Method and apparatus for alias suppressed digitizing of high frequency analog signals

[0001] 1. Field of the Invention

[0002] The present invention relates generally to the field of analog signal digitizing. More specifically the present invention is addressed to a problem of alias suppressed signal sampling at a frequency considerably lower than the upper frequency of the said analog signal. A method and apparatus covered by the invention are adapted to digitizing the high frequency signals in applications related to versatile digital signal processing (e.g. oscilloscope).

[0003] 2. Description of Prior Art

[0004] For many types of the analog-to-digital (A/D) converters, the upper frequency boundary F_BW of the front-end bandwidth exceeds the allowed maximum sampling rate F_S considerably. For example, the A/D converter AD9433 from “Analog Devices, Inc.”, at the bandwidth F_BW=700 MHz, is characterized by the maximum sampling rate F_S=125 MHz. Therefore, in the cases when such A/D converters are incorporated into electronic designs based on exploitation of the conventional equidistant (uniform or periodic) sampling, the upper frequency of the analog signals at the input of the said converter has to be limited down to the level equal to F_S/2 (Nyquist limit) to protect the digitized signal against aliasing.

[0005] The aliasing phenomenon is well known. The essence of it is the fact that in the cases where the original analog signal has frequencies exceeding half of the sampling rate the frequency overlapping occurs so that different frequencies might be represented by exactly the same data set and, generally, the digitized output signal bears no real relationship to the actual input signal. Therefore, to avoid aliasing in the case of the mentioned example, the upper frequency of the input analog signals has to be limited down to 62.5 MHz, to the frequency 11 times lower than the offered bandwidth of this A/D converter. In other words, the traditional approach to sampling and processing of the equidistantly sampled signals often leads to substantial under-exploitation of the bandwidth resource of the involved A/D converter.

[0006] It is known that signal sampling at predetermined non-uniformly spaced time instants can be applied for sampling an analog signal with aliasing substantially suppressed even when the mean sampling frequency F_SA is essentially lower than F_BW (see [1–2]). In this case, the upper frequency F_BW of input analog signal spectra is limited by the minimal digital increment D of the sampling interval variation, so that F_BW<½D.

[0007] However the said known methods and means for non-uniform signal sampling at predetermined time instants are oriented to fully digital implementations based on application of a special digital device controlling the sampling process at a given clock frequency F_clk in such a way that the achievable minimum value of D, determining the upper frequency of the input signal spectrum, is directly related to the maximum of the clock frequency F_clk at which the involved logic circuits still can function properly. For example, this maximum clock frequency F_clk should be at least 1.4 GHz to achieve the possibility of exploiting the whole 700 MHz bandwidth of the mentioned A/D converter. As bandwidths of the present-day A/D converters can be even much wider than 700 MHz, suppression of aliasing within the whole frequency range of their bandwidths, if attempted on the basis of the said non-uniform sampling method, would require usage of substantially higher clock frequencies measured in GHz. That clearly is not acceptable as the corresponding electronic implementations then, even when possible, would be very complicated and expensive.

[0008] Thus, to achieve alias suppression at digitizing wideband analog signals, the problem of reducing the value of the smallest clock interval digital increment D to sufficiently small values has to be resolved in a way providing that the value of D does not directly depend on the clock frequency F_clk used. For digitizing signals with the upper frequencies in hundreds of MHz and even GHz, the values of D, evidently, have to be as small as a few tens or hundreds of picoseconds.

[0009] The present invention addresses this problem of increasing the upper frequency of the analog signal being digitized in an alias suppressing way by sampling that input signal at time instants corresponding to the pseudo-randomly selected clock pulses which are additionally delayed for pseudo-random values. This results in the signal upper frequency restricted only by the minimum increment of the pulse delay variation.

[0010] This invention provides, a method and apparatus for alias suppressed digitizing of high frequency analog signals, comprising variable-variance sampling of the said signals at mean frequencies considerably lower than the upper frequencies in spectra of the said signals. A clock produces a sequence of electrical pulses at a predetermined frequency F_clk. This sequence is divided by a pseudo-random integer A to select one pulse from every series of A clock pulses. A digitally controllable delay block delays the selected pulse by a pseudo-random delay D×B, where D is a constant increment of the delay and B is a pseudo-random integer. An analog-to-digital (A/D) converter samples the input signal at the time instant of the said delayed pulse. A memory receives and stores the digital sample value from the output of the A/D converter. Said integers A and B are changed after every sampling event.

[0011] A primary object of the present invention is to provide a method and apparatus to substantially increase the upper frequency of alias-free analog signal digitizing.

[0012] Another object of the present invention is to provide a method and apparatus to adapt the primary object to input signal digitizing at the highest sampling rate.

[0017] FIG. 2 provides a simplified block diagram illustrating the manner in which the input signal is sampled and the obtained digital sample values stored in the memory of the prior art embodiment used for implementation of the conventional approach to alias suppressed signal sampling.

[0018] To obtain a sampling pulse, the sequence of the clock pulses is digitally divided by a predetermined pseudo-random integer A so, that one last pulse is selected from each series of A clock pulses. The variable A assumes a value from K to K+N, where K is the constant part of the integer A determining the minimum K/F_clk of the interval between any pair of two adjacent pulses at the output of the divider unit. In the specific embodiment of such a priori art solution, the value K is defined in a way ensuring that the interval K_clk exceeds the minimum 1/F_S of the sampling interval specified for the used A/D converter. The value N determines the random variation range for these sampling intervals. This range is adapted to the conditions of the specific application. For example, the value of N is set up near to K when the signals to be digitized are broadband. Each of the input signal sample values, obtained in the process of sampling and formed into a given digital format, is stored in a memory.

[0019] The drawback of this priori art approach is the limited input signal bandwidth for which such sampling could be realistically used. Even in the best case, the clock frequency F_clk, determining the said sampling interval time increment D, could not exceed a few hundreds of MHz. Therefore this limits the bandwidth of the digitized analog signal down to half of that figure.

[0020] The block diagram of the present invention is presented in FIG. 1. The major difference, in comparison with the block diagram of the priori art case, is usage of the digitally controllable delay block. This block additionally delays the pulses at the output of the divider block by a time interval D×B, where D is the delay increment equal to 1/F_clk(M+1) and B is a pseudo-random integer ranged from zero to M. The integer B is generated in parallel with said integer A. The analog input signal is sampled at the time instants determined by said divided and delayed clock pulses. The obtained signal sample values are converted into a predetermined digital format and stored in a memory. FIG. 3 illustrates how the sampling pulse sequence is formed in the case of the present invention (for K=3, N=3 and M=3).

[0021] As can be seen, the minimum digital increment D of the sampling interval variation, in this case, is quarter of the clock interval 1/F_clk. In general, the value of D can be reduced in this way down to tens of picoseconds if take M>>1. In practice that can be achieved, for example, by using the Programmable Delay Chip MC10EP195 from “ON Semiconductor”. Consequently, the bandwidth of the input analog signal, which can be digitized in this alias-free way, can be extended at least up to a few GHz. That is substantially more than can be provided by any of the present-day A/D converters. On the other hand, the minimum value of sampling intervals can be easily matched to the allowed sampling rate of specific A/D converters.

[0022] The maximum sampling rate F_S of high-speed A/D converters, in many practical cases, is near to the highest clock frequency F_clk allowable for thechnical implementation of the sampling pulse formers. In these cases, to sample the analog signal at the highest rate, said integer A should be varied in the minimum range (from one to two). However this results in the minimum sampling interval much smaller than 1/F_clk. In other words, the frequency F_clk should be considerably lower than F_S. To adapt the present invention to sampling of analog input signals at the highest rate, said integer A is taken as equal to one, if said integer B is greater than its preceding value; otherwise it is equal to two. FIG. 4 illustrates the performance of this additional rule for the case when M=3.

[0023] As shown, the minimum value of sampling intervals is always equal to 1/F_clk (independently from the value of M).

[0024] Implementation of the described approach to alias-free signal sampling is based on application of standard electronic blocks. Specifically, widely used pseudo-random number generators can be used to control the pulse divider and pulse delay blocks. However, to ensure that the parameter of the generated sampling pulse sequence meet the requirements of a specific application, a data memory (containing data of the required sampling pulse sequence) could be used as the source of the required random numbers. In this case, the present invention allows synthesizing a wide range of various type of sampling pulse sequences adapted to specific tasks of digital signal processing.

[0025] The above disclosure sets forth a number of embodiments of the present invention. Other arrangements or embodiments could be practiced under the teaching of the present invention and a set forth in the following claims.