backport from #1055 and #1012

[sss-create]

backported from: #1055

As the title says + testing :)
I just wonder what happens when shifting time:

tidal> s "bd sn bd bd bd cp sn" # nudge ( 1002 <~ rand)
  (0>⅐)|nudge: 0.26717657037079334f, s: "bd"
  (⅐>²₇)|nudge: 5.507626384496689e-2f, s: "sn"
  (²₇>³₇)|nudge: 8.829252980649471e-2f, s: "bd"
  (³₇>⁴₇)|nudge: 0.12956790439784527f, s: "bd"
  (⁴₇>⁵₇)|nudge: 0.4096888657659292f, s: "bd"
  (⁵₇>⁶₇)|nudge: 0.1880505084991455f, s: "cp"
  (⁶₇>1)|nudge: 0.9954968336969614f, s: "sn"

tidal> s "bd sn bd bd bd cp sn" # nudge ( 1002 <~ normal)
  (0>⅐)|nudge: 1.0577018423203732f, s: "bd"
  (⅐>²₇)|nudge: 1.3265466381271827f, s: "sn"
  (²₇>³₇)|nudge: 1.2562830059776906f, s: "bd"
  (³₇>⁴₇)|nudge: 1.1939579886268699f, s: "bd"
  (⁴₇>⁵₇)|nudge: 0.9585745094129575f, s: "bd"
  (⁵₇>⁶₇)|nudge: 1.1275300958307755f, s: "cp"
  (⁶₇>1)|nudge: 0.5326128959276948f, s: "sn"

Is it correct that values are not clamped between 0 and 1 anymore even with rand -- or are these values normalized internally then?

[yaxu]

Tagging @patrickt

[yaxu]

Agreed that it should be scaled to the 0..1 range.

Saying that, in the 2.0 branch I did start organising things so all the signals had a 2 version that was bipolar i.e. from -1 to 1.. We can backport that in a separate PR though.

Tidal/src/Sound/Tidal/Signal/Waveform.hs

Lines 32 to 37 in 2263c2e

	-- | Sawtooth waveform
	saw :: (Fractional a, Real a) => Signal a
	saw = waveform $ \t -> mod' (fromRational t) 1
	saw2 :: (Fractional a, Real a) => Signal a
	saw2 = toBipolar saw

Also looking at the normal function I also see that rot 1 is used to try to get a different random stream. I think that should be rotL 1.

[sss-create]

I'll further investigate the scaling issues and take a look at the bipolar implementation in the tidal 2 branch. Looks interesting!

[sss-create]

It looks like the bipolar version is already in the dev branch:

Tidal/src/Sound/Tidal/Core.hs

Lines 69 to 79 in d05c3b2

	-- | @saw@ - unipolar ascending sawtooth wave. A pattern of continuous values
	-- following a sawtooth with frequency of one cycle, and amplitude from
	-- 0 to 1.
	saw :: (Fractional a, Real a) => Pattern a
	saw = sig $ \t -> mod' (fromRational t) 1
	-- | @saw2@ - bipolar ascending sawtooth wave. A pattern of continuous values
	-- following a sawtooth with frequency of one cycle, and amplitude from
	-- -1 to 1.
	saw2 :: (Fractional a, Real a) => Pattern a
	saw2 = sig $ \t -> mod' (fromRational t) 1 * 2 - 1

[yaxu]

It looks like the bipolar version is already in the dev branch:

Oops, sorry for the distraction!

[sss-create]

These now look way more like normal distributed values:

tidal> s "bd sn bd bd bd cp sn" # nudge ( 1002 <~ normal)
  (0>⅐)|nudge: 0.4614305436669617f, s: "bd"
  (⅐>²₇)|nudge: 0.1837912906828634f, s: "sn"
  (²₇>³₇)|nudge: 0.21515425648364675f, s: "bd"
  (³₇>⁴₇)|nudge: 0.17697178051416398f, s: "bd"
  (⁴₇>⁵₇)|nudge: 0.470017897933857f, s: "bd"
  (⁵₇>⁶₇)|nudge: 0.3055551880905872f, s: "cp"
  (⁶₇>1)|nudge: 0.5f, s: "sn"

tidal> s "bd sn bd*7cp sn" # nudge ( 1002 <~ normal)
  (0>⅕)|nudge: 0.2527900002825703f, s: "bd"
  (⅕>⅖)|nudge: 0.5f, s: "sn"
  (⅖>³₇)|nudge: 0.5f, s: "bd"
  (³₇>16/35)|nudge: 0.5f, s: "bd"
  (16/35>17/35)|nudge: 0.5f, s: "bd"
  (17/35>18/35)|nudge: 0.17697178051416398f, s: "bd"
  (18/35>19/35)|nudge: 0.4679114687283475f, s: "bd"
  (19/35>⁴₇)|nudge: 0.47189795787381655f, s: "bd"
  (⁴₇>⅗)|nudge: 0.5f, s: "bd"
  (⅗>⅘)|nudge: 0.5f, s: "cp"
  (⅘>1)|nudge: 0.5f, s: "sn"

Do you think this can be merged now or do you have a testing method in mind for this? I calculated the mean outside Tidal and got values very close to 0.5f.

[yaxu]

Looks good! I'm not sure exactly what's going on, but it's probably a good idea to change rotL 1 to more like rotL 1000 to sample from a separate part of the random stream..

[yaxu]

You should have access now to merge when ready