Analog-to-Digital Conversion (ADC)

• Fundamentals of Signal Processing covers topics such as source coding.
• These notes are for Vehicular Multimedia (MPA-VHM).

Analog-to-Digital Conversion (ADC)

• Radio communication systems rely on the digital form of source signals.
• Analog source signals must be converted using analog-to-digital converters (ADCs).
• Analog-to-digital conversion follows sampling, quantization, and binary encoding.
• Sampling and holding involves sampling a continuous signal and freezing its value.
• The goal of sampling and holding is to maintain a steady value for a short time.
• Sampling and holding removes variations in the input signal with limited capabilities and power supply.
• Quantization partitions the reference signal into discrete quanta and matches input to quantum.
• Encoding assigns a unique digital code to each quantum.
• Then allocates the input signal with its determined digital code.
• Narrow discrete impulses are created from analog continuous signals.
• he frequency (fs) is at least double the signal’s maximum frequency (fm or fmax).
• Amplitude corresponds to the continuous signal's voltage in sampling time instants.
• Quantization involves assigning the closest quantization level to the original pulse amplitude.
• Quantization levels are discrete voltage levels, with the boundary being the decision level.
• Binary encoding assigns a binary number to the discrete quantization voltage level.
• The Sampling Theorem states if the highest analog signal frequency is fmax, sampling at a rate fs ≥ 2 × fmax allows exact signal recovery.
• Successive Approximation ADC compares input and output signals, which is expensive.
• Dual Slope ADC offers high accuracy via high sampling rate/resolution but is slow.
• Delta-Sigma ADC provides high accuracy but operates slowly due to oversampling.
• Flash ADC is the fastest type but very expensive.
• Digital-to-Analog Converters (DAC) use oversampling.
• DAC generates new values between input signal values through interpolation.
• Difference between interpolated values decreases, enabling quantization with 1-bit resolution when oversampling by 256 times.
• A 1-bit DAC (high linearity at small amplitudes) is controlled by PDM signal generated from a 1-bit sequence.
• The following analog integrator smooths the output signal form.

Source Coding

• Source coding is used in signals like voice, sound, video to reduce redundant/irrelevant parts.
• Leads to lower bit rate and channel bandwidth.
• Compression Ratio (CR) measures relative data size reduction via compression.
• It's the ratio of uncompressed to compressed size, such as (20 Mbit/s) / (2 Mbit/s) = 10 (or 10:1).
• Redundant data is repetitive data that doesn't add information.
• It allows lossless recovery at the receiver, common in Morse code, variable length coding (VLC), and Huffman coding.
• Irrelevant data can't be perceived by human senses or detected by the eye/ear.
• An example of irrelevant data is "sharpness" in color reduction, or fine structures in picture vs. coarse structures.
• Raw speech signal, the analog form requires ADC of the signal.
• Speech signals have a maximum frequency (fmax) of about 4 kHz.
• Speech source coders are cost-effective with CR typically between 2:1 and 25:1.
• The encoded output signal is evaluated using the Mean Opinion Score (MOS) on a scale from 5 (excellent) to 1 (worst).
• Vocoders are used for digital speech coding/voice simulation (1.2 to 64 kbps).
• They capture audio's characteristic elements to affect other audio signals.
• Vocoders were initially meant to synthesize speech.
• The main component is the formant, describing fundamental frequency and noise [2].

MPEG Audio Compression

• The human ear's dynamic range is about 140 dB with a hearing bandwidth up to 20 kHz.
• Sampling at 48 kHz with 16-bit resolution yields 786 kbit/s per channel (or 1.5 Mbit/s for stereo).
• Audio compression aims to reduce this rate to between 100 and 400 kbit/s.
• Sub-band coding splits the audio signal into sub-bands for irrelevancy reduction and is used MASCAM, MUSICAM, and DAB.
• Transform Coding transforms the audio signal, using Discrete Cosine Transform (DCT), removing irrelevant components in the time domain to the frequency domain.
• MPEG includes sub-band coding and transform coding.
• Psychoacoustic model redundancy reduction (lossless) and irrelevancy reduction (lossy) to lower data rate by 90%.
• Irrelevancy reduction relies on the psychacoustic model: audio components inaudible to human ears are not transmitted.
• Sensitivity depends on frequency, peaking at 3-4 kHz, dropping off outside this range.
• Sounds under a certain threshold are not audible.
• Masking involves applying a carrier to the ear with a constant frequency for a test.
• The masking threshold depends on the frequency.
• Components below the threshold don't need transmitting.
• MPEG audio encoding splits the source into two coder branches and is filtered to then be taken to a frequency analyzer.
• Analyzers perform spectrum analysis via FFT that determines audio components (low time resolution, high frequency resolution).
• Psychoacoustic model knowledge identifies irrelevant frequency components (masking effect).
• The audio signal is filtered into sub-bands, where complete sub-bands may be masked (signal level below audible threshold).
• If signals are slightly above masking thresholds, quantization is reduced till quantization noise is below masking threshold.
• Quantization is controlled by a sub-band quantizer for irrelevance reduction.
• Data coding is known as redundancy reduction.

Image and Video signal

• Color images (SD) involve analog luminance signal Y (0-6 MHz) and chrominance signals Y-R and Y-B, (0-1.3 MHz).
• According to ITU-R BT.601, the Y signal's sampling frequency is 13.5 MHz, while the chrominance signal is 6.75 MHz during digitalizing via PCM
• Individual samples quantized with 8 bits and up to 256 quantization levels. The Y signal's bit rate at 108 Mbit/s, each chrominance signal is 54 Mbit/s.
• Combined stream is 216 Mbit/s, requiring a reduction to 2-15 Mbit/s.
• Transform coding helps in reducing bit rate reduction for still images like JPEG.
• The image is divided into 8 × 8 blocks (pixels) as a compromise between quality/complexity.
• The Y, Y-R, and Y-B signals are processed separately within the blocks.
• 2D-DCT helps with result and complexity.
• An 8 × 8 pixel block transforms into the frequency domain via 2D-DCT.
• The first coefficient is the DC component. Column contains top-to-bottom energies to vertical direction.
• Quantization divides coefficients by appropriate quantization factors.
• After Quantization the 8 × 8 matrix is read with Zig-Zag scanning, giving adjacent zeroes.
• Huffman and Variable Length Coding (VLC) eliminate repetitive data.
• MPEG reduces bitrate of video frame sequences. The signal converts into macroblocks that consist of 16 x 16 samples with 4 blocks of Y that compliments a block of chrominance signal.
• MPEG-1 encoder divides at input and into Group of Pictures (GOP) that repeats every 12 frames (0.5 s).
• Transmitted in order, the reference frame transmitted without prediction is called Intra frame.
• Other predicted frames can be:
  • P (Predict frame): it transmits the difference, going one way, reduction rate is 2x
  • B (Bidirectional frame): it transmits the difference with reduction of rate at an 8x multiplier.
• Motion estimation, estimates those vectors from delta frame to get decoded. uses the block matching with forward predictions.
• If found front/back in coding, vector are made and sent. Block delta, coded apart does quantization.

Codecs

• CODEC describes encoding and decoding devices or (EN)CODER – (DE)CODER.
• MPEG-1 Layer I, the simplest, cuts bit rate slightly.
• MPEG-1 Layer II, based on the MUSICAM, lowers compared to I, yet increases sound quality with audio
• MPEG-1 Layer III, it uses the DCT (MDCT) at a lower bit rate for a complex decoder, (MP3).
• MPEG-2 transmits audio channels and five/six Dolby Channels.
• MPEG-2 Part 7 uses MDCT than filters and its supported via multi-channel coding which allows MP3 to better the complexity.
• Low Complexity Communication Codec (LC3) (LE)
• audio specified at group (SIG) for LE. it was developed by Fraunhofer participants; LC3plus codec is successor.
• audio in sections used in the 7.5 or 10 ms range which compressed sections one at a time
• Uncompressed has process by LC (MDCT) is made by the frequency of time.
• tools
  • SNS minimazes to the quantization being minimally audible.
  • TNS reduces an echo.
  • quantizes for the bits to encodes.
  • leveling reduces quantization.
• MP3: used, coding is set-up and constant rate variable with CD at kbit/s.
• WMA.
• Vorbis: open source audio codec: high quality
• WAV IBM is simple.
• FLAC has compression.
• ALAC has rate with open source capabilities.
• AAC lossy audio codec designed to replace and has successors called + codec.
• Filterbank uses an an MDCT.
• Win switches from samples to improve for length channels.
• Coding is improved for the long channel.
• and coding uses huffman based coding.
• and it transform (L, R) channel.
• High Efficiency (v2) is successor with high standard.
• Audio enter with split bandwidth
• has stereos and encoder used is used to the signal and with encoder has stereo.
• The AAC coder codes from the properties.
• the repilica enrich with frequencies so you can frequencies.
• Properties in the brain use higher to accuracy that can transmit bands,
• audio encoder does all the coding with several psycho-acoustic features. The replication helps encoding easily.
• stereo makes channel more signal that mix so its more effective.
• Its channel by 3 parameters
  • intensity difference (IID) is individual.
  • cross relation.
  • phase is of delay.

FLAC

• An audio allows noquality loss: It consist main blocks.
• divides a block in size. if it’s too increase in effect.
• inter channel (M,S) removed signals that bits will (L+R) and the L.R without any with is good compression.
• coding, if codes exact will the required signal codes to is that code methods that set-up or channels and is.

Learn about Analog-to-Digital Conversion (ADC) in radio communication systems. This involves sampling, quantization, and binary encoding. Understand how sampling and holding maintain a steady value for a short time, while quantization partitions the reference signal into discrete quanta and matches the input.