Noise & Interference
Introduction
Noise is a pervasive factor that can impact the stability and performance of electronic systems. It is important to distinguish between interference and noise, with interference being artificial noise and noise being either natural or man-made. Understanding the basics of noise sources, their impact, and their correlation with external factors is crucial for effectively debugging and identifying the root causes of issues. In this article, we will explore common hardware system design and testing issues related to noise and interference.
Common Electronic Noise
Analog Noise:
Thermal Noise: This type of noise increases linearly with temperature and is also known as Johnson-Nyquist noise. It is critical in radio receivers as it raises the noise floor and degrades signal quality.
Flicker Noise (1/f/Pink Noise): This noise's power density is dominant in low frequencies and inversely proportional to frequency. It is caused by inherent fluctuation processes in electronic materials and exhibits equal energy for each octave frequency band.
Shot Noise: This noise is a result of the random movement of charge carriers during current conduction. It is a phenomenon of quantum physics and becomes apparent in low current and low-light settings. Shot noise exhibits a flat power spectrum density and is independent of temperature. Example of shot noise is salt peper noise seen exhibit on digital cameras at low night conditions.
Digital Noise:
Quantization Noise: This is the digital noise/error that occurs during analog-to-digital conversion (ADC). It limits the precision level of sampled data, thereby reducing the signal-to-noise ratio.
Shot Noise Waveform
This waveform represents shot noise, characterized by random noise spikes.
This kind of noise is more evident in low current and low-light conditions.
Quantization Noise
This graph depicts a 5 Hz sine wave (Original Signal) and its 3-bit quantized version (Quantized Signal). The difference between the original and quantized signals represents quantization noise.
As you can see, the quantization noise is bounded within a certain range, which, in the case of ideal uniform quantization, is ± 1/2 of the step size of the step size (1 LSB = 0.25 amplitude) .
Interference
Common Mode Noise: This type of noise is the common electronic radiation noise that creates an equal offset voltage in both the signal transmission and return paths. The presence of this offset voltage prevents signal and return current cancellation, resulting in an alternate current return loop that generates unwanted noise radiation. Common mode noise can originate from various sources such as light flickers, phase misalignment in differential signal transmission, ground bounce, or AC main coupling from the power supply.
Differential Mode Noise: This noise is produced by power supplies and includes voltage ripples as a type of differential mode noise. The noise current in this case has equal and opposite polarity on the transmit and return paths.
Intersymbol Interference (ISI): ISI is a signal distortion observed in digital communication receivers. It occurs when the previous digital symbol pulse overlaps with the new symbol pulse due to channel distortion (i.e, delay and smearing effect), and the most common cause is parasitic capacitance of the line and load, meaning that the parasitic capacitance has left over charge that has not been competely drained before the next digital symbol is sent by the transmitter, hence causing interference with the later symbol and impacting signal integrity of the communication channel.
Common Mode Noise
The top graph shows both the positive and negative parts of the differential signal, each affected by 60 Hz common mode noise.
The bottom graph demonstrates how the differential nature of the signal can reject the common mode noise, resulting in a cleaner output.
Differential Mode Mode Noise
The red-shaded region highlights the Differential Mode Noise affecting the transmit signal.
Inter Symbol Interference
The top graph shows both the original pulses (solid lines) and their smeared counterparts (dashed lines) after passing through a communication channel, modeled as a first order low pass filter).
The bottom graph showsthe received signal, which is a superimposition of the individual smeared pulses. The red-shaded regions highlight where the smearing effect of one symbol extends into the time frame of the next symbol, causing ISI.
Note: For the example, we spaced out each bits to visually demonstrate the smearing effect into next symbol window. However in practice, the symbols are peseorandomly generated (i.e, 11011000. )
Quantifying Noise & Interference levels Using Power Spectral Density (PSD) Estimation
In a practical setting, time-series data often serves as the initial point of analysis, offering a preliminary view of how noise and interference impact signal quality within an electronic system's communication network. To delve deeper into quantifying these effects, spectral estimation techniques based on time-series data are commonly employed. One widely-used method for this purpose is Welch's method for Power Spectral Density (PSD) estimation.
In essence, Welch's method allows us to examine the power of the signal in the frequency domain, providing a more comprehensive understanding of noise characteristics. Analyzing the signal in the Fourier domain is particularly useful because it helps isolate individual frequency components, making it easier to identify and mitigate sources of noise and interference.
An illustration demonstrating the impact of added noise levels on spectral esitimation is provided below:
Noise Data
The top plot shows the original signal with multiple frequency components (5 Hz, 50 Hz, and 150 Hz).
The bottom plot shows the same signal but with added noise.
PSD Estimation
The top plot shows the PSD estimation for a clean signal with multiple frequency components (5 Hz, 50 Hz, and 150 Hz).
The bottom plot shows the PSD estimation for the same signal but with added noise.
As we can see, the spectral noise floor is increased in the noisy signal by looking at PSD plot. This provide a quick estimation for signal to noise ratio.
The Signal-to-Noise Ratio (SNR) is a measure that quantifies how much a signal has been corrupted by noise. It is usually defined as the ratio of the power of a signal to the power of background noise. Here are some common ways to calculate i
Steps for SNR calcualtion:
Identify the Peak Frequencies: Use a peak-finding algorithm to identify the frequencies where peaks occur.
Select Bandwidth: Manually select a bandwidth around each peak (first example assume 5 Hz for low frequency signals, or use adaptive bandwidth method to estimate all peaks)
Estimate Signal Power: Sum PSD Values under Peaks: Sum the PSD values within this bandwidth for each peak.
Estimate Noise Power: Sum the PSD Values without Peaks: this is equalivalent of partial sum of the PSD plot with the peak region masked/ removed, which equals to the noise power.
Calculate SNR: 10log(Signal Power/ Noise Power)
Noise Mitigation
One we identify the noise and inteference source and qualify it’s impact on signal quality. We can implement following migtigations in noise suppresion.
Thermal Noise: Mitigation techniques include using thermal pads and other solutions to limit the maximum IC package temperature, using smaller resistance parts, and employing wider copper traces to limit conductor temperature rise on the PCB.
Quantization Noise: Oversampling the analog signal spreads the noise power across a wider spectrum, reducing the noise floor and improving the signal-to-noise ratio. Dithering and noise shaping techniques involve adding random analog noise to the signal before quantization, effectively reducing quantization noise. In image processing, this ditering technique is also called pixel shifting to increase resolution for an imaging system.
Intersymbol Interference (ISI): the left over charges on the transmission line can be counteracted using equalization technique on the receiver side.
Summary and Conclusion
Thermal, pink, and quantization noise are the three dominant types of noise in electronics. Reducing the operating temperature of a product helps reduce the impact of dominant analog noise. Quantization noise limits the resolution of analog-to-digital conversion. Both analog and digital noise can significantly affect the performance of electronic systems. Understanding the sources and impact of noise is crucial for implementing the appropriate noise mitigation solutions in hardware system design.
For further reading
"Understand Common Mode Noise", https://www.pulseelectronics.com/wp-content/uploads/2016/12/G019.pdf