. That's the grat thing on this diy
list,
that you get much more ideas that you have ever thought of yourself ...

And here's another idea of my own which has to do with vertical clocking.
I'll call it "Event Queue".
I already wrote about a delayed, or synced, action of the sequencer to
manual
inputs. Example: I have 4 sequences set up (4 rows), and the 1st is
currently
running. I press a button to change to sequence #2. But it won't jump to #2
immediately (in the midst of one note), but rather wait for an impulse from

currently running sequence. Most obvious choices would be the end of one
full
sequence, or the end of the currently played note, but I'd take it as far
as
I allow any next pulse from a selected gate bus. Just as you set your skip
or reset points, set a "jump" point. So you press a button to tell the
sequencer
that you want to change the pattern, and the sequencer waits for the next
"jump"
event to carry out this order. Implementation is simple, one D-flipflop.

Now you could repeat this manually triggered action to jump into other
sequences,
but the "Event Queue" idea goes a little farther:
You don't press the manual button once before each jump event, but you can
preprogram several jumps in advance. Simply press the manual button
repeatedly,
and a number of logical "1"s are written into a shift register. Every time
the sequencer
runs across a "jump" point, it carries out this jump operation, and clears
one of these
"1"s in the shift register, until all orders are carried out.
This may be of some (limited) use to cycle thru the rows for a limited
number of times,
and then go on reapeating the last sequence. But the real fun starts when
you have 
several "jump" points in one sequence. Then the sequencer might play notes
1 and 2
from row #1, then notes 2, 3 and 4 from row #2, and continue with note 5 of
row #3.
Mixing the notes of several sequences into completely new patterns.


This is only an idea so far. No soldering today - I spent a beautiful day
in Salzburg
(Austria) with my wife instead. Back to the lab tomorrow (;->).

JH.