Pulse or digital communications – Receivers – Particular pulse demodulator or detector
Reexamination Certificate
1999-02-25
2002-09-17
Chin, Stephen (Department: 2634)
Pulse or digital communications
Receivers
Particular pulse demodulator or detector
C375S262000, C714S794000, C714S795000
Reexamination Certificate
active
06452984
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to a method and apparatus for decoding convolutionally encoded signals. Such signals are commonly used in communications and recording systems that employ error correction to combat signal corruption.
2. Description of the Related Art
The designers of mobile electronic devices such as cellular phones strive to reduce the power consumption of such devices. Lower power consumption leads to longer battery lives and thus greater convenience, greater reliability and reduced expenses for users of such devices. Therefore, manufacturers seek to differentiate themselves from competitors on the basis of reduced power consumption.
A major determinant of power consumption in cell phones (“mobile stations”) is the amount of time the cell phone is actively receiving signals versus the amount of time the cell phone is in “standby mode,” in which the mobile's receiver is shut down. It is desirable to increase the amount of time the cell phone is in standby mode.
To reduce power consumption, certain types of cellular phone systems have a “slotted” mode, in which the mobile station “knows” that any messages addressed to it will only be sent during specific frames. (A frame is a group of input symbols that are interleaved). These active frames are herein called the slot, and a slotted mode is one where this information to a specific mobile is transmitted only in that mobile's assigned slot. Thus, the mobile can shut off its receiver for all other frames (“standby mode”). This is a crucial part of power management; therefore, it is highly desirable to increase the time the mobile is in standby mode.
Increasing the time the mobile is in standby mode depends upon, among other things, how much information the mobile must receive, and therefore the amount of time the receiver is on, to correctly process a given piece of data. In turn, the amount of information a mobile must receive to process data is critically dependent upon the coding/decoding scheme employed by the transmitter/receiver system; the mobile's power consumption increases as the amount of information needed by the mobile to decode a given piece of data increases.
Coding/decoding schemes are employed in mobile communications systems to allow a mobile to detect and/or correct errors. (Additional layers of encoding/decoding may be applied for other reasons, such as security). These schemes function by adding some redundancy to information that is to be transmitted.
The redundancy, however, has a cost, in that the mobile must have more information to determine any particular symbol. Therefore, given a particular coding scheme, it is highly desirable to design a mobile that can estimate the desired symbol sequence based upon as little received information as possible, without significantly sacrificing the accuracy of such estimates.
One increasingly popular redundant coding scheme is convolutional coding, in which the coding of a sequence of precoding bits is passed through a shift register, which, at a particular point in time, outputs a certain number of encoded bits based upon the value of that portion of the sequence that is then in the register.
It would be desirable to improve upon conventional maximum likelihood sequence estimation (“MLSE”) decoders such as Viterbi decoders, which are commonly used to decode convolutionally encoded data. If s
k
is a vector that represents a possible transmitted sequence of symbols and r is a vector that represents the actual signals received by a mobile, a Viterbi decoder effectively tests all of the possible values of s
k
and selects the s
k
that maximizes the summation &Sgr;
n
c[n]r[n]s
k
[n]=s
k
T
Cr, where c[n] defines the channel gain for a transmitted symbol for sample n (and thus C is a diagonal matrix with these values). (The s
k
that maximizes the cross correlation is the s
k
that minimizes the “distance” between r and s
k
.) More generally, MLSE's selects the s
k
that maximizes a function of the probability of s
k
given r. For more details regarding Viterbi decoders, see, for example, “Digital Communications”, John G. Proakis (3d edition 1995).
The best s
k
may be represented as a path through nodes in a diagram (known as a trellis), where lines (“transitions”) between nodes in adjacent time steps represent whether an input bit (preceding bit) was a 0 or a 1 (for binary coding). The nodes (“states”) in vertical columns represent the values of prior input symbols. A “path” through the trellis therefore represents a particular sequence of input symbols.
FIG. 1
shows an example of a trellis with only two paths
10
and
12
shown. The time steps indicate a decoder trellis beginning at time t=0 to time t=4. At each time step, the decision units of the decoder contain the value of the cross correlation of the most likely path to the state of interest. Thus, the decoder trellis contains multiple paths from an initial state to a given state several time steps later.
The path that has the highest cross correlation corresponds to the most likely sequence. The most likely path to a particular state within a given time step is found by starting at that particular state at that given time step and tracing backward (a “traceback”) along the chosen transitions. Precoding bits that correspond to the transitions along the path are the decoded data. A preceding bit equal to 0 is shown is a solid line; a preceding bit equal to 1 is shown as a dotted line. Thus, path
10
shown in
FIG. 1
corresponds to a precoding bit sequence equal to 0100 while path
12
corresponds to a precoding bit sequence equal to 1110.
Given a particular sequence received by a decoder, a metric is generated for possible test sequences, where each test sequence corresponds to a path through a trellis. For a path ending at a particular state at a particular time (“current node”), the corresponding metric is generated by adding (a) the metric for that portion of the path ending at a particular state at the previous time period (“prior node”), to (b) the metric for the transition from the prior node to the current node. (According to the Viterbi algorithm, metrics are generated for all possible paths ending in a current node and the path with the highest metric is the survivor path; the other paths are discarded.) For example, assume that the portion of the path
10
that ends at node
22
in
FIG. 1
, which corresponds to encoder state B at time t=3, has a metric of 4. If the metric for the transition from node
22
to a node
20
is 2, the metric for the entire path
10
is 6 (=4+2). The path at any given time with the highest metric is the most likely path and is therefore the path that is “traced back.”
To achieve good noise performance from the decoding process, the traceback length must generally be several times the constraint length (the number of input bits upon which an output depends; this is one plus the length of a shift register that may be used to perform the encoding) of the code. For example, assume a code with a constraint length of 9 that outputs 2 coded bits for each input bit. If a decoder for such a coding scheme has been running from t=−∞ to t=0 and a traceback is then performed, the last 64 bits might not achieve adequate error performance, while those prior to those 64 will have been accurately decoded. This discrepancy occurs because for the most recent bits there is still subsequent information (that can help to decode those most recent bits) in the not yet arrived signal.
On the other hand, if the decoder must wait for subsequent information, the power consumption of the mobile may be increased. Similarly, power consumption may be increased if, at the beginning of an encoded data stream that a decoder must decode, the decoder must know information transmitted prior to such data. For example, assume there are three frames, and the actual information) intended for the user of the mobile is only in the
Banister Brian
Rick Roland R.
Chin Stephen
Fan Chieh M.
LSI Logic Corporation
LandOfFree
Metric biasing for maximum likelihood sequence estimators does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Metric biasing for maximum likelihood sequence estimators, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Metric biasing for maximum likelihood sequence estimators will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2898604