Skip to main content

Identifying Noise Sources to Reduce the Noise Floor in Your PCB

Man yelling through a megaphone at employee


As a self-proclaimed audiophile, I still have an older stereo system that was gifted to me from my father. This old system is great for annoying my neighbors while listening to hair metal, but turn down the volume and you can hear the faint hum of background noise through the speakers. I won’t be upgrading my stereo system, even though newer systems provide much better noise suppression.

A lot of folks like to talk about noise and the noise floor in their system in terms of audio because you can actually “hear” any electronic noise in the system. Some designers talk about noise specifically in terms of EMI, either radiated from an external source or conducted between components. When considered alongside some intrinsic noise sources and unique components, EMI is certainly at play. 

However, there are multiple effects that can manifest noise in your PCB, creating problems in measurement and operation in severe cases. There a number of strategies you can take to reduce the noise floor throughout your system, ranging from something as simple as shielding cans to simple circuits that help decrease the noise floor seen by critical components.

Noise Sources and Noise Floor

Some components, such as crystal oscillators for generating clock signals or RF filters/amplifiers, will output multiple overtones in addition to the desired signal. Other switching components, such as switching regulators, will output a ringing signal that can contribute to the noise level in your board, depending on its amplitude. These noise sources are well-known and it is better to plan on addressing these noise sources in your design. Other noise sources, such as white noise (also called Johnson noise), pink noise, phase noise, or simply unpredictable EMI from other devices, may also be present as background noise.

These noise sources cannot be eliminated completely, but you can devise some strategies for reducing them if you can identify prominent noise sources. The sum power spectral density of all noise sources in your system is your noise floor, and this will determine the minimum signal level that you can measure or detect in the system. By identifying and suppressing various noise sources, you can decrease the noise floor to the lowest possible value.

Something like ringing due to switching is usually easy to identify; it will appear as a damped oscillation that is superimposed on a digital signal, and this can be seen in the time domain. Minor ringing may have low amplitude and can still contribute to noise in the system. Some noise sources like phase noise (for analog signals), timing jitter (for digital signals), intense ringing, and residual ripple can be easily identified in the time domain if you know what to look for in an oscilloscope trace. However, looking at a noisy oscilloscope trace may hide some noise sources that can only be distinguished in the frequency domain.


Spectrum analyzer output showing the noise floor

Output from a spectrum analyzer showing the noise floor in the desired spectrum


Identifying Noise Sources with Noise Floor Measurements

When you’re gathering signal measurements in your board, such as the current travelling to a component or the voltage across a particular portion of a circuit, your measurement will return some noisy signal in the absence of the signal you want to measure. With some clever frequency domain analysis, you can distinguish noise components in a background noise signal and determine the best design choices for reducing the effects of noise.

When you have a noisy background signal in the time domain, the simplest way to determine the source of this noise is by applying a Fourier transform to the data. Your data will show the contribution from all noise sources in your system. The output from a Fourier transform can tell you which noise sources are prominent in the system. What appears as a chaotic signal in the time domain may actually be created by some small number of discrete, uncorrelated frequencies arising from different components in the board. This is shown as an example below:


Fourier transform applied to time domain noise floor data

Determining components in the noise floor with a Fourier transform


From this graph we can see that there are actually five harmonic noise sources at different frequencies. The peaks at approximately 2.5, 5, 7.5, and 10 kHz may be the first four harmonics in the output from a particular component. If the board is part of a multi-board system, these peaks could also arise in the board under test due to radiated EMI, although it would be surprising to see such strong peaks at low frequency. These peaks are more likely conducted EMI from an upstream component.

The broad background shows the noise floor and is likely due to Johnson noise, although we cannot discount the effects of pink noise at such low frequencies. Note that you can use a spectrum analyzer to gather an accurate measurement in the frequency domain directly.

Gathering an accurate noise floor measurement requires a probe system an extremely low noise floor (normally measured in dBm). For example, if your probe has a noise floor of -120 dBm, but you are trying to measure a -140 dBm signal, then you will not be able to see the signal above the noise floor. The noise floor must be reduced by at least 20 dBm in order to have a chance at seeing the signal. In this type of situation, you will need to increase the signal strength, decrease the noise floor, or both, such that the signal is detected above some minimum signal-to-noise ratio.

Reducing Your Noise Floor

The strategy you should take to reduce noise really depends on the particular sources. In the case of spurious harmonics on the output of some component, filtration may be sufficient to decrease the strength of these signals such that they become unnoticeable against other noise sources. Johnson noise is dependent on the temperature of your components, and decreasing this noise source requires cooling down the device. This is why sensitive systems for measuring weak optical signals are often Peltier or liquid cooled.

When you examine the effects of Johnson noise in a particular component, you need to consider the component’s bandwidth. The minimum equivalent input noise due to Johnson noise will normally be specified at a particular temperature (usually in units of dBm/Hz) and will contribute to the noise floor you would measure in the output from the component. The component’s bandwidth and the noise figure between the input and output can then be used to determine the noise floor:


Equation for noise floor

Noise floor equation due to Johnson noise, in units of dBm


When working with RF components or other analog circuits, techniques like harmonic balance analysis can help you determine how spurious harmonics in your noise spectrum will propagate a current into different circuits. When you compare the results with the specified noise floor in your system, you can then determine whether spurious harmonics should be filtered in order to meet your measurement requirements. You can also determine the strength of any spurious harmonics that propagate into downstream components (e.g., RF amplifiers) by looking at the bandwidth and noise figure of the downstream components.

If you need to take real steps to reduce the noise floor in your PCB, you can determine the right impedance matching conditions throughout your board with the right PCB design and analysis software. Allegro PSpice Simulator and Cadence’s full suite of analysis tools can help you examine the effect of noise in your circuits and help you identify the right design choices for reducing your noise floor.

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