System and methods for data compression and nonuniform quantizers

0Associated
Cases 
0Associated
Defendants 
0Accused
Products 
0Forward
Citations 
0
Petitions 
1
Assignment
First Claim
1. A method for differentiatorbased compression of digital data, comprising:
 generating a predicted signal using an adaptive filtering method, wherein generating the predicted signal using the adaptive filtering method comprises multiplying a tapweight vector by an original data vector to generate the predicted signal, the original data vector comprising N sequential samples of an original signal, N being an integer greater than or equal to one;
using a subtraction module, subtracting the predicted signal from a sample of the original signal to obtain an error signal;
using a nonuniform quantizer, quantizing the error signal to obtain a quantized error signal; and
updating the tapweight vector by adding to the tapweight vector a product of (a) a step size, (b) the error signal, and (c) a complex conjugate of the original data vector.
1 Assignment
0 Petitions
Accused Products
Abstract
A method for differentiatorbased compression of digital data includes (a) using a subtraction module, subtracting a predicted signal from a sample of an original signal to obtain an error signal, (b) using a quantization module, quantizing the error signal to obtain a quantized error signal, and (c) generating the predicted signal using a least means square (LMS)based filtering method.
2 Citations
No References
Adaptive reduction of noise signals and background signals in a speechprocessing system  
Patent #
US 20070043559A1
Filed 08/21/2006

Current Assignee
Entropic Communications Incorporated

Sponsoring Entity
Trident Microsystems Limited

CMOS IMAGE SENSOR AND IMAGING METHOD IMPLEMENTING CORRELATED DOUBLE SAMPLING AND COMPRESSION  
Patent #
US 20160119570A1
Filed 10/29/2015

Current Assignee
Sony Corporation, Pixim Inc.

Sponsoring Entity
Sony Corporation, Pixim Inc.

17 Claims
 1. A method for differentiatorbased compression of digital data, comprising:
generating a predicted signal using an adaptive filtering method, wherein generating the predicted signal using the adaptive filtering method comprises multiplying a tapweight vector by an original data vector to generate the predicted signal, the original data vector comprising N sequential samples of an original signal, N being an integer greater than or equal to one; using a subtraction module, subtracting the predicted signal from a sample of the original signal to obtain an error signal; using a nonuniform quantizer, quantizing the error signal to obtain a quantized error signal; and updating the tapweight vector by adding to the tapweight vector a product of (a) a step size, (b) the error signal, and (c) a complex conjugate of the original data vector.  View Dependent Claims (2, 3, 4, 5, 6)
 7. A method for receiving a signal, comprising:
receiving a transmission signal from a transmission medium; demodulating the transmission signal to obtain a demodulated signal; mapping the demodulated signal from V to U bits to obtain a quantized error signal, U and V each being respective integers, and U being greater than V; decompressing the quantized error signal to obtain a quantized original signal by; generating a predicted signal using an adaptive filtering method, and using an addition module, adding the predicted signal to a quantized error signal to obtain the quantized original signal; and converting the quantized original signal into an analog output signal.  View Dependent Claims (8, 9, 10)
 11. A method for transmitting data, comprising:
sampling an analog original signal to obtain a sample of the original signal; generating a first predicted signal using a first least means square (LMS)based filtering method; subtracting the first predicted signal from the sample of the original signal to obtain an error signal; quantizing the error signal to obtain a quantized error signal; mapping the quantized error signal from U bits to V bits to obtain a mapped error signal, U and V each being respective integers, and U being greater than V; modulating a carrier signal according to mapped error signal to obtain a transmission signal; transmitting the transmission signal from a first location to a second location using a transmission medium; demodulating the transmission signal to obtain a demodulated signal; mapping the demodulated signal from V to U bits to obtain the quantized error signal; generating a second predicted signal using a second LMSbased filtering method; using an addition module, adding the second predicted signal to the quantized error signal to obtain a quantized original signal; and converting the quantized original signal into an analog output signal.  View Dependent Claims (12, 13, 14, 15, 16, 17)
1 Specification
This application is a continuationinpart of U.S. patent application Ser. No. 16/153,692, filed Oct. 5, 2018, which claims the benefit of and priority to U.S. Provisional Patent Application Ser. No. 62/568,679, filed Oct. 5, 2017, and to U.S. Provisional Patent Application Ser. No. 62/573,780, filed Oct. 18, 2017. This application also claims benefit of and priority to U.S. Provisional Patent Application Ser. No. 62/693,704, filed Jul. 3, 2018. The disclosures of each of these prior applications are incorporated herein by reference in their entireties.
The field of the disclosure relates generally to optical communication systems and networks, and more particularly, to data compression, coding, and quantizers for analog and digital optical systems and networks.
Conventional hybrid fibercoaxial (HFC) architectures deploy long fiber strands from an optical hub to a fiber node, and typically many short fiber strands to cover the shorter distances from the HFC nodes to a plurality of end users. Conventional Multiple Service Operators (MSOs) offer a variety of services, including analog/digital TV, video on demand (VoD), telephony, and high speed data internet, over these HFC networks, which utilize both optical fibers and coaxial cables, and which provide video, voice, and data services to the end user subscribers. The HFC network typically includes a master headend, and the long optical fiber carries the optical signals and connects the link between the headend, the hub, and the fiber node. Conventional HFC networks also typically include a plurality of coaxial cables to connect the fiber nodes to the respective end users, and which also carry radio frequency (RF) modulated analog electrical signals.
The HFC fiber node converts optical analog signals from the optical fiber into the RF modulated electrical signals that are transported by the coaxial cables to the end users/subscribers. Some HFC networks implement a fiber deep architecture, and may further utilize electrical amplifiers disposed along the coaxial cables to amplify the RF analog signals. In the conventional HFC network, both the optical and electrical signals are in the analog form, from the hub all the way to the end user subscriber'"'"'s home. Typically, a modem termination system (MTS) is located at either the headend or the hub, and provides complementary functionality to a modem of the respective end user.
The continuous growth of optical intra/interdatacenter link, 5G mobile fronthaul (MFH) and backhaul (MBH), nextgeneration distributed HFC architectures and access networks, passive optical networks (PONs), and highspeed optical shortreach transmission systems with advanced multilevel modulation formats require an equivalent growth in the development of advanced digital multilevel modulation formats to process the vastly increased amount of data transmitted over the various networks. Presently, conventional deployments of 1G/10G PON systems using nonreturn to zero (NRZ) modulation are unable to meet the growing capacity demand to deliver future highspeed data and video services.
The growth of the optical link and network architectures has been matched, and in many ways outpaced, by a continuouslygrowing demand on highspeed Internet, highdefinition TV, and realtime entertainment services, which has created an additional challenge for future broadband access networks. These emerging new services, which include virtual reality and 5G, are rapidly depleting the bandwidth resources of existing PONs, MFH, and HFC networks, where upgrades to system capacity and spectral efficiency is urgently needed.
Some conventional communication networks operate according to DOCSIS 3.1 specifications, which are standardized, and feature orthogonal frequencydivision multiplexing (OFDM) and higher order of modulations (>4096 QAM). However, although the DOCSIS specification provides higher flexibility and spectral efficiency, it also presents new technical issues and challenges. For example, the demanding carriertonoise ratio (CNR) specified by DOCSIS 3.1 specifications for high order modulations cannot not be supported by legacy digitaltoanalog (D/A) or analogtodigital (A/D) converters having resolutions and the range of 810 digits. Replacement of every legacy D/A or A/D converter at the customer premises of a user would incur a substantially high cost, and therefore there is a need in the field new algorithms that are capable of suppressing quantization noise without the McKinley impairing D/A and A/D performances. Additionally, the continuous envelope and high peaktoaverage power ratio (PAPR) of OFDM signals render the signals vulnerable to nonlinear distortions in analog HFC networks.
Recent progress in advanced A/D and D/A quantizers and data compression techniques for transmitting OFDM signals is encouraging improvement to the transmission techniques for digital RFoG (DRFoG) systems. Some quantizer or compression techniques utilize partial bit sampling (PBS), but results in rapid increases to the quantization noise when reducing the number of digits. Fitting based nonlinear quantization (FBNQ) operations are recommended by the standard of Open Radio Equipment Interface (ORI). However, the FBNQ algorithm is complex and time consuming because it needs to estimate the statistical characteristics from a large number of samples, which increases the system delay. Moreover, the accuracy of FBNQ seriously degrades when quantizing the amplitudes distributed outside the interquartile range (IQR). Some matured companding methods, including μlaw and Alaw, for encoding acoustic signals have been used in tuning the quantization levels. However, the logarithmic compression function of these methods is not optimal to suppress the quantization noise of Gaussian distributed OFDM signals. Accordingly, new algorithms are needed for optimizing the nonuniformly distributed quantization levels.
The growing performance requirements of 5G newradio (NR), highspeed Internet access, and highresolution multimedia entertainment with virtual reality create even further challenges to future fiberwireless integrated MFH. Recent 5GNR specifications feature OFDM and higher order modulations (e.g., 256 and 1024QAM). However, proposals to integrate these two technologies have resulted in new difficulties, such as high sensitivity to nonlinear distortions and increased requirements on highresolution D/A converters, which limit the quality and transmission distance of analog radiooverfiber (ARoF) optical networks in MFH. However, digital RoF (DRoF) networks have demonstrated greater compatibility with different formats, as well as greater suitability for the 5GNR environment that includes a more diverse spectrum and services. DRoF also utilizes high immunity to nonlinear distortions, and better capability for errorfree transmissions through use of forward error coding/correction (FEC) techniques. Furthermore, by the additional use of data compression and advanced modulation formats, DRoF is better able to mitigate bandwidth and efficiency, and better digitally transport highquality wireless signals between the baseband unit pool (BBUpool) and radio access units (RAU) with increased transmission distance and improved power budgets. Accordingly, the results would need to develop algorithms to improve the compression efficiency of DRoF MFH for 5GNR specifications.
In an embodiment, an optical network includes a transmitting portion configured to (i) encode an input digitized sequence of data samples into a quantized sequence of data samples having a first number of digits per sample, (ii) map the quantized sequence of data samples into a compressed sequence of data samples having a second number of digits per sample, the second number being lower than the first number, and (iii) modulate the compressed sequence of data samples and transmit the modulated sequence over a digital optical link. The optical network further includes a receiving portion configured to (i) receive and demodulate the modulated sequence from the digital optical link, (ii) map the demodulated sequence from the second number of digits per sample into a decompressed sequence having the first number of digits per sample, and (iii) decode the decompressed sequence.
In an embodiment, an analogtodigital converter includes a sampling unit configured to sample an analog voltage signal into a sequence of discrete samples, a nonuniform quantizing unit configured to quantize the discrete digital samples with 2× quantization levels, where x represents a first number of digits per sample. The analogtodigital converter further includes a mapping unit configured to map the quantized discrete digital samples from the first number of digits per sample into a second number of digits per sample, wherein the second number is greater than the first number.
In an embodiment, a data compression method is provided for an input digital sequence of discrete signal samples having an input number of digits per sample. The method includes steps of applying a companding function to the input signal samples according to the input number of digits per sample, calculating a companded output number of digits per sample that is less than the input number of digits per sample, quantizing the input signal samples according to the companded output number of digits per sample, and outputting a digital sequence of signal samples having the companded output number of digits per sample.
In an embodiment, a method for differentialbased compression of digital data includes (a) using a subtraction module, subtracting a predicted signal from a sample of an original signal to obtain an error signal, (b) using a quantization module, quantizing the error signal to obtain a quantized error signal, and (c) generating the predicted signal using a least means square (LMS)based filtering method.
In an embodiment, a method for differentialbased decompression includes (a) using an addition module, adding a predicted signal to a quantized error signal to obtain a quantized original signal, and (b) generating the predicted signal using a least means square (LMS)based filtering method.
These and other features, aspects, and advantages of the present disclosure will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters represent like parts throughout the drawings, wherein:
Unless otherwise indicated, the drawings provided herein are meant to illustrate features of embodiments of this disclosure. These features are believed to be applicable in a wide variety of systems including one or more embodiments of this disclosure. As such, the drawings are not meant to include all conventional features known by those of ordinary skill in the art to be required for the practice of the embodiments disclosed herein.
In the following specification and the claims, reference will be made to a number of terms, which shall be defined to have the following meanings.
The singular forms “a,” “an,” and “the” include plural references unless the context clearly dictates otherwise.
“Optional” or “optionally” means that the subsequently described event or circumstance may or may not occur, and that the description includes instances where the event occurs and instances where it does not.
Approximating language, as used herein throughout the specification and claims, may be applied to modify any quantitative representation that could permissibly vary without resulting in a change in the basic function to which it is related. Accordingly, a value modified by a term or terms, such as “about,” “approximately,” and “substantially,” are not to be limited to the precise value specified. In at least some instances, the approximating language may correspond to the precision of an instrument for measuring the value. Here and throughout the specification and claims, range limitations may be combined and/or interchanged; such ranges are identified and include all the subranges contained therein unless context or language indicates otherwise.
As used herein, the terms “processor” and “computer” and related terms, e.g., “processing device”, “computing device”, and “controller” are not limited to just those integrated circuits referred to in the art as a computer, but broadly refers to a microcontroller, a microcomputer, a programmable logic controller (PLC), an application specific integrated circuit (ASIC), and other programmable circuits, and these terms are used interchangeably herein. In the embodiments described herein, memory may include, but is not limited to, a computerreadable medium, such as a random access memory (RAM), and a computerreadable nonvolatile medium, such as flash memory. Alternatively, a floppy disk, a compact discread only memory (CDROM), a magnetooptical disk (MOD), and/or a digital versatile disc (DVD) may also be used. Also, in the embodiments described herein, additional input channels may be, but are not limited to, computer peripherals associated with an operator interface such as a mouse and a keyboard. Alternatively, other computer peripherals may also be used that may include, for example, but not be limited to, a scanner. Furthermore, in the exemplary embodiment, additional output channels may include, but not be limited to, an operator interface monitor.
Further, as used herein, the terms “software” and “firmware” are interchangeable, and include any computer program storage in memory for execution by personal computers, workstations, clients, and servers.
As used herein, the term “nontransitory computerreadable media” is intended to be representative of any tangible computerbased device implemented in any method or technology for shortterm and longterm storage of information, such as, computerreadable instructions, data structures, program modules and submodules, or other data in any device. Therefore, the methods described herein may be encoded as executable instructions embodied in a tangible, nontransitory, computer readable medium, including, without limitation, a storage device and a memory device. Such instructions, when executed by a processor, cause the processor to perform at least a portion of the methods described herein. Moreover, as used herein, the term “nontransitory computerreadable media” includes all tangible, computerreadable media, including, without limitation, nontransitory computer storage devices, including, without limitation, volatile and nonvolatile media, and removable and nonremovable media such as a firmware, physical and virtual storage, CDROMs, DVDs, and any other digital source such as a network or the Internet, as well as yet to be developed digital means, with the sole exception being a transitory, propagating signal.
Furthermore, as used herein, the term “realtime” refers to at least one of the time of occurrence of the associated events, the time of measurement and collection of predetermined data, the time for a computing device (e.g., a processor) to process the data, and the time of a system response to the events and the environment. In the embodiments described herein, these activities and events occur substantially instantaneously.
According to the embodiments described herein, digitization transmission scheme may advantageously transmit digitized radio signals through the optical fiber link and convert the signal back into analog format at remote fiber nodes. This digitized radio transmission scheme is further useful for utilization in MFH networks using a common public radio interface (CPRI). The transmission scheme implements a sufficient number of quantization digits and FEC, such that digitized radio carriers may be reconstructed by a D/A converter with no quality degradation. Additionally, the digital compression algorithm techniques described herein enable further reduction of the number of digits, and with a suppressed quantization noise floor. According to these advantageous systems and methods, the transmission efficiency (TE, defined as the ratio between the effective bandwidths of the encapsulated analog signals and the converted digital signals) of the network is substantially improved.
NonUniform Quantizers for Analog and Digital Systems
The following embodiments describe algorithms for optimizing nonuniformly distributed quantization levels, and to resolve the challenges arising from use of PBS, FBNQ, and the logarithmic compression function of companding levels. A first such algorithm is based on the Kparameter fast statistical estimation (FSE), also referred to as Klaw. This algorithm embodiment proves to be computationally efficient, and particularly useful for OFDM signals with Gaussian distributed amplitudes. A second algorithm is based on the relaxed Lloyd algorithm (RLloyd), and is useful for determining the decision threshold and quantization levels based on minimum square error criteria. This RLloydbased technique (i) is compatible with multiple formats, (ii) is not limited to Gaussian distributions, and (iii) exhibits lower quantization noise level with fewer digits (sometimes though, at the expense of increased computing complexity).
The respective processing modules of nonuniform A/D converter 102 are substantially similar to those of nonuniform D/A converter 100, but function substantially in reverse. That is, an input analog voltage 114, or x_{A}(k), is sampled by a sampler 116 (i.e., instead of an interpolator). At a nonuniform quantizer 118, the sampled signal is quantized with 2^{8 }quantization levels and output as 8digitspersample chips. The 8digitspersample chips are received at map 120, where they are mapped to 15digits chips and output as an output signal 122, or x_{r}(k). Similar to D/A converter 100, quantizer 118 and 8to15digits lookuptables may be determined or controlled by the present Klaw or RLloyd algorithm techniques.
Process 200 begins at step 202, in which the 15 digit samples are input to D/A converter 100,
where Δ=1−2Φ(−K), Φ is the cumulative distribution function (CDF) of the standard Gaussian distribution, and x is the normalized modulus of the input signal. In Klaw, K is the key parameter to be configured, and the modulus of the OFDM signal is assumed to be distributed on [0+Kσ], where σ is the standard deviation.
In step 206, the normalized modulus x of the input signal is swept from 12^{15 }times. In step 208, the companded output y=C(x) is calculated. In step 210, the output y is subject to 8bit quantization. In step 212, process 200 implements the 8to15bit lookup table (or the 15to8bit lookup table for the other converter). The practical value of the Kvalues calculated according to process 200 are described further below with respect to
These Klaw techniques enable fast statistical estimation. Nevertheless, a more accurate technique for selecting quantization levels uses a Lloyd algorithm, and further based on minimum meansquare error criterion (MMSE). Prior to performance of the Lloyd algorithm, the probability density function (PDF) of the signal amplitudes is divided into multiple segments and the borders of each segment are given by the thresholds [t_{i}t_{i+1}]. After quantization, the amplitude falls within each segment is quantized as level l_{i}. The MISE between the quantized and original signals, {t_{i}} and {t_{i}}, may then be minimized according to the following equations:
and:
where f(x) is the PDF of signal modulus.
In step 508, the quantization thresholds (t_{1,1}, t_{2,1}, . . . t_{N+1,1}) are similarly obtained, using (l_{1,1}, l_{2,1}, . . . l_{N,1}), and according to Eq. 3. Step 510 is a decision step. In step 510, process 500 determines whether the value for j is less than a predetermined value M of the iteration index. If, in step 510, j<M, process 500 returns to step 506. In this manner, process 500 is able to process a repeatable loop such that the quantization thresholds and quantization levels are repeatedly updated, based on the former value of their counterpart, until the iteration index reaches M (i.e., j=M), upon which process 500 proceeds to step 512. In step 512, process 500 is configured to interpolate minor quantization levels (l_{K (1)}, l_{K(2)}, . . . l_{(q)}) such that the minor quantization levels are uniformly inserted between l_{K }and l_{K+1}, where q represents the number of minor digits. Process 500 completes after step 512.
In an exemplary embodiment of process 500, the selection of p and q may be optimized according to a tradeoff between the quantization accuracy and computational complexity. Typically, a large p value leads to improved precision at the expense of increased number of iterations to be converged. In contrast, if the value for p is too large, strong quantization noise may result within smallprobabilitydensity regions due to the number of training samples falling into that region being insufficient, thus reducing the confidence level of the estimation. Considering all of these factors, given a total number of digits of D, the value for p may be set at 5, and the value for q may be determined according to q=(D−5). A comparison of preliminary results between the different algorithms is described further below with respect to
As described above, the nonuniform quantizer embodiments described herein are capable of reducing the number of sampling digits, and also of suppressing the quantization noise levels. As described further below, the present nonuniform quantizer is also particularly useful with respect to DRoF and DRFoG systems. As described above, DRoF interfaces have been used in MFH systems, such as CPRI and the open base station architecture initiative (OBSAI), which converts the analog radio signals into digital format and delivers digitized baseband radio carriers from a radio equipment controller (REC) to radio equipment (RE). According to the embodiments described above, using a sufficient number of quantization digits and FEC, the digitized radio carriers may be advantageously reconstructed by a D/A converter with no quality degradation.
The present embodiments are still further useful for achieving significant benefits, in comparison with conventional techniques, over a DRFoG link between a hub and distributed remote fiber nodes. The present techniques are formatagnostic, with simple hardware implementation, at the distributed remote fiber nodes. The present techniques are still further capable of taking advantage of the digitization benefits of the link, including the high immunity to nonlinear distortions from power amplification. Essentially errorfree transmission may therefore be achieved through use of FEC, enabled by the high capacity from the fiber network. In combination with the nonuniform quantizer the systems and methods described herein, the required number digits and quantization noise floor for converting each analog sample may be further reduced, thereby significantly increasing the capacity of the DRFoG systems in comparison with conventional techniques.
DataCompression for Digital Mobile Fronthaul with Algorithm and Differential Coding
The differentialcoded Lloyd algorithms described above are also particularly useful as a datacompression technology for improving bandwidth efficiencies in digital MFH networks. The following embodiments demonstrate proof of concept with respect to experimental results demonstrating milestone transmissions of 180 Gbps over 80km fronthaul links, and encapsulating 64×100MHz 1024QAM 5GNR carriers with lowerthan0.5% EVM.
In the exemplary embodiment, V<U, and the bandwidth efficiency may therefore be improved according to the relationship (U−V)/V×100%. As described above, the Lloyd algorithm is based on MMSE criterion, and divides the PDF of the signal amplitudes into multiple segments with boundaries defined by the thresholds [t_{i}, t_{i+1}] as shown in
In the case of a compressed AxC chip 1006 having V digits 1010, an RLloyd method may be further implemented to reduce the complexity of the conventional Lloyd algorithm, where 2^{P }out of 2^{V }levels are computed using the conventional Lloyd algorithm first as the major levels and 2^{(V−P) }minor quantization levels are uniformly interpolated between [l_{i}, l_{i+1}]. The selection of P and V may therefore be determined by a tradeoff between the quantization accuracy and convergence speed, similar to the embodiments described above.
From illustrations 1200, 1206, it can be seen that, in comparison other algorithms 1204 (e.g., μLaw, ALaw, FSE, etc.), RLloyd algorithm 1202 is capable of realizing the most optimal EVM performance, and using fewer quantization digits. Additionally, although some algorithmic methods (e.g., FSE) are specially designed for OFDM signals having Gaussian distributed amplitudes these and other algorithms are not applicable to nonGaussian signals, such as SCFDM. According to the present embodiments though, performance of the present Lloyd algorithm techniques is independent of the statistical property of the given signal. That is, the same algorithmic techniques may be applied to Gaussian and nonGaussian signals, as demonstrated by illustration 1206. As shown in
The present systems and methods are additionally capable of advantageously applying DPCM techniques to further improve the signal quality after compression. In comparison with conventional PCM, which digitizes the original radio signal x(k), DPCM may be additionally utilized to predict and digitize the differential signal, namely, x(k)−x(k−1). For DPCM, most source signals exhibit some correlations between successive samples. Through differential precoding, the correlationinduced redundancy may be reduced to enable representation of the information with fewer digits. To reduce the corresponding complexity, a firstorder differentiator may be used, and having precoded signals, which are denoted here as d(k)=x(k)−βx(k−1).
Under simulated test conditions, the value of β was found to be optimized at the value of 0.6. Nevertheless, DPCM implementations may exhibit specific challenges with respect to the quantization error resulting from the compression process, and the decision error exhibited at the DPCM decoder. Those two types of errors propagated and accumulated from the beginning to the end of the whole frame, which may seriously degrade the quality of the reconstructed analog RF signals. The embodiments described further below resolve these challenges by providing a feedback loopbased differential quantizer that mitigates the quantization error transfer issue.
In a similar manner, receiver 1302 provides a signal from MFH network 1314 to a demodulator 1316, and may further include one or more of a bit map 1318, a DPCM decoder (or PCM decoder) 1320, and a D/A converter 1322 configured to output 5GNR analog carriers 1324 to one or both of an enhanced mobile broadband (eMBB) network 1326 and a massive MIMO network 1328. In this example, receiver 1302 further includes a decompression unit 1330 in operable communication with one or both of bit map at 1318 and DPCM decoder 1320.
Under test conditions, a simple IM/DD link tests the performance of the compressed DRoF link of transmitter 1300 and receiver 1302. At transmitter 1300, for example, one 20MHz LTE OFDM component is sampled with a resolution of 15digits/sample, and at a sampling rate of 30.72MHz. DPCM (or PCM) encoder 1308, integrated with a Lloydbased according to the present embodiments, converts the quantized samples into binary AxC chips. Bit map 1310 (e.g., a 15to8bit map) then compresses each AxC chip from 15bits/chip to 8bits/chip. The compressed AxC chips may then be interleaved and mapped into NRZ symbols.
At receiver 1302, a similar process, but in reverse, may be implemented to reconstruct the analog components by decompressing the received digital signals. The recovered analog component carriers (e.g., analog carriers 1324) may then be sent to wireless antennas in RAUs in networks 1326, 1328. The EVM performances of recovered analog signals 1324 may then be compared between DPCM and PCM encoded schemes, according to one or more of the techniques described above.
Frames 1514 fed into a dualpolarization IQ modulator (DPIQM) 1516, which modulates four streams 1518 of DRoF signals for I and Q tributaries on both polarizations (e.g., x and y) based on, for example, QPSK modulation formats. Transmitting portion 1502 then transmits the signals to a coherent receiver 1520 of receiving portion 1504 over medium 1506 (e.g., an 80km SSMF), and then sampled by, for example, a 4channel realtime sampling oscilloscope 1522 before application of digital signal processing (e.g., offline) by a coherent digital signal processor 1524 for signal decompression by a decompression unit 1526 and recovery. In this exemplary embodiment, the testing operation was accomplished using DPQPSK with 128 and 180Gbps data rates applied, which enabled encapsulation of 48 and 64, respectively, 100MHz 5GNRlike OFDM components with 1024QAM.
The present systems and methods provide an innovative combination of enhanced datacompression algorithms, which may be based on both a Lloyd algorithm and DPCM, to improve the SQNR and bandwidth efficiency in at least DRoF systems for nextgeneration 5GNRcompatible digital MFH, as well as the other types of communication networks described herein. With 8digit quantization, the present DRoF link the embodiments are capable of supporting up to at least 4096QAM OFDM or SCFDM formats, with up to at least 6dB SQNR improvement. Furthermore, the present embodiments have demonstrated how 128 and 180Gbps highcapacity MFH links based on coherent transmission technology may more efficiently transmit 48×100 and 64×100MHz 5GNRlike OFDM components with highorder 1024QAM format.
HighOrder DifferentiatorBased Data Compression Systems and Methods
DPCM is not limited to firstorder differentiator implementations; instead, DPCM can be implemented with highorder differentiators, i.e. with secondorder or higherorder differentiators. Disclosed herein are new data compression systems and methods which may be implemented with highorder differentiators. These new data compression systems and methods use a least means square (LMS)based filtering method to generate a predicted signal for use with DPCM, which achieves significant advantages compared to conventional techniques.
For example, the new data compression systems and methods may achieve accuracy that is comparable to that of highorder differentiatorbased data compression systems using MMSEbased filtering methods (“MMSEbased DPCM systems”), but with significantly greater computational efficiency than is possible with MMSEbased DPCM systems. For instance, certain embodiments do not require solving for an inverse of a correlation matrix, which is necessary in MMSEbased DPCM systems. Additionally, the data compression systems and methods promote high SQNR and bandwidth efficiency. For example, Applicant has found that in some embodiments, the new systems and methods may achieve a SQNR improvement of around 10 dB, as well as achieve a bandwidth efficiency improvement of around 40% to 60%, compared to conventional compression techniques. Furthermore, certain embodiments of the new data compression systems are adaptive, e.g. tap weight vectors are automatically adjusted according to changing statistical properties of original signal waveforms, thereby promoting high performance over varied operating conditions. Moreover, some embodiments of the new data compression systems and methods are conducive to integration in a digital interface of a 5G mobile fronthaul system.
For example,
Referring again to
Subtraction module 1802 subtracts a predicted signal {circumflex over (x)}(k) from a sample x(k) of an original signal (not shown in
The LMSbased filtering method of LMSbased filtering module 1808 is characterized by generating the least mean square of an error signal, where the error signal is a difference between a desired signal and an actual signal. In some embodiments, LMSbased filtering module 1808 operates according to equations 46, as follows:
x{circumflex over ( )}(k)=c(k)*v(k) (Eq. 4)
err(k)=x(k)−x{circumflex over ( )}(k) (Eq. 5)
c(k+1)=c(k)+μ·err(k)·conj[v(k)] (Eq. 6)
In equations 4 and 6, v(k) is an original data vector defined as follows: v(k)=[x(k−1), x(k−2), . . . , x(k−N)]^{T}. Thus, original data vector v(k) includes N sequential samples of the original signal. For example, x(k) is a current sample of the original signal, and x(k−1) is an immediately preceding sample of the original signal. N is an integer greater than or equal to one, and the value of N may be referred to as the “order” of LMSbased filtering module 1808. For example, if N is equal to three, LMSbased filtering module 1808 implements a thirdorder filter. c(k) in equations 4 and 6 is a tapweight vector defined as follows: c(k)=[c_{1}(k), c_{2}(k), . . . , C_{N}(k)]. μ in equation 6 is a step size, and conj[v(k)] in equation 6 is a complex conjugate of original data vector v(k).
In particular embodiments, data compression system 1800 executes equation 4 to generate predicted signal {circumflex over (x)}(k), and data compression system 1800 executes equation 6 to update tapweight vector c(k), i.e. to obtain updated tapweight vector c(k+1) for sample k+1, where error signal err(k) is obtained from equation 5. Updating tapweight vector c(k) advantageously enables LMSbased filtering module 1808 to adapt to realtime operating conditions of a communication system including data compression system 1800, which may be particularly beneficial if the communication system carries dynamic data, such as bursty data. Applicant has found that embodiments of LMSbased filtering module 1808 implementing Equations 46 will quickly converge to obtain a least mean square of error signal err(k) when initial tapweights of tapweight vector c(k) and step size μ are properly set. Additionally, it should be noted that equations 46 do not require calculating an inverse of a correlation matrix, which promotes computational efficiency of data compression system 1800, especially in highorder embodiments of LMSbased filtering module 1808.
Each tap weight module 2004 multiplies a respective sample of the original signal by a respective tapweight of tapweight vector c(k) to obtain a product 2012. For example, tap weight module 2004(2) multiples x(k−2) by c_{2}(k) to obtain a product 2012(2), which is equal to x(k−2)c_{2}(k). Addition module 2006 sums products 2012 to obtain predicted signal {circumflex over (x)}(k), and subtraction module 2008 implements equation 5 to obtain err(k). Adaption module 2010 implements equation 6 to update tapweight vector c(k) and provide respective updated tapweights to tap weight modules 2004.
Referring again to
One possible application of data compression system 1800 is in a transmitter. For example,
Receiver 2106 includes a demodulator 2120, a data decompression system 2124, and a digitaltoanalog converter (DAC) 2126. Demodulator 2120 demodulates transmission signal 2118 to obtain a demodulated signal 2128. Data decompression system 2124 maps demodulated signal 2128 from V bits to U bits to obtain quantized error signal err_{q}(k), and data decompression system 2124 generates a quantized original signal x_{q}(k) from quantized error signal err_{q}(k). In some alternate embodiments, mapping of demodulated signal 2128 from V bits to U bits is performed outside of data decompression system 2124. DAC 2126 converts quantized original signal x_{q}(k) into an analog output signal 2130. Analog output signal 2130 represents analog original signal 2114.
For example,
Referring again to
Mapping module 2201 maps demodulated signal 2128 from V bits to U bits to obtain quantized error signal err_{q}(k). Addition module 2202 adds a predicted signal {circumflex over (x)}(k) from LMSbased filter 2204 to quantized error signal err_{q}(k) to obtain quantized original signal x_{q}(k). In some embodiments, LMSbased filtering module 2204 is implemented in the same manner as, or in a similar manner to, LMSbased filtering module 1808, e.g. as illustrated in
Applicant has conducted simulations to evaluate performance of certain embodiments of data compression system 1800. For example,
It can be determined from graph 2700 that there is around 6dB improvement in SQNR by adding one more quantization digit. Compared with PCM, firstorder and secondorder DPCM can bring about an extra 1.4dB and 1.24dB improvement, respectively. By combining secondorder DPCM and an RLloyd quantization algorithm, an 8.24dB improvement can be achieved in SQNR. Considering that nearly a 6dB SQNR improvement can be traded for one quantization digit reduction, it is evident that data compression system 1800 can improve quantization efficiency relative to conventional data compression systems.
Exemplary embodiments of systems and methods for nonuniform quantizers and data compression are described above in detail. The systems and methods of this disclosure though, are not limited to only the specific embodiments described herein, but rather, the components and/or steps of their implementation may be utilized independently and separately from other components and/or steps described herein.
Although specific features of various embodiments of the disclosure may be shown in some drawings and not in others, this is for convenience only. In accordance with the principles of the disclosure, a particular feature shown in a drawing may be referenced and/or claimed in combination with features of the other drawings.
Some embodiments involve the use of one or more electronic or computing devices. Such devices typically include a processor or controller, such as a general purpose central processing unit (CPU), a graphics processing unit (GPU), a microcontroller, a reduced instruction set computer (RISC) processor, an application specific integrated circuit (ASIC), a programmable logic circuit (PLC), a field programmable gate array (FPGA), a DSP device, and/or any other circuit or processor capable of executing the functions described herein. The processes described herein may be encoded as executable instructions embodied in a computer readable medium, including, without limitation, a storage device and/or a memory device. Such instructions, when executed by a processor, cause the processor to perform at least a portion of the methods described herein. The above examples are exemplary only, and thus are not intended to limit in any way the definition and/or meaning of the term “processor.”
This written description uses examples to disclose the embodiments, including the best mode, and also to enable any person skilled in the art to practice the embodiments, including making and using any devices or systems and performing any incorporated methods. The patentable scope of the disclosure is defined by the claims, and may include other examples that occur to those skilled in the art. Such other examples are intended to be within the scope of the claims if they have structural elements that do not differ from the literal language of the claims, or if they include equivalent structural elements with insubstantial differences from the literal language of the claims.