GEN15 — Creates two tables of stored polynomial functions.
This subroutine creates two tables of stored polynomial functions, suitable for use in phase quadrature operations.
size -- number of points in the table. Must be a power of 2 or a power-of-2 plus 1 (see f statement). The normal value is power-of-2 plus 1.
xint -- provides the left and right values [-xint, +xint] of the x interval over which the polynomial is to be drawn. This subroutine will eventually call GEN03 to draw both functions; this p5 value is therefor expanded to a negative-positive p5, p6 pair before GEN03 is actually called. The normal value is 1.
xamp -- amplitude scaling factor of the sinusoid input that is expected to produce the following spectrum.
h0, h1, h2, ... hn -- relative strength of partials 0 (DC), 1 (fundamental), 2 ... that will result when a sinusoid of amplitude
xamp * int(size/2)/xint
is waveshaped using this function table. These values thus describe a frequency spectrum associated with a particular factor xamp of the input signal.
phs0, phs1, ... -- phase in degrees of desired harmonics h0, h1, ... when the two functions of GEN15 are used with phase quadrature.
Notes | |
---|---|
GEN15 creates two tables of equal size, labeled f # and f # + 1. Table # will contain a Chebyshev function of the first kind, drawn using GEN13 with partial strengths h0cos(phs0), h1cos(phs1), ... Table #+1 will contain a Chebyshev function of the 2nd kind by calling GEN14 with partials h1sin(phs1), h2sin(phs2),... (note the harmonic displacement). The two tables can be used in conjunction in a waveshaping network that exploits phase quadrature. Before version 5.16 there was a bug (pointed out by Menno Knevel and fixed by François Pinot) on the number of pfields transmitted to gen13 and gen14 by gen15. The consequence is that all the csd, or orc and sco files that used gen15 before this bug was fixed, are likely to sound different now. |
Here is an example of the GEN15 routine. It uses the files gen15.csd.
Example 1109. An example of the GEN15 routine.
<CsoundSynthesizer> <CsOptions> ; Select audio/midi flags here according to platform -odac ;;;realtime audio out ;-iadc ;;;uncomment -iadc if realtime audio input is needed too ; For Non-realtime ouput leave only the line below: ; -o gen15.wav -W ;;; for file output any platform </CsOptions> <CsInstruments> ;example from the Csound Book, page 85 sr = 44100 ksmps = 32 nchnls = 2 0dbfs = 1 instr 1 idur = p3 iamp = p4 ifrq = cpspch(p5) ;pitch iswp1 = p6 iswp2 = p7 kswp line iswp1, p3, iswp2 ;amplitude sweep values acosi oscili kswp*.5, ifrq, 2 ;f2=cosine wave asine oscili kswp, ifrq, 1 ;f1=sine wave atab1 tablei acosi, 33, 1, .5 ;tables a1 to GEN13 atab2 tablei acosi, 34, 1, .5 ;tables a1 to GEN14 knrm1 tablei kswp, 35, 1 ;normalizing f35 knrm2 tablei kswp, 36, 1 ;normalizing f36 anrm1 = atab1*knrm1 ;normalize GEN13 signal anrm2 = atab2*knrm2*asine ;normalize GEN14 signal amix = anrm1+anrm2 ;mix GEN13 and GEN14 kenv expseg .001, idur*.1, iamp, idur*.1, iamp*.8, idur*.8, .001 asig = amix*kenv outs asig, asig endin </CsInstruments> <CsScore> f 1 0 8193 10 1 ;sine wave f 2 0 8193 9 1 1 90 ;cosine wave ; Note that all the f33 tables in the following sections are defined with p4=-15, ; which means that tables 33 and 34 will not be normalized. Thus if we display ; tables when running this example, we'll get correct diagrams even if one table ; has very small values instead of 0 values, due to cpu approximations in processing ; sin(180), as in sections 2, 4, and 5. This has no consequence on the audio result, ; because of the use of amp normalization (tables 35 and 36). f 33 0 8193 -15 1 1 1 0 1 180 .8 45 .6 270 .5 90 .4 225 .2 135 .1 315 ;makes function tables 33 and 34 f 35 0 4097 4 33 1 ;amp normalization for f33 f 36 0 4097 4 34 1 ;amp normalization for f34 i 1 0 5 .6 8.00 0 1 i 1 + . .6 8.00 1 0 s ;even harmonics with no phase shift, odd harmonics with phase shift f 33 0 8193 -15 1 1 1 0 1 0 1 180 1 180 1 0 1 0 1 180 1 180 1 0 1 0 1 180 1 180 f 35 0 4097 4 33 1 ;amp normalization for f33 f 36 0 4097 4 34 1 ;amp normalization for f34 i 1 0 5 .6 8.00 0 1 i 1 + . .6 8.00 1 0 s ;different harmonic strenghts and phases f 33 0 8193 -15 1 1 1 0 1 0 .9 180 .5 270 .75 90 .4 45 .2 225 .1 0 f 35 0 4097 4 33 1 ;amp normalization for f33 f 36 0 4097 4 34 1 ;amp normalization for f34 i 1 0 5 .6 8.00 0 1 i 1 + . .6 8.00 1 0 s ;lower harmonics no phase shift, upper harmonics with phase shift f 33 0 8193 -15 1 1 1 0 1 0 .5 0 .9 0 .3 0 .75 0 .2 180 .6 180 .15 180 .5 180 .1 180 f 35 0 4097 4 33 1 ;amp normalization for f33 f 36 0 4097 4 34 1 ;amp normalization for f34 i 1 0 5 .6 8.00 0 1 i 1 + . .6 8.00 1 0 s ;lower harmonics with phase shift, upper harmonics no phase shift f 33 0 8193 -15 1 1 1 180 1 180 .5 180 .9 180 .3 180 .75 180 .2 0 .6 0 .15 0 .5 0 .1 0 f 35 0 4097 4 33 1 ;amp normalization for f33 f 36 0 4097 4 34 1 ;amp normalization for f34 i 1 0 5 .6 8.00 0 1 i 1 + . .6 8.00 1 0 e </CsScore> </CsoundSynthesizer>
These are the diagrams of the waveforms of the GEN15 routine, as used in the example
(in each diagram, the curve in blue is for ftable 33 and the curve in green is for
table 34):