[Csnd] conditional, nearest quantising

[Csnd] conditional, nearest quantising

Date	2012-12-15 15:40
From	peiman khosravi
Subject	[Csnd] conditional, nearest quantising
	Dear all, I was wondering if there is a simple method for quantising an arbitrary input value to a 'grid' defined in a table. Thanks Peiman

Date	2012-12-15 15:53
From	Justin Smith
Subject	Re: [Csnd] conditional, nearest quantising
	I would write a line of code or maybe UDO that maps from input value to 0-1, then use that number to do a non-interpolated table lookup (where the table has the quantized results you want) something like this (this is untested code) - this should get you a series of values in the set (1 2 3 4) with equal probability of each at any time iquant ftgentmp 0, 0, 5, 2, 1, 2, 3, 4, 4 ; guard point? kinput noise 0dbfs, 0 kinput limit kinput, 0, 1 ; just to be safe klookup = (kinput + 0dbfs)/(2*0dbfs) ; number between 0 and 1 kquant table klookup, iquant, 1 On Sat, Dec 15, 2012 at 7:40 AM, peiman khosravi <peimankhosravi@gmail.com> wrote: Dear all, I was wondering if there is a simple method for quantising an arbitrary input value to a 'grid' defined in a table. Thanks Peiman

Date	2012-12-15 15:54
From	Justin Smith
Subject	Re: [Csnd] conditional, nearest quantising
	oops, it should have been: iquant ftgentmp 0, 0, 5, 2, 1, 2, 3, 4, 4 ; guard point? kinput noise 0dbfs, 0 klookup = (kinput + 0dbfs)/(20dbfs) ; number between 0 and 1 klookup limit klookup, 0, 1 ; just to be safe kquant table klookup, iquant, 1 like I said, untested code On Sat, Dec 15, 2012 at 7:53 AM, Justin Smith <noisesmith@gmail.com> wrote: I would write a line of code or maybe UDO that maps from input value to 0-1, then use that number to do a non-interpolated table lookup (where the table has the quantized results you want) something like this (this is untested code) - this should get you a series of values in the set (1 2 3 4) with equal probability of each at any time iquant ftgentmp 0, 0, 5, 2, 1, 2, 3, 4, 4 ; guard point? kinput noise 0dbfs, 0 kinput limit kinput, 0, 1 ; just to be safe klookup = (kinput + 0dbfs)/(20dbfs) ; number between 0 and 1 kquant table klookup, iquant, 1 On Sat, Dec 15, 2012 at 7:40 AM, peiman khosravi <peimankhosravi@gmail.com> wrote: Dear all, I was wondering if there is a simple method for quantising an arbitrary input value to a 'grid' defined in a table. Thanks Peiman

Date	2012-12-15 16:01
From	peiman khosravi
Subject	Re: [Csnd] conditional, nearest quantising
	Nice idea. Let me test it out and get back to you. Thanks Peiman On 15 December 2012 15:54, Justin Smith <noisesmith@gmail.com> wrote: oops, it should have been: iquant ftgentmp 0, 0, 5, 2, 1, 2, 3, 4, 4 ; guard point? kinput noise 0dbfs, 0 klookup = (kinput + 0dbfs)/(20dbfs) ; number between 0 and 1 klookup limit klookup, 0, 1 ; just to be safe kquant table klookup, iquant, 1 like I said, untested code On Sat, Dec 15, 2012 at 7:53 AM, Justin Smith <noisesmith@gmail.com> wrote: I would write a line of code or maybe UDO that maps from input value to 0-1, then use that number to do a non-interpolated table lookup (where the table has the quantized results you want) something like this (this is untested code) - this should get you a series of values in the set (1 2 3 4) with equal probability of each at any time iquant ftgentmp 0, 0, 5, 2, 1, 2, 3, 4, 4 ; guard point? kinput noise 0dbfs, 0 kinput limit kinput, 0, 1 ; just to be safe klookup = (kinput + 0dbfs)/(20dbfs) ; number between 0 and 1 kquant table klookup, iquant, 1 On Sat, Dec 15, 2012 at 7:40 AM, peiman khosravi <peimankhosravi@gmail.com> wrote: Dear all, I was wondering if there is a simple method for quantising an arbitrary input value to a 'grid' defined in a table. Thanks Peiman

Date	2012-12-15 16:04
From	peiman khosravi
Subject	Re: [Csnd] conditional, nearest quantising
	Hi Justin, Could you please explain this bit? klookup = (kinput + 0dbfs)/(20dbfs) ; number between 0 and 1 Thanks Peiman On 15 December 2012 16:01, peiman khosravi <peimankhosravi@gmail.com> wrote: Nice idea. Let me test it out and get back to you. Thanks Peiman On 15 December 2012 15:54, Justin Smith <noisesmith@gmail.com> wrote: oops, it should have been: iquant ftgentmp 0, 0, 5, 2, 1, 2, 3, 4, 4 ; guard point? kinput noise 0dbfs, 0 klookup = (kinput + 0dbfs)/(20dbfs) ; number between 0 and 1 klookup limit klookup, 0, 1 ; just to be safe kquant table klookup, iquant, 1 like I said, untested code On Sat, Dec 15, 2012 at 7:53 AM, Justin Smith <noisesmith@gmail.com> wrote: I would write a line of code or maybe UDO that maps from input value to 0-1, then use that number to do a non-interpolated table lookup (where the table has the quantized results you want) something like this (this is untested code) - this should get you a series of values in the set (1 2 3 4) with equal probability of each at any time iquant ftgentmp 0, 0, 5, 2, 1, 2, 3, 4, 4 ; guard point? kinput noise 0dbfs, 0 kinput limit kinput, 0, 1 ; just to be safe klookup = (kinput + 0dbfs)/(2*0dbfs) ; number between 0 and 1 kquant table klookup, iquant, 1 On Sat, Dec 15, 2012 at 7:40 AM, peiman khosravi <peimankhosravi@gmail.com> wrote: Dear all, I was wondering if there is a simple method for quantising an arbitrary input value to a 'grid' defined in a table. Thanks Peiman

Date	2012-12-15 16:10
From	Justin Smith
Subject	Re: [Csnd] conditional, nearest quantising
	if 0dbfs was for example 8, kinput would be a number between -8 and 8 if kinput is between -8 and 8, a value of 0 should map to .5 on a 0 to 1 scale (0 + 8)/(28) = 8/16 = 0.5 so regardless of 0dbfs, that line of code will come up with a mapping of the input (between -0dbfs and 0dbfs) to a number between 0 and 1 of course you can adapt this to your own needs depending on how the number is generated On Sat, Dec 15, 2012 at 8:04 AM, peiman khosravi <peimankhosravi@gmail.com> wrote: Hi Justin, Could you please explain this bit? klookup = (kinput + 0dbfs)/(20dbfs) ; number between 0 and 1 Thanks Peiman On 15 December 2012 16:01, peiman khosravi <peimankhosravi@gmail.com> wrote: Nice idea. Let me test it out and get back to you. Thanks Peiman On 15 December 2012 15:54, Justin Smith <noisesmith@gmail.com> wrote: oops, it should have been: iquant ftgentmp 0, 0, 5, 2, 1, 2, 3, 4, 4 ; guard point? kinput noise 0dbfs, 0 klookup = (kinput + 0dbfs)/(20dbfs) ; number between 0 and 1 klookup limit klookup, 0, 1 ; just to be safe kquant table klookup, iquant, 1 like I said, untested code On Sat, Dec 15, 2012 at 7:53 AM, Justin Smith <noisesmith@gmail.com> wrote: I would write a line of code or maybe UDO that maps from input value to 0-1, then use that number to do a non-interpolated table lookup (where the table has the quantized results you want) something like this (this is untested code) - this should get you a series of values in the set (1 2 3 4) with equal probability of each at any time iquant ftgentmp 0, 0, 5, 2, 1, 2, 3, 4, 4 ; guard point? kinput noise 0dbfs, 0 kinput limit kinput, 0, 1 ; just to be safe klookup = (kinput + 0dbfs)/(20dbfs) ; number between 0 and 1 kquant table klookup, iquant, 1 On Sat, Dec 15, 2012 at 7:40 AM, peiman khosravi <peimankhosravi@gmail.com> wrote: Dear all, I was wondering if there is a simple method for quantising an arbitrary input value to a 'grid' defined in a table. Thanks Peiman

Date	2012-12-15 16:13
From	peiman khosravi
Subject	Re: [Csnd] conditional, nearest quantising
	Thanks Justin, This is great. I need to convert input frequency values to 8th tone scale degrees and then do some more mapping from there. I think it should be easy enough if I start from your example. Thanks again, Peiman On 15 December 2012 16:10, Justin Smith <noisesmith@gmail.com> wrote: between

Date	2012-12-15 16:38
From	Justin Smith
Subject	Re: [Csnd] conditional, nearest quantising
	There are opcodes designed for this sort of thing (octcps for example), which would be easier than table lookup (though of course table lookup is more general). It may work best to convert via octcps and then use that as the table lookup index. speaking of more general, here is a demo of an opcode that does the range mapping with arbitrary ranges usage looks like this (you can type in instrument invocations at the command line to test it, I recommend the rlwrap program for this, to allow scrolling and editing history): i 1 0 1 0 -8 8 0 1 ... map 0.000000 from (-8.000000 : 8.000000) to (0.000000 : 1.000000) = 0.500000 it even works when the ranges get reversed i 1 0 1 0.25 0 1 0 -1 ... map 0.250000 from (0.000000 : 1.000000) to (0.000000 : -1.000000) = -0.250000 <CsoundSynthesizer> <CsOptions> -Lstdin -n </CsOptions> <CsInstruments> opcode mapto, k, kkkkk kin, ki_min, ki_max, ko_min, ko_max xin ko_offset = ko_min ki_offset = ki_min ko_range = ko_max - ko_min ki_range = ki_max - ki_min kscale = ko_range / ki_range kres = (kin - ki_offset)*kscale+ko_offset xout kres endop instr 1 kout mapto p4, p5, p6, p7, p8 printf "map %f from (%f : %f) to (%f : %f) = %f\n", 1, p4, p5, p6, p7, p8, kout endin </CsInstruments> <CsScore> i 1 100000 0 0 0 0 0 0 </CsScore> </CsoundSynthesizer> On Sat, Dec 15, 2012 at 8:13 AM, peiman khosravi <peimankhosravi@gmail.com> wrote: Thanks Justin, This is great. I need to convert input frequency values to 8th tone scale degrees and then do some more mapping from there. I think it should be easy enough if I start from your example. Thanks again, Peiman On 15 December 2012 16:10, Justin Smith <noisesmith@gmail.com> wrote: between

Date	2012-12-15 17:00
From	Justin Smith
Subject	Re: [Csnd] conditional, nearest quantising
	since this is an interesting question to me, I just put together this opcode that quantizes to n pitches to the octave: <CsoundSynthesizer> <CsOptions> -Lstdin -n </CsOptions> <CsInstruments> ;; qnth - arguments are value to quantize, and number of tones per octave ;; (ie. for a standard piano scale, 12, for quarter tones, 24, etc.) opcode qnth, k, kk kcps, kquant xin koct = octcps(kcps) kquantized = round(koctkquant)/kquant ;printk2 kquantized xout cpsoct(kquantized) endop instr 1 kout qnth p4, p5 printk2 kout endin </CsInstruments> <CsScore> i 1 100000 0 0 </CsScore> </CsoundSynthesizer> On Sat, Dec 15, 2012 at 8:38 AM, Justin Smith <noisesmith@gmail.com> wrote: There are opcodes designed for this sort of thing (octcps for example), which would be easier than table lookup (though of course table lookup is more general). It may work best to convert via octcps and then use that as the table lookup index. speaking of more general, here is a demo of an opcode that does the range mapping with arbitrary ranges usage looks like this (you can type in instrument invocations at the command line to test it, I recommend the rlwrap program for this, to allow scrolling and editing history): i 1 0 1 0 -8 8 0 1 ... map 0.000000 from (-8.000000 : 8.000000) to (0.000000 : 1.000000) = 0.500000 it even works when the ranges get reversed i 1 0 1 0.25 0 1 0 -1 ... map 0.250000 from (0.000000 : 1.000000) to (0.000000 : -1.000000) = -0.250000 <CsoundSynthesizer> <CsOptions> -Lstdin -n </CsOptions> <CsInstruments> opcode mapto, k, kkkkk kin, ki_min, ki_max, ko_min, ko_max xin ko_offset = ko_min ki_offset = ki_min ko_range = ko_max - ko_min ki_range = ki_max - ki_min kscale = ko_range / ki_range kres = (kin - ki_offset)kscale+ko_offset xout kres endop instr 1 kout mapto p4, p5, p6, p7, p8 printf "map %f from (%f : %f) to (%f : %f) = %f\n", 1, p4, p5, p6, p7, p8, kout endin </CsInstruments> <CsScore> i 1 100000 0 0 0 0 0 0 </CsScore> </CsoundSynthesizer> On Sat, Dec 15, 2012 at 8:13 AM, peiman khosravi <peimankhosravi@gmail.com> wrote: Thanks Justin, This is great. I need to convert input frequency values to 8th tone scale degrees and then do some more mapping from there. I think it should be easy enough if I start from your example. Thanks again, Peiman On 15 December 2012 16:10, Justin Smith <noisesmith@gmail.com> wrote: between

Date	2012-12-15 18:34
From	Louis Cohen
Subject	Re: [Csnd] conditional, nearest quantising
	I wrote and use this UDO often. This assumes that the table starts with an entry count, then contains entries, all sorted. The table can be pretty big because the UDO uses a binary search to find the closest match. I hope it's useful. -Lou opcode findentry, k, kk ;finds nearest table entry to the input value, using binary search ktestvalue, ktable xin ;get input value, table number ;does not test for illegal table number, but if table number == -1, returns the input value kreturn = ktestvalue if(ktable=-1)then kgoto returnvalue endif kentrycount tablekt 0, ktable ;assumes table is sorted order, table values are in entries 1 through kentrycount klorangestart = 1 klorangeend = int(kentrycount/2) khirangestart = klorangeend + 1 khirangeend = kentrycount whichrange: ;is ktestvalue in bottom half of range? klowest tablekt klorangestart, ktable ;lowest value in range khighest tablekt klorangeend, ktable ktopofbottom = khighest if(klowest<=ktestvalue && ktestvalue<=khighest)then ;ktestvalue is somewhere in this range. ;divide the lower range into two parts klorangestart = klorangestart khirangeend = klorangeend klorangeend = klorangestart + int((klorangeend-klorangestart)/2) khirangestart = klorangeend + 1 ;we are done if the new range is very small if( (khirangeend - klorangestart) < 2) then kreturn tablekt klorangestart, ktable ; printk 1,767676 kgoto returnvalue endif kgoto whichrange endif ;if not in low range, it must be in the hirange, so adjust the new high range klowest tablekt khirangestart, ktable khighest tablekt khirangeend, ktable kbottomoftop = klowest if(klowest<=ktestvalue && ktestvalue<=khighest)then ;adjust ranges for high half of the current search area ;divide the higher range into two parts khirangeend = khirangeend klorangestart = khirangestart khirangestart = klorangestart + int( (khirangeend-klorangestart)/2 ) klorangeend = khirangestart - 1 ;we are done if the new range is very small if( (khirangeend - klorangestart) < 2) then kreturn tablekt khirangestart, ktable ; printk 1,77777 kgoto returnvalue endif kgoto whichrange endif if( (ktopofbottom wrote: > Dear all, > > I was wondering if there is a simple method for quantising an arbitrary input value to a 'grid' defined in a table. > > Thanks > Peiman Lou Cohen www.jolc.net www.opensound.org

Date	2012-12-16 00:49
From	peiman khosravi
Subject	Re: [Csnd] conditional, nearest quantising
	This is all great info for me. Thanks very much! Peiman On 15 December 2012 18:34, Louis Cohen <loucohen@jolc.net> wrote: I wrote and use this UDO often. This assumes that the table starts with an entry count, then contains <entry count> entries, all sorted. The table can be pretty big because the UDO uses a binary search to find the closest match. I hope it's useful. -Lou opcode findentry, k, kk ;finds nearest table entry to the input value, using binary search ktestvalue, ktable xin ;get input value, table number ;does not test for illegal table number, but if table number == -1, returns the input value kreturn = ktestvalue if(ktable=-1)then kgoto returnvalue endif kentrycount tablekt 0, ktable ;assumes table is sorted order, table values are in entries 1 through kentrycount klorangestart = 1 klorangeend = int(kentrycount/2) khirangestart = klorangeend + 1 khirangeend = kentrycount whichrange: ;is ktestvalue in bottom half of range? klowest tablekt klorangestart, ktable ;lowest value in range khighest tablekt klorangeend, ktable ktopofbottom = khighest if(klowest<=ktestvalue && ktestvalue<=khighest)then ;ktestvalue is somewhere in this range. ;divide the lower range into two parts klorangestart = klorangestart khirangeend = klorangeend klorangeend = klorangestart + int((klorangeend-klorangestart)/2) khirangestart = klorangeend + 1 ;we are done if the new range is very small if( (khirangeend - klorangestart) < 2) then kreturn tablekt klorangestart, ktable ; printk 1,767676 kgoto returnvalue endif kgoto whichrange endif ;if not in low range, it must be in the hirange, so adjust the new high range klowest tablekt khirangestart, ktable khighest tablekt khirangeend, ktable kbottomoftop = klowest if(klowest<=ktestvalue && ktestvalue<=khighest)then ;adjust ranges for high half of the current search area ;divide the higher range into two parts khirangeend = khirangeend klorangestart = khirangestart khirangestart = klorangestart + int( (khirangeend-klorangestart)/2 ) klorangeend = khirangestart - 1 ;we are done if the new range is very small if( (khirangeend - klorangestart) < 2) then kreturn tablekt khirangestart, ktable ; printk 1,77777 kgoto returnvalue endif kgoto whichrange endif if( (ktopofbottom<kbottomoftop) && (ktopofbottom<ktestvalue) && (ktestvalue<kbottomoftop))then kreturn = ktopofbottom kgoto returnvalue endif returnvalue: xout kreturn endop On Dec 15, 2012, at 10:40 AM, peiman khosravi <peimankhosravi@gmail.com> wrote: > Dear all, > > I was wondering if there is a simple method for quantising an arbitrary input value to a 'grid' defined in a table. > > Thanks > Peiman Lou Cohen www.jolc.net www.opensound.org Send bugs reports to the Sourceforge bug tracker https://sourceforge.net/tracker/?group_id=81968&atid=564599 Discussions of bugs and features can be posted here To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"

Date	2012-12-16 00:58
From	peiman khosravi
Subject	Re: [Csnd] conditional, nearest quantising
	But can octcps be limited to the 8th tone? P On 15 December 2012 16:38, Justin Smith <noisesmith@gmail.com> wrote: octcps

Date	2012-12-16 01:37
From	Justin Smith
Subject	Re: [Csnd] conditional, nearest quantising
	as I mention in the comments, second arg to the qnth opcode I showed is number of pitches per octave, so an argument of 48 will quantize to 8th tones On Sat, Dec 15, 2012 at 4:58 PM, peiman khosravi <peimankhosravi@gmail.com> wrote: But can octcps be limited to the 8th tone? P On 15 December 2012 16:38, Justin Smith <noisesmith@gmail.com> wrote: octcps

Date	2012-12-16 01:48
From	peiman khosravi
Subject	Re: [Csnd] conditional, nearest quantising
	I yes yes sorry I'd missed that one. This is a neat UDO, thanks! Best, Peiman On 16 December 2012 01:37, Justin Smith <noisesmith@gmail.com> wrote: as I mention in the comments, second arg to the qnth opcode I showed is number of pitches per octave, so an argument of 48 will quantize to 8th tones On Sat, Dec 15, 2012 at 4:58 PM, peiman khosravi <peimankhosravi@gmail.com> wrote: But can octcps be limited to the 8th tone? P On 15 December 2012 16:38, Justin Smith <noisesmith@gmail.com> wrote: octcps