Signal Noise Attenuation is About More Than Filter Design

These noise cancelling headphones provide one type of noise attenuation

Even as I sit writing this post about noise attenuation in PCBs and electronics, I’m listening to music on my phone through noise-cancelling headphones. Noise attenuation measures are just as important in other areas of circuit design as they are in audio, but implementing an effective noise attenuation scheme depends on properly identifying noise sources. 

If you can identify noise sources in different electronic components and circuits, then you can take the right steps to reduce the effects of noise in different circuits. As we’ll see shortly, some noise sources are not related to the typical types of noise we tend to discuss in electronics. In addition to filtering, there are other steps that should be taken to remove noise.

Signal Noise Attenuation

The goal with signal integrity, of course, is to be able to preserve and predict the actual behavior of your electronic device’s signal(s). Therefore, when I discuss signal noise attenuation for circuit design, I’m looking into the amount of conflicting sources of noise that can affect the overall signal quality and health within an electronic device. 

While expected noise can still be a pain to balance or accommodate for in any printed circuit board or IC, noise attenuation becomes more tricky when you are tasked with removing random sources of noise. 

Random Noise Attenuation

One of the more difficult situations you may find yourself in while working on signals is trying to preserve the signal while still removing potential random sources of noise and interference. Many filtering techniques, while effective at removing random sources of noise, will come with a necessary cost of taking with it some useful signal. 

Below, I’ll discuss the different types of random noise attenuation and some ways to consider designing around them.

Thermal Noise

Random noise due to thermal fluctuations can only be attenuated by cooling down your components. For most components, such as logic gates with moderate input/output impedance, thermal noise is not a major problem as the noise margin in these components is much larger than the thermal noise power spectral density. Thermal noise fluctuations are normally on the order of nV for components and circuits with low input impedance and narrow bandwidth.

Thermal noise becomes a real problem when working with broad bandwidth components that have high input impedance. Even in extreme cases where thermal noise fluctuations reach near mV-levels, this noise source still may not interfere with logic circuits running that have sufficiently high noise margin. TTL components that run at 5 V are a good example. If you require high precision, you’ll need to use components with smaller bandwidth. There is no reason to work with a 1 GHz bandwidth component when 1 kHz will do; note that this produces a factor 1000 reduction in thermal noise fluctuations.

Analog-to-Digital Converters

Noise sources become particularly problematic when working with high precision ADCs. In ADCs with low resolution, the spacing between digital output levels can be larger than thermal noise fluctuations, so error rates will be quite low. At very high resolution, noise on the input signal can be comparable to the resolution, which increases quantization error in the output. One solution here is to increase the sampling rate as this spreads the noise power across a broader Nyquist sampling bandwidth, followed by passing the output through a digital bandpass filter.

ADC noise attenuation by oversampling

Shot Noise and Phase Noise

The other noise component that becomes important at very high frequencies and low temperatures is shot noise, which results from quantization of electrons that make up an electric current. This is another unavoidable noise source, although it is typically masked by thermal noise in most systems.

Phase noise (or timing jitter in digital circuits) arises from variations in clock sources in addition to contributions from thermal noise. If you’re generating a stream of clock pulses from a reference voltage using a comparator, timing jitter will be directly proportional to the thermal noise. With crystal oscillators, you’ll need to use electrical and mechanical compensation to reduce variations in the output.

Spurious Harmonic Content as Noise

Components and circuits like RF filters/amplifiers or other nonlinear components can produce spurious harmonic content on the output, where multiple harmonics can be seen in the frequency spectrum in addition to the desired signal. This arises due to harmonic generation in nonlinear components (i.e., transistor-based components). As an example, this problem arises in RF power amplifiers for frequency-modulated signals.

These spurious harmonics can act like noise on downstream components with broad bandwidth. Removing spurious harmonic content requires filtration. If you are inputting a single harmonic into an amplifier, the harmonic content on the output will be present at integer multiples of the input frequency, thus simple low pass or bandpass filtration is sufficient. With frequency-modulated signals, you can attempt to reduce intermodulation products with a very high order bandpass filter that is centered at the carrier frequency. Alternatively, you can run the input signal at a lower level.

Noise on Power Rails

Noise on a power rail comes in two forms: ripple or switching noise from a switching regulator, and transient oscillations due to switching. Normally, a capacitor across the output from the regulator acts as a low pass filter to regulate the DC voltage, but switching elsewhere in the regulator can still induce noise on the power rail, which is downstream from the regulator. Placing a high order, very narrow bandstop filter on the regulator output with center frequency exactly equal to the switching frequency can greatly suppress switching noise.

This power supply filter can provide noise attenuation at the switching frequency, but it cannot properly address transients on a PDN.

Other noise problems, such as phase noise (i.e., timing jitter) in a downstream component, originate from a power integrity problem on the power rails. The PDN in your PCB is really a complicated RLC network with multiple resonances and anti-resonances in the impedance spectrum, and the structure of the PDN’s impedance spectrum depends on the PDN topology (i.e., on the geometry of your PDN). 

As such, a transient response can be induced on the PDN whenever a component switches between two output signal levels. This is especially problematic in PLDs with high gate count and low operating voltage (~1 V); these components draw large currents and can induce large ripple on an improperly designed PDN, which leads to high BER.

The transient response on a power rail is difficult to predict, although it can be measured on a test coupon. One might think that the solution here is to place a bandstop filter at exactly the transient oscillation frequency on the power and ground pins on each component, but this inelegant solution increases component count and takes up board space. The better solution is to do the following:

• Reduce the impedance in the relevant bandwidth below some target value. This can be rather complicated as it requires knowing the self-resonance frequencies of any capacitive elements and parasitic capacitances and inductances in the PDN. The idea here is to get the lowest possible impedance value and try to move any impedance anti-resonances outside the relevant bandwidth.

• Attempt to critically damp or overdamp the transient response. This is one reason why high speed circuits require large power and ground planes for sufficient decoupling in addition to standard decoupling capacitors. In addition to proper decoupling, you can implement an RLC decoupling network that includes your decoupling capacitors. The goal is to bring the transient response closer to the critically damped regime.

If you know something about your noise sources, then you can implement the right noise attenuation measures in your PCB with the right PCB design and analysis software. Allegro PSpice Simulator and Cadence’s full suite of analysis tools allow you to simulate the behavior of different noise sources in your board and experiment with different noise attenuation measures in your design.

If you’re looking to learn more about how Cadence has the solution for you, talk to us and our team of experts.