Oscilloscope Basics

1 Background

    Modern electronic design faces more and more challenges. In the digital field, the integration scale of circuits is getting larger and larger, the number of IOs is increasing, and the density of single-board interconnection is increasing; at the same time, the clock rates inside and outside the chip are getting higher and higher, and the signal edges are getting faster and faster; new technologies are constantly emerging, such as: PCIe3, SATA3, USB3.0, HDMI, Fibre Channel, RapidIO, MHL, 10G~28G high-speed backplanes, etc., so the high-speed problems, signal integrity problems, and electromagnetic compatibility problems at the system and board levels are more prominent. In the field of RF and microwave, the emergence of new technologies and the expansion of bandwidth, such as: UWB, high-precision, broadband radar, bring more and more challenges to our system design.

    As the most commonly used test and analysis tool, the oscilloscope has also made great progress. There are two trends in the development of oscilloscopes. One is the improvement of performance. Since Agilent launched the digital oscilloscope in the 1980s, digital oscilloscopes have continued to develop, from hundreds of megabytes to thousands of megabytes, and then to 6GHz bandwidth, 20GSa/s sampling rate, and now to the ultra-high performance oscilloscopes with tens of GHz bandwidth and hundreds of GSa/s real-time sampling rate. The performance of oscilloscopes has achieved a leap forward. On the other hand, with the application of Windows operating system on oscilloscopes, the availability and software analysis capabilities of oscilloscopes have also achieved great development. For example, most of the current oscilloscopes use the open Windows 7 operating system and are equipped with a variety of test and analysis software, such as: jitter test and analysis software, serial data test and analysis software, PCIe and other consistency test and analysis software, and vector signal analysis software 89601B that can expand the analysis of oscilloscopes from time domain to frequency domain, demodulation domain, and digital domain. This has become another development trend of digital oscilloscopes.

Figure 1 Internal structure of a digital oscilloscope

 

   Figure 1 is a diagram of the internal structure of a digital oscilloscope. The internal structure of an oscilloscope mainly includes the following parts:

      1. Signal conditioning part: mainly composed of attenuator and amplifier;

      2. Acquisition and storage part: mainly composed of analog-to-digital converter ADC, memory controller and memory;

      3. Trigger part: mainly composed of trigger circuit;

      4. Software processing part: consists of a computer.

 

    After the signal enters the oscilloscope, it is first attenuated and then amplified. Why is this?

    It turns out that the attenuator is an adjustable attenuator. When the attenuation ratio is adjusted to a larger value, we can test large-amplitude signals. When the attenuation ratio is adjusted to a smaller value or 0dB attenuation, the amplifier can be used to test small-amplitude signals. When we usually adjust the vertical sensitivity of the oscilloscope, we are actually adjusting the attenuation ratio of the attenuator (some oscilloscopes have attenuators that are linked to amplifiers, with coarse adjustment being the attenuator and fine adjustment being the amplifier, to achieve better signal conditioning results).

    The signal conditioning circuit enables the ADC to perform analog-to-digital conversion ideally, which is reflected on the oscilloscope screen as the waveform displayed should be able to reach more than 2/3 of the screen (but not exceeding the screen).

    On the one hand, the amplifier amplifies the signal (bias adjustment is also achieved through the amplifier), and on the other hand, it provides a matching circuit to drive the ADC and trigger circuit. The amplifier determines the analog bandwidth of the oscilloscope, which is the first important indicator of the oscilloscope.

    After the signal passes through the ADC, the points need to be stored in the memory first. When the set memory is full, the sample points are then transferred to the computer. Why is this?

    Originally, the sampling rate of ADC is relatively high (for example, 20G samples per second), and each sample is represented by 8 bits (the ADC of modern digital oscilloscopes is usually 8 bits). The bus bandwidth behind the ADC reaches 160Gbps, which makes it impossible to transfer the samples to the computer in real time. Therefore, it is necessary to adopt the Block working mode, store the points first, and then slowly transfer the data to the computer after the storage is full. Moreover, this time is generally longer than the acquisition time, so the dead time of the digital oscilloscope is still relatively large (generally up to more than 95%). So how to ensure that the oscilloscope captures the signal we are interested in? This depends on triggering, which can solve the contradiction between acquisition and transmission.

    The second most important indicator of an oscilloscope is the real-time sampling rate, which is determined by the ADC. The third most important indicator is the memory depth, which is determined by the memory controller and memory. The fourth most important indicator is the trigger capability, which is determined by the trigger circuit.

Figure 2 90000A series oscilloscope signal acquisition board

 

    Figure 2 is the capture board of Agilent 90000A series oscilloscope (90000A oscilloscope includes 2 capture boards). The signal is connected to the front end of the capture board through an SMA coaxial cable. The front end includes an attenuator, an amplifier and a part of the trigger circuit. These devices are bare sealed on an MCM chip. The front end circuit drives two ADC chips. The sampling rate of each ADC chip is 20GSa/s. The two use cross-collection to achieve a sampling rate of 40GSa/s. Behind the ADC is the memory controller IDA, which performs data storage allocation and some operations, such as amplitude, phase compensation, trigger jitter compensation, time correlation operation, FFT operation, etc. IDA is connected to the computer through the PCI Express bus.

    So what other processing needs to be done after the data is transferred to the computer? Figure 3 is a block diagram of the computer processing structure.

Figure 3: Computer data processing structure diagram of oscilloscope

 

After the collected data is transferred to the computer, it must first be reconstructed by Sin(x)/x sinusoidal interpolation or linear interpolation. The reconstructed waveform can be used for various parameter measurements, signal operations and analysis, etc. The final result or the original sample point can be directly displayed on the screen.

Bandwidth and sampling rate are the most critical indicators for selecting and using an oscilloscope. So how do we quantify and calculate bandwidth and sampling rate? Please refer to Table 1.

Table 1 Quantitative criteria for selecting oscilloscope bandwidth and sampling rate

 

Figure 4: Gaussian frequency response of a traditional oscilloscope and Flat frequency response of an Infiniium oscilloscope

 

    So, what bandwidth should be selected for testing digital signals or pulses? First of all, we need to consider the bandwidth of the digital signal or pulse to be tested. The bandwidth of the digital signal or pulse to be tested is mainly determined by its edge. The calculation formula is:

Signal bandwidth BWsignal = 0.5/Tr (10% to 90%) or BWsignal = 0.4/Tr (20% to 80%)

Reference: Howard Johnson and Martin Graham, "High-Speed ​​Digital Design: A Handbook of Black Magic", Prentice Hall, 1993 (ISBN:0133957241), page 13. The text uses Fknee to represent the bandwidth of the signal, and Figure 5 is copied from it.

 

Figure 5 Calculation of Fknee frequency or bandwidth of a digital signal or pulse

 

    So how high a bandwidth does the oscilloscope need to capture all signals within the Fknee frequency? For Gaussian frequency response, because the -3dB bandwidth has a certain impact on the spectrum, in order to capture signals within the Fknee frequency, a bandwidth of more than 2 times the Fknee frequency (or more than 2 times the signal bandwidth) is required to ensure a small spectrum error and a small edge error. For a flat frequency response oscilloscope, because the impact within the -3dB bandwidth is relatively small, the oscilloscope bandwidth needs to be more than 1.4 times the signal bandwidth or the Fknee frequency.