Sample Rates

From Audacity Wiki
Revision as of 06:22, 10 October 2007 by NT (talk | contribs) (Category:Digital Audio Technology)
Jump to: navigation, search

Bandwidth

Recording sample rate sets the maximum audio frequency that can be recorded.


Sample Rate vs Frequency limit

* 8 k           3.6 kHz
* 11 k          5 kHz
* 22 k          10 kHz
* 32 k          14.5 kHz
* 44.1 k        20 kHz
* 48 k          21.8 kHz
* 64 k          29.1 kHz
* 88.2 k        40 kHz
* 96 k          43.6 kHz


Which Sample Rate To Use

44.1k

20kHz is the full human audible range, so 44.1k sample rate is the logical choice for most audio material, giving the full 20 kHz bandwidth.

44.1kHz is the sample rate of the original .cda CD audio format.


Reduced Bandwidth Recording

Audio may be recorded at below 20kHz bandwidth for any of a few reasons:

  • To reduce file size
  • To reduce CPU use
  • Because the source material is of less than 20kHz bandwidth.
  • To reduce hiss (but see notes on this further down)


32k, 14.5 kHz

  • 14.5kHz is above the frequency limit of many medium quality sources, such as many cassette recordings, etc.

14.5kHz cut-off also has a slight effect on speech, so 32k will give good quality speech recording, though not the best.

14.5kHz is a little below FM radio bandwidth, so FM radio can be recorded with only a little quality loss.

32k sample rate is thus good for:

  • Some cassette recordings
  • Speech
  • All other audio where smaller files than 44.1k are desired with only slight compromise on sound quality.

22k, 10 kHz

22k has been a reasonably popular sample rate for low bit rate mp3s (eg 64k) in years past. Audio quality is significantly affected, with higher frequency content missing. With the general rise in the availability of large file storage space and faster data links, 22k is now of more limited use.

  • For speech recording where perceived quality is unimportant, but clarity must be maintained.
  • AM radio
  • To squeeze mp3 music onto floppy disc & very small mp3 players.


11k, 5 kHz

11k sample rate gives 5kHz max recorded frequency. Very poor sound quality.


8k, 3.6kHz

3.6kHz matches the bandwidth of telephone sytems. Microcassette dictaphones have also used a similar bandwidth. 8k sample rate is thus suitable for:

  • Telephone speech
  • Microcassette recording
  • To squash long speeches into small file sizes

Sound quality is terrible, with speech as clear as a mud bath. Any further drop in sample rate makes intelligibility a challenge.


Ultrasonic Recording

Recording and playing with a bandwidth higher than 20kHz gives no audible benefit (assuming replay is not speed altered). It can be used:

  • to maintain full bandwidth when the recording will be slowed down afterwards
  • for analysis of signals above 20 kHz
  • for some oscilloscope waveform recording
  • to record sig gen waveforms up to 40 kHz
  • for recording some wildlife
  • and other ultrasonic applications.

When recording ultrasonic input it should be borne in mind that not all audio sources can output signals above 20 kHz. To capture ultrasound, all items in the chain of audio must support the full frequency range to be recorded.

Some sound cards support sampling at up to 96kHz, some don't.


Custom Rates

When recording a signal of known limited bandwidth, you can minimise file size by using a recording sample rate of 2.2x signal bandwidth. This is seldom necessary.


Notes

Bandwidth

Bandwidth is the frequency difference between the highest and lowest signals in an audio stream.

For audio work, bandwidth is normally about 20Hz less than the highest recorded frequency, so for practical purposes they can be treated as the same thing. Hence the terms are used interchangeably here.

The term bandwidth may be applied to the frequency content of an audio signal stream, or the frequency ability of audio hardware or software.

Bandwidth of Human Hearing

Human hearing in 18 year olds extends to around 20kHz. As age progresses this upper limit falls. A percentage of people can not hear above 14.5kHz, and will gain no benefit in recording at above 32k sample rate. This is primarily true for older people.

Hiss Reduction

Reducing sample rate can be used to remove the highest frequency hiss present in a noisy signal. In practice though one often does not know exactly what the signal bandwidth is, and thus a better option is to use noise gating, which has much more effect and is less likely to compromise the recorded signal. The Audacity 'noise removal' filter is a multi-channel noise gate.

Adjusting sample rate to reduce hiss should only be considered where every drop of filesize reduction is needed. The good news is that as long as the bandwidth of the sample rate stays above the audio signal bandwidth, there is no loss of quality. Thus this approach can sometimes be used to squeeze a little more run time onto a CD etc without affecting sound quality.


See Also

Bit Depth