Deliverable 3
Yüklə 1,61 Mb.
|
- Bu səhifədəki naviqasiya:
- 3.1Time interleaving proposal for NGH
- 3.1.2I/Q shift for rotated constellations
- 3.1.3Hybrid convolutional/ block time interleaver structure
- 3.1.4Scheduling of PLPs
- 3.1.5Frequency Interleaving
3Time interleavingDVB-NGH being dedicated to mobile applications, some research has been undertaken to optimize the DVB-T2 channel interleaver in order to reduce the time interleaver memory in the receiver and to take into account the hybrid satellite/terrestrial mode of NGH. Section 3.1 presents the results of the studies that led to the adoption of a combination of block and convolutional interleaving in the DVN-NGH baseline. Section 3.2 gives the summary of an analysis of time interleavers in Land Mobile Satellite conditions, which is presented in details ENGINES Deliverable 2.4. 3.1Time interleaving proposal for NGHTeracom was deeply involved in the design of a time interleaver structure appropriate to DVB-NGH. This section presents a synthesis of their work. 3.1.1IntroductionThe time interleaving (TI) proposal for NGH was originally heavily based on the block interleaving (BI) used in the DVB-T2 standard, including the possibility to interleave over several frames. For complexity reasons it was however found desirable to reduce the TI memory by 50%. This would however severely limit the possible time interleaving depth, especially since lower order constellations like QPSK, which are quite likely to be used for NGH, allows a shorter interleaving depth for a given number of cells in the TI memory and for a given PLP bit rate. For these reasons, but also to allow for shorter zapping time (especially in connection with the hybrid satellite/terrestrial mode of NGH) convolutional interleaving (CI) was suggested as an alternative to block interleaving. CI had however already been studied and rejected for DVB-T2, due to certain problems (e.g. TFS and VBR). However, after some analysis a solution was agreed: to combine BI and CI in such a way that BI is done internally in an NGH frame in a similar way as for T2, but the interleaving across NGH frames is done by CI. In this way the previously known problems with CI would vanish but there would still be significant memory gains and gains in zapping time. 3.1.2I/Q shift for rotated constellationsWhen rotated constellation is used it is important for the original I and Q components of a cell to be transmitted with large separation in time and frequency, so that e.g. a bad RF channel or a badly received NGH frame (due to fading or interference) will not contain both components. In T2 this was not ensured, but for NGH it has been suggested to use a shift of one Interleaving Unit (IU) for the Q component before cell interleaving. This will ensure that the I and Q are always transmitted in different NGH frames, see Figure : 
Figure : Proposed I/Q shift for DVB-NGH 3.1.3Hybrid convolutional/ block time interleaver structureWith CI each FEC block needs to be spread over several frames. The way this is achieved is to divide each FEC block into a number of so-called Interleaving Units (IUs), with number of IUs equal to the time interleaving depth expressed in number of NGH frames. In order to minimise power consumption the highest possible IU size should be used. Figure shows the way the CI works. The upper figure shows the situation before CI and the lower figure after CI. There is one column per NGH frame and four IUs per FEC block. There are two PLPs – one “red PLP” and one “green PLP”.
Figure : Illustration of the proposed IU-based CI process The starting point for the interleaving process is that we have an integer number of 16200 bit FEC blocks per PLP for each NGH frame. Unfortunately it is not always possible to divide a FEC block into equally large IUs using all the constellations QPSK, 16-QAM, 64-QAM and 256-QAM. In some combinations of constellation and interleaving depth (number of NGH frames N) some IUs get one cell more than the others (these cases are: 16-QAM N=4, 256-QAM, N=2, 4, 6) . This does not have any effect on the CI as such, but will affect the number of cells per frame for a given PLP. In each NGH frame BI is performed for each PLP individually. The BI is shown in .
Figure : IU-based BI For a given PLP each IU is inserted vertically in a column. The result is read out horizontally, row-by-row. Due to the slightly different IU lengths (for a given configuration) the IUs will in general not perfectly fit a column – some IUs will, but others will be “one cell short”. There will thus be some empty space at the end of the BI. These empty positions are simply discarded when the data is read out, as can be seen from Figure . 
Figure : Insertion of IUs into the BI 3.1.4Scheduling of PLPsAfter the BI the PLPs are put on top of each other in a cell-based matrix, according to Figure .
Figure : PLP arrangement after the BI The number of columns in the matrix equals “number of RF channels” multiplied by “number of sub-slices per RF channel”. When this condition is fulfilled (including any necessary padding cells in the top row), a deterministic scheduling of the cells into the NGH frame can be done in the way outlined below. First step: Divide the matrix into “number of sub-slices per RF channel” (here two) and put one half on top of the other, see Figure :
Figure : Result of the first scheduling step Second step: Divide the resulting structure into “number of RF channels” (here three in Figure ): 
Figure : Result of the second scheduling step Third step: Perform time shifting and folding back, according Figure . The time shift makes the PLPs evenly distributed in order to allow for good time diversity and time/frequency hopping. 
Figure : Result of the third scheduling step End result of scheduling process
Figure : End result of scheduling It should be noted that the previous process has not yet filled the scheduled cells with data – only allocated the positions in the frame. Time interleaved PLP cells are introduced into sub-slices in the natural time sequence, independently of RF channel The first time interleaved cell is therefore introduced in the first cell position of the first sub-slice of the PLP (in whichever RF channel it appears). A receiver performing the frequency hopping will therefore receive the time interleaved cells in the right order.
Yüklə 1,61 Mb. Dostları ilə paylaş:
|
