Editor’s Note:
In this DI, high school student Tommy Liu modifies a popular low-cost DIY oscilloscope to enhance its input noise rejection and ADC noise with anti-aliasing filtering and IIR filtering.
Part 1 introduces the oscilloscope design and simulation.
This part (Part 2) shows the experimental results of this oscilloscope.
Experimental Results
Three experiments were conducted to evaluate the performance of our precision-enhanced oscilloscope using both analog and digital signal processing techniques.
First, we test the effect of the new anti-aliasing filter described in Part 1. For this purpose, a 2-kHz sinusoidal signal is amplitude modulated (AM) with a 961-kHz sinusoidal waveform by a Rigol DG1022Z signal generator (Rigol Technologies, Inc., 2016) and is used as the analog input to the oscilloscope.
In this scenario, the low-frequency (2 kHz) sinusoidal waveform is our signal, while the high-frequency tones caused by modulation with 961 kHz sinusoidal represent high frequency noises at the signal source. In the experiment, a 10% modulation depth is used to make the high frequency noise easily identifiable by sight. The time division is set at 20 µs with the ADC sampling frequency of 500 KSPS.
Wow the engineering world with your unique design: Design Ideas Submission Guide
Results of anti-aliasing filter
The original DSO138-mini lacks anti-aliasing filter capability due to its insufficient -3-dB cut-off frequencies (around 500 kHz to 800 kHz). As a result, the high-frequency noise tones caused by modulations pass through the analog front-end, without much attenuation, and are sampled by the ADC at 500 KSPS. This creates aliasing noise tones at the ADC output and can be clearly seen in the displayed waveform on the DSO128-mini (Figure 1).
Figure 1 The aliasing noise tones at the ADC output on the DSO138-mini.
Our new anti-aliasing filter provides a significant lower -3-dB cut-off frequency of around 100 kHz, and effectively filters away most of the out-of-band high frequency noises, in this case, the noise tones caused by the signal modulation with 961 kHz sinusoidal. Figure 2 is a screenshot with the new anti-aliasing filter, indicating a significant reduction in the aliasing noise.
Figure 2 Reduction of the aliasing noise with the new anti-aliasing filter.
Detailed analysis on the captured data with the new anti-aliasing filter indicates a 10 dB to 15 dB (3.2x to 5.6x) improvement over the original DSO138-mini on noise rejection at frequencies higher than the oscilloscope’s signal bandwidth.
In practical applications, high frequency noises with a magnitude of a few millivolts RMS are not uncommon. A 5-mV RMS noise at near 900 kHz is attenuated to 0.73 mV (RMS) with our new anti-aliasing filter versus 2.48 mV (RMS) with the original DSO138-mini. With an ADC full-scale input range of 3.3 V, 0.73 mV RMS is of an effective resolution well above 10 bits (ENOB). With the original DSO138-mini, the ENOB would be at only an 8-bit level.
Results of digital post-processing filter
The second test evaluates the performance of the digital post-processing filter. As explained in Part 1, besides the noises at the analog input, other noise sources in oscilloscopes, such as noises on ADC inside the MCU damage the measurement precision. This is evident in Figure 3, which is a screenshot of the DSO138-mini with its Self-Test mode turned on. In Self-Test mode, an internally generated pulse signal—less susceptible to the noises from the external signal source—is used to test and fine tune the oscilloscope. We can see that there are still ripple noises on the pulse waveform.
Figure 3 Ripples on internally generated pulse signal during self-test mode on the DSO138-mini.
It is not easy to identify the magnitude of these ripples due to the limited pixel resolution of the DSO138-mini’s LCD display (320 x 240). We transferred the captured data to a PC via DSO138-mini’s UART-USB link for precise data analysis. Figure 4 shows the waveform of the captured self-test pulses on a PC. The ripple noises are calculated and shown in Figure 5.
Figure 4 Captured self-test pulse signal waveform on PC for more precision data analysis.
Figure 5 Magnitude of noises on self-test pulse with no digital post-processing.
Considering the voltage division setting (1 V, -20 dB on Input) and attenuation setting (x1), the ripple on the self-test pulse has a peak-peak magnitude of 8 mV. This error is about 10 LSB and the calculated RMS value is about 3 mV, yielding an effective resolution of 8.3 bits. Digital post-processing can be used to suppress some of these noises.
Figure 6 is the waveform after first-order infinite impulse response (IIR) digital filtering (α = 0.25) is performed on the PC, and Figure 7 shows the noises on the self-test pulse.
After IIR filtering, the noise RMS value reduces to about 0.75 mV, or by a factor of 4. This brings back the effective resolution from 8.3 bits to 10.4 bits. We notice that the rise and fall transition edges of the pulse look a bit less sharp than the signal before post-processing.
This is due to the low-pass nature of the IIR filter. With α=0.25, the passband (-3 dB) is at around 23 kHz, covering an input bandwidth up to audio frequencies (20 kHz). For tracking faster signals, such as fast transition edges of a pulse signal, we can relax α to a higher value allowing for more input bandwidth.
Figure 6 Self-test pulse with first-order IIR digital filter where α = 0.25.
Figure 7 Noises on self-test pulse with first-order IIR filter where RMS noise reduces to ~0.75 mV.
The effects of both filters
Finally, we test the overall effect of both the new anti-aliasing filter and the digital post processing by inputting a sinusoidal input of 2 kHz from a signal generator to our new oscilloscope. We can see from Figure 8 that even with the new anti-aliasing filter, there are still some noises on the waveform, due to the ADC noises inside the MCU. The RMS value of the noises is about 2.8 mV and the effective resolution is limited to below 9 bits.
Figure 8 Noises on a 2 kHz sinusoidal input waveform despite having the new anti-aliasing filter.
As shown in Figure 9, with the first-order IIR filter in effect, the waveform cleans up. The RMS noise reduces to 0.7 mV and, again, this brings up the effective resolution from below 9 bits to above 10 bits. Other input frequencies, up to 20 kHz (audio), have also been tested and an overall effective resolution of 10 bits or more was observed with the new anti-aliasing filter and the digital post-processing algorithm.
Figure 9 A 2 kHz sinusoidal input waveform after digital post-processing where the RMS noise reduces to 0.7 mV.
Low-cost oscilloscope
Many traditional low-cost DIY type digital oscilloscopes have two major technical drawbacks, namely inadequate anti-aliasing capability and large ADC noises. As a result, these oscilloscopes can only reach an effective resolution of 8 bits or less, even though most of them are based on an MCU, equipped with built-in 12-bit ADCs.
These problems limit DIY oscilloscopes from more demanding professional high school projects. To address these issues, a well-designed first-order analog low-pass filter at the analog front-end of the oscilloscope, plus a programmable first-order IIR digital post-processing filter, are implemented on a popular low-cost DIY platform (DSO138-mini).
Experimental results verified that the new oscilloscope could maintain an overall effective resolution of 10 bits or above with the presence of high frequency noises at its analog input, up to an input bandwidth of 20 kHz and real-time sampling of 1 MSPS. The implementations are inexpensive—the BOM cost of the new anti-aliasing filter is just the cost of a ceramic capacitor (far less than a dollar), and the digital post-processing program is completely implemented in the PC software.
Costing less than fifty dollars, this precision digital oscilloscope can be used in many high schools. This includes high schools without the funds for pricey commercial models and, thus, enable students to perform a wide range of tasks: from the first-time electrical signal capture and observation to the more demanding precision measurement and signal analysis for complex electrical and electronic projects.
Tommy Liu is currently a junior at Monta Vista High School (MVHS) with a passion for electronics. A dedicated hobbyist since middle school, Tommy has designed and built various projects ranging from FM radios to simple oscilloscopes and signal generators for school use. He aims to pursue Electrical Engineering in college and aspires to become a professional engineer, continuing his exploration in the field of electronics.
Related Content
- Building a low-cost, precision digital oscilloscope—Part 1
- Build your own oscilloscope probes for power measurements (part 1)
- Build your own oscilloscope probes for power measurements (part 2)
- Basic oscilloscope operation
- FFTs and oscilloscopes: A practical guide
The post Building a low-cost, precision digital oscilloscope – Part 2 appeared first on EDN.
