Design and Construction of an Arbitrary.docx
《Design and Construction of an Arbitrary.docx》由会员分享,可在线阅读,更多相关《Design and Construction of an Arbitrary.docx(6页珍藏版)》请在淘文阁 - 分享文档赚钱的网站上搜索。
1、398 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. IM -32, NO. 3, SEPTEMBER 1983 Design and Construction of an Arbitrary Waveform Generator DANIELE D. CAVIGLIA, MEMBER, IEEE, ALESSANDRO DE GLORIA, STUDENT MEMBER, 丨 EEE, GIULIANO DONZELLINI, GIANCARLO PARODI, AND DOMENICO PONTA Abstract 一
2、 The purpose of this paper is to describe the design and characteristics of an arbitrary waveform generator, built at the Biophysical and Electronic Engineering Division of the Electrical Engineering Department of the University of Genoa, which exhibits the following features: a) The signal waveform
3、 is any band-limited periodic signal, whose frequency content ranges from 0.125 mHz up to 30 MHz. The low- frequency limit has been chosen in order to meet the needs of bioengineering, mechanical, and chemical applications. The high-frequency limit covers, for instance, TV and image processing appli
4、cations. b) The noise level is at least 66 dB below the maximum signal level. c) Typical distortion is less than 66 dB. d) The generators performance can be easily improved, depending on the availability of up-to-date solid-state devices, without changing its design. I. INTRODUCTION T HE AVAILABILIT
5、Y of arbitrary waveform generators is a frequent need in several applications. Most available waveform generators are designed for special-purpose appli- cations, hence they can deliver a limited class of signals only. The main requirement of a truly general-purpose waveform generator is full progra
6、mmability of the waveform, without the need for any special expertise by the user. The easiest way to attain such a goal is to use a microcomputer programmed to control each instrument function and the operator dialogue. In our case, after the user has entered, through a TTY/CRT terminal, the signal
7、 function f ( t ) that must be generated, a software routine computes the signal samples f ( t i ) over the signal period at a suitable sampling rate (max 40 MHz). The signal samples are stored at adjacent locations of a high-speed memory, whose contents are cyclically addressed to drive a D/A conve
8、rter, which generates the output waveform. Accordingly, the system can be divided into four blocks: a microcomputer, which provides the computing facility and controls the whole operation of the generator; a high-speed memory; a memory addressing system, and a D/A converter and an output amplifier (
9、Fig. 1). Manuscript received April 27, 1982. The authors are with the Biophysical and Electronic Engineering Division of the Electrical Engineering Department, University of Genoa, Genoa, Italy. Fig. 1. General block diagram of the arbitrary waveform generator. II. SYSTEM DESIGN The two main error s
10、ources in the output signal u ( t ) , as compared with the input function/( , are due to quantization and sampling. The corresponding effects on u ( t ) are evaluated separately in order to obtain, from distortion and noise requirements, the characteristics of the signal synthesizer which allow us t
11、o meet them. We express, as usual, the quantized signal as the sum of the ideal signal /(/) plus a quantization error signal e ( ( ) . If the conditions described in 1, 2, are fulfilled (i.e., moderate correlation through successive samples; large number of output points close to the midpoint of the
12、 corresponding quantization intervals), it can be proven that e ( t ) is almost uncorrelated to /(/), and that its spectrum is quasi-white. The influence of this error on the output is measured by the signal-to-noise ratio due to quantization SNR = 0 o g a / ( j l ) (1) where aj is the variance of t
13、he input signal and a 2 e is the variance of the error signal, i.e., the output mean square distortion due to quantization. The signal characteristics are fully known, so that, at least 0018-9456/83/0900-0398S01.00 1983 IEEE CAV1GLIA el ai: DESIGN AND CONSTRUCTION OF A WAVEFORM GENERATOR 399 in prin
14、ciple, the evaluation of (1) is feasible for each/(/), but it is not practical. On the other hand,/( can be any arbitrary waveform so that we cannot give a general expression for (1) unless we limit the class of functions to be synthesized 1. Therefore we introduce a more practical approach letting
15、SNR, SNRP = 20 log (Ypp/A) (2) where Ypp is the peak-to-peak output signal amplitude and A is the quantizing interval, which is assumed to be constant over the whole dynamic range of the instrument. For an Ai-bit quantization it is A = Ym/2n, where Ym is the output full scale. The output amplitude i
16、s a fraction 1 /r = Ypp/Ym 1 of the full scale so that SNRp = 6Mn - 20 log r. (3) In order to show the usefulness of relationship (3) we compute SNR according to (1) for a sinusoidal waveform f(t)= 0. 5 Ym sin (oot). We can obtain the corresponding signal- to-noise ratio due to quantization from the
17、 corresponding input variance, which becomes Fm2/8, and from the output mean square distortion. For its evaluation we can consider the error noise as a uniformly distributed random variable over the interval -A/2 to +A/2; this leads to - A2/12 = Ym2/2 22n). Hence we obtain SNR, = 6.02n + 1.76 (4) sh
18、owing that the expression (2) is a very good approximation of (1). Of course, the minimum of (3) is reached for n constant and for r = 1. In any case, the dependence of SNR on n is linear 1. For the readers convenience, we have computed, in Table 1, the value of SNRp for some values of n and r of pr
19、actical interest. It becomes apparent that, in order to obtain, for instance, SNR 60 dB, it is necessary to employ at least 12 bits. The contribution of sampling to output noise can be estimated using again a sinuosoidal waveform f(t) = Ys sin cot + p). Sampling theory offers the theoretical backgro
20、und for the synthesis of any band-limited signal: the sampled signal with a zeroth-order holding 3 is given by KA(,) = (Ky/7) E m TABLE I. SIGNAL TO QUANTIZATION NOISE RATIO FOR VARIOUS BIT NUMBERS AND REDUCTION FACTORS r SA (dB) 8 10 12 1 48 60 72 2 42 54 66 5 34 46 58 TABLE II HARMONIC AMPLITUDES
21、OF A SAMPLED SIGNAL FOR VARIOUS SAMPLE NUMBERS N AND FOURIER INDEXES m m N-10 haem* | Am| N-100 harm* | Am| N-1000 harm. |Am| 0 fund* .9836 fund* .9998 fund* 99998 -1 9 1093 99 0101 999 00101 1 11 .0894 101 0099 1001 00099 -2 19 .0518 199 0050 1999 .00050 2 21 .0468 201 0050 2001 00050 When using 40
22、96 memory words, which require 12 bits for the address, the most important harmonic is the 4095th, with an amplitude coefficient A- = 2.4420 X 10-4. This is the main contribution to distortion, together with the 4097th harmonic (|+i | = 2.4408 X 10-4). The corresponding sig- nal-to-noise ratio is SN
23、R, = 20 log (lol/l.j |) = 72 dB. (7) With such an oversampling, the harmonic distortion is not larger than the quantization noise so far obtained. This result has been achieved by using as many bits for the memory words where the samples are stored as for their addresses. We conclude that the implem
24、entation of a system which satisfies the above SNR condition, requires that the waveform to be synthesized be defined in a “square lattice” of 2” X dimensions, where n is the number of bits needed for both the sample word and the address word. fTh sin (co + moos)(t - r) + ipdi) (5) where Ts = lir/oo
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- Design and Construction of an Arbitrary
限制150内