Title of Invention


Abstract This invention is on adaptive iterative Turbo Decoder for UMTS. In this invention, channel condition is mapped into number of iteration groups and subsequently each iteration's requirement number is fed into the Turbo Decoder there by meeting required Eb/N or BER. More importantly channel condition is directly computed from a set of measurements that were specified in UMTS system for other purposes. Thus by a small increase in the analysis of channel condition and obtaining the iteration requirement for Turbo decode, MIPS requirement for Turbo Decoder is saved and thereby reduces the average computation power required for decoding a block with Turbo decoder in UMTS UE and Node-B receivers at no extra overhead.
This invention relates to the Turbo decoder in UMTS. Further, this invention relates to the broad category of Channel Coding and Forward Error Correcting codes. More particularly this invention encompasses a method for reducing power consumption in an adaptively iterative turbo decoder for UMTS.
A Turbo code is a parallel-concatenated convolution code with near Capacity performance. A typical Turbo Encoder comprises of two constituent encoders operating in parallel, with inputs to each being the interleaved/randomized version of the input of another [3]. Each constituent encoder is an RSC (Recursive Systematic Convolution) encoder. Generally convolution encoders are feed forward, however they result in large number of code words with lower code-weight. However for good error correction capabilities input bit words should be mapped on to code-words with higher code-weights meaning they are far apart in code vector space. Hence a RSC encoder is the solution. In this encoder structure the generator polynomials are divided by one of the generator polynomials and this changes the mapping of input to output vectors. But in this way even though some input vectors get mapped on to code words with good weight, some other input vectors may result in poor weighted code words. Hence an interleaver is provided which randomizes the input bit vectors such that even though it may give a low weight code word from first constituent RSC encoder, still the interleaved pattern may give a good weight code word from the second constituent RSC encoder. This is shown in Figure 1.
Turbo coder encodes same information twice but in different order. This makes it possible for the exchange of information between the two decoders for constituent convolutional decoders. The more scrambled the information, the more un-correlated the information exchange is. This is one of the keys that allow a continuous improvement in correction capability when decoding process is iterated.
Soft output decision algorithms provide as an output a real number, which is a measure of the reliability of the decoder's hard decision. This extra information is

very important Tor tne next decoder in tne iterative process. There are two categories of soft decision algorithms. The first one includes the maximum likelihood decoding, which minimizes the probability of bit error such as MAP (maximum aposteriori) algorithm. The MAP algorithm calculates a posteriori probabilities of each state transition, a message bit b(i) and/or code symbol produced by a Markov process(the RSC encoder is a Markovian process), given the received noisy code word Y'. Log likelihood ratio (LLR) which is the logarithm of the ratio of a-posteriori probabilities of b(i) being 1 given "r" and b(i) being 0 given "r". Generally the LLR of a bit is 1. The second category includes maximum likelihood sequence decoding, which minimizes the probability of word or sequence error. In this most probable transmitted sequence is found by computing the most probable path a code-trellis will follow till the allowed depth, given the received bit sequence.
Turbo decoder is constructed from simple constituent decoders which share bit reliability measures. This sharing of information is made possible through the interieaving involved in the encoder, which creates two weakly correlated parity streams. The constituent decoders are optimal decoders for the component codes used by the turbo encoder. For convolutional codes, the soft output viterbi algorithm (SOVA), an implementation of Maximum Likelihood Sequence decoder and the Bahl, Cocke, Jelinck and Rajiv (BCJR) algorithm, an implementation of MAP algorithm are applicable. BCJR has been most commonly used in turbo codes since it is the MAP symbol -decoding rule and gives strikingly low bit error rates. A typical Turbo decoder comprises of two constituent convolution decoders based on a MAP algorithm. Each decoder is fed with the streams of received data bit and the received output bit of the corresponding constituent encoder at transmitter and a confidence weight for each bit which is nothing but the a-posteriori probability (APP) computed for that bit by the other decoder. This way each decoder is giving the feedback of its decision (APP) about a bit to the other decoder which is useful for it to further narrow down on to the decision. The more iteratively this is done more error correction capability results even at low Eb/No values, albeit at a cost of more computation and power consumption. The same is shown in Figure 2.
Decoding continues for a set of iterations. Usually performance improves with the number of iterations but follows law of diminishing returns.

A typical Turbo decoder consumes many MIPS (Million Instructions Per Second) per iteration.
A UE/Node-B receiver comprises of a RF-front end that receives a very high frequency signal on-air and translates it into a baseband signal, a base band equalizer that minimizes the inter-symbol interference inflicted by the channel, a demodulator that detects the digital symbols from the equalized analog waveform and a channel decoder that is capable to correct, to a specific extent, the errors remaining in the detected digital bit stream. Hence whenever such a receiver is designed, the target Bit Error Rate (BER) for different channel conditions is considered before fixing the error detection or error correction capabilities [1]. The turbo decoder and the number of iterations for which it executes are decided considering achieving the target BER at worst channel conditions (represented by SIR and Eb/No) [1]. By increasing the number of iterations, BER performance can be improved.
However the fading channels are highly non-stationary and the fading could be constructive or destructive at random. These properties of channel are not taken into account in the present turbo decoders, thus making it inefficient and power consumed is more.
The primary object of this invention is to invent a method for reducing MIPS requirement in an adaptively iterative turbo decoder for UMTS utilizing pre-specified measurements.
It is another object of the invention to exploit the hidden information about channel and use it to reduce the average MIPS requirement in real time for Turbo decoder implementation in UMTS UE and Node-B receivers thus making the Turbo Decoder implementation more efficient.

This invention exploits the random behaviour of channel and takes in to account the fact that, at times channel can be better, at times it is worse and at times it can be worst.
This invention estimates the condition of channel not by any extra processing or computation but utilizes the information that was already computed for other specified purposes.
Thus considering the above, the Turbo Decoder can be modified such that it results in reduction in the amount of computation required and consequently reduces the power consumption at no extra overhead.
This invention is applicable equally at UE as well as at Node-B.
Accordingly this invention relates to a method for reducing MIPS requirement in an adaptively iterative turbo decoder for UMTS utilizing pre specified measurements for other purposes like different power control measurements and monitoring purposes etc wherein the average computation power required for decoding a block with Turbo decoder in UMTS UE and Node-B receivers is reduced in that the number of decoder iterations are made adaptive to UMTS specific channels and decoder runs for designed minimum iterations based on the channel condition computed from the pre specified measurements.
The other objects, features and advantages of the present invention will be apparent from the accompanying drawings and the detailed description which are as follows.
Figure 1 is a schematic block diagram of a typical Rate 1/3 Turbo encoder. This can be extended to Turbo code of other rates.
Figure 2 is a schematic block diagram of a typical Turbo decoder with information exchange between the two constituent decoders.
Figure 3 illustrates a typical data flow in a UMTS Physical layer.

Figure 4 illustrates the procedure that comprises the invention.
The preferred embodiments of the present invention will now be explained with reference to the accompanying drawings. It should be understood however that the disclosed embodiments are merely exemplary of the invention, which may be embodied in various forms. The following description and drawings are not to be construed as limiting the invention and numerous specific details are described to provide a thorough understanding of the present invention, as the basis for the claims and as a basis for teaching one skilled in the art on how to make and/or use the invention. However in certain instances, well-known or conventional details are not described in order not to unnecessarily obscure the present invention in detail.
A typical data block submitted to turbo decoder at the receiver in UMTS comprises of data received over 'n' slots of a data channel. Usually for different power control measurements and monitoring purposes, the Physical layer of the receiver will measure the RSCP or Eb/No /or SIR or some measure of signal strength which it receives in all those n slots on the control channel corresponding to the above data channel [1] , [4]. Data Channel and corresponding control channel will always be transmitted simultaneously [2]. Thus signal strength measure taken on control slots will represent the channel condition during the data block transmission. These measurements that were already taken for a different purpose can be used to estimate the channel condition. Physical layer can take the average of signal strength of all slots that comprise a data block and this block average signal strength can be compared against a pre-determined threshold to decide whether the channel is good or bad. This decision of good or bad channel can be used to control the number of iterations of turbo decoder per block. If the channel condition is good, target BER can be achieved with lesser number of iterations than the earlier implementation and if the channel condition is bad, turbo decoder would require same number of iterations it has in its earlier implementation to meet the target BER. Thus the average computation required would be reduced a lot assuming the channel is good at least for half the time. When the channel is worst continuously,

the decoding scheme is more or less same as before and in any case, this new decision-making does not add much overhead.
This idea can be extended and improvised to a much larger extent. Instead of categorizing a channel as merely good or bad, a multi-threshold statistical model can be fitted to the received signal strength and thus many states of a channel can be defined and thus the number of iterations can be controlled in a better manner.
The thresholds for channel conditions can be imposed either by observation or by simulation of BER performance in various channel conditions for different iterations of Turbo decoder.
With reference to Figure 3 here the data block is received at the receiver in UMTS over n slots of a data channel. At first the data is fed to the physical layer of the receiver which performs the signal strength and/or signal quality measurement on all those n slots of the control channel corresponding to this data channel for power control purposes and takes the average of it [1], [2], [4]. This average value is compared to a pre-determined threshold value to decide if the channel is good or bad. Next the data is demodulated and equalized in DSP Base band and fed to the turbo decoder for decoding.
Operation of the method:
The steps of the invention are as follows:
1) The data channel slotsi to n of the block are received one by one and the signal strength measure ss1 till ssn is calculated on the corresponding slots of associated control channel, which is simultaneously transmitted. The data is processed and stored in the data buffer;
2) The average value of the signal strength is calculated over the slots that comprise a data block. The state of the channel is found by comparing the average value with the threshold value, which is pre-computed based on simulations with different channel conditions. According to the state of channel, the number of iterations is calculated which is then fed into the turbo

3) The turbo decoder processes the data and thus the decoded data block is
Method of averaging works out well and it can be practically explained as below: Suppose the actual bit X1 is transmitted then the channel condition is poor. Received estimation of X1 bit at receiver y1 is poor. Further, during transmission of repeated bit X1' channel condition might have improved resulting in a better estimation of the received bit y2'. Similarly all the repeated bits may experience different channel condition since the channel is fast fading. Taking average of all the estimated bits (y1. y1'.) gives better estimation of the transmitted bit X1 and hence improving performance of Viterbi decoder.
Simulation parameters:
Data Block size in 3G varies based on the application type. Data block size as much as 640 bits have been observed [3]. However, data block size of 150 bits is the most frequently used. Therefore data block size of 150 bits is chosen for the present simulation.
Performance of the proposed scheme depends on the percentage of bits repetition and not on the block size.
A percentage of repetition ranging from 0% to 30% is considered, as this is the typical range of repetition in 3G systems.
Simulation results:
It is clear that improvement (in terms of BLER) is dependent on the percentage of repetition. At 30% repetition, BLER improvement of Approx 60% is observed and for the case of 0% repetition the old and the new approach behave similarly (as expected).

1. The metric with which channel condition can be estimated can be varied.
2. The number of channel states and the statistical model to which the channel can be fit can be varied.
It will also be obvious to those skilled in the art that other control methods and apparatuses can be derived from the combinations of the various methods and apparatuses of the present invention as taught by the description and the accompanying drawings and these shall also be considered within the scope of the present invention. Further, description of such combinations and variations is therefore omitted below. It should also be noted that the host for storing the applications include but not limited to a computer, printer or a multi function device.
Although the present invention has been fully described in connection with the preferred embodiments thereof with reference to the accompanying drawings, it is to be noted that various changes and modifications are possible and are apparent to those skilled in the art. Such changes and modifications are to be understood as included within the scope of the present invention as defined by the appended claims unless they depart therefrom.

3G: Third Generation
BCJR: Bahl, Cocke, Jelinck and Rajiv algorithm
BER: Bit Error Rate
BLER: Block Error Rate
ORG: Cyclic redundancy Check as discussed in ref [3]
CPICH: Common Pilot Channel
GCTRCH: Coded composite transport Channel as discussed in ref [3]
DTX: Discontinous Transmission
DMC: Discrete Memoryless channel
Eb/No: Energy per bit to Noise spectral density Ratio
MIPS: Million Instructions Per Second
RSC: Recursive Systematic Convolution Code
RSSI: Received Signal Strength Indicator
SIR: Signal to Interference Ratio
SOVA: Soft Output Viterbi Algorithm
TrGh: Transport Channel as discussed in Ref. [2]
TTI: Transmission time Interval as discussed in Ref. [1]
TrBk: Transport Block as discussed in ref [1]
UE: User Equipment
UMTS: Universal Mobile Telecommunication System
WCDMA: Wideband Code division multiple access

[1] 3GPP TS 25.201: "Physical layer - General Description".
[2] 3GPP TS 25.211: "Physical channels and mapping of transport channels onto
physical channels (FDD)".
[3] 3GPP TS 25.212: "Multiplexing and channel coding (FDD)".
[4] 3GPP TS 25.214: "Physical layer procedures (FDD)".

1. A method for reducing the power consumption of a Turbo decoder in UMTS UE and Node-B receivers characterized in that the number of decoder iterations are made adaptive to UMTS specific channels and decoder runs for designed minimum iterations based on the channel conditions are computed from the pre specified measurements.
2. A method as claimed in claim 1, wherein channel condition is computed from pre specified measurements by suitable averaging as herein described, and the turbo decoder iterations are varied.
3. A method as claimed in claim 1, wherein the state of the channel is estimated and consequently the number of iterations of turbo decoder to decode a block is decided in that by averaging the signal strength measurements over a period of data transmission and comparing with pre-computed thresholds obtained from simulations, state of the channel is determined which in turn is used to decide the number of iterations of turbo decoder in UMTS.
4. A method for reducing power consumption in an adaptively iterative turbo decoder for UMTS utilizing pre specified measurements as substantially described herein particularly with reference to figure 4 .



1075-che-2003-claims duplicate.pdf

1075-che-2003-claims original.pdf

1075-che-2003-correspondence others.pdf

1075-che-2003-correspondence po.pdf

1075-che-2003-description complete duplicate.pdf

1075-che-2003-description complete original.pdf


1075-che-2003-form 1.pdf

1075-che-2003-form 26.pdf

1075-che-2003-other documents.pdf

Patent Number 201493
Indian Patent Application Number 1075/CHE/2003
PG Journal Number 08/2007
Publication Date 23-Feb-2007
Grant Date 02-Aug-2006
Date of Filing 31-Dec-2003
# Inventor's Name Inventor's Address
PCT International Classification Number H04L027/00
PCT International Application Number N/A
PCT International Filing date
PCT Conventions:
# PCT Application Number Date of Convention Priority Country
1 NA