1. Understanding the signal source
The instrument used to generate various test signals is called a signal generator or simply a signal source. It can be used as various simulation signals or excitation signals and is widely used in the testing of various complete machines, systems, components, and devices. For example, a signal source is used to generate a certain modulated signal and input it to the receiver to test the receiver performance; when we demonstrate the LeCroy oscilloscope to customers, we often bring some signal sources, such as using LeCroy's arbitrary waveform generator ArbStudio to generate some special signals, which is convenient for demonstrating the various functions of the oscilloscope. There are
many types of signal sources. In terms of the characteristics of the generated signals, there are sine signal generators, function generators, arbitrary waveform generators, etc. The importance of sine signals to the frequency domain analysis of linear systems makes sine signal generators widely used. Users usually require this type of signal source to have a wide frequency range, high frequency accuracy and stability, high spectrum purity, and low phase noise. For example, the sine signal generator required in the communication system test generally requires that the frequency can be extended to the radio frequency band and have various modulation functions. The implementation principles of the sine signal generator are generally phase-locked technology and frequency synthesis technology.
A function generator is a signal source that can generate sine waves, square waves, and triangular waves. A traditional function generator uses a constant current source to charge and discharge a capacitor, and the voltage across the capacitor is a triangular wave. If the triangular wave is sent to a comparator, a square wave can be generated. The triangular wave can also generate a sine wave through a waveform shaping circuit. By changing the current and the size of the capacitor, the signal frequency can be adjusted. This signal source generally has a low output frequency, low frequency accuracy and stability. With the development of digital technology. The implementation of function generators has gradually evolved from analog to digital. For example, the DDS technology to be introduced later is used to generate signals, which also makes the function generator gradually merge into an arbitrary waveform generator.
Both sine signal generators and function generators can only generate regular signals. To generate irregular signals, an arbitrary waveform generator (AWG, Arbitrary Waveform Generator) is required. The basic design concept of AWG is: to intercept a cycle of the signal waveform to be reproduced for uniform sampling and save it in the memory. The waveform data in the memory is read out in sequence, converted by DAC, and then filtered to obtain the required waveform. In principle, AWG and digital storage oscilloscope can be considered as an inverse process. Digital storage oscilloscope digitizes analog waveforms through ADC, and AWG converts digitized waveform data into analog waveforms through DAC. Both types of instruments are subject to the Nyquist law, and the highest frequency component that can be measured/output does not exceed half of the ADC/DAC sampling rate. You can use LeCroy oscilloscope to collect a waveform and save it as a file. Import the waveform file into LeCroy arbitrary waveform generator ArbStudio to restore the analog waveform. In practical applications, the waveform data used by AWG is not all obtained by real sampling, but is usually generated with the assistance of software.
2. AWG principle
According to the specific implementation principle, AWG can be divided into two categories: DDS-based and True Arb.
1. DDS-based AWG
DDS, namely Direct Digital Synthesis, is a method of frequency synthesis. As mentioned above, sine signal generators generally use frequency synthesis technology. So what is frequency synthesis? The
signal source needs to use an oscillator. Generally, the output frequency range of the oscillator is limited, and it is difficult to obtain high stability in a wide frequency range. Then it is necessary to derive various required frequencies from an oscillator source (such as a crystal oscillator) with a single frequency but high accuracy and stability. This method of deriving multiple frequency signals from one frequency is called frequency synthesis. For example, a single frequency signal output by an oscillator can be used to implement frequency addition, subtraction, multiplication and division operations using frequency multiplication, frequency division and frequency mixing techniques to synthesize a series of frequency signals. These frequencies are all related to the oscillator frequency (called the reference frequency). This method is called direct analog synthesis. In addition, there is indirect frequency synthesis based on PLL.
DDS is another important frequency synthesis technology. Let's first look at how to generate a sinusoidal signal using DDS technology. We know that the frequency of a sine wave can be expressed as f=ω/2π=Δθ/(2πΔt). Δθ is the phase increment of a sinusoidal signal after a period of time Δt. In the same time Δt, the larger the phase increment, the higher the angular velocity and the higher the oscillation frequency. DDS synthesizes the desired frequency based on this relationship between frequency and angular velocity. Please see the following DDS principle block diagram:
[page]
The phase accumulator is used to generate a sequence representing the phase. It uses the reference clock fs as the beat and K as the accumulated value to generate a linearly growing phase sequence. For example, the phase represented by the initial value of the phase accumulator is 0, and the accumulated value corresponding to K is π/10. Then, under the stimulation of the reference clock, the accumulator generates a sequence representing the phases of 0, π/10, 2π/10, 3π/10…18π/10, 19π/10, 2π, 0, π/10,…, etc. The sequence
representing the phase is used as the address of a sine wave lookup table. The lookup table realizes the conversion from phase to amplitude and outputs the sine wave amplitude value corresponding to the phase. At this time, the amplitude is still a digital sequence. After passing through the DAC and the low-pass filter, a smooth sine wave is output. The reference clock is also the output clock of the lookup table and the sampling clock of the DAC.
K is the phase increment Δθ, and the reference clock cycle 1/fs is the time increment Δt, that is, f=Δθ/(2πΔt)=K*fs/2π. In other words, the output signal is coherent with the reference clock fs, and the frequency is controlled by the phase increment K. For example, when the phase increment is π/10, 20 reference clock cycles can output a sinusoidal signal of one cycle, that is, the frequency is fs/20; if the phase increment is increased to π/5, a cycle of scanning can be completed in a shorter time, and the frequency is therefore increased, that is, fs/10.
Extended to the case of arbitrary waves, if the lookup table stores the quantized data of any waveform, then the output signal will be a repetitive signal with the waveform as one cycle. The repetition frequency is also controlled by the phase increment.
From the principle of DDS, it can be seen that if the output frequency is changed while the reference clock remains unchanged, not all data points in the memory can be output. The higher the output frequency, the larger the phase increment required, and the more data points are skipped. This may affect the signal fidelity.
For example, if the reference clock is 100MHz, the memory capacity is 100pts, and the sampling point data of one cycle of the output signal is stored.
If the output frequency is required to be 1MHz, all sampling points can be output in sequence at
each sampling clock beat to meet the requirement. If the output frequency is required to be 2MHz, the phase increment must be increased by 1 times, that is, one sampling point needs to be skipped at each sampling clock beat.
If the data point reflecting the signal transient happens to be in the skipped sampling point, the output signal fidelity will be impaired. See the figure below:
If the output frequency is required to be 300kHz, which is not an integer multiple of the reference clock. Then not only will some data points be skipped, but the data points output in this cycle will be different from those output in the next cycle, as shown in the figure below. The output signal may have phase truncation and frequency leakage. Although the AWG based on the DDS principle has some of the above disadvantages, DDS has the advantages of low cost, low power consumption, high resolution and fast frequency conversion speed, and is suitable for outputting frequency modulation, phase modulation and swept frequency signals. Moreover, by reasonably setting the relationship between the reference clock, phase increment and output frequency, the signal fidelity can meet the test requirements. 2. True Arb True Arb means true arbitrary waveform generator, which is relative to the characteristic of DDS that some data may be lost. The figure below is the principle block diagram of True Arb: [page] True Arb reads quantized data from the waveform memory with a variable sampling clock, and then outputs it after passing through the DAC and low-pass filter. If the output signal frequency needs to be changed, the frequency of the sampling clock is adjusted. True Arb is closer to the reverse process of a digital storage oscilloscope: 1) The sampling rate of a digital storage oscilloscope can be changed, and the sampling clock frequency of True Arb can also be changed. 2) Constraints of digital storage oscilloscope: Storage capacity = sampling rate × sampling time Constraints of True Arb: Storage capacity = sampling rate × signal repetition period For example, the storage capacity of AWG is 2Mpts. If the sampling rate is set to 100MS/s, the maximum repetition period of the output signal is 20ms, that is, the minimum repetition frequency is 50Hz. If you want to output a signal with a lower frequency, you must reduce the sampling rate. It can be seen that the method of changing the output frequency by changing the sampling clock can ensure that all data in the memory are output, and the signal fidelity is better than the DDS method. In addition, the True Arb memory can be logically divided into several segments, and the waveform data of each segment can be output separately and repeatedly, which is the so-called waveform sequence function. For example, if the memory depth is 1Mpts, it can be divided into 3 segments, which can store 500kpts, 400kpts, and 100kpts of data respectively. You can define that after the first segment waveform is repeated 3 times, the second segment waveform is output once, the third segment waveform is repeated 2 times, and then it returns to the first segment to start a new cycle... The advantage of this function is that you can design more complex waveforms and make the most of the memory capacity. For example, you need to generate a signal that first outputs a pulse with a width of 25us and an amplitude of 3V, keeps a low level for 20s, then outputs a pulse with a width of 25us and an amplitude of 5V, and then completes a repetition after 50s of low level, and starts the next cycle. For such a signal, a relatively high sampling rate must be used to reconstruct a 25us pulse, but the signal repetition period is long. The constraints of True Arb determine that under the limited memory capacity, it is impossible to save long-term waveform data at a high sampling rate. At this time, the waveform sequence function can be used. For example, the first segment stores a 3V pulse and outputs it once. The second segment stores a 50us low level and outputs it 400,000 times, with a total duration of 20s. The third segment stores a 5V pulse and outputs it once. The third segment stores a 50us low level and outputs it 1,000,000 times, with a total duration of 50s. As shown in the following figure: III. AWG performance and functions Generally, an AWG is evaluated by the following performance and functions:
Signal Integrity and Power Integrity Analysis (Eric Bogatin)
京公网安备 11010802033920号