Carlo
03-28-2007, 08:05 PM
As is known, the quantization function of the Clavinova in the Digital Recording/Song Creator environment serves to align a recorded musical phrase to a basic value of the beat (e.g. quarter, eighth note, etc.) in the bar, so as to remedy misalignments and imprecisions of the execution (in this discussion, we leave aside aesthetical considerations on the usage of "rubato" and the like).
At present, the quantization function applies to a WHOLE channel, and therefore the choice made by the user enforces a quantization based on a UNIQUE choice of the basic value of the beat, which represents the minimum level of resolution of the clock setting to apply to the entire channel (i.e. to a whole staff of the score of your song).
In practice, if you choose to quantize a channel to eighth notes, the minimum duration of the displacement between notes in that channel is one eighth, and all other displacements will be multiple thereof. In other words, the position within a bar of the resulting notes will not be more "fine-grained" than 960 clocks, a clock being the minimal resolution quantity supported by the CVP (remember that 1 quarter equals 1920 clocks).
Now, since within a channel (corresponding to a score staff) musical phrases can use notes of various duration, if you wish to use quantization without distorting and corrupting the melody course, you are compelled to choose the shortest note value employed therein. This is the advice that the Yamaha manuals give to the user.
But, if you use the minimal duration note, the longer notes, after quantization, may happen to be relocated in the "wrong places" of the bar, because their natural quantization would have to have been computed on a longer basic value of the beat (say, for instance, half note instead of eighth note).
In short, the final result obtained by using the technique suggested by the manual may be quite messy. This, at least, is my experience.
The usage of the "composite" basic values of the beat (e.g. eighth note along with eighth-triplet, and the like), recommended by the manual in case of "heterogeneous" musical phrases, leads to even more imprecise and casual results.
The best solution, which is my wish, is that quantization should be applicable, in principle, not only on a whole channel, but also on a user-selected set of channel-events, i.e. a group of bars that the user can manually select by means of the "multi-select" feature already available in the editing channel environment (the whole-channel option coincides, of course, with the "select-all" case).
In this way, you could quantize a musical phrase consisting of, say, four contiguous measures of eighth triplets in the appropriate way, whereas a little further it might be convenient to quantize the phrase by using a basic value of 16th notes (of course the second multi-selection would leave the first quantization unmodified).
I think this would be a very, very useful feature to obtain highly-precise recorded songs, even starting from a rather poor first recording.
At present, I use a trade-off work-around that tries to minimize the manual editing interventions to be performed after executing the quantization.
In practice, I try to evaluate the "prevailing" (most frequent) subdivision in the recorded song to choose the right basic value of the beat and let the quantization operate accordingly (crossing the fingers, after pushing the button... ;) ).
Afterwards, based on the obtained result, I mend the mis-quantized bars by manually editing them :( .
If anyone has analyzed this problem more deeply, I would greatly appreciate comments and suggestions.
Regards,
At present, the quantization function applies to a WHOLE channel, and therefore the choice made by the user enforces a quantization based on a UNIQUE choice of the basic value of the beat, which represents the minimum level of resolution of the clock setting to apply to the entire channel (i.e. to a whole staff of the score of your song).
In practice, if you choose to quantize a channel to eighth notes, the minimum duration of the displacement between notes in that channel is one eighth, and all other displacements will be multiple thereof. In other words, the position within a bar of the resulting notes will not be more "fine-grained" than 960 clocks, a clock being the minimal resolution quantity supported by the CVP (remember that 1 quarter equals 1920 clocks).
Now, since within a channel (corresponding to a score staff) musical phrases can use notes of various duration, if you wish to use quantization without distorting and corrupting the melody course, you are compelled to choose the shortest note value employed therein. This is the advice that the Yamaha manuals give to the user.
But, if you use the minimal duration note, the longer notes, after quantization, may happen to be relocated in the "wrong places" of the bar, because their natural quantization would have to have been computed on a longer basic value of the beat (say, for instance, half note instead of eighth note).
In short, the final result obtained by using the technique suggested by the manual may be quite messy. This, at least, is my experience.
The usage of the "composite" basic values of the beat (e.g. eighth note along with eighth-triplet, and the like), recommended by the manual in case of "heterogeneous" musical phrases, leads to even more imprecise and casual results.
The best solution, which is my wish, is that quantization should be applicable, in principle, not only on a whole channel, but also on a user-selected set of channel-events, i.e. a group of bars that the user can manually select by means of the "multi-select" feature already available in the editing channel environment (the whole-channel option coincides, of course, with the "select-all" case).
In this way, you could quantize a musical phrase consisting of, say, four contiguous measures of eighth triplets in the appropriate way, whereas a little further it might be convenient to quantize the phrase by using a basic value of 16th notes (of course the second multi-selection would leave the first quantization unmodified).
I think this would be a very, very useful feature to obtain highly-precise recorded songs, even starting from a rather poor first recording.
At present, I use a trade-off work-around that tries to minimize the manual editing interventions to be performed after executing the quantization.
In practice, I try to evaluate the "prevailing" (most frequent) subdivision in the recorded song to choose the right basic value of the beat and let the quantization operate accordingly (crossing the fingers, after pushing the button... ;) ).
Afterwards, based on the obtained result, I mend the mis-quantized bars by manually editing them :( .
If anyone has analyzed this problem more deeply, I would greatly appreciate comments and suggestions.
Regards,