News for hackers. Stuff that matters.
Re:The Roland TR-808: the drum machine that revolutionised m... by crb3 on Tuesday May 12, 2015 @09:19PM (#9536133)
> Surely the reproduction of an ancient hardware drum sequencer in today's software should be perfect.

CD-quality is 44.1KHz sample-rate 16-bit stereo. That's 22us per sample. That's a lot of perceptible quantization error when, it turns out, our pitch, timbre and spacial-placement sensory gear is sensitive down to the microsecond level (google "microsecond phase sensitivity" on that). A drum synthesizer done on a PC is limited to the soundcard's samplerate at output, so it's always going to be perceptibly less real than continuous analog.

Re:The Roland TR-808: the drum machine that revolutionised m... by nitrogen on Wednesday May 13, 2015 @03:54AM (#9537412)
The phase information contained in a digitally sampled signal is not limited by the sample rate[2]. It's not (just[0]) the audio sample rate that makes simulations inaccurate, but inadequate detail in the simulation.

A digital recording of an analog drum machine will sound indistinguishable from the original, but obviously lacks the flexibility of changing settings on the original machine.

[0] An analog circuit simulation is basically a numeric integration and differentiation engine. Having a coarse simulation timestep, especially with a simple integration method, can affect the accuracy of a simulation. See [1] for information about how this affects video game physics.

[1] http://gafferongames.com/game-physics/integration-basics/

[2] Proof by example (not written by me): "Jack_delay -- This is a small command line JACK app you can use to measure the latency of your sound card. It uses a phase measurements on a set of tones to measure the delay from the output to the input. Accuracy is about 1/1000 of a sample." -- http://kokkinizita.linuxaudio.org/linuxaudio/