To me itās much more unclear how sound is first encoded into a digital signal, transmitted as a digital signal through wires and radio waves, and then translated back into sound in a phone. I mean itās essentially the same physics as the analog electronics, just with a bunch of extra steps added.
Yeah, this is where sample rate and bit depth come into play. In case youāre curious, digital audio is possible due to the Nyquist-Shannon Sampling Theorem. The TL;DR is that you donāt record a continuous stream of audio data; You just sample the wave at regular intervals by recording the current amplitude. And then you can recreate it on the other end. The theorem states that an analog wave can be perfectly recorded and replicated, as long as you have a sufficiently high sample rate and bit depth. Since human hearing generally tops out at 20kHz, we need to sample the audio signal at least 40k times per second; Most consumer-grade audio equipment uses 44.1 or 48kHz. Phones actually use a much lower sample rate for calls, but more on that later.
Again, as long as your sample rate is at least 2x the rate of the highest frequency being recorded, youāre able to perfectly recreate the wave. For an example, hereās a gif:
The image on the left shows the wave being recorded, and the dots are samples. As you add more samples, the reproduced wave gets more accurate. By the time you have 2x the fastest frequency, there is only one possible wave that will fit every sample. Again, human hearing tops out around 20kHz, so we use a sample rate just above 40kHz.
Phone calls will often put a filter on the high and low ends, and only capture the mid-range. It gives that distinct āthis is shitty phone call qualityā sound, but means they can use a much lower sample rate; Since theyāre lopping off most of the high end with that filter, they may only need a sample rate closer to 15 or 20kHz. Because fewer samples means less data. The intelligibility happens in the mid-range, so thatās what the phone makers (and telecom companies) focus on. This low sample rate is also why hold music sounds so fucking awful. Itās essentially being passed through a āmake this sound as shitty as possible while still being intelligibleā filter.
And then bit depth simply determines how detailed each sample is. If you use 8 bits per sample, that gives you 256 potential values per sample. 12 bits gives you 4096. The trade-off is that a higher bit depth means each sample takes exponentially more data; Audiophiles will generally push for higher bit depths, so each sample is more accurate. In contrast, phone calls often use lower bit depths, (again, to save data).
As for how it actually transmits the data, thatās just 1ās and 0ās. Itās a little more complicated than that, (packets, for example) but in the digital realm, as long as the 1ās and 0ās get to where they need to be, youāre good to go.
Holy shit, so Iām not just uniquely terrible at understanding people on the phone? Iāve searched so long for a phone that does high-quality phone calls, and I canāt believe I never figured that it was a problem with both the phones and the carriers.
Yup itās also why hold music sounds terrible is the sample rate and ranges are so small thereās basically no music which would sound decent over the connection
Yeah, this is where sample rate and bit depth come into play. In case youāre curious, digital audio is possible due to the Nyquist-Shannon Sampling Theorem. The TL;DR is that you donāt record a continuous stream of audio data; You just sample the wave at regular intervals by recording the current amplitude. And then you can recreate it on the other end. The theorem states that an analog wave can be perfectly recorded and replicated, as long as you have a sufficiently high sample rate and bit depth. Since human hearing generally tops out at 20kHz, we need to sample the audio signal at least 40k times per second; Most consumer-grade audio equipment uses 44.1 or 48kHz. Phones actually use a much lower sample rate for calls, but more on that later.
Again, as long as your sample rate is at least 2x the rate of the highest frequency being recorded, youāre able to perfectly recreate the wave. For an example, hereās a gif:
The image on the left shows the wave being recorded, and the dots are samples. As you add more samples, the reproduced wave gets more accurate. By the time you have 2x the fastest frequency, there is only one possible wave that will fit every sample. Again, human hearing tops out around 20kHz, so we use a sample rate just above 40kHz.
Phone calls will often put a filter on the high and low ends, and only capture the mid-range. It gives that distinct āthis is shitty phone call qualityā sound, but means they can use a much lower sample rate; Since theyāre lopping off most of the high end with that filter, they may only need a sample rate closer to 15 or 20kHz. Because fewer samples means less data. The intelligibility happens in the mid-range, so thatās what the phone makers (and telecom companies) focus on. This low sample rate is also why hold music sounds so fucking awful. Itās essentially being passed through a āmake this sound as shitty as possible while still being intelligibleā filter.
And then bit depth simply determines how detailed each sample is. If you use 8 bits per sample, that gives you 256 potential values per sample. 12 bits gives you 4096. The trade-off is that a higher bit depth means each sample takes exponentially more data; Audiophiles will generally push for higher bit depths, so each sample is more accurate. In contrast, phone calls often use lower bit depths, (again, to save data).
As for how it actually transmits the data, thatās just 1ās and 0ās. Itās a little more complicated than that, (packets, for example) but in the digital realm, as long as the 1ās and 0ās get to where they need to be, youāre good to go.
Holy shit, so Iām not just uniquely terrible at understanding people on the phone? Iāve searched so long for a phone that does high-quality phone calls, and I canāt believe I never figured that it was a problem with both the phones and the carriers.
Yup itās also why hold music sounds terrible is the sample rate and ranges are so small thereās basically no music which would sound decent over the connection
Iāve heard that the pikachu cry in pokemon yellow has a bit depth of 1.