Recommended architectures for microtonal synthesizers

The bare minimum architecture:

The moderate architecture

The advanced architecture

Special considerations for all instruments

Samples and parameters should be mapped according to the microtuning table note pitch being played, not the MIDI note number. To do otherwise makes full keyboard microtuning tables virtually useless, as scales with more than 12 notes per octave would be "stretched" away from their natural sampled sound. This also applies to any filter or DSP parameters that track the keyboard.

Using sample keymaps to acheive tuning scales by creating a range for individual notes and and then retuning it is an extremely painstaking and error prone way to get microtonal scales. This is not support for microtuning, but rather just a byproduct of the sampling process that conveniently can be used for microtuning by those patient enough to do so. For this reason, samplers are not listed in the charts unless they have explicit support for tuning tables.

Pitch resolution should be displayed in cents in the actual resolution used by the instrument. To make it "easier" for users, most manufacturers display microtuning offsets in whole number cents: -2, -1, 0, +1, +2, etc. Unfortunately, the internal resolution rarely, if ever, matches these values. Many synthesizers have an internal pitch resolution, set in software, of 768 steps per octave or 64 steps per semitone. This is 1.5625 cents. So the actual microtuning offsets are 0, 1.5625, 3.125, 4.6875, 6.25, 7.8125, etc. This results in an additional amount of pitch tuning error for the user due to the amibiguity of the whole number offset to internal offset resolution. Let's say the user wants to set a note to +7.4 cents over the equal tempered value. The displayed microtuning value selected would be +7 cents, but the actual tuning offset would probably be +6.25 cents since that is closest to + 7 cents. In actuality +7.81 cents would have been far more accurate. Showing the cents value to two decimal places (0, 1.56, 3.12, 4.69, etc.) would be sufficient, and much more accurate.

Pitch resolution of 1 cent is generally considered to be sufficient resolution for equal temperament. But this is not the case for just intonation, and here's why. Just intonation is based upon eliminating the beats we hear in all tempered harmonies. When you listen to an a cappella choir or a string quartet, their natural instinct is to play the pure harmonies of just intonation. These harmonies can only be achieved by eliminating the beats between any intervals sounded. While it is generally agreed that pitch perception is limited to 1-2 cents for melodies, harmonies in just intonation can be corrupted by this pitch resolution because anyone can still hear the beats between notes. A limited polling of practicing microtonalists has indicated that 0.1 cent pitch resolution is a suitable value for most music.

One useful feature for all sampler users, suggested by Patrick Ozzard-Low, would be an onboard absolute pitch meter. This could be implemented non-realtime like many of the DSP functions such as mixing and crossfade looping. Setting pitch offsets would be incredibly simplified with this feature onboard. Of course, this would only be expected to work with solo instruments, not choruses or vaguely pitched instruments like most percussion.

Patrick Ozzard-Low also suggested the following interface. It serves as a good example of a synthesizer microtuning user interface.

Click on a button called 'Tuning';
Up comes a dialog box, which says:
New or Load or Edit
New is new, Load loads predefined tuning configuration; Edit edits one. Suppose we choose 'New' -
Up comes a dynamic dialog box:
Global or Instrument or Instrument-Element or MIDI Channel
As you click different options, radio buttons/check boxes become available or not etc:
Define '1/1' to keyboard : key and pitch (conventional name plus or minus cents, and in Hz)
Choose Equal-Tempered or Non-Equal-Tempered;
Depending on what you choose, dialogue box changes etc:
Define terms: Cents or Hz or Ratios;
Choose cyclic interval (1200=default, or XXXX, or none);
If ET - choose number of scale steps;
If non-ET - choose number of scale steps and exact sequence of intervals;
Assign synth sound(s) - synth deals automatically;
Assign sampler sounds - user chooses from a list of samples,
(each of which contains the original sampled pitch in Name +/- cents/Hz as part of the sample field. Module assigns program or samples accordingly, to nearest acurate pitch etc. User is then offered opportunity to test and override sample mapping if necessary).

Obviously no one wants to do this for every instrument - so once a template is saved it can be recalled etc.

Questions? Contact