WO2018126197A1 - A method for multichannel signal search and demodulation and tech nique to demodulate and detect dbpsk fdma ultra-narrow band signal - Google Patents

Abstract

Described is a method of searching of multichannel signal and technique of demodulating and detecting DBPSK frequency division multiple access (FDMA) ultra-narrow band signal. A search is based on algorithm encompassing a signal-processing signal, and technique to demodulate and detect FDMA ultra-narrow band together with a method to increase time-frequency resolution.

Description

A METHOD FOR MULTICHANNEL SIGNAL SEARCH AND DEMODULATION AND

TECHNIQUE TO DEMODULATE AND DETECT DBPSK FDMA ULTRA-NARROW BAND

SIGNAL

Inventors Andrey Orlov and Vasiliy Anisimov

Priority Claim/Incorporation by Reference

[0001] The present application claims priority to U.S.

Provisional Patent Applications: 62/440,906 filed on

December 30, 2016 entitled "A method for multichannel signal search and demodulation"; and hereby incorporates by reference, the entire subject matter of this Provisional Application.

Background Information

[0002] Many applications require sufficient multichannel

reception of signals. Search for signals could easily be

accomplished by series of Fourier Platforms, but its time and frequency resolution is insufficient to demodulate found signals effectively. Using narrowband signals is one of simplest ways to improve efficiency and reliability of transmission, but it is limited by carrier frequency uncertainty resulting in carrier frequency offset in receiver. In most systems, carrier frequency must be relatively low compared to signal bandwidth.

[0003] In some prior art solutions, signal bandwidth is

increased, and noise tolerance is managed using some sort of error-correction coding of just repetitions. For example, LoRa wireless protocol is widely known. It uses chirp spread

spectrum, utilizing very wide bands compared to information speed of transmission. LoRa uses CDMA multiple access, which is different from the solution proposed in the present invention.

[0004] SigFox is another prior art solution employing

multichannel receivers with carrier tracking. This solution is limited by computational inefficiency of carrier tracking systems. Also, it is algorithmically and computationally

difficult to employ error correction coding in such systems, as they are limited by noise tolerance of carrier tracking systems. So, some sort of code-aided tracking might be necessary, which may be further complicating such solutions.

[0005] The present invention solves two main problems related to wireless communications: receiving many simultaneous messages and receiving narrowband messages with frequency offset greater than signal bandwidth. These problems are common for many m2m and Low-Power Wide Area Network (LPWAN) communication systems. LPWAN systems must transmit and receive many signals

simultaneously and efficiently.

Summary

[0006] The present invention relates to method of multichannel signal search based on algorithm comprising a choosing Fourier transform size according to symbol length, calculating frequency resolution, calculating number of frequency shifts to achieve time resolution, performing multiple Fourier transforms for each unique frequency shift and starting time offset pair. In

addition, the algorithm uses direct quadrature components of phase information of resulting frequency domain signal to search and demoduluate signals. [0007] Furthermore, method involves signal processing scheme and technique to demodulate and detect (FDMA) ultranarrow band (UNB) signal together with a method to increase time-frequency

resolution .

Detailed Description

[0008] The present invention maybe further understood with reference to the following description. An exemplary embodiment of the present invention provides multichannel signal search allowing demodulation based on very computationally efficient Fourier transform with improved frequency and time resolution. Also, it allows demodulation using three independent channels in frequency domain.

[0009] The algorithm sets length of Fourier transform equal to or lower than the length of a transmitted symbol. This allows direct usage of obtained frequency domain data for signal modulation, if time data is fed to consecutive Fourier

transforms each of which starts close to beginning of each new consecutive symbol.

[0010] Then, as time and frequency resolution of such transform do not allow consistent data transfer, additional series of transforms with different starting time offsets and frequency shifts need to be performed.

[0011] At least two additional frequency shifts are required, in the case of two shifts each shifting time domain data by one and two thirds frequency resolution value of Fourier transform.

[0012] At least three additional initial offsets must be

considered, in the case of three offsets constituting of one, two, and three quarters of transform size. [0013] Each unique shift-offset pair forms three independent channels, allowing signal decoding and demodulation based on in- phase and quadrature components and on phase information.

[0014] The exemplary embodiment includes several signal

processing modules or blocks. Block to perform time and

frequency shift, block of a fast Fourier transform (FFT), module of differential binary phase-shift keying (DBPSK) demodulator, module of signal detection.

[0015] The time-frequency shift block comprises the following parts. First, memory block for storing input time-domain

information. In some cases, real input is used, in other cases quadrature input is used. Second, frequency shift block

comprising a sine wave generation block and a complex

multiplier. In some cases, sine frequency generation block is a DDFS block. This block generates a given frequency and shifts input data by this frequency. Third, time shift block, this blocks shifts data in time domain using the memory block, where data is stored.

[0016] This time- and frequency shift block is used to increase time and frequency resolution of proposed method of

demodulation. In some cases, used frequency shifts are equal to one third and two thirds of single FFT bin.

[0017] FFT block performs Fourier transform. In some cases, this is a real Fourier transform. In other cases, a complex Fourier transform is performed. This block takes time-frequency shifted time-domain data as input and outputs frequency-domain

quadrature data.

[0018] DBPSK demodulator can be implemented either by complex conjugate multiplication of current and preceding samples or via phase subtraction of current and preceding samples. [0019] Detection process of signal's preamble is performed along both axis of signal's 2d vector space if complex conjugate multiplication in DBPSK demodulator block, or it is performing preamble search with several phase rotations if phase

subtraction is used in DBPSK demodulator.

[0020] The components are connected serially, and input

information is fed to a block performing a number time-frequency shifts. The output of this module is connected to the FFT module. Transformed data from FFT module goes to a parallel structure of identical modules performing DBPSK demodulation in frequency domain and signal detection for each unique time- frequency shift. Each demodulated stream corresponding to each time-frequency shift is also stored in a unique cyclic buffer for further processing. In some cases, this buffer is connected serially to the output port of signal detection block. In other cases, this buffer is connected parallel with preamble search block to the output of DBPSK demodulation block.

[0021] In some cases, all output data from FFT block for every unique FFT channel, time and frequency shift combination is stored in these cyclic buffers.

[0022] In some cases, only one data stream per unique FFT channel and time shift is stored.

[0023] The choice as to which unique Fast Fourier Transform (FFT) channel, time and frequency shift combination to choose, is made based on output of preamble detection block.

[0024] The exemplary embodiment involves manner in which

components are connected and used and the following of features of algorithm. First, one FFT transform is used for both signal detection and demodulation, preamble detection block is performing header search without interruptions even when messages are being demodulated and stored in cyclic buffers.

[0025] Both preamble detection and demodulation blocks as well as cyclic buffers can process any number of header detection events in different FFT channels simultaneously.

[0026] It is possible to process many header detection events in the same FFT channel, is many data streams per channel are stored .

[0027] In some cases all headers detected in one channel at different moments in time even in case of overlap.

[0028] In other cases only one active header event per channel is processed at time. The decision which event to process is made in preamble search block.

[0029] The decision if message is valid or not is made only after whole message is received. All blocks in receiver work always so there is no risk of message loss in case of false alarm. Many messages can be received simultaneously at any time.

[0030] There is no difference between a message with huge carrier offset and message on different channel and therefore, there is no difference in receiving messages with any kind of frequency offset. There are methods to receive many

simultaneous messages with unknown carrier offset but most of these methods require some rough header search with FFT, and then demodulation of found signals, usually using phase locked loops or other algorithms with carrier tracking capabilities. All these methods are not as computationally effective as the proposed herein method.

Claims

What is claimed is:

1. The method of signal search, time and frequency of

synchronization and demodulation comprising an algorithm:

setting length of Fourier transform equal to or lower

than the length of a transmitted symbol;

additional series of transforms with different starting time offsets and frequency shifts to be performed;

for two additional frequency shifts, each shifting time domain data by one and two thirds frequency resolution value of

Fourier transform;

for three additional initial offsets, one, two and three

quarters of transform size;

each unique shift-offset pair forms three independent channels, allowing signal decoding and demodulation based on in-phase and quadrature components and on phase information.

2. The time-frequency shift block of claim 1 comprising:

memory block for storing input time-domain information;

frequency shift block comprising of a sine wave generation;

block and a complex multiplier where said block generates

a given frequency and shifts input data by this frequency; time shift block that shifts data in time domain using the

memory block where data is stored.

3. One FFT platform of claim 1 is used for signal detection and demodulation where preamble detection block performs header search without interruptions even when messages are being demodulated and stored in cyclic buffers;

4. Preamble detection and demodulation block of claim 3 and cyclic buffers of claim 3 process any number of header detection events in different FFT channels simultaneously;

5. Same FFT channel of claim 4 processes many header detection events with many data streams per channel are stored;

6. The decision of validity of message of claim 3 is made after whole message is received with simultaneous receipt of messages at any time and with any kind of frequency offset.