Sample Rates - help on choosing the appropriate sample rate to work with
Bit depth is the number of bits used to carry the data in each sample of audio.
The bit depth chosen for recording limits the dynamic range of the recording. (Other factors in the audio chain may also limit this, so more bits often will not produce a better recording.)
Effects on file size and CPU use
Bit depth (sometimes called also resolution) affects file size. All other things being equal, a 32 bit file is twice the size of a 16 bit file, and an 8 bit file half the size of a 16 bit one.
Bit resolution also affects CPU use. Recording in 32 bit quality, which is Audacity's default behaviour, takes a lot more work than recording 16 bits, and a slower computer may not be able to keep up, with the result being lost samples. If you are recording for immediate export without editing, 32 bit recording may offer no advantage over 16 bit recording if you only have a 16 bit sound device. Most built-in consumer sound devices on computers (and even many cheaper sound cards) are only 16 bit. You can change the default sample format Audacity records at to 24 bit or 16 bit by going to the.
Bit Depth of Various Sources
Figures vary a little for analogue sources, depending on quality of equipment and recording.
Some figures were measured using above average quality domestic hifi equipment (where applicable), and will be typical for professional recordings played on good domestic equipment. Other figures were obtained from audio engineers' written material.
Edison cylinders, microcassettes, picocassettes and hifi are mutually alienating concepts, and typical low fidelity equipment was measured (or will be (eventually (one day))).
Digital recorders don't record half bits of resolution, thus dynamic range can only be chosen in increments of 6dB for digital recording. Where a half bit figure is given, the dynamic range is half way between the 2 digitally recordable values. Analogue can of course record at such intermediate dynamic range values, unlike digital.
Some sources have been measured, some haven't yet. Some are a bit hard to come by. If you have a cylinder player, please measure and post the result.
Companders are often used with cassette and open reel tape to improve the signal to noise ratio (SNR). Compansion is a combination of dynamic range compression and expansion, in this case (and almost always), compression in recording or transmission followed by expansion on playback.
The use of compansion can effectively increase bit depth during low level passages. How much extra bit depth is added at these times depends on the amount of compansion used. Increased bit depth is only achieved if the compansion is used during playback.
- Dolby A adds
- Dolby B can adds nearly 2 bits of extra dynamic range
- Dolby C adds just over 3 bits at low levels
- Dolby SR adds up to 4 bits
- Dolby S adds 2-4 bits
- DNL and DNR can add less than 2 bits of dynamic range at most
- Dbx type I adds...
- Dbx type II adds...
- Dbx type B adds up to nearly 2 bits
- HighCom Telefunken adds.. (btw. - is there any HighCom compander plugin out there? )
- Standard audio CDs (cda): the digital recording format is 16 bit, but a fair bit of the source material is of much lower dynamic range. Also high dynamic range source material is often deliberately compressed, so the resulting real dynamic range is well below 16 bit in many cases. Some CD recordings do have 16 bits of dynamic range, but far from all.
- Other file types on data CDs: Resolution depends on the file format and audio content, and can vary widely.
- Ferric: 6 bit
- Chrome: 6 bit
- Some home recordings have less dynamic range
- Low quality equipment or recordings will generally have less dynamic range.
- A 1970s portable cassette player recording live via microphone, with plenty of hiss, a little motor noise and low sound quality, yielded just 2.5 bit depth for the whole record & play cycle.
- 16 rpm (broadcast record):
- 33 rpm 12" (album): 9 - 10 bit
- 45 rpm 7" (single):
- 45 rpm 12" (12" single): 10 bit
- 33 rpm 7" (EP double single):
- 45 rpm flexible single (paper thin plastic film):
- 78 rpm (shellac):
- 78 rpm aluminium disc (home recordings):
Note that 33 rpm 12" compilation albums that have 10 tracks recorded per side usually have noticeably reduced dynamic range. This is needed to achieve the extended play time.
- Single measurement only: 4.5 bit. Exact bit depth is dependant on signal strength and tuner specs.
MiniDiscs can record in 24 bit format, but the dynamic range will often be less than 24 bits due to the limitations of the source material.
- VHS analogue audio:
- VHS digital audio:
- Umatic analogue:
Until figures come in, analogue recording on VHS & Betamax will give similar to poorer bit depth than cassette.
TV systems vary, and several systems are in use. As a general guide:
- Today's mono broadcast tv sound is FM, and not far removed from FM radio
- Digital stereo sound formats tend to be of significantly better spec
- Historic formats (eg 405 line black & white), used AM sound, and are similar in quality to AM radio.
- Standard optical recording:
- Variable density recording: I have no sample to test, but its much poorer quality than the standard system, and is not often encountered
Telephone is historically in a 3000Hz range from 300Hz to 3300Hz, which covers the human voice, though higher/lower range subtleties of the human voice are lost, which is what makes someone on the phone sound different than they do in person. Most Interactive Voice Response systems demand 8 bit, 8000 Hz sampling to conserve hard drive space, which is more than adequate to deal with this range.
- Typical 1970s player: 5.5 bit depth
Reel to Reel
aka Open Reel:
- 15/16th ips:
- 1 7/8th ips:
- 3.75 ips:
- 7.5 ips:
- 15 ips:
- 30 ips with Dolby A: 11 bit
Used in many dictation machines & answerphones. Unsuitable for music.
Used in few dictation machines
1940s dictation machine technology
Probably the least popular home audio format!
- Prerecorded: comparable to 78 rpm discs
- Home recording: not tested, but probably in the region of 2 bit resolution.
How Many Bits To Use?
If there will be no adjustment of gain after recording and no effects applied, the recording bit depth can be 1 bit more than the audio source bit depth without losing any quality.
If gain will be reduced after recording, recording with 1 bit more depth will avoid degradation.
When Gain Changes
If gain will be increased after recording, recording without degradation will occur if the number of bits equals the source bit depth plus 1 plus another one for each 6dB or part of 6dB of gain that will be added.
Where multiple operations are to be carried out on the recorded signal, each operation can be assessed in terms of how much gain it adds, allowing a margin of one bit for each 6dB or part of 6dB of gain, and another bit for each operation.
Using more bits than these figures will not give you more quality, but will enable more processing of the signal to occur before any degradation of quality occurs due to the number of bits in use.
using less bits than the above will cause some signal degradation. Whether it is noticeable will depend on the details of each case and the listener.
8 bit resolution is good for most recording of medium and low quality sources, such as:
- Medium quality:
- FM radio
- Low quality:
- AM radio
- 8 track
- Telephone audio
- Reel to reel at 3.75ips or slower
- Pocket recorders at live events
If medium quality sources are to be manipulated before saving the recording, it may be preferable to record in 16 bit to avoid any possible quality loss during application of effects. This does not apply to simple editing, and does not apply to low quality sources, whose resolution is below 8 bit.
Audacity 1.2 does not support 8 bit recording, 16 bit is the nearest option. It is possible to export projects in an 8 bit format, though Audacity defaults to exporting as 16 bit.
16 bit matches audio CDs, and is thus suited where the better dynamic range and S/N ratio of CD quality audio is required. 16 bit is a good general purpose high quality setting. 16 bit recording is suitable for vinyl records.
24 bit recording may be used for signals that will be manipulated but still need to maintain the full 16 bit quality of CD audio. 24 bit is good for mastering.
If you're merely listening to thousands of pounds of expertly chosen high end audio kit, and not doing large amounts of editing, there may be no real reason to exceed 24 bit depth.
Some see 32 bit recording as taking things to extremes. Although 32 bit recording can in theory have better technical specs than less bits, it is not often such great bit depth is needed. General purpose recording does not need 32 bit depth for the same reason clothing sizes do not come in increments of 1/1000th of an inch.
Finding audio sources capable of providing signals with better dynamic range than than 24 bit resolution is a demanding task. A 32 bit data stream records 65,000 times the dynamic range of 16 bit CD audio. In real world applications, a lot of those bits will be normally recording nothing but very low level background noise.
Also bear in mind that in many cases you will exporting to a 16 bit format (there are not many computer media players that support playing 32 bit files, and if you are burning to a standard audio CD, that format is by definition 16 bit).
But if you want the highest standards (for example, operate a recording studio), expect to do a large amount of manipulation of the data before export, and have audio source equipment with an extremely low noise floor, 32 bit recording (which is the default setting in Audacity) will give the best possible quality and avoid the bit depth having any effect on the sound even after heavy manipulation of the audio.
Much of the reason for this is that Audacity uses "float" format for 32 bit recording instead of fixed integer format. Normalised floating point values are quicker and easier to process on computers than fixed integer values and allow greater dynamic range to be retained even after editing. This is because intermediate signals during audio processing can have very variable values. If they all get truncated to a fixed integer format, you can't boost them back up to full scale without losing resolution (i.e. without the data becoming less representative of the original than it was before). With floating point, rounding errors during intermediate processing are negligible.
The (theoretically audible) advantage of this is that 32-bit floating point format retains the original noise floor, and does not add noise. For example, with fixed integer data, applying a compressor effect to lower the peaks by 9 dB and separately amplifying back up would cost 9dB (or more than 2 bits) of signal to noise ratio (SNR). If done with floating point data, the SNR of the peaks remains as good as before (except that the quiet passages are 9dB louder and so 9dB noisier due to the noise they had in the first place).
The advantage of using 32 bit float to work with holds even if you have to export to a 16 bit format. Using Dither on the will improve the sound quality of the exported file so there are only minimal (probably non-audible) effects of downsampling from 32 bit to 16 bit.