Systems and methods for training sequence selection, transmission and reception

[0001] The present application claims the advantage of an earlier temporary US 5 application. 61/089 712, filed August 18, 2008.

Field of the Invention [0002] The present invention relates to systems and methods for selecting, transmitting and receiving a training sequence.

Background [0003] Mobile communication systems use signal processing techniques that counteract the impact of time-varying and frequency-selective mobile radio channels to improve link performance. Ecualization is used to minimize inter-symbol interference (ISI) caused by multi-path signal fading in frequency-selective channels. Since the mobile radio channel is random and changes over time, the equalizer must identify the time-varying characteristics of the mobile channel adaptively through training and tracking. Wireless systems with time division multiple access (TDMA), such as the Global Mobile Communications System (GSM), transmit data in time slots of a fixed length and the training sequence is contained in the slot (impulse),

[0004] The GSM system is a successful digital cellular technology used throughout the world. Currently, GSM networks provide both voice and data services for billions of subscribers and are still subject to expansion. The scheme of access of GSM technology is TDMA. As illustrated in FIG. 1, in the 100 Hz frequency band, the downlink 102 and uplink 104 are separated and each has a bandwidth of 25 MHz including 124 channels. Carrier separation is 200 kHz. The TDMA 106 frame consists of 8 time slots 108 corresponding to one carrier frequency. The duration of the slot is 577 ms. For a normal pulse, one GSM time slot contains 114 data bits, 26 bits of training sequence, 6 end bits, 2 substitution bits, and 8.25 bits of guard interval.

[0005] The 3GPP specification defines eight training sequences for normal GSM pulses (see TS 45.002, "GERAN: Multiplexing and multiple access on the radio path") and they are widely used in practice for pulse synchronization and channel estimation in current GSM systems. / EDGE Radio Access Network (GERAN).

[0006] With the increase in the number of subscribers and traffic, there is a great emphasis on GSM operators, especially in countries with high population density. In addition, it is desirable to efficiently use hardware resources and spectrum, as prices for voice services are falling. One approach to increasing voice capacity is to multiplex more than one user in a single time slot.

[0007] Voice services on adaptive multi-user channels in one slot, Adaptive Multi-user channels on One Slot (VAMOS), (see GP-081949, 3GPP Work Item Description (WID): Voice services over Adaptive Multi-user channels on One Slot) (compare: Multi-User Reusing-One-Slot (MUROS) (see GP-072033, "WID": MultiUser Reusing-One-Slot) is an appropriate research work)) are an ongoing element of work in GERAN, which aims to increase GERAN voice capacity by a factor of two for an uplink base transceiver and downlink through multiplexing of at least two users simultaneously on the same physical radio resource, i.e. multiple users share the same radio frequency and same time slot .The three candidate MUROS techniques are Orthogonal Sub Channel (OSC) (see GP-070214, GP-071792, "Voice capacity evolution with orthogonal sub channel"), co-TCH (see GP-071738, "Speech capacity enhancements using DARP"), and Adaptive Symbol Constellation (see GP-080114 "Adaptive Symbol Constellation for MUROS (Downlink)").

[0008] In the upstream OSC technique, the co-TCH techniques and the Adaptive Symbol Constellation technique of two users sharing the same time slot use Gaussian minimum shift keying with different training sequences. The base station uses signal processing techniques such as diversification and / or interference reduction to separate the data of two users. Similarly to the uplink, the co-TCH downlink uses two different training sequences for DARP (Downlink Advanced Receiver Performance) devices to separate two users. In the downward link between OSC or Adaptive Symbol Constellation, two subchannels are mapped to subchannels I and Q of QPSK or Adaptive QPSK (AQPSK), in which the ratio of sub-channel I and sub-channel Q can be adaptively controlled. The two subchannels also use different training sequences.

[0009] Figure 2 lists eight 26-bit GSM training sequence codes, each of which has a cyclic sequence structure, i.e. a reference sequence with 16 bits is in the middle, and 10 guard bits (5 guard bits each on each side of the sequence) reference). The 5 most-significant bits and the 5 bits of the least-significant reference sequences are copied and arranged, respectively, to join and precede the training sequence. Protective bits can cover inter-symbol interference times and make a training sequence immune to timing errors. Each GSM training sequence has perfect periodic autocorrelation properties for non-zero shifts in the range [-5, 5], when only the 16-bit reference sequence is considered.

[0010] In GP-070214, GP-071792, "Voice capacity evolution with orthogonal sub channel", a new set of eight training sequences with a length of 26 bits for the OSC technique is proposed in which each of the new training sequences is optimized for cross-correlation properties with the corresponding primary GSM training sequence. The new sequences are listed in Figure 3. It can be seen that these new training sequences do not retain the cyclic sequence structure as the original GSM training sequences.

[0011] In GP-080785 and GP-080641, "New series of training sequence codes for MUROS", new training sequence codes are proposed. New training sequence codes with GP-080785 have 16 central bits containing complete information about the 26 bit code of the training sequence. The remaining 10 bits are placed in the header and end sections for protection. Furthermore, it is described that "the design rules are as follows: First, the new training sequence code must have a small cross-correlation with the existing training sequence code that has been paired with it. Second, it is desirable that the new training sequence code has a small cross-correlation with other new training sequence codes. Thirdly, it is required so that the new training sequence codes have a good correlation in general with all existing training sequences, not only with its pair "(sic). In addition, GP-070620, "New training sequences for RED HOT and HUGE" describes that "for new RED HOT and HUGE working items, new training sequence codes should be considered that they both have good autocorrelation and cross-correlation properties to improve channel estimation" . Similarly, GP-070707, "Training Sequences for High Symbol Rate" reveals the training quota for RED HOT and HUGE, which are evaluated and compared with other proposed training sequences. that they both have good autocorrelation and cross-correlation properties to improve channel estimation. " Similarly, GP-070707, "Training Sequences for High Symbol Rate" reveals the training quota for RED HOT and HUGE, which are evaluated and compared with other proposed training sequences. that they both have good autocorrelation and cross-correlation properties to improve channel estimation. " Similarly, GP-070707, "Training Sequences for High Symbol Rate" reveals the training quota for RED HOT and HUGE, which are evaluated and compared with other proposed training sequences.

Summary [0012] Various aspects of the present invention are set out in the appended claims.

Brief description of the drawings [0013] Embodiments of the application will now be described with reference to the accompanying drawings, in which:

FIG. 1 shows a schematic diagram of band allocation and TDMA frame definitions for GSM;

figure 2 is a table listing the original GSM training sequences; figure 3 is a table including a set of training sequences with optimized cross-correlation properties compared to the original GSM training sequences;

Figure 4A is a table containing a collection of training sequences; Figure 4B is a schematic diagram of a computer readable medium comprising the training sequences of Figure 4A;

Figure 5A is a table containing a set of training sequences;

Figure 5B is a schematic diagram of a computer readable medium comprising the training sequences of Figure 5A;

Figure 6A is a table containing a set of training sequences;

Figure 6B is a schematic diagram of a computer readable medium comprising the training sequences of Figure 6A;

figure 7 shows several sets used to define a set of training sequences;

Figure 8 is a flowchart of the first method for determining training sequences;

Figure 9A is a flow chart of a first method of assigning training sequences;

Figure 9B is a flow chart of a second method of assigning training sequences;

Figure 10A is a block diagram of a transmitter for downlink transmission

OSC;

Figure 10B is a block diagram showing a pair of OSC sub-channel receivers;

Figure 11A is a block diagram of a co-TCH transmitter for downlink transmission;

Figure 11B is a block diagram of a pair of co-TCH receivers for downlink transmission;

Figure 12A shows a pair of OSC or ω-TCH transmitters for uplink transmission; and figure 12B is a block diagram of a receiving device constructed from two receivers for receiving respective transmissions from a pair of transmitters from figure 12A.

DETAILED DESCRIPTION [0014] Degradation of signal to noise ratio (see B. Steiner and P. Jung, & quot; Opti-mum and suboptimum channel estimation for the uplink CDMA mobile radio systems with joint detection & quot ;, European Transactions on Telecommunications, volume 5, January-February 1994, pages 39-50, as well as M. Pukkila and P. Ranta, "Channel estimator for multiple co-channel demodulation in TADM mobile systems", Proc. of the 2nd EPMC, Germany) is used here to evaluate properties correlation training sequences and / or to design new training sequences. In MUROS / VAMOS, the interference comes from the second sub-channel of the same MUROS / VAMOS pair in the same cell as well as from co-channel signals from other cells.

[0015] The SNR degradation can be determined as follows. Let the training sequence with length N be S = {sj, s₂, ..., s_N}, p_n e. {-1, + 1}, n = 1, ..., N. Let us divide two synchronous co-channel signals or MUROS / VAMOS with L independent complex pulsed channel responses hm = (hm, 1, hm, 2, ... ., hm, L), m = 1, 2. The combined impulse channel response is h = (h1, h2). Let samples of the received signal in the receiver be: y = Sh^t + n, where the noise vector is n = (n1, n2, ... nN_L+ l), and S = [S₁S₂] is the matrix (N - L + 1) X 2L, and S_m (m = 1, 2) is defined below ^SmL " ^S m, 2 $ m, lv · · v ę ,, _ ³ iii-ii Δ ^Jand "3 ^jih, 2 • · * · & m.N ^ mN-L + 2 ^ "iN-L + l (1) which is the appropriate training sequence (sm, 1, sm, 2, ..., sm, N) (should be note that S1 and S2 can be constructed with two different training sequences, respectively, either from the same training sequence set or from different training sequence collections).

[0016] The channel estimate in the sense of the smallest mean square error is:

<img file="PL2157752T3_D0001.tif" />

Degradation of the SNR coefficient of training sequences is defined as:

d_sm = 101og₁₀(l + / r [(S'S) - ']) (dB) (3) where 1r [X] is the trace of the matrix X, while Q = [q_andi]_2Lx2L = S¹ S is a correlation matrix containing S1 and S2 autocorrelations and a cross correlation between S1 and S2 with the calculation of elements as:

<img file="PL2157752T3_D0002.tif" />

[0017] Based on the definitions (1) - (3) values ​​of the SNR degradation in pairs between the GSM training sequences are calculated and detailed in Table 1.

Table 1 SNR degradation values ​​in pairs of existing GSM training sequences (dB)

<td>RSE # TSC&gt; \</td><td>0</td><td>1</td><td>2</td><td>3</td><td>4</td><td>5</td><td>6</td><td>7</td>

<td>0</td><td>|</td><td>6.91</td><td>3.24</td><td>3.08</td><td>4.75</td><td>4.87</td><td>4.85</td><td>3.88</td>

<td>1</td><td>6.91</td><td>-</td><td>3.08</td><td>? 79</td><td>5.03</td><td>4.70</td><td>4.70</td><td>3.67</td>

<td>2</td><td>3.24</td><td>3.08</td><td>-</td><td>6.91</td><td>5.57</td><td>3.97</td><td>5.12</td><td>7.16</td>

<td>3</td><td>3.08</td><td>2.72</td><td>6.91</td><td>-</td><td>4.06</td><td>4.99</td><td>4.79</td><td>6.91</td>

<td>4</td><td>4.75</td><td>5.03</td><td>5.57</td><td>4.06</td><td>-</td><td>11.46</td><td>5.87</td><td>6.11</td>

<td>5</td><td>4.87</td><td>4.70</td><td>3.97</td><td>4.99</td><td>11.46</td><td>-</td><td>3.73</td><td>5.03</td>

<td>6</td><td>4.85</td><td>4.70</td><td>5.12</td><td>4.79</td><td>5.87</td><td>3.73</td><td>-</td><td>5.72</td>

<td>7</td><td>3.88</td><td>3.67</td><td>7.16</td><td>6.91</td><td>6.11</td><td>5.03</td><td>5.72</td><td>-</td>

[0018] Mean, minimum and maximum values ​​of SNR degradation in pairs between different GSM training sequences are equal to 5.10 dB, 2.72 dB and 11.46 dB, respectively. Table 1 shows that some pairs of GSM training sequences result in good SNR degradation values, while certain pairs of GSM training sequences are strongly correlated. It seems useless to use all existing GSM training sequences for MUROS / VAMOS. It would be desirable to have new training sequences for MUROS / VAMOS, each of which would have very good autocorrelation properties and very good cross-correlation properties with the corresponding GSM training sequence.

[0019] Tables 2 and 3 show in pairs the effect of SNR degradation of the sequence in Figure 3 between any pairs of these sequences and GSM training sequences, as well as between any pairs of these sequences themselves.

Table 2 SNR degradation values ​​in pairs between any sequence pairs from FIG. 3 and GSM training sequences (in dB)

<td>TSC&gt; \</td><td>0</td><td>1</td><td>2</td><td>3</td><td>4</td><td>5</td><td>6</td><td>7</td>

<td>0</td><td>2.14</td><td>3.38</td><td>3.20</td><td>3.03</td><td>2.43</td><td>2.31</td><td>2.25</td><td>2.71</td>

<td>1</td><td>4.87</td><td>2.13</td><td>2.59</td><td>3.30</td><td>2.58</td><td>2.36</td><td>2.26</td><td>2.79</td>

<td>2</td><td>3.20</td><td>3.03</td><td>2.14</td><td>3.38</td><td>2.26</td><td>2.34</td><td>2.51</td><td>2.38</td>

<td>3</td><td>2.59</td><td>3.30</td><td>4.87</td><td>2.13</td><td>2.48</td><td>2.31</td><td>2.53</td><td>2.29</td>

<td>4</td><td>2.71</td><td>2.55</td><td>2.40</td><td>2.78</td><td>2.05</td><td>2.38</td><td>2.24</td><td>2.41</td>

<td>5</td><td>2.33</td><td>2.77</td><td>2.74</td><td>2.86</td><td>2.21</td><td>2.11</td><td>2.41</td><td>2.38</td>

<td>6</td><td>2.78</td><td>2.68</td><td>2.69</td><td>2.70</td><td>2.26</td><td>2.93</td><td>2.06</td><td>2.28</td>

<td>7</td><td>2.50</td><td>3.93</td><td>2.79</td><td>2.41</td><td>2.21</td><td>2.31</td><td>2.20</td><td>2.12</td>

Table 3 SNR degradation values ​​in pairs between any pairs of sequences from Figure 3 (in dB)

<td>go# TSC&gt; \</td><td>0</td><td>1</td><td>2</td><td>3</td><td>4</td><td>5</td><td>6</td><td>7</td>

<td>0</td><td></td><td>2.37</td><td>2.35</td><td>2.52</td><td>3.23</td><td>2.80</td><td>2.32</td><td>3.64</td>

<td>1</td><td>2.37</td><td>-</td><td>2.52</td><td>2.74</td><td>3.23</td><td>3.49</td><td>2.93</td><td>2.60</td>

<td>2</td><td>2.35</td><td>2.52</td><td>-</td><td>2.37</td><td>3.41</td><td>2.69</td><td>2.86</td><td>3.17</td>

<td>3</td><td>2.52</td><td>2.74</td><td>2.37</td><td>-</td><td>3.10</td><td>3.71</td><td>6.89</td><td>2.71</td>

<td>4</td><td>3.23</td><td>3.23</td><td>3.41</td><td>3.10</td><td>-</td><td>3.66</td><td>3.71</td><td>3.79</td>

<td>5</td><td>2.80</td><td>3.49</td><td>2.69</td><td>3.71</td><td>3.66</td><td>-</td><td>3.33</td><td>3.93</td>

<td>6</td><td>Λ ·· \</td><td>2.93</td><td>2.86</td><td>6.89</td><td>3.71</td><td>3.33</td><td>-</td><td>3.32</td>

<td>7</td><td>3.64</td><td>2.60</td><td>3.17</td><td>2.71</td><td>3.79</td><td>3.93</td><td>3.32</td><td>-</td>

[0020] In Table 2, the SNR degradation values ​​in pairs on the diagonal of the table are the results of the sequences in Figure 3 and the corresponding GSM training sequences. In this document, the respective sequences are defined as two sequences with the same training sequence number in two separate sequence tables. The average value from the diagonal in Table 2 is 2.11 dB. Mean, minimum and maximum values ​​from SNR degradation between any pairs of sequences from FIG. 3 and GSM training sequence codes are 2.63 dB, 2.05 dB and 4.87 dB, respectively. [0021] Table 3 shows that mean, minimum and maximum values ​​for SNR degradation between any pairs of different sequences in Figure 3 are 3.19 dB, 2.32 dB and 6.89 dB, respectively.

[0022] Both tables 2 and 3 show that the average SNR degradation in pairs between any sequence pairs from table 2 and GSM training sequences, as well as any pairs of different sequences from table 2 is good. However, peak SNR degradation values ​​in pairs presented in Tables 2 and 3 may affect co-channel interference reduction when introducing MUROS / VAMOS.

New training sequences for MUROS / VAMOS

A. Training Sequences Best Paired With Appropriate GSM Training Sequence Codes [0023] In an embodiment of the present disclosure, a set of eight 26-bit sequences is obtained as a result of a computer-guided search which is best paired with the corresponding GSM training sequences, respectively, in terms of SNR degradation calculated using equations (1) (3). Figure 4A shows these best matched sequences, referred to as the Training Sequence Collection A, generally designated by number 120 in the data structure 122 stored on a computer readable medium 124. The search was conducted as follows:

1) start from the first GSM training sequence;

2) comprehensively search the set of all candidate sequences for the sequence with the lowest SNR degradation, then add the found sequence to the new set, and delete the found sequence from the set of candidates;

3) repeat steps 1 and 2 for the sequences that are best paired with each GSM training sequence from the second to the eighth.

[0024] Figure 4B shows a computer readable medium, generally designated by 128, on which the data structure 125 is stored. Data structure 125 contains a set of standard GSM training sequences 126, and also contains a set of training sequences A 127. Between training sequences GSM 126 and the set of training sequences A 127 there is a one-on-one relationship.

Table 4 The SNR degradation values ​​in pairs between the sequences in the figure

4A and GSM training sequence codes (in dB)

<td>TSC&gt; \</td><td>0</td><td>1</td><td>2</td><td>3</td><td>4</td><td>5</td><td>6</td><td>7</td>

<td>0</td><td>2.09</td><td>3.05</td><td>4.10</td><td>3.08</td><td>2.88</td><td>2.52</td><td>2.49</td><td>2.64</td>

<td>1</td><td>2.60</td><td>2.09</td><td>2.67</td><td>3.18</td><td>2.56</td><td>2.44</td><td>2.57</td><td>2.67</td>

<td>2</td><td>4.10</td><td>3.08</td><td>2.09</td><td>3.05</td><td>2.36</td><td>2.80</td><td>2.57</td><td>2.35</td>

<td>3</td><td>2.67</td><td>3.18</td><td>2.60</td><td>2.09</td><td>2.62</td><td>2.37</td><td>2.66</td><td>2.19</td>

<td>4</td><td>2.47</td><td>2.53</td><td>2.27</td><td>2.69</td><td>2.04</td><td>2.34</td><td>2.26</td><td>2.27</td>

<td>5</td><td>2.15</td><td>2.15</td><td>2.64</td><td>2.53</td><td>2.18</td><td>2.07</td><td>2.42</td><td>2.22</td>

<td>6</td><td>2.38</td><td>2.30</td><td>2.48</td><td>2.50</td><td>2.28</td><td>2.42</td><td>2.05</td><td>2.24</td>

<td>7</td><td>2.55</td><td>2.72</td><td>2.32</td><td>2.25</td><td>2.37</td><td>2.51</td><td>2.19</td><td>2.07</td>

The average, minimum and maximum SNR degradation values ​​between any of the sequence pairs from FIG. 4A and the GSM training sequences are 2.52 dB, 2.04 dB and 4.10 dB, respectively. The average of the diagonal values ​​in Table 4 is 2.07 dB. Based on the results presented in Table 4, the new training sequences of Figure 4A are well designed to be paired with the corresponding GSM training sequences.

[0026] Table 5 shows the SNR degradation values ​​between the sequences detailed in Figure 4A. The values ​​of SNR degradation in pairs, average, minimum and maximum, between sequences best paired with GSM training sequences are 3.04 dB, 2.52 dB and 4.11 dB, respectively.

Table 5 The SNR degradation values ​​in pairs between the sequences in the figure

4A (in dB)

<td>T§C # tsc £ \</td><td>0</td><td>1</td><td>2</td><td>3</td><td>4</td><td>5</td><td>6</td><td>7</td>

<td>0</td><td>+</td><td>2.73</td><td>2.93</td><td>. ··&gt; $ ,? ·· '&gt; . · »» »F ·&gt; .. · *. &lt;».</td><td>-&gt; 9</td><td>2.85</td><td>2.56</td><td>4.11</td>

<td>1</td><td>2.73</td><td>-</td><td>9&gt; &lt;&gt;&gt; Α », v» * d ».</td><td>2.90</td><td>3.29</td><td>3.15</td><td>2.87</td><td>2.76</td>

<td>2</td><td>2.93</td><td>»'\» " . · *. ·· . &lt;»» F</td><td>-</td><td>2.73</td><td>3.76</td><td>2.92</td><td>3.58</td><td>3.09</td>

<td>3</td><td>2.52</td><td>2.90</td><td>2.73</td><td>-</td><td>2.58</td><td>3.17</td><td>2.80</td><td>2.95</td>

<td>4</td><td>&gt; £ ?? * »'F K»? - &lt; »-</td><td>3.29</td><td>3.76</td><td>2.58</td><td>-</td><td>2.69</td><td>3.73</td><td>4.11</td>

<td>5</td><td>2.85</td><td>3.15</td><td>2.92</td><td>3.17</td><td>2.69</td><td>-</td><td>3.64</td><td>2.70</td>

<td>6</td><td>2.56</td><td>2.87</td><td>3.58</td><td>2.80</td><td>3.73</td><td>3.64</td><td>-</td><td>2.89</td>

<td>7</td><td>4.11</td><td>2.76</td><td>3.09</td><td>2.95</td><td>4.11</td><td>2.70</td><td>2.89</td><td>-</td>

B. Training sequences with a cyclic structure with optimized properties of autocorrelation and cross-correlation.

[0027] As a result of a computer-based search, a set of training sequences with optimized autocorrelation and cross-correlation properties was determined using the method described in detail below. The set of training sequences is shown in Figure 5A, referred to as the Training Sequence Collection B, generally designated by number 130, in the data structure 132 stored on a computer readable medium 134.

[0028] Figure 5B shows a computer readable reference generally labeled 138 on which the data structure 135 is stored. The data structure 135 comprises a set of standard GSM training sequences 136, and also comprises a set of training sequences B 137. Between the training sequences GSM 136 and the set training sequences A 137 there is a one-to-one relationship.

[0029] The SNR degradation values ​​between any pairs of the new training sequences in Figure 5A and the GSM training sequences are shown in Table 6. The average, minimum and maximum SNR degradation values ​​in Table 6 are 2.43 dB, 2.13 dB and 2, respectively. , 93 dB. The mean value of the diagonal in Table 6 is 2.22 dB.

Table 6 SNR degradation values ​​in pairs between the new training sequences in FIG. 5A and GSM training sequences (in dB)

<td>xc # TSC&gt; \</td><td>0</td><td>1</td><td>2</td><td>3</td><td>4</td><td>5</td><td>6</td><td>7</td>

<td>0</td><td>2.27</td><td>2.67</td><td>2.77</td><td>2.59</td><td>2.43</td><td>2.68</td><td>2.30</td><td>2.31</td>

<td>1</td><td>2.35</td><td>2.26</td><td>2.39</td><td>2.78</td><td>2.59</td><td>2.75</td><td>2.28</td><td>2.93</td>

<td>2</td><td>2.88</td><td>2.55</td><td>2.25</td><td>2.27</td><td>2.51</td><td>2.39</td><td>2.51</td><td>2.74</td>

<td>3</td><td>2.53</td><td>2.37</td><td>2.32</td><td>2.26</td><td>2.67</td><td>2.27</td><td>2.75</td><td>2.29</td>

<td>4</td><td>2.17</td><td>2.34</td><td>2.21</td><td>2.64</td><td>2.20</td><td>2.20</td><td>2.34</td><td>2.42</td>

<td>5</td><td>2.24</td><td>2.26</td><td>2.28</td><td>2.73</td><td>2.36</td><td>2.14</td><td>2.55</td><td>2.32</td>

<td>6</td><td>2.61</td><td>2.26</td><td>2.48</td><td>2.14</td><td>2.52</td><td>2.93</td><td>2.13</td><td>2.45</td>

<td>7</td><td>2.40</td><td>2.23</td><td>2.23</td><td>2.30</td><td>2.30</td><td>2.36</td><td>2.58</td><td>2.21</td>

[0030] Table 7 shows the SNR degradation values ​​in pairs between the sequences detailed in Figure 5A. The values ​​of SNR degradation in pairs, mean, minimum and maximum in table 7 are respectively: 3.17 dB, 2.21 dB and 4.75 dB.

Table 7 SNR degradation values ​​in pairs between the new training sequences of FIG. 5A (in dB)

<td>IN TSC&gt; \</td><td>0</td><td>1</td><td>2</td><td>3</td><td>4</td><td>5</td><td>6</td><td>7</td>

<td>0</td><td>|</td><td>3.20</td><td>2.57</td><td>2.39</td><td>3.34</td><td>3.12</td><td>2.71</td><td>3.37</td>

<td>1</td><td>3.20</td><td>-</td><td>3.29</td><td>3.44</td><td></td><td>4.75</td><td>3.59</td><td>4.64</td>

<td>2</td><td>2.57</td><td>3.29</td><td>-</td><td>2.70</td><td>3.29</td><td>4.41</td><td>2.50</td><td>3.51</td>

<td>3</td><td>2.39</td><td>3.44</td><td>2.70</td><td>-</td><td>2.49</td><td>2.49</td><td>2.95</td><td>4.00</td>

<td>4</td><td>3.34</td><td></td><td>3.29</td><td>2.49</td><td>-</td><td>2.43</td><td>2.46</td><td>2.37</td>

<td>5</td><td>3.12</td><td>4.75</td><td>4.41</td><td>2.49</td><td>2.43</td><td>-</td><td>3.31</td><td>4.40</td>

<td>6</td><td>2.71</td><td>3.59</td><td>2.50</td><td>2.95</td><td>2.46</td><td>3.31</td><td>-</td><td>2.89</td>

<td>7</td><td>3.37</td><td>4.64</td><td>3.51</td><td>4.00</td><td>2.37</td><td>4.40</td><td>2.89</td><td>-</td>

C. Training Sequences Without a Cyclic Structure [0031] In contrast to the set of training sequences B, the third set of training sequences, referred to herein as a set of training sequences C, consists of sequences that do not retain a cyclic structure. To generate the set of training sequences C, only the optimization procedure II-IV for the set of training sequences B is considered below. To optimize the degradation of SNR between new training sequences and GSM training sequences, the set of W1 sequences is obtained by selecting | W1 | sequence from 2²⁶sequences with minimal mean SNR degradation values ​​between the sequences in | W1 | and all GSM training sequences. The set of training sequences C is detailed in Figure 6A, generally designated by number 140 in the data structure 142 stored on a computer readable carrier 144. [0032] Figure 6B illustrates a computer readable medium designated generally as 148 on which the data structure 145 is stored. Data structure 145 includes a set of standard GSM training sequences 146 and includes a set of training sequences C 147. There is a one-on-one relationship between the GSM 146 training sequences and the C 147 training sequence set.

[0033] The SNR degradation values ​​in pairs between any pairs of the new training sequences in Figure 6A and the GSM training sequences are shown in Table 8. The average, minimum and maximum SNR degradation values ​​in Table 8 are 2.34 dB, 2.11 dB, respectively. and 2.87 dB. The mean value of the diagonal in Table 8 is 2.16 dB.

Table 8 SNR degradation values ​​in pairs between the new training sequences in FIG. 6A and GSM training sequences (in dB)

<td>New \ Rs # GSIVK TSC # \</td><td>0</td><td>1</td><td>2</td><td>3</td><td>4</td><td>5</td><td>6</td><td>7</td>

<td>0</td><td>2.18</td><td>2.50</td><td>2.55</td><td>2.64</td><td>2.23</td><td>2.37</td><td>2.31</td><td>2.62</td>

<td>1</td><td>2.37</td><td>2.20</td><td>2.60</td><td>2.39</td><td>2.41</td><td>2.67</td><td>2.67</td><td>2.61</td>

<td>2</td><td>2.52</td><td>2.29</td><td>2.18</td><td>2.31</td><td>2.87</td><td>2.40</td><td>2.27</td><td>2.33</td>

<td>3</td><td>2.52</td><td>2.61</td><td>2.41</td><td>2.16</td><td>2.49</td><td>2.29</td><td>2.48</td><td>2.18</td>

<td>4</td><td>2.26</td><td>2.23</td><td>2.17</td><td>2.37</td><td>2.16</td><td>2.27</td><td>2.30</td><td>2.24</td>

<td>5</td><td>2.17</td><td>2.25</td><td>2.26</td><td>2.32</td><td>2.23</td><td>2.18</td><td>2.45</td><td>2.20</td>

<td>6</td><td>2.37</td><td>2.33</td><td>2.31</td><td>2.28</td><td>2.24</td><td>2.28</td><td>2.13</td><td>2.38</td>

<td>7</td><td>2.34</td><td>2.24</td><td>2.25</td><td>2.25</td><td>2.13</td><td>2.29</td><td>2.11</td><td>2.12</td>

[0034] Table 9 shows the SNR degradation values ​​in pairs between the sequences specified in Figure 6A. The values ​​of SNR degradation in pairs, mean, minimum and maximum in table 9 amount to 3.18 dB, 2.44 dB and 4.19 dB, respectively.

Table 9. SNR degradation values ​​in pairs between the new training sequences of FIG. 6A (in dB)

<td>\ Rsc # Tfeę #</td><td>0</td><td>1</td><td>2</td><td>3</td><td>4</td><td>5</td><td>6</td><td>7</td>

<td>0</td><td>-</td><td>3.62</td><td>2.74</td><td>3.24</td><td>3.80</td><td>2.90</td><td>2.65</td><td>2.81</td>

<td>1</td><td>3.62</td><td>-</td><td>3.52</td><td>2.86</td><td>4.19</td><td>2.56</td><td>3.51</td><td>3.21</td>

<td>2</td><td>2.74</td><td>3.52</td><td>-</td><td>3.42</td><td>3.00</td><td>2.99</td><td>3.06</td><td>4.02</td>

<td>3</td><td>3.24</td><td>2.86</td><td>3.42</td><td>-</td><td>2.52</td><td>2.69</td><td>3.53</td><td>3.80</td>

<td>4</td><td>3.80</td><td>4.19</td><td>3.00</td><td>2.52</td><td>-</td><td>2.44</td><td>2.92</td><td>3.12</td>

<td>5</td><td>2.90</td><td>2.56</td><td>2.99</td><td>2.69</td><td>2.44</td><td>-</td><td>3.43</td><td>3.42</td>

<td>6</td><td>2.65</td><td>3.51</td><td>3.06</td><td>3.53</td><td>2.92</td><td>3.43</td><td>-</td><td>3.00</td>

<td>7</td><td>2.81</td><td>3.21</td><td>4.02</td><td>3.80</td><td>3.12</td><td>3.42</td><td>3.00</td><td>-</td>

Sequence Search Procedure - First Method [0035] Figure 7 illustrates the procedure for searching for the second set of | W3 | W3 training sequences, which have the properties of autocorrelation and cross-correlation referring to the target set of training sequences Y. This procedure is related to the determination of the sequence sets W 152, W1 154, W2 156 and W3 158, where W o W1 o W2 o W3, and W | W3 | = number of sequences to find, where | · | represents the number of elements in a set. W 152 is a subset of the set of all possible sequences 150 determined by the first optimization step. W1 154 is a subset of W 152 determined by the second optimization stage. W2 156 is a subset of W1 154 determined by the third optimization stage. W3 158 is a subset of W2 156 determined by the fourth stage of optimization.

[0036] The method is computer-implemented and will be described with reference to the flowchart of Figure 8. The method starts in block 8-1 from the autocorrelation optimization for the candidate set of training sequences to create the first set of training sequences. The method is continued in block 8-2 for optimizing the SNR degradation between the sequences of the first training sequence collection and the target set of training sequences Y to create a second set of training sequences. The method is continued in block 8-3 to optimize the SNR degradation between the sequences of the second set of training sequences to form a third set of training sequences. The method continues in block 8-4 to optimize the SNR degradation between the training sequences of the third set and the corresponding sequences of the target training sequence Y. The result of block 8-4 is the fourth set of training sequences and is a new set of training sequences. This set is the result for use in a multi-user system. Another embodiment is a computer readable medium having stored instructions that, when executed by a computer, implement the method of FIG. 8.

[0037] In some embodiments, steps 8-2 and 8-3 are performed in reverse order to what is shown and described above. This results in a computer method comprising: optimizing cross-correlation between sequences from the first set of training sequences to create a second set of training sequences; optimizing cross-correlations between the sequences of the second set of training sequences and the target set of training sequences to form a third set of training sequences; optimizing cross-correlations between the sequences of the third set of training sequences and the corresponding sequences of the target set of training sequences to create a fourth set of training sequences for use in a multi-user transmission system.

[0038] The first optimization step: autocorrelation optimization: Consider all binary sequences of the desired length. Optionally, copy some of the last bits of the sequence to the beginning in order to give the sequence a bit cyclicality. Search for sequences with zero autocorrelation values ​​for the range of non-zero offsets.

[0039] In order to obtain zero autocorrelation, the sequences on which autocorrelations are determined must have an even length. If a sequence of an odd length is required, then an additional bit is added to the sequence set with zero autocorrelation values. This can be done, for example, by copying the first bit of the original sequence or by including -1 or +1 to further optimize the cross-correlation properties.

[0040] The result of this step is a set of W sequences with optimized autocorrelation properties. For sequences to which an additional bit has been added, the maximum magnitude of the autocorrelation coefficient will be 1 in the correlation matrix Q = S^t S in equation (3);

[0041] The second optimization step: optimization of the SNR degradation between the new sequences and the target set of training sequences Y: the subset W, W1 is obtained by selecting | W1 | sequences from W with minimum mean values ​​of SNR degradation between the sequences in | W1 | and the target pool of training sequences Y. The average degradation of the SNR for a given sequence from W is determined by calculating the degradation for this sequence and each of the training sequences from the target set and averaging the result.

[0042] Third optimization stage: optimization of the SNR degradation between the new sequences: a subset of the set W1, W2 with the minimum average values ​​of SNR degradation between the sequences in | W2 | is selected. The following example of the third stage of optimization is given:

1) select the first sequence from the set W2 and delete it from W2;

2) examine all the remaining sequences in the set W2 to find one with the lowest SNR degradation with this first sequence and select it as the second sequence and then delete it from the W2 file;

3) examine all remaining sequences in the set W2 for one with the lowest mean SNR degradation with the first sequence and the second sequence and delete it from the set W2;

4) and so on until the desired number of sequences is identified. Calculate the average SNR rate deviation between the sequences thus identified;

5) repeat steps 1 to 4 using a different first sequence from set W2 to generate a corresponding set of sequences and corresponding average SNR degradation;

6) from all sequence sets so generated, select a set of sequences with a minimum average SNR degradation.

[0043] The fourth optimization step: optimization of the SNR degradation between the new training sequences and the respective sequences of the target set of training sequences Y: The | W3 | sequences from the W2 sequence set. This stage is used to determine pairs of training sequences that contain one of the target set and one of the new set. Next is an example of this stage:

a) select the training sequence from the target set;

b) find the training sequence in a new set that has the lowest SNR degradation with the training sequence of the target set and pair that training sequence with the first training sequence from the target set, and remove this training sequence from the set of available sequences;

c) repeat steps a) and b) until all sequences from the target set are selected.

[0044] The optimization procedure described above was used to develop the set of training sequences of Figure 5A from | W1 | = 120 and | W2 | = 12. In particular:

The first stage of optimization: autocorrelation optimization: Consider all binary sequences with a length of 20 (the size of the set is 2²⁰). Similar to the GSM training sequence codes, for each such sequence, copy the last 5 bits of the sequence and precede these 5 bits at the most significant points to generate a sequence of 25; search for sequences of 25 with zero autocorrelation values ​​for a non-zero shift [-5, 5] by using the au definition, and ·· ^ (A)⁼ = -5, ..., - 1 .... ..

tokorelation. In total, 5,440 such sequences are available.

[0045] In order to be compatible with the current training sequence code format, the length of the new training sequence codes must be 26. 26-bit full-length sequence (length 26) can be obtained either by copying the first bit of the respective sequence of 20, or by attaching -1 or +1 to further optimize cross-correlation properties. Therefore, a set of W sequences with optimized autocorrelation properties is generated. Both methods will limit the maximum size of autocorrelation coefficients to 1 in the correlation matrix Q = S^t S in equation (3).

[0046] The second optimization step: optimization of the SNR degradation between the new sequences and the GSM training sequence codes: a subset of the set W, W1 is obtained by selecting | W1 | sequences with W with minimum mean values ​​of SNR degradation between the sequences in | W1 | and all GSM training sequence codes. The average SNR degradation for a given sequence with W is determined by calculating the degradation for this sequence and each of the GSM sequences and averaging the result.

[0047] Third optimization stage: optimization of the SNR degradation between the new sequences: a subset of either W1, W2 with minimum mean values ​​of SNR degradation between the sequences in | W2 | is selected.

[0048] The fourth optimization step: optimization of the SNR degradation between the new training sequences and the corresponding training sequence codes:

It is determined | W3 | = 8 sequences from the sequence set W2. The result is the set B of the sequence in Figure 5A.

Sequence finding procedure - Method Two [0049] According to another sequence search method, a solution similar to the "first method" described above is omitted, in which the first optimization step is omitted. In this case, the method starts with the second optimization step of the first method and the W1 sequence is obtained by selecting | W1 | sequences from all possible sequences with minimum mean values ​​of SNR degra- dation between the sequences in | W1 | and all sequences from the target Y set. It should be noted that in this embodiment, the sequences are not required to be cyclic. By using the language of the flowchart of Figure 8, block 8-1 is skipped and the "first set of training sequences" becomes the candidate set of training sequences.

[0050] When applied to the MUROS / VAMOS problem, in the second optimization step, the set of W1 sequences is obtained by selecting | W1 | sequences from all 2²⁶26 possible sequences with minimum average SNR degradation values ​​between the sequences in | W1 | and all GSM training sequences. The result is a set of sequences C from Figure 6A above.

ASSIGNMENT OF TRAINING SEQUENCES After defining a new set of training sequences or parts of a new training sequence set for use with a target training sequence set, e.g. a set A defined above or part of the A set together with the original GSM training sequences, a B set or parts defined above set B together with the original GSM training sequences or the C set or part of the C set defined above, together with the original GSM training sequences, various mechanisms are used to assign the training sequences. It should be noted that these mechanisms are not specific to the examples given in the present description. A specific example of a multi-user work is, for example, the MUROS / VAMOS operation described above, whose specific implementations include the OSC or co-TCH technique or their implementations of Adaptive Symbol Constellation. [0052] In a cell in which multi-user transmission is performed, interference from at least two sources occurs in interference-limited scenarios. This includes interference from another (different) user (s) on the same physical transmission resource within the cell, as well as interference from mobile stations of the same physical transmission resource in other cells. Conventional mobile stations are already equipped to deal with interference from mobile stations that use the same physical transmission resource in other cells. [0053] A mobile station that is particularly aware of multi-user work will be referred to as "multi-user aware". In a specific example, a mobile station that is aware of working in the VAMOS mode may be referred to, for example, as a mobile station aware of the VAMOS mode. Such mobile devices are configured to be able to use each training sequence from the target set and each training sequence from the new set. Mobile stations that are not particularly aware of multi-user work will be referred to as "multi-user unconscious". Such mobile devices are configured to be able to use only training sequences from the target set. It should be noted that multi-user unconscious mobile stations can still be handled in a multi-user context;

[0054] Likewise, networks may have or not the ability to work with a multi-user. A network that has the ability to work multi-user works with the use of a target file and a new set of training sequences, while a network that does not have the ability to work multi-user, uses only the target set of training sequences.

[0055] In some embodiments, the assignment of training sequences to base stations is performed during network configuration and does not change until a reconfiguration is performed. A multi-user aware network element, e.g. a base station, is configurable with a training sequence from a target set and a training sequence from a new set. In the event that the base station performs a multi-user transmission, the base station is configured with the corresponding training sequence from the target set and the corresponding training sequence derived from the new set for each carrier frequency it uses. The training sequence from the new set is a training sequence that is best paired with the training sequence from the target set and vice versa. In this case, the training sequences assigned to the mobile stations will be a function of the previously performed network configuration. When the mobile station moves between the coverage areas, the assigned training sequences will change. In some embodiments, the same training sequence is assigned to a given mobile station for both uplink and downlink transmissions. In other embodiments, different training sequences may be assigned. In some embodiments, the same training sequence is assigned to a given mobile station for both uplink and downlink transmissions. In other embodiments, different training sequences may be assigned. In some embodiments, the same training sequence is assigned to a given mobile station for both uplink and downlink transmissions. In other embodiments, different training sequences may be assigned.

[0056] Network behavior can be divided into two types: behavior, when the time slot is to be used only for one user, and behavior when the time slot is to be used by many users.

[0057] Behavior in case when the time slot is to be used for one user:

A) When a multi-user aware mobile station is to be served by a network with no multi-user capability, one training sequence from the target set will be assigned to that mobile station, namely the training sequence assigned to the serving base station during network configuration or otherwise. A multi-user aware mobile station becomes aware and able to use both the set of target training sequences and the new set of training sequences.

B) When a multi-user aware mobile station is to be served by a network capable of multi-user operation, if there is a free time slot, then this mobile station does not have to share the time slot with another mobile station. As the average new training sequences have been designed to have better correlation properties than the training sequences from the target set, one new training sequence will be assigned to this mobile station, namely a new training sequence assigned to the serving base station during network configuration or otherwise.

[0058] Behavior when a timeslot is to be used for multiple users:

A) When the first multi-user aware mobile station, MS-A, is served by a multi-user network, as described above, a new training sequence is assigned to the MS-A mobile station, namely a new training sequence assigned to the the serving base station during network configuration or otherwise. If there is a request to share the same time slot with the second MS-B mobile station, regardless of whether the MS-B mobile station is a multi-user aware mobile station, then the training sequence from the target set will be assigned to the MS-B mobile station. which is best paired with this new training sequence used by the MS-A mobile station,

B) When the first multi-user un-aware MS-A is supported by a multi-user network, a training sequence from the target set is allocated to the MS-A mobile station, namely the training sequence from the target set allocated to the serving base station in during network configuration or otherwise. If there is a request to share the MS-A mobile station with a second mobile station that is multi-user aware, MS-B, in the same time slot, then a new training sequence will be assigned to the MS-B mobile station that is best paired with the sequence. training of the target set used by the MS-A mobile station, namely the training sequence assigned to the serving base station during network configuration or otherwise.

[0059] Two flowcharts of an exemplary method of assigning a training sequence using multi-user slots are illustrated in Figures 9A and 9B. Figure A relates to case A) described above, and figure 9B relates to case B) described above.

[0060] Referring now to Figure 9A, when the first multi-user aware mobile station is to share with the second mobile station, block 9A-1 assigns a new training sequence to the first mobile station. Block 9A-2 assigns to the second mobile station a training sequence from the target set that is best paired with the new training sequence used by the first mobile station, regardless of whether the second mobile station is a multi-user aware mobile station or not. ;

[0061] Referring to Figure 9B, when the first multi-user unaware mobile station is to share with a second mobile station that is multi-user aware, block 9B-1 assigns to the first mobile station a training sequence from the target set. Block 9B-2 is associated with assigning to the second mobile station a training sequence from a new set that is best paired with the training sequence used by the first mobile station.

Example implementations of the transmitter and receiver [0062] Detailed example implementations of the transmitter and receiver will now be described. Figure 10A shows an OSC transmission (orthogonal subchannels) (or Adaptive Symbol Constellation) for downlink transmission. Most of the components are standard and will not be described in detail here. The transmitter includes a training sequence repository 200, comprising a target training sequence set and a new training sequence set generated using the above-described methods. Usually, during the network configuration or otherwise, the trainer is assigned a training sequence from the target set and a training sequence from the new set. In some embodiments for multi-user slots, the training sequences thus assigned are used in accordance with the method of Figure 9A or 9B to provide several specific examples. The remaining part of the drawing is a specific example of a signal generator de generating a multi-user signal.

[0063] Figure 10B shows a pair of OSC receivers (orthogonal subchannels) for downlink transmission. Most components are standard and will not be described in detail. Each receiver includes a corresponding memory 210 including a target training sequence set and a new training sequence set generated using one of the methods described above. One of them, referred to as the training sequence A, is used by the upper receiver to perform the channel timing / estimation and DARP processing. The training sequences used by the receivers are assigned by the network and must match the transmitted training sequences. When the mobile station moves to another coverage area to which other training sequences have been assigned, then the mobile station appropriately changes the training sequence,

[0064] Figure 11A shows a co-TCH transmitter (common talk channel) for downlink transmission. Most components are standard and will not be described in detail here. The transmitter includes a training sequence repository 200 containing the target training sequence set and a new training sequence set generated using one of the above-described methods. Usually, during the network configuration or otherwise, the training sequence from the target set and the training sequence from the new set are assigned to the transmitter. In some embodiments for multi-user slots, the training sequences thus assigned are used in accordance with the method of Figure 9A or 9B, listing only a few specific examples.

[0065] Figure 11B shows a pair of co-TCH receivers (common talk channel) for downlink transmission. Most components are standard and will not be described in detail. Each receiver includes a corresponding memory 210 containing a target training sequence set and a new training sequence set generated using one of the methods described above. One of them, referred to as training sequence A, is used by the upper receiver to perform channel timing / estimation and DARP processing, while the second one, referred to as training sequence B, is used by the lower receiver to perform channel timing / estimation and DARP processing. . The training sequences used by the receivers are assigned by the network and must match the transmitted training sequences.

[0066] Figure 12A shows a pair of OSC or co-TCH transmitters (common talk channel) for uplink transmissions. Most components are standard and will not be described in detail here. Each transmitter includes a training sequence repository 210 containing a target training sequence set and a new training sequence set generated using one of the methods described above. Both transmitters use the same carrier frequency and the same time slot. This is similar to the multi-user signal generated by the network, but in this case the corresponding components are generated in the appropriate mobile stations, not in a single transmitter. The training sequences used for uplink transmissions are assigned by the network and will change,

[0067] Figure 12B shows a reception device constructed from two receivers for receiving respective transmissions from a pair of mobile stations of Figure 12A in OSC or co-TCH for uplink transmission. Most components are standard and will not be described in detail here. The receivers include a training sequence repository 100 containing a target training sequence set and a new training sequence set generated using one of the above-described methods. One of them, referred to as training sequence A, is used by the upper receiver to perform the timing estimation, combined channel estimation / detection, and another of them, referred to as training sequence B, is used by the lower receiver to perform the estimation of timing, combined estimation / channel detection.

[0068] In some embodiments, the approaches described herein are used to create training sequences for the GSM frame format described with reference to Figure 1. More generally, these approaches can be used to transmit frame formats in which content for a given user includes at least a corresponding one. training sequence and data block, wherein the data block simply contains any content that does not include a training sequence.

[0069] In all of the described embodiments, the SNR degradation was used as a criterion for optimizing the cross-correlation properties of the sequence. More generally, another optimization criterion can be used to optimize the sequence correlation properties of a sequence. Specific examples include:

1) parameters related to the amplitude of cross-correlation coefficients (maximum value, average value, variance and etc.);

2) optimization based on simulations;

3) other criteria for correlation optimization;

[0070] In some embodiments, each base station is coded with the entire target training sequence set and the entire new set of training sequences. For example, for the purposes of transmitting a training sequence repository, the base station 200 may be configured with the entire target training sequence set and the entire new set of training sequences.

[0071] In another embodiment, each base station, or more generally speaking, each transmitter, is encoded with at least one training sequence derived from the target training sequence set (e.g. one training sequence among all training sequences), and from the least one training sequence derived from a new set of training sequences (e.g. one training sequence from all training sequences). The training sequence (se23 training sessions) from the new set of training sequences may include a training sequence (training sequences) from a new set of training sequences that is best paired with the training sequence (training sequences) from the target training sequence set. In some embodiments, each base station is configured with such training sequences when setting the network. [0072] A transmitter or receiver encoded with at least one training sequence derived from a set consisting of a target set and at least one training sequence derived from a set consisting of a new set of training sequences may also be encoded with one or more of the sequences. training sequences other than the target set of training sequences and a new set of training sequences.

[0073] In some embodiments, each receiver, e.g. each mobile station, is coded with at least one (e.g. one or all) training sequence from the target training sequence set and at least one (transceiver one or all) sequence training from a new set of training sequences. For example, for the purpose of transmitting the training sequence repository 210, the mobile station may be configured with at least one target training sequence and at least one new training sequence. In the case of encoding with all training sequences from the target set of training sequences and a new set of training sequences, this will allow the mobile station to perform suspensions between the base stations,

[0074] The transmitter or receiver that is encoded with the training sequence is a transmitter or receiver that has a training sequence that is somehow stored and usable by this transmitter or receiver. A transmitter or receiver, e.g. a base station or a mobile station having a specific training sequence, is a transmitter or receiver that is capable of using a particular training sequence. This is not associated with the active step of storing the training sequence on the mobile station, although this may be preceded by such an active stage. It can be previously stored, for example, during the configuration of the device.

[0075] In view of the above description, numerous modifications and variations of the disclosure are possible. Therefore, it should be understood that embodiments other than those described herein in particular may be implemented within the scope of the appended claims.