Digital Signal Processing -EC504- Module3 ( MAKAUT-Syllabus)

Today, our communication elements are: -  Effect of Finished Register Length in FIR FILTER DESIGN, Parametric and non-parametric spectral estimate, Introduction to Multidimensional Signal Processing, Applications of Digital Signal Processing (DSP).

Effect of Finished Register Length in FIR FILTER DESIGN: -

The filters for final impulse responses (FIR) are widely used in Digital Signal Processing (DSP) due to their stability, linear-step property, and direct implementation. However, when practising practical hardware or software from theoretical design, many seem invisible. One of the most important issues is the effect of the final register length. Sign-ups in digital systems can only save numbers with limited precision, showing review and rounding errors. These boundaries affect filter coefficients, arithmetic operations, and eventually the performance of FIR filters.

In this blog, we will detect the effect of four filter designs, their results, and methods to reduce these effects.

1. Final register length and its role:

In digital systems, signal and filter coefficients are represented using a certain number of bits in binary. The number of pieces available to represent these values is referred to as the length of the register. The length of a large register means high accuracy, while a small register leads to rough permigration in length.

When designing the FIR filter, three main aspects of the final register length are affected:

• Filter coefficients may be represented in the exact value of the coefficient obtained from the design methods (such as window or frequency sampling), but may not be absolutely represented in final accuracy, causing adjacent errors.
• Signal perishable – input and intermediate signals are rounded or cut to fit into the registers.
• In arithmetic operations, during rounding and overflow filtering, multiplication and addition can accumulate errors due to limited word lengths.

2. Effect of Coefficient Quantization

The FIR filters depend on the exact values of the coefficients to achieve the desired frequency response. For example, an ideally lower passport filter may require coefficients with infinite accuracy. However, in the implementation of hardware, the coefficient must be a volume to fit into the available registers.

• Deformation of quantity response: The penetration changes the real frequency reaction compared to the ideal design. It can move the cutting frequency or change the passenger tape and stopband properties.
• Increased ripple: Both passenger tape and stopping waves can increase due to review errors. This means that the filter no longer meets the exact design specifications.
• Loss of symmetry: In linear-phase FIR filters, coefficients are usually symmetrical. Celebration can break this symmetry, which can lead to phase deformation.

The degree of deformation depends on the number of pieces used. For example, a 16-bit representation retains much more accuracy than an 8-bit system.

3. Effect of signal abolition

Digital signals processed by FIR filters also suffer from the implementation, as the input signal values are stored with final precision. Each quantified value varies from an ideal signal by a small error, which behaves like an extra source of noise.

This implementation spreads into the noise frequency spectrum and improves the noise bottom. For signs with small dimensions, the review error can be relatively important, reducing the signal-to-noise ratio (SNR).

4. Rounding noise in arithmetic surgery

When you multiply and transport the FIR filter, the results often require more pieces than the length of the register. These results should be cut or rounded to fit back in the registers.

• Multi-error: Multiplying two final accurate numbers often produces double results along the word length. Running this result back to the length of the register introduces errors.
• Accumulation error: FIR filtration requires adding more words to each output test. Frequent joints with final sequential numbers accumulate rounding noise.

These errors appear as random noise at the exit, which reduces the performance of the filter, especially for long filters with multiple coefficients.

5. Overflow in fixed-point arithmetics:

In the implementation of a fixed point, another anxiety is overflow, which occurs when the results are higher than the registrar-represented limit. For example, adding large numbers can lead to results outside of the maximum or minimum representative values.

Common solutions to overflow include:

• Saturation variety: The values exceeding the area are cut from the maximum or minimum.
• Modulo Arithmetic (Rap-Around): Packing is packed around the registry area.

While saturation is usually preferred, both approaches deform the output signal.

6. Demonstration measures affected:

The length of the final register affects several performance measures of the FIR filter, including:

• Dimensions Deformation – Deviations from the desired frequency reaction.
• Face formation – especially when the symmetry of the coefficient is lost.
• Signal-to-show ratio (SNR) permitisation and rounding noise reduction.
• Stability and strength – although four filters are naturally stable, accurate errors can affect their strength in adaptive applications.

7. Ways to reduce the final register effect:

To ensure that the practical four-filter is close to its theoretical design, many strategies are used:

Increase the length of the word – using multiple pieces for coefficients and signals reduces the review error.

• Care coefficient scaling – scaling ensures that coefficients use the available dynamic area effectively.
• Error analysis during design – designers often follow the effect of the execution of meeting specifications.
• Floating-point arithmetic – floating-point representation provides a very large dynamic range, reducing review and overflow problems.
• Noisy Size and Dithering – Techniques spreading the review noise smoothly, improving subjective signal quality.

Parametric and non-parametric spectral estimate: -

In Digital Signal Processing (DSP), the frequency of the spectral estimated signals plays an important role in analysing the content. Whether it is for speech treatment, radar, biomedical signal analysis, or communication systems, it is fundamental to understand how energy is distributed in frequencies. The two extensive approaches to estimating the power spectral density (PSD) of a signal are parametric and non-parametric methods. Each approach has its own benefits, perceptions, and limitations, which makes it important to understand their differences and applications.

1. Introduction to spectral assessment:

Power Spectral Density (PSD) explains how the signal power is distributed with frequency. For stable random signals, PSD estimates allow us to identify the main frequency components, periodicity, and noise level. However, since we usually only have a final length sample of the signal, we cannot calculate the exact PSD. Instead, we assume that it uses calculation methods.

The spectral inference methods are widely classified:

• Do not consider an underlying model for non-pivotal methods or gyanals. They estimate PSD directly from data.
• Parametric methods – Think of a mathematical model (e.g., autoregressive) for signals and estimate PSD based on model parameters.

2. Non-purified spectral estimates:

a) Concept

Non-parametric methods estimate PSD directly from data without receiving any pre-model for signals. The most common technique is based on periodograms, which are derived from the amount of the uncertain Fourier transformation (DTFT) to the signal.

B) methods

Endless method

• PSD is calculated as a class size on the Discrete Fourier Transformation (DFT).
• Simple but exposed to high variance and spectral leakage.

Revised period

• Uses a window function to reduce the leakage effect.
• Frequency improves the resolution at the expense of some bias.

Method of Melth

• The signal of an overlapping segment uses a window for each and an average duration.
• The variance of estimates significantly reduces.

Blackman-Tuki Law

• First, it estimates the autocorrelation of the signal and then takes its Fourier transformation.
• Prejudice and variance provide better control over trades.

C) properties

Advantage:

• Easy to calculate.
• Pre-perceptions of the signal are not required.
• The search is good for data analysis.

Limits:

• Poor resolution for short data registrations.
• The variance of estimates may be higher.
• Spectral leakage from air.

3. Parametric spectral estimate:

a) Concept

Parametric methods believe that the signal can be modelled when using a parametric process, such as an autoregressive (AR), moving average (MA), or autoregressive moving average (ARMA) model. PSD is then estimated by mounting the data for the data and its theoretical range.

B) General Model:

Autoregressive (AR) model

• Each sample depends linearly on previous samples, and the white noise depends on the entrance.
• Provides high resolution in spectral estimates, especially for small data registrations.
• PSD is achieved from the AR coefficient.

MOVING AVERAGE (MA) model

• The previous noise represents the signal in the form of a linear combination of values.
• Suitable for dominant signs of components such as noise.

ARMA model

• AR and MA combine both processes.
• More flexible, but calculated and complicated.

C) Estimate Methods

• Estimate the AR coefficient from the Yule-Walker Equation Autocorrelation Function.
• Method rock – provides stable, high-resolution AR estimates.
• Methods of maximum probability – estimate the approximate parameters that maximise the possibility of the data shown.

d) properties

Advantage:

• Also high-frequency resolution with small data registrations.
• Smooth spectral estimate.
• Strong spectral peaks can be modelled effectively.

Limits:

• The correct model selection (AR, MA, ARMA) requires.
• Sensitive to model orders – understood Mrs Spectral Details, overestimating overestimation Spi as tops.
• Computationally more demanding than non-paternal methods.

Comparison of Parametric and Non-Parametric Methods:

Feature	Non-Parametric	Parametric
Assumption	No prior model	Assumes AR/MA/ARMA model
Data Requirement	Requires long data records for accuracy	Can work well with short data records
Frequency Resolution	Limited by record length	High resolution
Variance	High (especially for the raw periodogram)	Lower variance with a proper model
Complexity	Simple to compute	Computationally intensive
Applications	General-purpose, exploratory analysis	Speech, radar, and biomedical signals

Introduction to Multinational Signal Processing: -

In Digital Signal Processing (DSP), applications such as voice compression, audio improvement, image processing, radar, and communication systems are important when handling signals and providing flexibility when handling signals. A powerful concept that enables efficient processing is the multitasking of signal processing. As the name suggests, multitation treatment involves better performance, low complexity, or using different sampling speeds in the same system to achieve increased functionality.

This blog introduces multitation signal treatment, its basic operation, benefits, and basic principles of applications.

1. What is multitation signal treatment:

In a traditional DSP system, signals are usually tested and treated at a certain sampling speed. However, many practical systems need to work with several samples. For example, in a communications receiver, different stages such as filtration, equalization, and modulation can work more effectively at different speeds.

Multitasking of signal processing refers to techniques where the test speed for signals changes - either increases or sinks - at different stages of processing. There are two basic operations:

• Desimation (downsampling): Reduce the sample speed by an integer factor.
• Project: To increase the sampling speed by an integer factor.

By combining these tasks with care, the signal can be treated at the most appropriate speed for each task.

2. Basic Operations:

a) decimation (downsampling)

Decimation reduces the test speed of a character by a factor of M. The process includes two steps:

• A lower passport filter (anti-aliasing filter) is used to prevent filtration aliasing.
• Down Sampling - Keep each MTH sample and leave the rest.

B) projection (omission)

The interpolation increases the speed of sampling by a factor L. The process includes:

• UPSAMPLING - Insert l -1 zero between each sample of the original sequence.
• Apply a lower passport filter (interpolation filter) to remove the spectral images initiated by filtration-zero-mamilan.

C) Conversion of sample speed after rational factors

In many applications, the desired sampling rate is not an integer multiple of the original or original speeds. In such cases, rational factor conversion is done. The conversion factor is expressed as l/m, where:

• First the signal is launched by L,
• Then decomposed by M.

This technique allows flexible conversion between arbitrary sampling speeds.

3. Filter structures at several levels

When using multitation techniques, filter design plays an important role. There are two common structures:

• Polypse filter: Use projection and resolution effectively by decomposing the filter into small subfilters. It reduces the calculation load compared to direct filtration.
• Multistage Implementation: For large conversion factors, instead of applying projection/resolution in one step, the process for better efficiency is broken into several small stages.

These structures are widely adopted in practice to use talented digital filter banks, sound codes, and communications receivers.

4. Benefits of multinational signal processing

Computer efficiency - By reducing the sampling speed before filtration, less calculation is required. It is especially useful in systems that work with high-speed signals.

Flexibility - Different parts of the system can work with the most appropriate sampling speed.

Better performance - Multitate technology allows better low -cow -duty filter design by utilizing speed changes.

Effective memory use - Low sampling speed reduces the number of samples to be stored and processed.

Adaptability - enables spontaneous conversion between different standards or signal formats.

5. Application of multinational signal processing

Multidimensional processing receives applications in almost all fields of DSP:

Digital Sound System:

• Example speed conversion between different standards (eg, 44.1 kHz for CD for 48 kHz for professional sound).
• Sound compression and increase.

Communication system:

• Effective modulation and demodulation.
• Complexing and multicarrier systems such as OFDM.
• The adaptive receiver requires variable sampling speed.

Speech treatment:

• Compression technology such as CELP and Vocoders.
• Reduced noise in speech growth systems.,

Biomedical signal processing:

• ECG, EEG signal analysis, where multi-technology improves the solution and reduces data size.

Image and video processing:

• Upscaling/downscaling.
• Multispectral analysis when using waves.

6. Challenges in multinational treatment

While multination treatment provides great benefits, it also introduces some challenges:

• Filter Design Complexity - The filter should be designed to prevent carefully (in decimation) or to eliminate spectral images (estimated).
• Computer weighing yourself on effective overall, polypage, and multi-stage design requires careful implementation.
• In synchronization problems in the world's time systems, it can be complicated to maintain exactly time in many places.

Applications of Digital Signal Processing (DSP): -

Digital Signal Processing (DSP) is a powerful branch of engineering science related to the analysis, change, and manipulation of digital signals. It has become an essential part of modern technology because of the ability to treat the signal properly, effectively, and in real time. Today, DSP techniques are used in almost all fields where information is represented as signals - audio, images, videos, biomedical data, radar, and communication.

1. Sound and speech treatment

One of the broadest applications of DSP is in sound and speech.

• Noise reduction: DSP algorithm removes unwanted background noise in mobile phones, hearing equipment, and voice assistants.
• Compression: When you retain standard quality, such as MP3 and AAC, use DSP to reduce file size.
• Speech recognition and synthesis: virtual auxiliary texts such as Siri, Alexa, and Google rely on DSP to convert speech and generate human reactions to human reactions.

2. Image and video processing

DSP plays a key role in the multimedia system.

• Image enlargement: Technology such as filtration, edge detection, and sharpening improves the quality of the image in cameras and medical image systems.
• Compression: Forms such as JPEG, MPEG, and H.264 use DSP to reduce storage and transmission requirements.
• Data vision: Applications such as face identification, object detection, and monitoring of DSP-based algorithms for precise analysis.Use of Digital Signal Processing (DSP)

3. Communication system

Modern communication without DSP will not be possible.

• Modulation and demodulation: DSP enables effective transmission and receipt of signals in wireless, satellite, and optical systems.
• Error detection and correction: Provide reliable communication on noise channels.
• Adaptive filtration: ECO is used in ECO cancellation for telecommunications systems.

4. Biomedical application

DSP has revolutionized the healthcare system by improving the signal analysis in medical equipment.

• ECG and EEG analysis: Detects abnormalities in heart and brain activity.
• Medical imaging: Techniques such as CT, MRI, and ultrasound depend on DSP for reconstruction and growth.
• Portable equipment: Smartwatch uses DSP to monitor the heartbeat, oxygen level, and physical activity.

5. Radar, control, and industrial system

• Radar and sonar: DSP increases detection and tracking in defense and navigation systems.
• Robotics and control: DSP-based controls improve system stability and accuracy.
• Industrial automation: Quality control, vibration analysis, and machine monitoring are used.

conclusion

Digital signal processing is the spine of modern technology. From smartphones to medical equipment, from entertainment to defense, DSP equipment provides information effectively and accurately. Versatility ensures that DSP will continue to innovate in industries in the coming years.

-------------------------------------------MODULE-4 NEXTPAGE-------------------------------------------