High speed ADC is an indispensable part of signal processor, and its performance is also very important to the overall performance of signal processing system. Usually, the technical parameters of ADC are provided by the manufacturer and can be used as an important basis for design. However, the performance of ADC module formed on the circuit board is also closely related to the peripheral circuit or input signal of ADC, such as reference voltage source, sampling clock, input operational amplifier, interference on power supply, ground wire and signal line. Therefore, it is necessary to evaluate the dynamic performance of high-speed ADC module online and analyze its impact on the performance of signal processor system. This paper introduces a method of on-line evaluating the dynamic performance of high-speed ADC module on the actual equipment of signal processor. This method uses the data acquisition ability of the signal processor itself, adds the standard test signal at the analog input, reads the conversion result by the DSP of the signal processor after AD conversion, reads the data into the PC through the DSP simulation system, then uses the MATLAB software to analyze the spectrum of the data, and finally calculates several dynamic parameters such as SNR and SENAD. This paper also gives the specific test results and analysis, and gives some opinions on the design of high-speed ADC module.
1、 Dynamic performance parameters of high speed ADC
The main parameters for evaluating the dynamic performance of ADC are defined as follows:
1. Signal to noise ratio (DB)
Where asignal is the root mean square value of full amplitude sinusoidal analog input signal, and anoise is the root mean square of the sum of all noise sources.
2. Signal to noise distortion ratio (SINAD)
Where aharmonic is the sum of the root mean square of the frequency components of each harmonic (except DC).
3. Significant bits (ENOB)
Where n is the quantization bit number of the conversion circuit, a measurement error is the average value of the measurement noise, and a measurement error is the average value of the quantization error.
4. Total harmonic distortion, where AF_ In is the root mean square value of the fundamental wave of the input signal, ahd_ 2 down to ahd_ N is the root mean square value of the second to Nth harmonic components in the frequency domain of the sampled signal.
5. Spurious free dynamic range, where AF_ In is the root mean square value of the fundamental component of the input signal, ahd_ Max is the root mean square value of the maximum distorted harmonic component or the maximum spurious signal in the spectrum of the sampled waveform.
Through the definition of these parameters, it can be seen that most of the dynamic parameters of high-speed ADC circuit can be expressed in frequency domain; Therefore, the corresponding dynamic performance parameters can be obtained by testing the high-speed ADC circuit in frequency domain.
2、 Traditional test methods
In the traditional test, a DAC whose accuracy is more than 2 bits higher than that of the tested analog-to-digital conversion circuit is used to generate a single frequency sine wave as the test signal of the tested analog-to-digital conversion circuit, and a DAC is connected at the back end of the tested circuit to recover the waveform. As shown in Figure 1.
The test structure is simple and intuitive. However, in engineering practice, in order to evaluate the analog-to-digital circuit, a DAC circuit must be added, which will make it difficult to connect with the actual module and introduce the error of DAC. Therefore, online evaluation is not easy to adopt.
3、 Test method based on DSP Technology
DSP technology can be used to form a simple and accurate test structure. At the back end of ADC circuit, digital signal processor is used to collect and save the output data, and then the JTAG interface of DSP simulation equipment is used to transmit the data to PC, and MATLAB software is used for relevant frequency domain processing, so as to obtain the actual conversion characteristic parameters of high-speed analog-to-digital conversion circuit. The structure is shown in Figure 2.
Here, the analog test instrument is replaced by the frequency domain analysis tool based on DSP, which can provide higher precision and repeatable test results. After the data is collected into PC, the discrete signal waveform in time domain is converted to frequency domain through DFT (FFT) algorithm on the platform of MATLAB. The relevant dynamic performance parameters are obtained according to the definition in the frequency domain. Under the ideal condition that the input signal has no distortion, when the input analog quantity is sine wave, the output spectrum should be the impulse function graph whose frequency is equal to the input frequency. In fact, the quantization error of ADC, various noises inside the converter, and even the noise of the test system will be reflected in the noise background on the spectrum. The basic function of signal analysis based on FFT is FFT itself and power spectrum. In the FFT algorithm, it is assumed that the discrete-time series can accurately carry out periodic continuation in the whole time domain. All signals containing the discrete-time series are periodic functions, and the period is related to the length of the time series. However, if the length of the time series is not an integral multiple of the signal period, that is, spectrum leakage will occur. Here is the input signal frequency; Fsample is the sampling frequency; Length of nwndows window function; Nrecord sampling signal data length. Hanning window function is generally used in the test to reduce spectrum leakage.
After the spectrum of the sampled output signal is calculated, the position of the fundamental frequency and each harmonic of the signal must be determined in order to obtain the relevant test parameter values. The relationship between the frequency component and its image is given below, as shown in Figure 3. Let fo be a frequency located in the Nyquist interval (DC ~ FS / 2). Let FK be the mirror image of fo in the frequency band, also known as the k-th mirror image of fo.
Fig. 3 Schematic diagram of frequency component and its image component
The relationship between all mirrors can be expressed as follows:
From the above relationship, it can be seen that for odd and even sub image components, their corresponding frequency components can be found in the Nyquist interval, that is, for any harmonic component beyond the Nyquist interval, its image components can be found in the Nyquist interval. If the harmonic is located in the Nyquist interval, the harmonic position can be calculated directly; If the harmonic exceeds the Nyquist interval, the image method above should be estimated to obtain the harmonic data.
4、 ADC module test based on signal processor
We have carried out on-line test on the high-speed ADC module of the circuit board of the signal processor. The part surrounded by the virtual frame in Fig. 4 is the analog-to-digital conversion circuit of the signal processor. Dsp21065l stores the collected data in the internal storage area. After the collection, the data is transmitted to PC through the JTAG interface of DSP simulation system. Finally, the relevant digital signal processing is carried out in MATLAB, and the conversion performance of the conversion circuit is evaluated and analyzed according to the processing results. According to the performance index of the analog-to-digital conversion circuit, the influence of the conversion circuit on the signal processor can be analyzed.
In the figure, the AD conversion chip ad9225 has 12 bit accuracy and the maximum conversion rate is 25msps. The chip is equipped with a sample and hold circuit and a reference voltage source. The basic performance is as follows:
Accuracy inl: DNL:
4.1 partial test data
The test signal is full amplitude sine wave. 8192 points of data are sampled each time, and 8192 points of FFT is done to obtain the spectrum of sampled data. Under different sampling frequency and signal frequency, do 5 experiments respectively and take the average value.
4.2 test data analysis
The theoretical signal-to-noise ratio of the conversion circuit with 12 bit accuracy should reach 74dB, while it is only 60 ~ 70dB in the actual test. It indicates that there are other noise sources except quantization noise. In the experiment, it is found that if the analog power supply and digital power supply in the conversion circuit are powered by the same power supply, the signal-to-noise ratio will be reduced by about 5 ~ 6dB than when they are powered respectively. Improving the grounding condition can also effectively improve the conversion dynamic performance. For example, in the case of poor grounding condition, the signal-to-noise ratio of 5MHz sampling at 1.54mhz is only 59.7db, while it is increased to 61.64db and 1.94db after improving the grounding condition. The main reason is that the purity of the harmonic signal is relatively low. The test results are close to the parameters given in ad9225 technical manual, which shows the reliability of this method. From the SNR value, because the test is aimed at the whole high-speed analog-to-digital conversion circuit, the results reflect the dynamic conversion characteristics of the whole circuit. Due to the influence of each auxiliary circuit, the performance of the conversion circuit is lower than the parameters given in the device manual. According to the analysis of test results, the following reference opinions are put forward for the design of high-speed analog-to-digital conversion circuit:
L the data latch is used for isolation drive between the back-end of the converter and the data bus to reduce the noise on the back-end data bus, which is cascaded into the analog circuit and reduce the conversion performance.
L analog power supply and digital power supply shall be isolated to reduce the noise at the analog end.
L select appropriate operational amplifier to reduce harmonic distortion.
L analog input signal line shall be as wide as possible to reduce signal distortion.
L during device selection. The conversion accuracy of the selected device shall be more than 2 significant digits higher than the required accuracy.
L the conversion rate of the selected ADC chip should be much higher than the sampling speed.
The on-line evaluation method for the dynamic performance of high-speed ADC module of signal processor introduced in this paper has the characteristics of simplicity, high reliability and easy to be used in engineering. It has a good reference value for the evaluation and analysis of the performance of the signal processor system.
Responsible editor: GT