Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
28 Cards in this Set
- Front
- Back
PCM quantizes analog audio into steps
|
These steps are:
- 65,536 steps for 16-bit audio - 16,777,216 steps for 24-bit audio |
|
Dynamic Range
|
- The difference in dB between noise floor and maximum output, before distortion
- Wordlength x6 - 6dB of gain per bit |
|
Lowest Signal
|
- 16-bit audio = -96dB
- 24-bit audio = -144dB - 48-bit audio = - 288dB - 56-bit audio = -336dB |
|
PCM quantizing creates
|
- Stepped Distortion
|
|
PCM has difficulty
|
- encoding signal near the noise floor
|
|
Dither
*(is white noise, equal energy per frequency) |
- randomized noise which averages 0 volts DC (having an equal number of 1's and 0's)
|
|
Dither is added
|
- to the analog input stage to toggle or alternate the LSB* in the wordlength
|
|
LSB*
(Least significant bit) |
- represents the lowest level signal at the converter
|
|
Modulating the LSB
|
- allows signals at or below the noise floor to be encoded by averaging the dither gain with the low level signal
|
|
Cheap low-pass filters (LPF)
|
- create higher distortion, ripple (ringing) and aliasing.
|
|
Ripple/Aliasing artifacts are heard
|
- as time-smearing of the audio and short (millisecond) echoes
|
|
Pass Band
|
- Unfiltered range below LPF
|
|
Transition band
|
- beginning of cutoff frequency to stop band
- steepness determines slope of filter |
|
Stop band
|
- point at which maximum loss of the filter is reached
- at or below Nyquest frequency to prevent aliasing |
|
Aliasing or ripple in the passband is caused by
|
- the LPF not attenuating enough in the stop band, allowing excess signal above the Nyquist frequency to fold back (or alias) into the passband
|
|
Aliasing can be prevented
|
- with a very steep filter, or a gentle filter at a high sample rate
|
|
For ease and cost, most manufacturers -
|
- use cheap filter/high sample rate approach
|
|
Filter distortion issues are off-ten fixed by...
|
- Oversampling (A/D) or Downsampling (D/A) processing
- Where the front or back end of the converter operates at double (or greater than) the sample rate |
|
EXAMPLE
|
- A 44.1kHz source with a 128x oversampling converter is actually sampling at 5.64MHz
|
|
Oversampling...
|
- moves the filter distortion above the audible range, which itself is then filtered out with a "decimator" anti-aliasing filter in the downsampling phase
|
|
This LPF decimator...
|
- is applied AFTER the oversampling circuit and allows poor filters to not become a liability in the converter's performance
|
|
The biggest impact on our perception of clarity in digital audio...
|
- The LPF designs used in different converters - their performance impacts the sound more than higher sample rates alone
|
|
High sample rates sound better
|
- because they push the filter's distortion out of the audible band into the area above 20kHz where it cannot effect musical content
|
|
The ideal response for an LPF
|
- has to be at or above 50kHz to avoid pre-echo artifacts from being created in the cochlear filter of the human ear
|
|
192kHz sample rates may be unnecessary...
|
- as most 192kHz ADC's are not really doing more than a double 96kHz conversion
|
|
This higher sample rate...
|
- generates little usable content while introducing extra noise, raising the potential for calculation errors
- and creating storage/processing challenges |
|
The use of 88.2kHz and 96kHz rates for sampling and reproduction
|
- are justified as they add the benefits of the higher filter setting while maintaining reasonable size files
|
|
The computational simplicity of down sampling 88.2kHz content into 44.1kHz
|
- makes it an option preferred by many professionals
|