|
The year 2003 was, for me, a turning point in the field of computer composition. For some decades, composing programs of various sorts had been available, and put to good use by many human composers. But, apart from some specialist single-purpose tools, the machine was generally an all-or-nothing solution: you either composed by hand, or you carried out a computer search. This meant that, for anything other than fairly simple problems with small search spaces, a computer was useless: only a human could make progress. How much better it would be, I thought, if man and machine could work together on the tough jobs.
By the autumn of 2003 I had found a way to do this. I realised that human composers often broke composing problems down into separate steps, for example by first creating small blocks of densely musical courses, then trying to link the blocks together into a full peal length. Why couldn't the machine help with one of these two tasks? It was easy to computer-search for blocks: just specify a start and end change other than rounds. But what about the potentially more difficult and tedious job of linking blocks?
It turns out that this is easy to automate, too. In the table-build phase of a computer search, which is really just the initalisation, it is possible to "force" a desired sequence of calls from a certain lead head, by removing all other links between the leads in the block. You can also go further, and delete any other lead head (or node) in the search space which is false against your favoured block. If you do this, the main search algorithm proceeds perfectly as normal; but if it ever reaches the starting change of your block, it is forced to include the whole thing. What's more, the search will never include changes which would preclude your block from being rung. Allow multiple blocks to be specified in this fashion, and mandate their starting change, and you have the basis of SMC32's "linkage search", introduced in October of 2003.
By early 2004, the linkage search was bedded in, and starting to enable the symbiosis between man (or MBD) and machine that I had hoped for. One of the first, and still the most successful applications, was the Bristol Major one part. And so, the birth of "Series 1", whose four compositions are set out below with the original composer's notes, which describe in my usual generous manner the architecture of each arrangement. The starting points for these peals were blocks composed by hand, designed to join up all of the 5678/8765 courses: a good example is the "super-efficient" block described fully in the notes for the 5088. Using this and other blocks I would try various types of machine linkage search, in an attempt to join them optimally, with the right length and the highest music counts, particularly focussing on little-bell music. Sometimes this would fail to produce anything worthwhile, and I would move on to other possibilities, but occasionally I would break into the mother lode of music and elegance.
5154 no.1 was the first successful result, but this was rapidly bettered as I discovered superior blocks, and cleverer ways of linking them. 5026 no.1 is the only peal in this series to include all 72 5678/8765, and is still very popular in some quarters, but 5088 no.3 and 5120 no.1 are arguably superior, with higher little-bell counts. (For the fullest description of the musical goals of my "short-course" genre, see the 5088).
In my view these Series One peals are still excellent arrangements, and at the time I did not think they could be bettered. But the continuing development of more sophisticated tools and techniques has allowed me to revisit the genre again and again, each time finding more to squeeze out of the method: see in particular Series Two, Series Five and Series Six. In the great cordillera of change ringing resources, Bristol is proving to be a rich and, so far, inexhaustible mine.
Series One:
| | | | | |
5026 no.1 | 72, 17, 120 | Martin Cansdale, October 2004 |
5088 no.3 | 70, 23, 134 | Matthew Hilling, April 2004 |
5120 no.1 | 70, 22, 132 | Pauline Champion, October 2004 |
5154 no.1 | 64, 12, 136 | Benjamin Constant, May 2007 |
5026 Bristol Surprise Major (no.1)
Composed by Mark B Davies
23456 M B W H
--------------
43526 5 2 -
26354 2
43652 - -
54632 -
24536 - -
43265 -
45362 2 -
62453 - 2
34256 - -
32546 - 2
(42536) 5 s
--------------
Contains:
72 5678/8765 front and back
17 6578 back
120 LB4
36 LB5
| |
Bristol is not a sledgehammer method; the way in which its music
develops, although well-known, is clever and subtle. However it
seems to me that too many compositions ride rough-shod over this
delicacy, hammering home rollups with multiple calls at the
course end, and so ignoring the interplay of music between
front and back, and lengthening the courses to create great
static chunks of work.
Beginning with my 5154 (also available on this site) I sought
a way of treating the method better, whilst using standard,
tenors-together calls. These new peals use blocks of Befores,
with very few course-end calls, to give lots of relatively
short courses; all these arrangements have twenty courses to
the peal.
The 5026 here starts and finishes with blocks of five Befores.
These provide 56 or 65 rollups but also have wonderful little-
bell music at their heart. The middle of the composition uses
the same BBMHWBWB block as the previous round-block
arrangements, which is beautifully efficient in pulling in the
the xxx56/56xxx and 6xxx5 courses not included in the rotations
of 42356.
There is a sprinkling of little-bell music throughout the
composition, including courses of 53462 and 63542 as well
as the 46532, 24653, 23564 and 35642 courses within the Before
blocks. Finally, the entire plain course is of course present,
together with both full courses 52436 and 54326, which appear
just before the final section of Befores.
|
5088 Bristol Surprise Major (no.3)
Composed by Mark B Davies
23456 H M B W
------------------
54326 - -
26354 2 2
36452 -
54632 - -
24536 - -
43265 -
62345 - 5 -
56342 - - -
23645 2 2
34256 -
53462 - 2
63254 - - -
23456 -
------------------
Contains:
70 5678/8765 front and back
23 6578 back
134 LB4
40 LB5
| |
What is there new to be added to the field of Bristol Major
composition? There are many hundreds of existing compositions
for this most popular of methods, but a brief survey led me to
conclude that the majority concentrate on back-bell
combinations, with the occasional diversion to Queens music
with split-tenors arrangements. Crucially, almost all extant
compositions contain very few courses - on average perhaps 15
or 16, but some as low as 13 - because the emphasis is on the
rollups around the course end.
The main exception to this rule is the handbell composition,
where blocks of Befores are often used to keep the inside pairs
coursing for several uninterrupted courses. Now it struck me
that not only does this have the advantage of letting the music
develop off the front, and generally producing a simpler,
faster-moving structure, but also that many of the best
coursing orders that we're now used to from Royal and Maximus
compositions are, in the same way, easily linked by Befores. For example,
from the 64235 coursing order we have:
64235
B 56423 (generates 5678 and 8765 front and back)
B 35642 (the waterfall course, 5-bell little-bell music)
B 23564 (more little-bell music, including two superb LB6 runs)
B 42356 (5678 and 8765 plus lots more little-bell music)
B 64235 (back to 64235: 65s plus LB5)
This is an absolutely classic cascade of musical courses which
can be had in five full courses with no calls around the course
end.
|
Taking this idea further, I decided that the ideal modern
composition of Bristol Major should fulfill these goals:
- Maximise the number of courses within a standard length,
in order to give a dynamic feel to the peal and provide the
same music off the front as the back.
- Pull in as many of the classic higher-numbers little-bell
courses as possible, in particular the LB5 courses 53246,
64235, 35642, 24653 and 53462. Other excellent little-bell
courses include 23564 and 46532 (654321xx and xx123456 runs)
and the LB4 courses with the 6th home: 42356, 24536 and 35426.
- Use whole courses of 56xxx, 5xxx6 and xxx56 to get as many
5678 and 8765 runs front and back as possible.
- Finally, provide an interesting and well-paced development,
with different types of music spread through the peal, lots of
little gems tucked away within the structure, and making sure
to finish on a real high.
The 5088 above fulfills all these criteria. It contains 20
courses, extremely high for any but a handbell composition;
there are only two paired calls. It has all but two of the
72 5678/8765 rollups possible front and back, and these are
spread evenly throughout the peal. Little-bell music of various
kinds is also interspersed throughout, including most of the
possible LB5 runs. I didn't expressly aim for 6578 rollups, but
there's all but one of these. Finally, the peal finishes with
the last six leads of the plain course - what could be better.
Structure of the Peal:
The peal opens with two settling courses with the 6th fixed,
both providing 56 rollups. The first Before then transitions
into a long block which very efficiently threads together both
6xxx5 courses other than 64235, together with two-thirds of the
56xxx and xxx56 courses - again, the ones that aren't rotations
of 64235. In the middle of this block, there's a whole course of
the five-bell run generator 53462.
A bob Home transitions into the next block, which is the five
Befores from 64235. This wraps up the remaining 56xxx/xxx56
courses as well as providing large numbers of LB5 runs.
We're now over halfway through, but there's plenty of music
left! Another short block of Befores wraps up the other
waterfall course, 24653, as well as 46532. Some more 6578s are
generated by the two Wrongs after this section (I think 6578s at
the back are probably worth having, whereas 8756s off the front
seem less worthwhile), and then immediately the 54236 course is
brought up - the second 5xxx6 course and the last before the
plain course.
The peal moves into the final four courses with some excellent
little-bell music, and the last few 65s, before transitioning
quickly and simply to the plain course for the last six
beautiful leads.
|
5120 Bristol Surprise Major (no.1)
Composed by Mark B Davies
23456 M B W H
--------------
42356 -
54326 -
42563 -
35264 - -
63254 -
43526 - 2 -
26354 2
43652 - -
54632 -
24536 - -
24365 - -
36245 5 - -
56342 - -
23645 2 2
23456 - -
--------------
Contains:
70 5678/8765 front and back
22 6578 back
132 LB4
40 LB5
| |
A similar idea to my one-part 5088, but with the plain course
appearing in the 5th course rather than being reserved to the
end. The bulk of the peal is identical, but if you'd like to
start with the little-bell courses 65324 and 24536/62453 rather
than finish with them (as the 5088 does, with 35426/63542 and
65234) then you might prefer this one.
|
5154 Bristol Surprise Major (no.1)
Composed by Mark B Davies
23456 M B W H
--------------
26354 - - -
35264 - -
45236 4 - -
34562 - 2
63425 - 2
24536 - -
65324 3
54326 - 2
34256 2 -
32546 - 2
(42536) 5 s
--------------
Contains:
64 5678/8765 front and back
12 6578 back
136 LB4
37 LB5
| |
The final peal (for now!) in my series of 20-course
arrangements. To recap, the targets were:
- Lots of whole, but short, courses, without many course-end
calls.
- Lots of little-bell music.
- Concentration on 5678 and 8765 rollups off front and back;
these provide the best contrast with the little-bell runs.
Apart from the odd 6578 at the back, other CRUs are considered
dull and worthless.
For further notes, see my one-part 5026 and 5088 peals, also on
this site.
In fact, this 5154 was the first 20-course arrangement I
produced, and pre-dated the discovery of the 'super-efficient'
BBMHWBWB block. In place of this it uses two separate sections
to pull in the 342 and 234 56/65 courses; the WHWH extension
at the beginning of the peal, and the block around the three
Befores. There is an excellent range of little-bell music in
this peal, including transitions 35426/63542 and 64352/26435
at Befores, plus whole courses of 25346 and 24536, in addition
to the LB5 music in the two "five-Befores" blocks.
|
|