Craig wrote:

> There's more to it than "smoking fingers."  Most player rolls
> are "arranged," in that they were modified to simulate what would
> otherwise require very high tempos on many more of them.

I'm not an old hand at this, but have done some work lately, so let me
give my 2c.

With a roll arrangement (and Doug Henderson is absolutely correct: they
were *all* ultimately arranged) the choice of roll speed is potentially
influenced by a number of factors:

  1.  Roll buildup
  2.  Granularity
  3.  Repetition

1.  I can't speak for arrangements from a recording piano (MIDI
nowadays) since everything I do is a "drafting board" arrangement.
Having said that, I don't pay much attention to roll buildup except
that it shouldn't be too horribly noticeable if left unattended.  To
achieve this, I must pay attention to roll length.   Roll length can
be adjusted by changing the "steps per beat" ratio.  Thus a 12/beat
arrangement will use more paper than a 10/beat than an 8, than a 6.

So assuming a fixed maximum paper length a fast two-step would be coded
at 8 or 6 per beat, a medium up-tempo number at 10, and a typical
fox-trot at 12.

None of this leaves much room for subtle rubatos or arpeggiation, which
brings up the next factor.


2.  Granularity, or the ability to play "in the cracks" of the beats is
ultimately affected by the rows/foot of the production puncher.  Doug
Henderson achieves high granularity from his ability to advance the
paper by very small steps while he is creating the roll.  (I have not
asked him about this, but I venture to guess that some choices he had
to make in the past were influenced by the granularity of the Play-Rite
production punchers.)

There are two ways of increasing granularity in production -- increase
the number of steps/foot or increase the roll speed to achieve more
steps/beat.   The former maintains paper length, the latter increases
it.

Ampico granularity was (I believe) around 360/foot on classical rolls
and 240 for earlier pop rolls (Wayne, help me here).  QRS advance is
264 (22/inch).   Custom Music Roll is ca. 540?


3.  There is a fixed standard measurement of "land" that must exist
between two successive strikes of the same pitch, but I've forgotten
it.   In practice, one leaves three rows of land.  assuming 240 rows
per foot.  This is to allow the player and piano mechanism enough
time to reset before the next note strikes.   Thus the *only* way to
increase the speed of repeated equal pitches is by increasing the roll
speed.

There are two instances of this.  One where a pair of notes is in
succession, and in marimba effects.  Marimba effects in octaves are more
successful then ones with the same pitch because the individual slots
can be longer.   With the same pitch, in order to keep the 3-row "land"
requirement the slot length is reduced and the note may not strike
well.  The end of "XMAS-1997" has a single note marimba effect that is
on the edge of unplayability because of this.   On well regulated
machines it succeeds, and on leaky ones it does not.

Marimba effects in octaves are the same problem as trills.  If the
effect is too fast, then either there is not enough land or the slot
is too short.  There are two solutions to this.   One is to increase
the roll speed, and the other is to execute the marimba/trill in a
dissimilar meter.  Thus in 4/4, execute the effect as a triple or
quintuple, and vice versa.   I've found that in 10/step arrangements,
the tremolo works very well because it is naturally a quintuple.


In the case of an Art Tatum arrangement, such as one of the ones that
J. L. Cook did, one must recognize that he had to target a "coarse"
perforator (264/foot), and keep the roll length within economic limits.
Thus "Get Happy" is an astonishing feat.  Careful study of the roll
(probably still available from QRS) will show that there are few, if
any, repetitions that violate the "land" requirements or the
slot-length -- they all sound.  Perhaps not with the speed and velvet
effect that Tatum could get, but --  it's not Tatum.   And the roll
sounds well at 90-100.

But ultimately, one does not need high roll speeds (>120) to get fast
note clusters since repetition is not involved, only succession, and
there one can get at least 240, 262, 360, 540 (Doug -- higher?) notes
per foot with no problem as long as your player has good lungs.

Also, Pauline Alpert's speed is in successive notes, not repeated
notes so I think the mechanism had little to do with it.  They were
pop rolls, and nobody spent much time with pop rolls.  They needed to
be produced like sausage.

I think I have most of this right, but please correct the errors in
the above and we can all learn.

George Bogatko