11:00
9. Noise Shaping - Digital Audio Fundamentals
Akash Murthy
Overview
This video explains noise shaping, a technique used in digital audio to manipulate the distribution of noise across different frequencies. It's not about reducing overall noise but shifting it to higher, less audible frequencies, thereby increasing the perceived dynamic range. Noise shaping is always used in conjunction with dithering, especially when reducing the bit depth of audio. The video details how noise shaping works by leveraging the psychoacoustic principle that human hearing is more sensitive to certain frequencies than others, and illustrates the concept with a simple algorithm and the POW-R standards.
How was this?
Save this permanently with flashcards, quizzes, and AI chat
Chapters
- Noise shaping manipulates the frequency distribution of white noise, moving it to higher frequencies.
- The goal is to make noise less audible and increase the effective dynamic range of audio.
- It is always used with dithering and is relevant when reducing audio bit depth.
- Noise shaping itself doesn't audibly change the sound significantly but preserves audio integrity.
Understanding noise shaping helps explain how digital audio can maintain a wide dynamic range even when using lower bit depths, which is crucial for efficient storage and processing.
Shifting noise to higher frequencies where the human ear is less sensitive, making the overall perceived noise floor lower.
- Human hearing sensitivity varies across different frequencies, as shown by equal-loudness contours.
- We are most sensitive to frequencies between 500Hz and 5kHz and less sensitive at very low and very high frequencies.
- Noise shaping exploits this by concentrating noise power in the less audible frequency ranges.
- This redistribution of noise power, based on psychoacoustic principles, lowers the perceived loudness of the noise.
This chapter explains the core principle behind noise shaping's effectiveness: it leverages the limitations and characteristics of human auditory perception.
Concentrating noise power above 5kHz, where the ear is less sensitive, rather than having it spread evenly across all frequencies.
- Noise shaping can be implemented by feeding the quantized audio signal's error back into a filter.
- This feedback loop shapes the spectral content of the quantization noise, not the original signal.
- A simple method involves a feedback loop that causes the noise samples to have a higher tendency to change polarity, concentrating noise at higher frequencies.
- More complex filters can be used to more precisely match the desired noise shaping curve, like the absolute threshold of hearing.
This section demystifies the technical implementation of noise shaping, showing how mathematical processes create the desired spectral manipulation.
Using a feedback loop where the quantization error signal is filtered and then subtracted from the input signal before quantization, shaping the error itself.
- Higher-order filters offer more control but can increase overall noise power, potentially causing distortion.
- A balance must be struck between spectral shaping precision and the total noise introduced.
- The POW-R (Psychoacoustically Optimized Word Length Reduction) algorithms are industry standards for dithering and noise shaping.
- Different POW-R variants (1, 2, 3) are optimized for various audio content and playback scenarios, from compressed music to classical music in cinemas.
This introduces practical, standardized implementations of noise shaping, highlighting that different algorithms are suited for different audio applications.
POW-R 3 uses aggressive noise shaping for complex audio like classical music intended for loud playback in cinemas, pushing noise to very high frequencies.
- Dithering, in a broader sense, involves introducing small, random vibrations or signals.
- Historically, aircraft computers used vibration (dither) to reduce mechanical errors and improve accuracy.
- In digital audio, dither is a small amount of random noise added to simulate analog behavior and improve quantization accuracy.
- Applying dither, like tapping a meter for a more precise reading, helps achieve more accurate digital representations.
This provides a relatable, historical analogy for dither, reinforcing its purpose of improving accuracy in measurement and digital processes.
The vibration from an airplane's flight reducing friction and errors in mechanical navigation computers.
Key takeaways
- Noise shaping redirects noise to less audible frequencies, enhancing perceived dynamic range without lowering total noise.
- The effectiveness of noise shaping relies on the psychoacoustic principles of human hearing sensitivity.
- Noise shaping is an integral part of bit depth reduction processes, working alongside dithering.
- Technical implementations involve feedback loops and filters to spectrally shape quantization noise.
- Industry standards like POW-R offer specific noise shaping profiles tailored to different audio genres and listening environments.
- While subtle, noise shaping and dithering are essential for maintaining audio fidelity in digital systems.
- Dither, historically and in audio, serves to improve accuracy by introducing controlled randomness or vibration.
Key terms
Noise ShapingWhite NoiseFrequency BandsDynamic RangeDitheringBit DepthQuantization ErrorEqual-Loudness CurvesAbsolute Threshold of HearingPsychoacousticsPOW-R Algorithms
Test your understanding
- How does noise shaping differ from simply reducing the overall noise level in an audio signal?
- Why is understanding the human ear's frequency sensitivity crucial for noise shaping?
- What is the fundamental mechanism by which noise shaping algorithms manipulate quantization noise?
- What are the trade-offs involved in designing a noise shaping filter, and how do standards like POW-R address them?
- How does the historical concept of 'dither' relate to its application in digital audio processing?