The sampling frequency is basically the number of times per second audio is sampled and stored as a number - CD audio is sampled at 44.1 KHz, which means 44,100 samples per second. CD audio uses 16 bit samples, so the bitrate of uncompressed CD audio = 44,100 x 16 bits per second (well x 2 actually, because it's stereo).
The "bitrate" on the other hand, when talking about MP3 files, refers to the transfer bitrate for which the files are encoded - i.e. an MP3 file encoded "at a bitrate of 128 kbps" is compressed such that it could be streamed continuously through a link providing a transfer rate of 128 thousand bits per second but most of us don't really use MP3 as a streaming medium (except for Shoutcast, etc.) so really what the MP3 "bitrate" is a measure of is how severely the files is being compressed. The lower the bitrate, the more the file has been compressed and the more you compress a file, the more of the original data is lost, and so the worse the playback sound quality will be. It's almost exactly analogous to compressing a JPG image with a higher compression ratio - you get a smaller file, but when you view it, it doesn't look as good.
PLEASE NOTE: The feedback comments below are considerably beyond the knowledge of those editing/revising FAQs but are left for the information they contain. However they would be far better in one of our forums where their exposure would be greater and others could contribute or learn from them.
NOTE2: The following information was left in a comment, but was removed due to language and attitude. Cleaned up and edited, here it is:
No, the author does not mean bit "depth", which is what you're describing. He means said bit rate and means bit rate, (one is time-based, one is, well depth-based, so it is a one-dimensional value with an arbitrary unit). And I'm sorry, but what you're saying about reverb is [incorrect] ... Most all reverb is a DELAY-based effect, except when you're talking about convolution reverb, and regardless, in either case it has [nothing] to do with oversampling because oversampling techniques are only ever employed to improve a signal's quality (because that's what they do in audio). And sampling, which concerns quantization in the TIME axis of the wave (i.e. "x") .... Again, reverb is delay-based, and if any errors exist in the audible spectrum, the reverb will process them and you will hear it. Changing the bit depth, which again is the quantization in the y-axis (so basically the resolution or number of bits or or range the amplitude of the wave can occupy (16 bit, there are 2^16 values, so 65,536 values, 24 bit, 2^24 values and so on, same as how a sampling rate of 44.1 kHz allows a resolution of 44,100 in one second in the x-axis of the wave)) will not change [anything] in the time domain.And saying those other effect's quality depends on the bit depth is incorrect, as is your comment about bitrate, which you describe incorrectly. The author was right in his definition.
Feedback received on this FAQ entry:
- one more thing please...I was thinking that the sample freq is what stores the time series data and the bitrate used for each sample controls how much dynamic contrast is available...which is what I think you've explained...
So, for the case of a cheap digital reverb, wouldn't the sample freq of the reverb affect the quality, versus the bit depth...? Suppose I get artificial harmonics from decimating a signal, that doesn't have to do with the bit depth...does it ? Can a reverb use a simple filter to block the new aliasing noise...?
- A thoughtful reply to a good topic....I want to ask a clarifying question about interpolation. I have used interpolation in my work, to add points to data, so that I may artificially gain a smoother curve, which enables me to use some tricks, like the first derivative test for peaks.
I was asking myself, does a play back device interpolate lower freq samples rates, and I thought to myself, "NO, because that would be more work for the processor and for what goal....why not just have more data points and play those back...?"
I assume that any interpolation used would be a spline interp....isn't that too computationally expensive...?
I'm a novice however with an undergrad physics background not a MS in EE.
- With "bitrate" he probably meant the amount of bits that audio data is encoded on the audio CD. Without knowing enough details about it the audio bitrate on an audio CD determines how much artificial noise will be generated due to (bad) interpolation.
Generally audio encoded at 16bit 44,1kHz is good enough to be compared to CD audio. Audio encoded in this format has about 96dB SPL, meaning on a perfect recording, you'll be able to record a pin or needle falling on the floor, while at the same time you can record a 4 cylinder engine with hood opened revving up.
Audio encoded at 24bit 44,1kHz is similar to audio DVD's, cinema audio, and studio audio. Audio encoded in this format, has a dynamic range of about 144dB SPL.
In a recording like this you'll be able to record a needle falling to the floor, as well as standing 200 feet from an F16 jet airplane landing track, and recording a jet take off or land. Music recorded at 24 bit is the highest dynamic range of volume that a human could ever perceive.
Music recorded at 32 bit is a much higher quality, often for use in the studio. Music and audio recorded at this bitrate are being recorded for processing the audio waves.
One of the most intensive of simple effects is a reverb. If a reverb is 16 bit, while regular audio encoded at 16 bit seems kind of flawless to us, a reverb is actually not only reverbing the 'nearly' perfect audio, but also the oversampling (the artifacts, the errors). A 16 bit reverb sounds cheap. A 24 reverb can still be perceived as a cheap reverb. A 32 bit reverb is the only reverb where the sound is interpolated so well, that audio artifacts are not audibly amplified (or repeated); thus the only reverb that sounds convincing is 32 bit.
Reverb is one of the few effects, together with a phaser (and perhaps a chorus), as well as pitch up/down, where choosing a higher bitrate makes a lot of sense.
Concerning MP3 bitrate, yes, this is the compression magnification in which audio is compressed. A lower bitrate results in a smaller filesize, with loss of quality from the input signal. Some codecs like WMA (in lower bitrates <48kbps), OGG(higher bitrates >88kbps), and AAC (across the whole spectrum), and flac (lossless, but usually comparable to >320kbps), can record higher quality data per same filesize, or, record same (or better) quality for a (way) lower bitrate setting.
Those are considered better encoders than the age old MP3 format, which is over 15 years old and should be put to rest.
by snapcase$ edited by KeysCapt
last modified: 2015-02-16 14:44:46