source file: mills3.txt Date: Wed, 14 Jan 1998 21:06:13 +0100 Subject: MIDI/Audio wish list From: "Patrick Ozzard-Low" John Loffink wrote: >> My point is that it is not > >necessary to do this by using more "Partials" (I hate Roland's usage of that > >term, which is why I always put it in quotes) or zones if the instrument > >supports microtuning tables. The microtuning table, if implemented properly, > >will do the mapping for you, automatically, without all the recompilation and > >excessive memory usage. Granted, you will not have precise control at a less > >than semitone level of how the samples are mapped, but is this really > >necessary? You will get "mickey mouse" effects no matter where your base > >sample was recorded at, this is the effect of shifting the formant (fixed > >resonant) frequencies of the sample. Regarding Mr Mickey Mouse: On the Roland, if you want to preserve a any kind of realistic quality of timbre the upper limit of transposing a sample from its original sampled pitch is about a minor third (up), and a major third down - depending on the timbre. In my own case, I typically use a sample mapped to no more than about 150 cents BELOW the original sampled pitch, but very seldom transposed upward. This is more than adequate for my purposes (local friends think I'm too fussy) and there is no 'mickey mouse' effect within these limits. (Of course, you can get away with more than this if realism is not important, - or you can do away with realism altogether and transpose up (as far a two octaves) - and down (as far as you like). The E4 is slightly better better at preserving realism over a wider range, but by no more than a semitone. It can also transpose upward over a much greater range, I believe). > >Increasing the number of "Partials" will not solve your problem, > >because you'll have a new one -- the sampler's internal RAM will > >not have enough storage to handle the increase. Program parameters > >are most likely to be stored in 64K to 512K local RAM inside the > >unit, not your 32M to 128M sample RAM. If you now ask > >manufacturers to increase the local RAM just for microtonal > >support, I can tell you that none of them will do it because that > >increases the price of the hardware. Sorry, John, this is wrong. Increasing the number of 'partials' has no effect on the used sample RAM on the Roland or AKAI systems. (It does on the EMU, but, as I understand, so little that it is negligible). On the Roland - suppose there are 4 samples in memory, of flute tones, Eb4, F#4, A4 and C5; and that they take up 2Mb. You then create a 'one octave 41-ET' flute, by creating 41 'partials' which map each of 3 samples to 10 divisions of the scale, and the 4th sample to the remaining 11 divisions, such that you can now play this scale over 3 and a bit standard-keyboard-octaves. The mapping from sample to 'partial' is one-many, and you've still only used 2Mb RAM. On the Roland there is no limit to the degree of one-many-ness, excepting the limits on samples and 'partials' (512 and 255 respectively). (Actually this is not quite right - the relation is many-many, except that the maximum number of samples that can be associated with any one 'partial' is 4.) For your interest, there is also a 'micro-key-follow' tuning facility in the Roland S7XX series, whereby you can 'interpolate' (to use Gary's 'Ensoniq' terminology) a 'partial' across the (any part of the) keyboard in 1/8ths of a semitone, 2/8 of same, 3/8, 4/8 (quarter tones), 5/8,7/8, 8/8 (12-ET), 9/8, etc up to 16/8 of a semitone (whole tones). > >I'm sorry if this is getting too technical again, but you can't really > >understand the ramifications of your requests without a little understanding > >on how today's synthesizers and samplers work. My points are 1) for > >microtonal scales to be implemented in mass produced instruments it must be > >economically realistic for the manufacturers to do so, and 2) for microtonal > >scales to be used by non-specialists it must be as easy as possible for them > >to be changed. You can't accomplish either of these by defining microtonal > >scales at the "Partial" or zone level. I agree entirely with the general points 1 and 2. I'm unconvinced however about the last sentence of this paragraph. Samplers, if used for anything approaching realism, will require significant amounts of RAM whatever the 'partial'/zone architecture etc, at least until some form of data compression of the 16 bit format is introduced (or something). The fact that EMU have implemented 128Mb on the E4 seems to confirm this - they didn't build it for composers working with ATS. More 'partials' does not mean more RAM; on the Roland (at least) only samples take up sample RAM. But I suspect you may have other (good) reasons for saying what you do (?). Gary Morrison wrote: > To set up a repeating pattern of pitches, such as 12-Toned Ptolemaic JI > in C, you enter the pitches for the keys C4 through C5. You would set, for > example, E4 14 cents flat of the 12TET reference. (Actually, that's not > exactly true; you specify pitches between 0 and 99 cents above whatever > 12TET pitch reference lies directly below it, so you'd define the E4 key > top play an Eb4 86 cents sharp.) Notice that you set the pitch of both of > the C4 and of the C5, the C5 being an octave higher than the C4. Thanks, Gary. I've seen this sort-of kind-of thing on various synths. Can I ask a further question? I understand what you have described, that is if you want one single sound (either a synthesised sound or a sample) mapped to this kind of 'key-table'. But, on the Ensoniq sampler, if you want to achieve some simulatory realism do you have to laboriously map each sample to a point or points on the key-table - or is that not something that can be done (or is nor worth the trouble etc)? (Please excuse me, I'm aware of being very biased toward wanting realism. I know that's not what everyone wants - and it should not bias this discussion. But it's what interests me about these machines!) > Now, 22TET... You have a very basic question to answer before you can > set it up. What keyboard mapping do you want? How do you want pitches in > the tuning to map to keys on the keyboard? A common way is to map them > "linearly", each key 1/22nd of an octave apart, mismatching octave > boundaries in the tuning with those on the keyboard. If that's the way you > want to map it, then set the pitches of the keys C4 through Bb5 to > successive 22nds of an octave, and extrapolate your table from C4 to Bb5. In my work I've created simulations of many of the standard orchestral instruments for every n-ET between 13 and 36, (and some others above, especially 41) in each case creating a single partial-sample relationship for each pitch (!). Creating them is one thing - learning to play them all is another! In most cases the keyboard mapping I've used is simply 'keyboard-chromatic', since I get used to re-remembering where the octaves (and everything else) are. Anything less than but near to a multiple of 12, I cheat a bit, and ditch a key somewhere. Patrick Ozzard-Low SMTPOriginator: tuning@eartha.mills.edu From: Aline Surman Subject: Surman/b5 ratios PostedDate: 15-01-98 00:34:56 SendTo: CN=coul1358/OU=AT/O=EZH ReplyTo: tuning@eartha.mills.edu $MessageStorage: 0 $UpdatedBy: CN=notesrv2/OU=Server/O=EZH,CN=coul1358/OU=AT/O=EZH,CN=Manuel op de Coul/OU=AT/O=EZH RouteServers: CN=notesrv2/OU=Server/O=EZH,CN=notesrv1/OU=Server/O=EZH RouteTimes: 15-01-98 00:34:22-15-01-98 00:34:22,15-01-98 00:34:09-15-01-98 00:34:09 DeliveredDate: 15-01-98 00:34:09 Categories: $Revisions: Received: from ns.ezh.nl ([137.174.112.59]) by notesrv2.ezh.nl (Lotus SMTP MTA SMTP v4.6 (462.2 9-3-1997)) with SMTP id C125658C.00817A61; Thu, 15 Jan 1998 00:34:43 +0100 Received: by ns.ezh.nl; (5.65v3.2/1.3/10May95) id AA02348; Thu, 15 Jan 1998 00:34:56 +0100 Date: Thu, 15 Jan 1998 00:34:56 +0100 Received: from ella.mills.edu by ns (smtpxd); id XA02386 Received: (qmail 17326 invoked from network); 14 Jan 1998 15:34:52 -0800 Received: from localhost (HELO ella.mills.edu) (127.0.0.1) by localhost with SMTP; 14 Jan 1998 15:34:52 -0800 Message-Id: <34BD5343.695@dnvr.uswest.net> Errors-To: madole@mills.edu Reply-To: tuning@eartha.mills.edu Originator: tuning@eartha.mills.edu Sender: tuning@eartha.mills.edu