input leaks past filter, when synth is replaced only

[telephon]

// problem in attack phase: input leaks past filter

t = Steno.push(2);

t.filter(\f, { |in| (100, 200..1000).sum { |freq| BPF.ar(in, freq, 0.01) } });
t.quelle(\n, { WhiteNoise.ar(0.4) });

--!nf // evaluate this line several times

// each time, the noise comes through

t.fadeTime = 1;

--!nf // evaluate this line several times …

[telephon]

@LFSaw do you have an idea where we have introduced this problem? I'll try and fix it, but maybe you have a hint.

[telephon]

There is a problem around here: https://github.com/musikinformatik/Steno/blob/master/Classes/StenoSignal.sc#L108

[LFSaw]

minor revelation it is not an order of execution (caused by the force-reload of everything):

s.boot;
t = Steno.push(2);

t.filter(\f, { |in| (100, 200..1000).sum { |freq| BPF.ar(in, freq, 0.01) } });
t.filter(\g, { |in| (100, 200..1000).sum { |freq| BPF.ar(in, freq, 0.01) } });
t.quelle(\n, { WhiteNoise.ar(0.4) });

t.settings.globalSettings // nothing set, also no proto or parent

--nf // evaluate alternate between these two lines
--ng

// each time, the noise comes through, still.

If I change the filter definition to something simple, the burst gets shorter

t.filter(\f, { |in| 0.5 * in });
t.filter(\g, { |in| 0.2 * in });
t.quelle(\n, { WhiteNoise.ar(0.4) });

--!ng

[telephon]

probably the silence is detected earlier and the synth freed earlier then …

[LFSaw]

why is the freeselfwhensilent happening before the mix? https://github.com/musikinformatik/Steno/blob/1288832a78249503a5370ffd5e3ca40dafde7b34/Classes/StenoSignal.sc#L96

[LFSaw]

for the sake of ease of understanding, I'd appreciate tackling naming issues first.

Specifically, I'd like to get a definition of what is associated with variables

inputSignal // input to filter (first argument in user-settable function)
oldSignal // previous signal on bus (not necessarily the same as inputSignal)
signal // signal routed to outBus
tailSignal // signal routed to tailBus

is this correct for all?

[telephon]

some corrections:

1. inputSignal – input to filter (first argument in user-settable function) (maps to in arg).
2. oldSignal – previous signal on bus (not necessarily the same as inputSignal, and not necessarily the filter input) (maps to out arg).
3. signal – signal routed to outBus (currently, no. currently, this is also routed partly to tailBus, in the method writeToBus)
4. tailSignal – signal read from tailBus (tailBus arg).

[telephon]

why is the freeselfwhensilent happening before the mix?

because after the mix of the previous bus signal (oldSignal), it might never become silent. Probably we need to combine the mix parameter with the envelope. Because otherwise, we add the oldSignal to the mix twice (but this is unrelated to the current issue).

[telephon]

btw. I've just moved the routing function from Steno to StenoSignal, for clarity.

[telephon]

This is now almost fixed …

just adding/removing parentheses leaks a little:

t = Steno.push(2);
t.filter(\f, { |in, e| BPF.ar(in, 600 * (e[\index] + 1), 0.01)  * 8 });
t.quelle(\n, { WhiteNoise.ar(0.4) });
--n[f f f]
--nf