IERI Procedia 4 ( 2013 ) 195 – 200

Available online at www.sciencedirect.com

2212-6678 © 2013 The Authors. Published by Elsevier B.V.Selection and peer review under responsibility of Information Engineering Research Institutedoi: 10.1016/j.ieri.2013.11.028

ScienceDirect

2013 International Conference on Electronic Engineering and Computer Science

Measurement Techniques for Spurious Free Dynamic Range of Analog to Digital Converter

Lijun Yang1, Xiaofeng Zhu1,*, Shuming Liu 2 1Beijing Institute of Graphic Communication,Beijing 102600,China

2Beijing Onsky Co., LTD, Beijing 100086, China

Abstract

SFDR is an important index to describe the ADC, but in practice it is difficult to accurately measure. The reasons are not only to understand the concept of problems also have test techniques, Measurement of experience for many years, with a comparatively easy to understand way describes the meaning of SFDR, is capable of accurately measuring the skill of SFDR, has the very good value for evaluating the quality of ADC, is very significant for evaluation of related products. © 2013. Published by Elsevier B.V. Selection and peer review under responsibility of Information Engineering Research Institute Key words: Spurious free dynamic range (SFDR); analog to digital converter (ADC); measurement techniques

1. The meaning of SFDR

SFDR(Spurious Free Dynamic range), Is a important index of ADC analog-to-digital conversion system, is the ratio of the useful signal power and maximum clutter signal power.

General use of DB, expressed as: SFDR=Ps-Pn, SFDR is the spurious-free dynamic range ( dB ), Ps is a useful signal power ( dBm ), Pn is the largest clutter signal power value ( dBm ).

* Corresponding author. Tel.: 010-60261166; fax: 010-60261160. E-mail address: zhuxiaofeng@bigc.edu.cn.

Available online at www.sciencedirect.com

© 2013 The Authors. Published by Elsevier B.V.Selection and peer review under responsibility of Information Engineering Research Institute

196 Lijun Yang et al. / IERI Procedia 4 ( 2013 ) 195 – 200

For the ideal system ADC, the useful signal can be considered as a full-scale input signal, no signal power to zero, so SFDR is infinite. The noise power of the actual system is not zero, interference various factors are finally reflected i†n the amplitude spectrum noise limited system, SFDR is of course limited.

For floating-point calculation by computer, can be thought of as a generalized data acquisition system, which floating process to the simulation data of the error, subsequent analysis of the digitized data calculation can introduce errors. Figure 1 is the FFT power spectrum of double precision floating-point sine data, from Figure 1 that the SF‡DR is about 274dB ( -6dB useful signal amplitude, maximum interference frequency range -280dB ), can also be 16 bits, 12 bit and 8 bit signed integer data corresponding to SFDR was about 104dB ( Figure 2), 78dB and 52dB[1-2].

Fig.1 FFT double precision floating point data normalized power spectrum

Fig.2 The sixteen bit signed integer data FFT normalized power spectrum

Data acquisition systems, sampling data to obtain the ADC analog-digital conversion not only contains quantization noise, also contains a variety of interference noise internal circuit system is introduced, the noise input signal may contain the external introduction, the total noise is above three kinds of noise as the result, therefore, the noise level of the real system to higher than the above software system, the spurious free dynamic range is reduced, figure 3 is a sixteen bit ADC data acquisition card collecting data FFT power spectrum, we can see the maximum noise level is about -65dB, SFDR is about 59dB [3-5].

The interference of ADC data acquisition system in practical application performance in particular spectral line, the noise spectrum are often higher than the white noise power level is much higher, level in Figure 3 times high spectral line is -65dB, much higher than the average noise platform -85dB. The noise in the -65dB spectrum, the noise power is not necessarily all from within the system, a part of power may come from external and independent, noise evaluation of a ADC digital-analog conversion system is one of the main task is to minimize the input signal contains a noise power, to calculate the SFDR index close completely from the conversion system.

197 Lijun Yang et al. / IERI Procedia 4 ( 2013 ) 195 – 200

Fig.3 ICS1555Sixteen bit data acquisition card FFT normalized power spectrum of the sampled data

In many data collecting and detecting system, the ratio of dynamic range of main parameters in system -- SFDR is the main signal and interference useful spectral line, describes the system one of the most important indexes. But the measurement is accurate, useful data is very difficult, such as the detection range of radar or sonar system. The results lead to related products quality inspection can not pass inspection, research work can not be completed successfully, the project can not be finished [6-7].

2. SFDR measurement techniques [8]

Methods and techniques presented here to solve the problem of several SFDR measurement should master, very practical value.

2.1. The input signal is close to full scale

The amplitude of the input signal should be as close as possible to the full range, dynamic range so as to effectively use the ADC sampling digit inherent, but not quite reach the full range, because nearly reached full-scale, analog-to-digital conversion of poor linearity, nonlinearity will produce noise is not necessary, the actual measurements, the need to repeatedly test the choice of an appropriate input signal level.

2.2. Selection of the test frequency

In order to accurately measure the useful signal power, the frequency of the input signal to accurately hit the FFT fixed line, otherwise the power of useful signal spread to many spectral line caused the useful signal power spectrum amplitude decreased. If you want to avoid the useful signal spectral diffusion, there is a constraint between the sampling frequency and the input signal and FFT points: Fs = N * Fclk NUM_FFT;

0 N NUM_FFT integer; Fs the frequency of the input signal; Fclk the sampling clock frequency; NUM_FFT FFT points. Meet the above relationship, the power input signal will be concentrated to a line, as shown in figure 4A is

a meet the above relationship for example, figure 4B is not satisfy the above-mentioned example. Tests using ICS1555 AD card, FFT for 1024 points, the input level is 1.5dBm, the sampling clock 160MHz, 4A input signal is fs=62.5MHz(N=400), 4B is fs=62MHz. We can see even the frequency difference between small, spectrum will obviously is diffusion, the spectrum of 4B is unable to obtain the accurate SFDR measurements.

198 Lijun Yang et al. / IERI Procedia 4 ( 2013 ) 195 – 200

Fig.4 The control spectrum concentration and diffusion

2.3. Signal generator using high quality as a test signal source

Using high quality signal generator as the input signal source, can effectively reduce the noise input is introduced, but due to the current signal source the best use of digital synthesis technology, white noise and harmonic noise so even signal generator best will be different levels of.

2.4. The use of an external clock source

General ADC board with fixed sampling or programmable clock, this clock source is generally composed of simple circuit, may bring more noise, to the analog-to-digital conversion so, to deal with internal and external clock in the actual test, if the external clock low noise is chosen the better performance of the external clock source.

2.5. The signal frequency and sampling clock to synchronize

The actual test, if the input signal and sampling clock is not homologous phase-locked, very hard to satisfy the above phase locking relationship, so the actual measurement, it is best to use the reference clock signal source (such as 10MHz )phase locking mechanism, let the signal and clock synchronization, it is possible to measure spectral amplitude values accurately.

2.6. On the power spectral readings

Even if the phase locking, there is slight frequency difference or contrast changes, then, the spectrum line is often not a line, observe the spectrum line, will think of a few dB loss, in fact, is due to the absence of the main spectral next to the line power calculation, cause, SFDR read, should be full of range calculation, for display bilateral spectrum, should calculate the useful signal power based on -3dB. Can also reduce FFT points, the power to concentrate to the single line.

199 Lijun Yang et al. / IERI Procedia 4 ( 2013 ) 195 – 200

2.7. Input signal increases filter to filter the input disturbance

Modern signal source is the digital synthesis, are digital signal source, noise mostly contains many harmonic components, the presence of these disturbances which cannot be accurate measurement of SFDR internal system, see the harmonic components of -65dB greatly reduces the recovery system of 16 ADC number SFDR in Figure 3, but the actual test results show this interference is introduced by the test signal input of the impure. The effective way is to join a narrowband filter, noise and harmonics removing signal source brings. Figure 5 is a comparison before and after the filter, wherein 5A is not add filters, FIG. 5B is increased by filter, noise suppression can be seen outside the band is about 18dB, the effect is very obvious. For clock source in using an external sampling clock source, the noise components will also give the sampling data into the noise outside, in the external clock circuit with narrowband filter similar can effectively filter out noise, improve the testing precision of SFDR.

Fig.5 Add filter and contrast without filter ( A without filter, B filter )

2.8. Reduced interference power fan and other auxiliary equipment

The system chassis backplane board power supply fan and other auxiliary equipment will produce harmful electromagnetic interference, which is harmful to system SFDR test, so it should be optimized for these devices, rational layout, take necessary measures against interference.

3. Conclusion

In actual measurement, the meaning of SFDR popular description of understanding of the measurement work, measurement techniques used in the front, can accurately measure the important index of data acquisition system SFDR. These methods are useful, practical, has the very big help to the relative measuring.

Acknowledgements

This work was financially supported by Institute Level key projects Funded by Beijing Institute of Graphic Communication(E-a-2012-31) and Funding Project for Academic Human Resources Development in Institutions of Higher Learning Under the Jurisdiction of Beijing Municipality [PHR(IHLB)] (PHR201107145).

200 Lijun Yang et al. / IERI Procedia 4 ( 2013 ) 195 – 200

References

[1] Qiongzhi Wu, Fang Nan, Feng Zhang.And Realization of DSP system software design parallel debugger [J]. Journal of Beijing Institute of Technology, 2011, 31(7): 855-858. [2] Xiaoyi Chen, Yingmin Wang. TigerSHARC Research and design of sonar signal processing system [J]. Underwater acoustic engineering, 2011,35(1): 60-63. [3] Xin Lu, Tongliang Zheng, Chao Xiong. Design and implementation of real-time accurate radar frequency measuring circuit [J]. Aerospace Electronic Warfare, 2011,27(2):32-34. [4] Gang Liu , Qihu Li. Application of TigerSHARC201 in sonar signal processing system [J]. Application of acoustics, 2009, 28(3): 208-213. [5] Song Yuan, Yun Liu, Jihong Wang. Research and improvement of DSP loading of nuclear TigerSHARC [J]. Computer and digital engineering, 2007, 35(5):53-55. [6] See CM S, Nehorai A. Source locat ion with partially cali-brated distributed electromagnetic component sensor array[ C]//IEEE Workshop on Statist ical Signal Processing, 2003: 458- 461. [7] Pesavento M, Gershman AB, Wong KM. Direction of arriv al estimat ion in partly calibrated time- varying senso rarrays[ C]//Pro c. ICASSP. Salt Lake City, UT, 2001:3005- 3008. [8] Lijun Yang, Xiaofeng Zhu, Fuping Liu, Shuming Liu. A Method for Testing Period or Rate of TigerSharc201 DSP Code Module Shortly and Accurately [J]. Journal of Beijing Institute of Graphic Communication, 2012, 20(2):51-54.