Optimal patent val: Difference between revisions

The '''optimal patent val''' for a [[regular temperament]] is the unique [[patent val]] that [[support]]s the temperament with the lowest [[error]].

Given any temperament, which is characterized by the [[comma]]s it [[tempering out|tempers out]], there is a finite list of [[patent val]]s that temper out all the commas of the temperament in the same [[subgroup]]. The list is not guaranteed to contain any members, but in most actual circumstances it will. If the list is not empty, then among these patent vals will be found the one which has the lowest [[TE error]]; this is the (TE) optimal patent val for the temperament. Note that other definitions of error lead to different results.

On this wiki, the optimal patent val for each temperament is given as the last patent val in the [[optimal ET sequence]], or stated explicitly in case it is not a member of the sequence.  

By tempering a JI scale using the ''N''-edo found on the list below we automatically temper it to the corresponding temperament. This can be done in [http://www.huygens-fokker.org/scala/ Scala] using the Quantize command: either type in "Quantize/consistent N" on the bottom, or use the pull-down menu under "Modify", check the box saying "Consistent" and type N (without a decimal point) into "Resolution".

To limit the search range when finding the optimal patent val a useful observation is this: given ''N''-edo, and an odd prime ''q'' ≤ ''p'', if ''d'' is the absolute value in cents of the difference between the tuning of ''q'' given by the [[POTE tuning]] and the POTE tuning rounded to the nearest ''N''-edo value, then d < 600/''N'', from which it follows that N < 600/d. Likewise, if ''e'' is the absolute value of the error of ''q'' in the patent val tuning, then ''e'' < 600/''N'' and so ''N'' < 600/''e''. If ''N''-edo defines an optimal patent val, then the patent val will be identical to the val obtained by rounding the POTE tuning to the nearest N-edo value. We have two distances from the patent val, one to the POTE tuning and one to the the JI tuning, both bounded by 600/''N'', and so by the triangle inequality the distance from the JI tuning to the POTE tuning, which is the error of the prime ''q'' in the POTE tuning, is bounded by 1200/''N''. Hence, ''N'' < 1200/error(''q''). If now we take the minimum value for 1200/error(prime) for all the odd primes up to ''p'', we obtain an upper bound for ''N''.

== Examples ==

Below are tabulated some values. In each case an identifier which uniquely identifies the temperament in question is given. In the codimension one case, where the temperament is defined by a single comma, the comma is given and used as a name. In other cases, for a temperament of rank ''n'', ''n'' independent vals are given. Normally this is by way of integers conjoined by ampersands, such as 2&10 for 7-limit pajara. This tells us we can use the 7-limit patent vals for 2 and 10 to define the temperament. In case ''n'' independent patent vals cannot be found, vals using the [[wart notation]] are given; this adjusts the nth prime mapping to its second-best value by appending the ''n''-th lower-case letter in alphabetical order. Thus, "12f" adjusts a patent val for 12 in the 13-limit or above, for instance {{val| 12 19 28 34 42 44 }}, to {{val| 12 19 28 34 42 45 }} (which is actually a better mapping, and hence more useful for this purpose.)

=== 5-limit rank two ===

Comma: ET w/ optimal patent val: 1000 * badness

16/15: [[8edo|8et]] 14.884

648/625: [[12edo|12et]] 47.231

27/25: [[14edo|14et]] 32.801

25/24: [[17edo|17et]] 13.028