Skip to content

Commit 2dade58

Browse files
ulul
committed
init
0 parents  commit 2dade58

27 files changed

+5377
-0
lines changed

‎.gitignore

+2
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,2 @@
1+
target/
2+
*.dylib

‎README.md

+84
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,84 @@
1+
# LiveCore
2+
3+
A hardcore livecoding system for realtime audio synth in [the spirit of Sound Garden](http://ul.mantike.pro/sound-garden-manifesto.html).
4+
WIP with
5+
[Work with the garage door up](https://notes.andymatuschak.org/Work_with_the_garage_door_up) ethos.
6+
7+
## Influences
8+
9+
- [clive](https://mathr.co.uk/clive/)
10+
- [Extempore](https://extemporelang.github.io)
11+
- Sound Garden [1](https://github.com/ul/sound-garden) & [2](https://github.com/ul/sound-garden-0x2)
12+
- [Ad Libitum](https://github.com/ul/ad-libitum)
13+
14+
## Dependencies
15+
16+
- [Nim](https://nim-lang.org) 1.4.2
17+
- [libsoundio](http://libsound.io) 2.0.0
18+
- [fswatch](http://emcrisostomo.github.io/fswatch/) 1.15.0
19+
- [Soundpipe](https://pbat.ch/proj/soundpipe.html) 233c9646
20+
- [libsndfile](http://www.mega-nerd.com/libsndfile/) 1.0.28
21+
22+
On macOS:
23+
24+
```
25+
$ brew install libsoundio fswatch libsndfile
26+
or
27+
$ port install libsoundio fswatch libsndfile
28+
```
29+
30+
For Nim try [choosenim](https://github.com/dom96/choosenim#choosenim),
31+
and it's easy to install Soundpipe from the sources.
32+
33+
## Configuration
34+
35+
Sample rate and channels count is hardcoded to `48000` and `2` respectively.
36+
If your device requires different values, edit `src/dsp/frame.nim`.
37+
38+
## Session workflow
39+
40+
NB: you need to run all the mentioned scripts from the repo root.
41+
42+
- Checkout starting point. For a fresh session the `main` branch is a good choice.
43+
- Make sure that the git tree is clean, as this script will be committing your
44+
changes as you make them.
45+
- Start server with `./start-server`
46+
- Run `./start-session` with a session name as an argument.
47+
It must be valid as a part of git branch name as the script will prefix it with
48+
`session/` and checkout this branch.
49+
- If the session with such name exists it will be resumed.
50+
- Make changes in `src/session.nim` and save file to compile and send to the server.
51+
- After every successful compilation there will be a commit.
52+
- Ctrl-C this script to stop and switch back to the starting point.
53+
- The system is ready for the next session!
54+
55+
## Render
56+
57+
To render a specific session duration in a non-interactive mode to a file:
58+
59+
```
60+
$ ./render <duration in seconds> <path/to/output.wav>
61+
```
62+
63+
## FAQ
64+
65+
### Why Nim?
66+
67+
C-like freedom, performance and fast compilation with heaps of syntactic sugar.
68+
clive very much aligns with the vision I wanted to implement and I'd just port
69+
clive to macOS if only C was terser. When I'm jamming I want the code to be
70+
clean, concise and close to my intention, all other necessary trade-offs
71+
considered.
72+
73+
### Does it run on Windows/Linux?
74+
75+
Maybe. Please try and let me know! I strive to write the code in a
76+
platform-independent way but I test it only on macOS.
77+
78+
### Why snake_case procs and vars?
79+
80+
Purely irrational, æsthetic choice. CamelCase seems to be prevalent in the Nim
81+
ecosystem, and majority of the code I write/read in other languages is camelCase
82+
too. However, I enjoy the most writing/reading OCaml and Rust in the regard of
83+
that particular convention, and I'm glad that Nim's compiler doesn't actually
84+
care, so I went with underscores.

‎doc/Audio-EQ-Cookbook.txt

+280
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,280 @@
1+
2+
Cookbook formulae for audio EQ biquad filter coefficients
3+
----------------------------------------------------------------------------
4+
by Robert Bristow-Johnson <[email protected]>
5+
6+
7+
All filter transfer functions were derived from analog prototypes (that
8+
are shown below for each EQ filter type) and had been digitized using the
9+
Bilinear Transform. BLT frequency warping has been taken into account for
10+
both significant frequency relocation (this is the normal "prewarping" that
11+
is necessary when using the BLT) and for bandwidth readjustment (since the
12+
bandwidth is compressed when mapped from analog to digital using the BLT).
13+
14+
First, given a biquad transfer function defined as:
15+
16+
b0 + b1*z^-1 + b2*z^-2
17+
H(z) = ------------------------ (Eq 1)
18+
a0 + a1*z^-1 + a2*z^-2
19+
20+
This shows 6 coefficients instead of 5 so, depending on your architechture,
21+
you will likely normalize a0 to be 1 and perhaps also b0 to 1 (and collect
22+
that into an overall gain coefficient). Then your transfer function would
23+
look like:
24+
25+
(b0/a0) + (b1/a0)*z^-1 + (b2/a0)*z^-2
26+
H(z) = --------------------------------------- (Eq 2)
27+
1 + (a1/a0)*z^-1 + (a2/a0)*z^-2
28+
29+
or
30+
31+
1 + (b1/b0)*z^-1 + (b2/b0)*z^-2
32+
H(z) = (b0/a0) * --------------------------------- (Eq 3)
33+
1 + (a1/a0)*z^-1 + (a2/a0)*z^-2
34+
35+
36+
The most straight forward implementation would be the "Direct Form 1"
37+
(Eq 2):
38+
39+
y[n] = (b0/a0)*x[n] + (b1/a0)*x[n-1] + (b2/a0)*x[n-2]
40+
- (a1/a0)*y[n-1] - (a2/a0)*y[n-2] (Eq 4)
41+
42+
This is probably both the best and the easiest method to implement in the
43+
56K and other fixed-point or floating-point architechtures with a double
44+
wide accumulator.
45+
46+
47+
48+
Begin with these user defined parameters:
49+
50+
Fs (the sampling frequency)
51+
52+
f0 ("wherever it's happenin', man." Center Frequency or
53+
Corner Frequency, or shelf midpoint frequency, depending
54+
on which filter type. The "significant frequency".)
55+
56+
dBgain (used only for peaking and shelving filters)
57+
58+
Q (the EE kind of definition, except for peakingEQ in which A*Q is
59+
the classic EE Q. That adjustment in definition was made so that
60+
a boost of N dB followed by a cut of N dB for identical Q and
61+
f0/Fs results in a precisely flat unity gain filter or "wire".)
62+
63+
_or_ BW, the bandwidth in octaves (between -3 dB frequencies for BPF
64+
and notch or between midpoint (dBgain/2) gain frequencies for
65+
peaking EQ)
66+
67+
_or_ S, a "shelf slope" parameter (for shelving EQ only). When S = 1,
68+
the shelf slope is as steep as it can be and remain monotonically
69+
increasing or decreasing gain with frequency. The shelf slope, in
70+
dB/octave, remains proportional to S for all other values for a
71+
fixed f0/Fs and dBgain.
72+
73+
74+
75+
Then compute a few intermediate variables:
76+
77+
A = sqrt( 10^(dBgain/20) )
78+
= 10^(dBgain/40) (for peaking and shelving EQ filters only)
79+
80+
w0 = 2*pi*f0/Fs
81+
82+
cos(w0)
83+
sin(w0)
84+
85+
alpha = sin(w0)/(2*Q) (case: Q)
86+
= sin(w0)*sinh( ln(2)/2 * BW * w0/sin(w0) ) (case: BW)
87+
= sin(w0)/2 * sqrt( (A + 1/A)*(1/S - 1) + 2 ) (case: S)
88+
89+
FYI: The relationship between bandwidth and Q is
90+
1/Q = 2*sinh(ln(2)/2*BW*w0/sin(w0)) (digital filter w BLT)
91+
or 1/Q = 2*sinh(ln(2)/2*BW) (analog filter prototype)
92+
93+
The relationship between shelf slope and Q is
94+
1/Q = sqrt((A + 1/A)*(1/S - 1) + 2)
95+
96+
2*sqrt(A)*alpha = sin(w0) * sqrt( (A^2 + 1)*(1/S - 1) + 2*A )
97+
is a handy intermediate variable for shelving EQ filters.
98+
99+
100+
Finally, compute the coefficients for whichever filter type you want:
101+
(The analog prototypes, H(s), are shown for each filter
102+
type for normalized frequency.)
103+
104+
105+
LPF: H(s) = 1 / (s^2 + s/Q + 1)
106+
107+
b0 = (1 - cos(w0))/2
108+
b1 = 1 - cos(w0)
109+
b2 = (1 - cos(w0))/2
110+
a0 = 1 + alpha
111+
a1 = -2*cos(w0)
112+
a2 = 1 - alpha
113+
114+
115+
116+
HPF: H(s) = s^2 / (s^2 + s/Q + 1)
117+
118+
b0 = (1 + cos(w0))/2
119+
b1 = -(1 + cos(w0))
120+
b2 = (1 + cos(w0))/2
121+
a0 = 1 + alpha
122+
a1 = -2*cos(w0)
123+
a2 = 1 - alpha
124+
125+
126+
127+
BPF: H(s) = s / (s^2 + s/Q + 1) (constant skirt gain, peak gain = Q)
128+
129+
b0 = sin(w0)/2 = Q*alpha
130+
b1 = 0
131+
b2 = -sin(w0)/2 = -Q*alpha
132+
a0 = 1 + alpha
133+
a1 = -2*cos(w0)
134+
a2 = 1 - alpha
135+
136+
137+
BPF: H(s) = (s/Q) / (s^2 + s/Q + 1) (constant 0 dB peak gain)
138+
139+
b0 = alpha
140+
b1 = 0
141+
b2 = -alpha
142+
a0 = 1 + alpha
143+
a1 = -2*cos(w0)
144+
a2 = 1 - alpha
145+
146+
147+
148+
notch: H(s) = (s^2 + 1) / (s^2 + s/Q + 1)
149+
150+
b0 = 1
151+
b1 = -2*cos(w0)
152+
b2 = 1
153+
a0 = 1 + alpha
154+
a1 = -2*cos(w0)
155+
a2 = 1 - alpha
156+
157+
158+
159+
APF: H(s) = (s^2 - s/Q + 1) / (s^2 + s/Q + 1)
160+
161+
b0 = 1 - alpha
162+
b1 = -2*cos(w0)
163+
b2 = 1 + alpha
164+
a0 = 1 + alpha
165+
a1 = -2*cos(w0)
166+
a2 = 1 - alpha
167+
168+
169+
170+
peakingEQ: H(s) = (s^2 + s*(A/Q) + 1) / (s^2 + s/(A*Q) + 1)
171+
172+
b0 = 1 + alpha*A
173+
b1 = -2*cos(w0)
174+
b2 = 1 - alpha*A
175+
a0 = 1 + alpha/A
176+
a1 = -2*cos(w0)
177+
a2 = 1 - alpha/A
178+
179+
180+
181+
lowShelf: H(s) = A * (s^2 + (sqrt(A)/Q)*s + A)/(A*s^2 + (sqrt(A)/Q)*s + 1)
182+
183+
b0 = A*( (A+1) - (A-1)*cos(w0) + 2*sqrt(A)*alpha )
184+
b1 = 2*A*( (A-1) - (A+1)*cos(w0) )
185+
b2 = A*( (A+1) - (A-1)*cos(w0) - 2*sqrt(A)*alpha )
186+
a0 = (A+1) + (A-1)*cos(w0) + 2*sqrt(A)*alpha
187+
a1 = -2*( (A-1) + (A+1)*cos(w0) )
188+
a2 = (A+1) + (A-1)*cos(w0) - 2*sqrt(A)*alpha
189+
190+
191+
192+
highShelf: H(s) = A * (A*s^2 + (sqrt(A)/Q)*s + 1)/(s^2 + (sqrt(A)/Q)*s + A)
193+
194+
b0 = A*( (A+1) + (A-1)*cos(w0) + 2*sqrt(A)*alpha )
195+
b1 = -2*A*( (A-1) + (A+1)*cos(w0) )
196+
b2 = A*( (A+1) + (A-1)*cos(w0) - 2*sqrt(A)*alpha )
197+
a0 = (A+1) - (A-1)*cos(w0) + 2*sqrt(A)*alpha
198+
a1 = 2*( (A-1) - (A+1)*cos(w0) )
199+
a2 = (A+1) - (A-1)*cos(w0) - 2*sqrt(A)*alpha
200+
201+
202+
203+
204+
205+
FYI: The bilinear transform (with compensation for frequency warping)
206+
substitutes:
207+
208+
1 1 - z^-1
209+
(normalized) s <-- ----------- * ----------
210+
tan(w0/2) 1 + z^-1
211+
212+
and makes use of these trig identities:
213+
214+
sin(w0) 1 - cos(w0)
215+
tan(w0/2) = ------------- (tan(w0/2))^2 = -------------
216+
1 + cos(w0) 1 + cos(w0)
217+
218+
219+
resulting in these substitutions:
220+
221+
222+
1 + cos(w0) 1 + 2*z^-1 + z^-2
223+
1 <-- ------------- * -------------------
224+
1 + cos(w0) 1 + 2*z^-1 + z^-2
225+
226+
227+
1 + cos(w0) 1 - z^-1
228+
s <-- ------------- * ----------
229+
sin(w0) 1 + z^-1
230+
231+
1 + cos(w0) 1 - z^-2
232+
= ------------- * -------------------
233+
sin(w0) 1 + 2*z^-1 + z^-2
234+
235+
236+
1 + cos(w0) 1 - 2*z^-1 + z^-2
237+
s^2 <-- ------------- * -------------------
238+
1 - cos(w0) 1 + 2*z^-1 + z^-2
239+
240+
241+
The factor:
242+
243+
1 + cos(w0)
244+
-------------------
245+
1 + 2*z^-1 + z^-2
246+
247+
is common to all terms in both numerator and denominator, can be factored
248+
out, and thus be left out in the substitutions above resulting in:
249+
250+
251+
1 + 2*z^-1 + z^-2
252+
1 <-- -------------------
253+
1 + cos(w0)
254+
255+
256+
1 - z^-2
257+
s <-- -------------------
258+
sin(w0)
259+
260+
261+
1 - 2*z^-1 + z^-2
262+
s^2 <-- -------------------
263+
1 - cos(w0)
264+
265+
266+
In addition, all terms, numerator and denominator, can be multiplied by a
267+
common (sin(w0))^2 factor, finally resulting in these substitutions:
268+
269+
270+
1 <-- (1 + 2*z^-1 + z^-2) * (1 - cos(w0))
271+
272+
s <-- (1 - z^-2) * sin(w0)
273+
274+
s^2 <-- (1 - 2*z^-1 + z^-2) * (1 + cos(w0))
275+
276+
1 + s^2 <-- 2 * (1 - 2*cos(w0)*z^-1 + z^-2)
277+
278+
279+
The biquad coefficient formulae above come out after a little
280+
simplification.

‎render

+8
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,8 @@
1+
#!/bin/sh
2+
3+
# ./render <duration in seconds> <path/to/output.wav>
4+
5+
# To render as fast as we can.
6+
NIM_FLAGS="--gc:arc -d:danger --passC:-ffast-math"
7+
8+
nim c $NIM_FLAGS -r -o:target/render src/render.nim $@

0 commit comments

Comments
 (0)