Detection of digital signals received from multipath channels

• US 5,005,189 A
• Filed: 01/23/1989
• Issued: 04/02/1991
• Est. Priority Date: 01/23/1989
• Status: Expired due to Fees

• Current Assignee: Charles M. Hackett Jr
• Original Assignee: Charles M. Hackett Jr
• Inventors: Hackett, Charles M. Jr.
• Primary Examiner(s): Olms, Douglas W.
• Assistant Examiner(s): Chin, Stephen
• Application Number:
  US07/300,785
• Time in Patent Office: 799 Days
• Field of Search: 375/82, 375/84, 375/101, 375/102, 375/94, 455/143, 455/303, 364/480, 364/485, 364/726, 370/23
• US Class Current: 375/348
• CPC Class Codes: H04B 7/005 Control of transmission; Equalising
• CPC Class Codes: H04L 27/38 Demodulator circuits; Receiver circuits

Abstract

Methods for detection or, or equivalently, demodulation of phase-and-amplitude-shift-keyed communication signals that may be received via multipath propagation or dispersion in a noisy channel are disclosed. The signal format has a block structure with a fixed pattern preceding and following the data portion of each block. Implementation is accomplished in real-time by obtaining samples of the complex-envelope of the received waveform, processing at the symbol rate mainly in the frequency domain, then returning to the time domain with a coagulated signal where ordinary detection of the symbols is completed according to the modulation in use. The signal format and computation in the frequency domain permit the estimation of the multipath-structure of the channel, which in turn permits the estimation of the data using a reasonable amount of processing. Performance of the basic method is substantially improved by generating a metric and searching for alterations of the data that reduce the metric. The search can be controlled by an indicant vector that identifies those symbols of a block that are likely to be in error. Synchronization is obtained by trial using the basic method to compare the detected pattern with the known pattern.

Claims

1. For any positive integers N, M, and K with N=K+2M, a method for detection or demodulation of a phase-and-amplitude-shift-keyed signal having a block format wherein each block has N symbols and is comprised of a known fixed pattern of M symbols preceding and following K symbols of data, where said signal is transmitted through a channel that may have multiple propagation paths and may add interfering noise, said method comprising the steps of:

a. obtaining from a received waveform complex samples of its complex-envelope at a rate which is an integer multiple of its symbol rate;

b. establishing block synchronization and symbol synchronization for said complex samples;

c. down-sampling to the symbol rate, and buffer-storing the N complex samples of the block to be detected and the first M complex samples of the next subsequent block;

d. computing the M-point fast Fourier transform (FFT) of the first M complex samples of said block to be detected, and computing the M-point FFT of the first M complex samples of said next subsequent block, where an M-point FFT is an efficient algorithm for evaluating an M-point discrete Fourier transform;

e. computing a weighted average of the two FFT'"'"'s computed in step d;

f. computing the quotient of said average formed in step e divided component-by-component by the precomputed M-point FFT of said known fixed pattern;

g. obtaining a complex vector that is an estimate of the multipath-structure of the channel by computing the M-point inverse fast Fourier transform (IFFT) of said quotient formed in step f, and subjecting the components of said IFFT to a threshold;

h. computing the N-point FFT of said estimate of the multipath-structure of the channel obtained in step g padded with K+M zeros;

i. scaling up the small components of said N-point FFT computed in step h and replacing any zero by a non-zero number;

j. computing the N-point FFT of said block to be detected;

k. dividing component-by-component said N-point FFT formed in step j by the result of step i;

l. forming analog estimates of said data symbols and pattern symbols by computing the IFFT of the result of step k;

m. detection of said analog estimates of said data symbols formed in step 1 by an operation appropriate for detection of a signal of the same modulation transmitted over a noisy channel having but a single propagation path.

9. The invention of claim 1 or claim 2 wherein bit synchronization and block synchronization are obtained by trial using the method of claim 1 but omitting step 21b to obtain an analog estimate of said fixed pattern, then computing a measure of the difference between said estimate and said known fixed pattern, and declaring synchronization when said measure is satisfactorily small.

2. For any positive integers N, M, and K with N=K+2M, a method for detection or demodulation of a phase-and-amplitude-shift-keyed signal having a block format wherein each block has N symbols and is comprised of a known fixed pattern of M symbols preceding and following K symbols of data, where said signal is transmitted through a channel that may have multiple propagation paths and may add-interfering noise, said method comprising the steps of:

a. obtaining from a received waveform complex samples of its complex-envelope at a rate which is an integer multiple of its symbol rate;

b. establishing block synchronization and symbol synchronization for said complex samples;

c. down-sampling to the symbol rate, and buffer-storing the N complex samples of the block to be detected;

d. computing the N-point Fast Fourier Transform (FFT) of said block to be detected;

e. obtaining a preliminary analog estimate of the K data symbols in said block;

f. obtaining an estimate of the multipath-structure of the channel;

g. obtaining a preliminary detection of said K data symbols;

h. performing analog adjustment of said estimated multipath-structure, this adjustment to be omitted on the second pass and subsequent even passes;

i. performing digital adjustment of the preliminary detected data symbols, this adjustment to be omitted on the first pass and subsequent odd passes;

j. forming the component-by-component product of the N-point FFT of the block comprising said current estimates of the K data symbols preceded and followed by said known fixed pattern of M symbols multiplied by the N-point FFT of the current estimate of the multipath structure padded with K+M zeros;

k. generating an error vector equal to the difference of said N-point FFT of said received block computed in step d and the result of step j;

l. using said error vector to compute a metric that provides a measure of the difference between said N-point FFT of said received block and the result of step j;

m. using said metric to determine whether said adjustment is an improvement and to control subsequent adjustments of said preliminary detected data symbols and of said estimate of the multipath structure;

n. retaining those M symbols that are regarded as the best current estimate of the multipath structure and those K symbols that are regarded best as the current estimate of the data;

o. iterating steps h through n as desired to obtain a refined estimate of the data.

4. The invention of claim 2 or claim 3 wherein said estimate of the multipath-structure is adjusted depending upon the gradient of said metric with respect to the multipath-structure.

5. The invention of claim 2 or claim 3 wherein said adjustment of said detected symbols is controlled by an indicant vector, said indicant vector being a vector whose components give some indication of the likelihood of error in corresponding data symbols.

6. The invention of claim 5 wherein said indicant vector is obtained from the absolute values of the analog estimates of the data components.

7. The invention of claim 5 wherein said indicant vector is obtained from the gradient of said metric with respect to said current detected data.

8. The invention of claim 5 wherein said indicant vector is computed by component-by-component division of said error vector by the N-point FFT of said estimate of the multipath-structure with small components of said FFT amplified and zero replaced by a non-zero number, forming the inverse fast Fourier transform (IFFT) of the quotient, and controlling the sign of each component by the sign of the corresponding component of said current detected data.

3. For any positive integers N, M, and K, a method for detection or demodulation of a phase-and-amplitude-shift-keyed signal having a block format wherein each block has N symbols and is comprised of a known fixed pattern of M symbols preceding and following K symbols of data, where said signal is transmitted through a channel that may have multiple propagation paths and may add interfering noise, said method comprising the steps of:

a. obtaining from a received waveform complex samples of its complex-envelope at a rate which is an integer multiple of its symbol rate;

b. establishing block synchronization and symbol synchronization for said complex samples;

c. down-sampling to the symbol rate, and buffer-storing the N complex samples of the block to be detected;

d. computing the N-point fact Fourier transform (FFT) of said block to be detected;

e. obtaining a preliminary analog estimate of the K data symbols in said block;

f. obtaining an estimate of the multipath-structure of the channel;

g. obtaining a preliminary detection of said K data symbols;

h. performing digital adjustment of the preliminary detected data symbols, this adjustment to be omitted on the second pass and subsequent even passes;

i. performing analog adjustment of said estimate multipath-structure, this adjustment to be omitted on the first pass and subsequent odd passes;

j. forming the component-by-component product of the N-point FFT of the block comprising said current estimates of the K data symbols preceded and followed by said known fixed pattern of M symbols multiplied by the N-point FFT of the current estimate of the multipath structure padded with K+M zeros;

k. generating an error vector equal to the difference of said N-point FFT of said received block computed in step d and the result of step j;

l. using said error vector to compute a metric that provides a measure of the difference between said N-point FFT of said received block and the result of step j;

m. using said metric to determine whether said adjustment is an improvement and to control subsequent adjustments of said preliminary detected data symbols and of said estimate of the multipath structure;

n. retaining those M symbols that are regarded as the best current estimate of the multipath structure and those K symbols that are regarded best as the current estimate of the data;

o. iterating steps h through n as desired to obtain a refined estimate of the data.