What is bits per sample?

[comrinec]

I've seen bits per sample and know that cds are 16 bit, hd audio 24 bit. But how does the bits per sample relate to quality?

[Guest]

More bits is better but the file size is larger. So there's a tradeoff.

[comrinec]

But what actually is it in terms of audio? What do the bits measure?

[kypec]

:search: http://en.wikipedia.org/wiki/Audio_bit_depth

[comrinec]

Thanks. I searched on wikipedia and google for it before but couldn't find much info on it.

[pest]

More bits is better but the file size is larger. So there's a tradeoff.

16 bits are enough if you don't want to further process the audio-data ;-)

[ursamtl]

In very simple terms, imagine you are filming a movie. The sampling rate is the numbers of frames per second and the bit depth (also known as bit resolution or word length) is the resolution of the frame.

However, audio is not entirely the same as video. An analog audio wave has to be sampled and converted to digitial format, then it has to be reconstructed as closely as possible to the original. If you can visualize the sound wave drawn on a graph, the sampling rate represents the horizontal axis or time component. The bit depth represents the vertical axis or amplitude. While the sampling rate is linear, the bit depth is exponential. In other words, doubling the sample rate from 48 to 96kHz will provide twice the possible number of values (high frequency limit of 48kHz...most of us can't hear it anyway). Increasing the bit depth from 16 to 24 gives an increase of 256 times the number of possible levels or from 65536 (2^16) to 16777216 (2^24).

However, since no sound can be louder than 0dB in digital audio, all these extra steps are there for sounds that are not the loudest. Each bit is the equivalent of 6dB, so if a sound peak is at -12dB, it's only rendered with 14 bits (on a 16-bit CD) or 22 bits on a hires recording. This isn't bad, but what happens when sounds fade, such as cymbal fades or reverb tails? When the sound reaches -48dB, on a 16-bit CD, the audio is actually only rendered at 8 bits (exactly the same as that old Commodore 64 or an old 80's video game). It may seem like a low level, but believe me, -48dB is very audible.

Anyway, I could provide a lot of examples, but to sum up, the more bits the better, if you have the storage. CD audio is fixed at a bit depth of 16, and that's not bad, providing you use something called dithering to convert down from higher bit rates. All the big audio editing programs operate internally at 32 bits (and some at up to 64 bits), so when you reduce to 16-bit resolution for audio CDs or mp3s, you should always use dithering (another somewhat complex subject). Otherwise, a lot of sound data is simply lost because it's chopped off or "truncated."

Here are a couple of really helpful articles on the subject:

16 bit and 24 bit audio (http://www.tweakheadz.com/16_vs_24_bit_audio.htm)

The 24-bit Field Recording FAQ (http://www.24bitfaq.org/)

Regards,
Steve.