Comments by "Lawrence D’Oliveiro" (@lawrencedoliveiro9104) on "Sound By Numbers: The Rise of Digital Sound" video.

  1. Such a filter on the output side is called the “reconstruction filter”, while on the input side it is called the “antialiasing filter”. In both cases, their job is to get rid of spurious frequencies above the Nyquist limit. The reconstruction filter is a “usually-good-enough” approximation to the ideal sinc reconstruction function.
    8
  2. Digital is robust against errors up to a point, then it completely falls apart. Analog suffers from the slightest error, but the degradation is gradual all the way down from there, rather than falling off a cliff. You pays yer money and you takes yer choice.
    7
  3. The ideal reconstruction involves plotting something called a “sinc function” centred at each sample, then adding up all these curves. The trouble with this ideal is that the sinc function has an infinite extent and only slowly fades to zero as you get further from the centre.
    4
  4. 0:32 Actually, both are transducers. And any transducer can work to some degree in either direction -- the two physical systems that are coupled (air versus wood/bone/electricity/whatever) can exchange energy either way.
    1
  5. Is electricity not “physical”?
    1
  6. But “semantics” means “meaning”. So if you want to ignore the meaning, then why are you saying anything at all?
    1
  7. So what does your supposed “audience” classify electricity as, if not “physical”, then? Is it something “spiritual”, perhaps? I’m trying to think non-semantically here, as you suggest.
    1
  8. Because the binary digits of the number making up each sample are recorded and transmitted as signal pulses. Electrically, these can be voltage or current pulses. On magnetic media, they are pulses in the magnetic field. On optical media, the pattern of “pits” and “lands” induces pulses in the photodetection system. And so on.
    1
  9. 09:22 More likely 65,535. This is because they are signed twos-complement integers. Since the largest value is +32767, it makes sense to omit the smallest value -32768 and only go down to -32767 for symmetry. Otherwise your signal could end up with a DC component. Which, as far as sound is concerned, is not helpful.
    1