Clik here to view.
Even in the best designs, noise and interference sneak in to reduce the signal-to-noise ratio (SNR), obscure desired signals, and impair measurement accuracy and repeatability. Digitizing instruments like oscilloscopes and digitizers incorporate many features to characterize, measure, and reduce the effects of noise on measurements.
Interfering signals
Every measurement includes the signal of interest and a collection of unwanted signals such as noise, interference, and distortion. Noise and interference are generally unrelated to the signal being measured. Distortion is an interfering signal or signals related to the signal of interest, such as harmonics.
Noise is a random signal that is described by its statistical characteristics. Interference includes signals that are coupled into the measurement system by processes like crosstalk. Interfering signals are usually periodic in nature. Figure 1 shows an example of an interfering signal containing random and periodic components and some tools for characterizing the signal. The oscilloscope is triggered on the periodic element.
Image may be NSFW.
Clik here to view.
Figure 1 An example of an interfering signal with random and periodic elements. Source: Arthur Pini
The interfering signal contains both random and periodic components. The periodic component consists of 10 MHz “spikes”. The frequency at level (freq@lvl) measurement parameter (P4 beneath the display grid) reads the frequency of the spikes at approximately 70% of the signal amplitude to avoid noise peaks. Additionally, the mean, peak-to-peak, and rms levels are measured. Digitizing instruments, including oscilloscopes and digitizers, have a variety of tools to measure the characteristics of noise signals like this. They also offer a range of analysis tools to reduce the effects of these unwanted signal elements.
Instrument noise
All digitizing instruments also add noise to the measurement. Generally, instruments are selected where the noise is much lower in level and does not affect the measurement. Based on the measurement application, oscilloscopes with 8-bit or 12-bit resolution and digitizers with 8-bit to 16-bit or higher amplitude resolution can be selected to keep instrument noise within reasonable bounds.
Differential connections
When reducing noise and interfering signals, the digitizing instrument’s input is the place to start. A good starting point is using differential connections to reduce common mode signals. Many digitizers and a few oscilloscopes have differential inputs, while oscilloscopes commonly offer differential probes to connect the device under test (DUT) to the instrument.
Differential signaling transmits a signal using two wires driven by complementary signals. Noise and interference common to both conductors (common mode signals) are removed when the voltage difference between the two lines is calculated. The common mode rejection ratio (CMRR) measures the extent to which common mode noise is suppressed. Note also that the differential signal also does not require a ground return. In some cases, this also helps minimize the pickup of interfering signals. An example of differential signaling is the controller area network or CANbus, shown in Figure 2.
Image may be NSFW.
Clik here to view.
Figure 2 The two differential components of the CANbus (left side) and the resultant difference showing a reduction in common mode noise. Source: Arthur Pini
The two CANbus signal components are complementary, and when one is subtracted from the other, the common mode signals, like noise and interference, cancel. Note that the difference between the two components is a voltage swing twice that of the individual signals, providing a 6 dB improvement in SNR.
The differencing operation, either in a differential probe or a difference amplifier, reduces the noise common to both lines, allowing longer cable runs. In addition to CANbus, differential signaling is common in RS-422, RS-485, Ethernet over twisted pair, and other serial data communications links.
Common mode noise and interference can be further reduced in differential signals by using twisted pairs or coaxial transmission lines which provide additional shielding from the source of the interference.
Digitizing instrument tools to reduce noise and interference.
Oscilloscopes and digitizers can perform a variety of measurements and analyses on the interfering signal. Averaging will reduce the amplitude of the random component, and background subtraction can remove the periodic component from the waveform. Figure 3 shows an analysis of the interfering signal shown in Figure 1 using these tools.
Image may be NSFW.
Clik here to view.
Figure 3 Using averaging and background subtraction to separate an interfering signal’s random and periodic elements. Source: Arthur Pini
The interfering signal appears in the upper left grid. To the immediate right is the Fast Fourier Transform (FFT) of the interfering signal. The vertical spectral lines are related to the periodic component. The periodic narrow pulse train has a fundamental component of 10 MHz, which repeats at all the odd harmonic frequencies at a near-constant amplitude. The random element, which is spectrally flat and has equal energy at all frequencies, appears as the baseline of the FFT spectrum. The top right grid holds the histogram of the interfering signal. The random component dominates the histogram, which appears to have a bell-shaped normal distribution.
Averaging the interfering signal will reduce the random noise component. If the noise component has a Gaussian or normal distribution, the signal amplitude will decrease proportional to the square root of the number of averages. The average waveform appears in the center-left grid; note the absence of the random component on the baseline. The FFT of the average waveform is in the center grid, second down from the top. Note that the amplitude of the spectral lines is still the same but that the baseline is down to about -80 dBm. The histogram has a much smaller bell-shaped response due to the noise reduction. The range measurement of the histogram reads the amplitude from the maximum peak amplitude to the minimum valley amplitude or the peak-to-peak amplitude.
Subtracting the averaged background waveform from the interfering waveform as it is acquired will remove most of the periodic waveform. This process is called background subtraction. It works where the background signal is stable, and the oscilloscope can be triggered from it. The resulting waveform appears in the bottom grid on the left. The FFT of this signal is in the occupied bottom center grid. Note that its spectrum is mostly a flat baseline with an amplitude of about -68 dBm, the same level as the baseline in the original FFT. There are some small spectral lines at the harmonic frequencies of the 10 MHz periodic signal that were not canceled by the subtraction operation. They are less than ten percent of the original harmonic amplitude. The histogram of the separated random component has a Gaussian shape. Its range is lower than the original histogram due to the absence of the periodic component.
Using background subtraction with a real signal requires that the background is captured and averaged before the signal is applied. The averaged background is then subtracted from the acquired signal.
Cleaning up a real signal
Let’s examine reducing noise and interference from an acquired signal. The signal of interest is a 100 kHz square wave, as shown in the top left grid of Figure 4.
Image may be NSFW.
Clik here to view.
Figure 4 Reducing noise and interference from a 100 kHz square wave using averaging and filtering. Source: Arthur Pini
The interference waveform that we have been studying has been added to a 100 kHz square wave. The oscilloscope is triggered on the 100 kHz square wave. The FFT appears in the upper right grid. The frequency spectrum consists of the square wave spectrum with a spectral line at 100 kHz and repeated at all its odd harmonics, with their amplitudes decreasing exponentially with frequency. The 10 MHz interfering signal contributes spectral lines at 10 MHz and all its odd harmonics, which have a uniform amplitude across the whole span of the FFT. The random component raises the FFT baseline to about -70 dBm.
Averaging the waveform (second grid down on the left) removes the random component but not the periodic one. The FFT of the average signal (second down on the right) shows the 100 kHz and 10 MHz components as before, but due to the reduction in the random component, the baseline of the FFT is down to about -90 dBm. Averaging does not affect the periodic component because it is synchronous with the oscilloscope trigger.
Filtering can reduce noise and interference levels. This oscilloscope includes 20 MHz and 200 MHz analog filters in the input signal path. It also included six finite impulse response lowpass digital filters known as enhanced resolution (ERES) noise filters. The third grid down on the left, shows the signal filtered using an ERES filter. This is a lowpass filter with a -3 dB cutoff frequency of 16 MHz. The signal appears to be quite clean. The effects of the filter can be seen in the FFT of the filtered signal to the right. The low-pass filter suppresses spectral components above 16 MHz. While this works, you must be careful, low-pass filtering suppresses the harmonics of the desired signal and can affect measurements like those for transition times.
The six bandwidths available with the ERES noise filter vary with the instrument sample rate, limiting their usefulness. This oscilloscope also has an optional digital filter package that provides a greater range of filter types and cutoff characteristics, permitting the optimization of noise and interference reduction.
By background subtracting the filtered waveform from the acquired waveform, we can see what was removed by the filter (bottom left grid). The FFT (bottom right grid) shows the missing 10 MHz and 100 kHz harmonics.
Minimizing the efforts of noise with digitizing instruments
The key techniques for minimizing the effects of noise in measurements with digitizing instruments include differential acquisitions, averaging to reduce broadband noise, background subtraction, and filtering to reduce both noise and periodic signal interference.
Arthur Pini is a technical support specialist and electrical engineer with over 50 years of experience in electronics test and measurement.
Related Content
- Not all oscilloscopes are made equal: Why ADC and low noise floor matter
- Reducing noise in oscilloscope and digitizer measurements
- View noisy signals with a stable oscilloscope trigger
- Oscilloscope rise time and noise explained
- Using oscilloscope filters for better measurements
- Basic jitter measurements using an oscilloscope
The post Combating noise and interference in oscilloscopes and digitizers appeared first on EDN.