GB2389757A - Multiplexing multiple users into a single multibit 8-PSK symbol for enhanced capacity radiocommunications - Google Patents

Abstract

A radiocommunication system supports multi-user multiplexing. Half-rate data communications with interleaving of data between two sources are described. This multiplexing technique provides, effectively, two sub-channels of information, which in turn provides an opportunity to improve decoding/demodulation of each sub-channel by re-encoding/remodulation information after successful decoding/demodulation of one sub-channel. In particular a scheme where data from different sources (A, B) is multiplexal into different bits of a multibit symbol is proposed. The symbol may be an 8-PSK symbol.

Description

( WO 99/59269 2389757 19/00798

METHOD AND SYSTEM FOR MULTIPLEXING Or MULTIPLE USERS FOR ENHANCED CAPACITY RAD IOCOMMUNICATIONS

BACKGROUND

5 Applicant's invention relates to electrical telecommunication, and more particularly to wireless communication systems, such as cellular and satellite radio systems, for various modes of operation (analog, digital, dual mode, etc.), and access techniques such as frequency division multiple access (FDMA), time divisional multiple access CIDMA), code divisional multiple access (CDMA), hybrid 10 FOMAITDMA/CDMA, for example. More specifically, this invention relates to methods and systems which detect multiple information streams which are transmitted as a composite signal in a Owner intended to improve the detection of the individual streams. In North America, digital communication and multiple access techniques such as 15 TDMA are currency provided by a digital cellular radiotelephone system called He digital advanced mobile phone service (DAMPS), some of the characteristics of which are specified in the interun standard TIA/EIA/IS-54 B. Dual-Mode Mobile Station Base Station Compatibility Standards, published by the Telecommunications Industry Association and Electronic Industries Association A/DA) and some of which are 20 specified by the later interim standard IS-136 (which describes, among other things a digital control channel) which standards are expressly incorporated herein by reference.

Because of a large existing consumer base of equipment operating only in the analog domain with frequency-division multiple access (PDMA), TIA/EIA/IS-5tB is a dual mode (analog and digital) standard, providing for analog compatibility together win 25 digital communication capability.

-2 In a TDMA cellular radiotelephone system, each radio channel is divided into series of time slots, each of which contains a burst of information from a data source, e.g., a digitally encoded portion of a voice conversation. The time slots are grouped into successive TDMA frames having a predetermined duration. The number of time slots in each TDMA frame is related to the number of different users that can simultaneously share the radio channel. If each slot in a TDMA frame is assigned to a different user, the duration of a IDMA frame is the minimum amount of time between successive tune slots assigned to the same user.

The successive time slots assigned to the same user, which are usually not 10 consecutive tune slots on the radio carrier, constitute the user's digital traffic channel (DTC), which may be considered a logical channel assigned to the user. In only one of many possible embodiments of a TDMA system as described above, the llA/ELA/IS-

54-B, IS-136 standards provided that each TDMA frame consists of six consecutive time slots and has a duration of 40 milliseconds (msec) as illustrated in Figure 1.

15 Thus, each radio channel can carry from three to six DTCs (e.g., three to SLY telephone conversations), depending on the source rates of the speech coder/decoders (codecs) used to digitally encode the conversations.

Such speech codecs can operate at either full-rate or half-rate. A fullrate DTC.; requires twice as many time slots in a given time period as a half-rate DTC, and in 20 TIA/ELA/IS-5=B and IS-136, each full-rate DTC uses two slots of each TDMA frame, i.e., the first and fourth, second and fifth, or third and sixth of a TDMA frame's six slots. Each half-rate DTC uses one time slot of each TDMA frame. During each DTC tune slot, as seen in Figure 2, 324 bits are transmitted, of which the major portion, 260 bits, is due to the speech output of the coder, including bits due to error correction 25 coding of the speech output, and the remaining bits are used for guard tunes and overhead signalling for purposes such as synchronization.

Once information has been output from the speech codec, it is then processed for transmission on a radio carrier. This processing can be generalized as illustrated in the upper branch of Figure 3. Therein, channel coding 30 and interleaving 32 are 30 provided to protect against channel errors which corrupt the information as it is

it transmitted over the radio channel 36. Channel coding, e.g., block coding or convolutional coding, adds redundancy to the information stream which can be used to identify and correct errors which occur during transmission of the information over the radio channel. Bit errors which occur due to transmission over the radio channel 5 frequently occur in bursts. However, certain types of channel coding are most effective at correcting single bit errors and are less effective at correcting long strings of erroneously recoined bits. Accordingly, interleaving is used to separate consecuti ie information bits and transmit them in a nonconsecutive manner. In this way, burst errors are effectively spread out so that when the received information is de 10 interleaved, the channel coding is more likely to be able to correct the errors which occurred during transmission.

Different systems use different types of channel coding and interleaving. For example, systems designed in accordance with the IS-136 standard system described above can provide for channel coding and interleaving, according to one vocoder 15 described therein, as illustrated in Figure 4. Therein, output of the speech coder 40 is separated into class 1 and class 2 bits, class 1 bits being more important than class 2 bits in terms of the perceived signal quality upon reproduction and, therefore, being more heavily protected against errors. In fact, to further protect the 12 most perceptually sigrufcant class 1 bits, a 7 bit cyclic redundancy check (CRC) is computed 20 over those 12 bibs at block 42 and added to the string of bits to be convolutionally encoded at block 44. In convolutional encoding, an output, coded bit depends not only on the bit value of a most recently input bit, the bit values of preceding bits, which provides a form of memory that can be used to detect errors in the received signal stream. The rate of convolutional coding, in this example 112, denotes the amount of 25 redundancy provided-in this case for every input Formation bit, two coded bits are produced. The coded class 1 bits and the uncoded class 2 bits are then ciphered (block 46) and interleaved (block 48) over two time slots as shown in Figure S. Thus, bits from each of two speech frames are transmitted in each time slot of D-AMPS systems to spread out burst errors as described above.

Returning to Figure 3, the output of interleaver 32 is sent to modulator 34 wherein the data is modulated onto the radio frequency carrier. In the D-AMPS example described above, the particular modulation which is currently used is J4 shited, differentially encoded quadrature phase shift keying (DQPSK). In this scheme, 5 as will be appreciated by those skilled in the art, information modulation is achieved by relative changes in phase of the modulating waveform. Grey coding is used in the constellation mapping (described below) of symbols to all-bits so that adjacent signal changes differ by only one bit. In this way, noise errors which result in the erroneous selection of a symbol associated with an adjacent phase create only a single bit error.

10 Once the information is modulated, some post-processing (e.g., filtering and amplification) may be performed and the information is then transmitted over the radio channel. For completeness, Figure 3 also indicates functional blocks associated with a receiver, e.g., in a mobile station, that process the received signal. Therein, 15 demodulator 38, deinterleaver 41 and decoder 43 effectively reverse the processes performed by modulator 34, interleaves 32 and channel encoder 30, respectively.

Those skilled in the art will be familiar with the operation of these devices and, therefore, they are not further described herein. An optional equalizer 39 (or RAKE -' receiver, e.g., for a DS-(:DMA system) may also be included in the signal processing 20 path. This device handles the effects of signal reflections which occur during transmission of the information over the radio channel by, e.g., creating a model of the channel and attempting to determine the most likely transTrutted sequence in view of the various echoes received during some reception interval.

As mentioned earlier, infonnation in IS- 136, as well as many other systems, can 25 be transmitted at full or half rate. Half rate communications provide an opportunity for additional capacity in terms of the number of connections, since each frame supports, for example, six channels instead of three. However, the two-slot interleaving described above is applicable only to full rate transmissions, since half rate transmissions only use one time slot per frame. Thus, current implementations of half

-5- rate communications in IS-136 systems provide for no inta-slot interleaving and, accordingly, suffer from burst error correction problems.

Accordingly, it would be desirable to identify solutions to provide for half rate communications which overcome these problems and drawbacks. More generally, it 5 would be desirable to provide systems and methods which consider multi-user detection wherein multiple users or sources transmit information in an overlaid or interleaved manner. SUMMARY

10 These and other drawbacks and limitations of conventional methods and systems for communicating information are overcome according to the present invention, wherein Applicants present techniques and systems for multiplexing two users or sources iron each time slot at half rate. This multiplexlug technique provides, effectively, two suchannels of information, which in turn provides an opportunity to 15 improve decoding/demodulating of each suchannel by re-encoding/remodulating information after successful decoding/demodulating of one sub-channel. Various modulation constellations are described which take advantage of the fact that certain bits in each symbol may be known.

BRIEF DESC}tIP1'ION OF 1l nit; DRAl7VINGS 20 These and other objects, features and advantages of the present invention will become more apparent upon reading from the following detailed description, taken in

conjunction with the accompanying drawings, wherein: FIG. 1 illustrates an exemplary frame format in a conventional IS-136; FIG. 2 illustrates both an uplink (upper) and a downlink (lower) tune slot 25 format for a conventional IS-136 system; FIG. 3 is a functional block diagram of an exemplary, conventional radiocommunication system; FIG. 4 depicts channel coding and interleaving according to a conventional IS 136 system;

( FIG. 5 illustrates two slot interleaving for full rate communications in a conventional IS-136 system; FIG. 6 is a block diagram of an exemplary mobile station and base station in a radiocornmunication system; 5 FIG. 7 illustrates half rate commurucation involving sub-channels from two users or sources according to exemplary embodiments of the present invention; FIG. 8 illustrates an exemplary symbol multiplexing embodiment of the present invention; FIG. 9 depicts an exemplary inter-symbol multiplexing embodiment of the 10 present invention; FIGS. 114(a) illustrate various constellation mappings associated with 8-PSK modulation; O FIG. 15 depicts speech and FACCH interleaving and chaMel coding according to one exemplary embodiment of the present invention; and 15 FIGS. 16(a) and 16(b) depict exemplary downlink formats for 8-PSK and 8 DPSK modulated transmissions, respectively.

DETAILED DESCRIPTION

The following exemplary embodiments are provided in He context of Tl)MA radiocommmucation systems. However, those skilled in the art will appreciate that this 20 access methodology is merely used for the purposes of illustration and that the present invention is readily applicable to various different types of access methodologies including, for example, hybrid TDMA/code division multiple access (CDMA).

Figure 6 is a block diagram of an exemplary cellular mobile radiotelephone system in which the present invention can be implemented, including an exemplary 25 base station 110 and mobile station 120. The base station includes a control and processing unit 130 which is connected to the MSC 140 which in turn is connected to the PSTN (not shown). General aspects of such cellular radiotelephone systems are known in flee art, as described by the above-cited U.S. patent applications and by U.S. Patent No. 5,175,867 to Wedge et al., entitled 'Neighbor-Assisted Handoff in a

l' -7 Cellular Communication System," and U.S. Patent Application No. 07/967,027 entitled "Multi-mode Signal Processing," which was filed on October 27, 1992, both of which are incorporated in this application by reference.

The base station 110 handles a plurality of voice channels through a voice 5 channel transceiver 150, which is controlled by the control and processing unit 130.

Also, each base station includes a control channel transceiver 160, which may be capable of handling more than one control channel. The control channel transceiver 160 is controlled by the control and processing unit 130. The control channel transceiver 160 broadcasts control information over the control channel of the base 10 station or cell to mobiles locked to that control channel. It will be understood that the transceivers 150 and 160 can be implemented as a single device, like the voice and control transceiver 170, for use with DCCs and DTCs that share the same radio carrier frequency. The mobile station 120 recedes the information broadcast on a control channel 15 at its voice and control channel transceiver 170. Then, the processing unit 180 evaluates the received control channel information, which includes the characteristics of cells that are candidates for the mobile station to lock on to, and determines on which cell the mobile should lock. Advantageously, the received control channel information not only includes absolute information concerning the cell with which it is associated, 20 but also contains relative information concenung other cells proximate to the cell with which the control channel is associated, as described in U.S. Patent No. 5,353,332 to Raith et al., entitled Method and Apparatus for Communication Control in a Radiotelephone System, n which is incorporated in this application by reference.

As mentioned earlier, full rate communications in exemplary IS-136 systems 25 provide for interleaving information from one user or source over each time slot assigned to that user (i.e., two for IS-136) within each TDMA frame as illustrated in Figure 5. To extrapolate this type of interleaving to half rate communication as it is defined for IS-136 systems, i.e., each half rate user has only one time slot per frame, would suggest that information transmitted at half rate be interleaved over time slots of 30 different TDMA frames. However, interleaving infonnation over time slots of

-8 different TDMA for half rate communications would add an additional delay in the reproduction of the signal for half rate communications as opposed to full rate communications. Accordingly, it would be desirable to provide some fo m of interleaving for half rate communications to spread out burst errors which occur du -! 5 the transmission of information between, for example, mobile station 120 and base station 110 without increased delay.

According to exemplary embodiments of the present invention, therefore, interleaving of information can be provided for half rate users sources between time slots of the same frame by interleaving information of different users in each time slot 10 on the downlink. This provides the same interleaving delay at the receiver as full rate communication. Consider the exemplary downlink frame format of Figure 7 to illustrate this concept.

Therein, information associated with portions of speech frames X, and YA for user A and portions of speech frames XB and YB for user B are multiplexed together 15 and transmitted in time slot 1. Portions of speech frames Ye and ZA for user A and portions of speech frames YB and ZB are multiplexed together and transmitted in time slot 4. Similar multiplexing of information from other users can be performed in time slots 2 and 5, and 3 and 6, such that six half-rate channels are still provided in each TDMA frame. By spreading the half rate information of both users A and B across 20 time slots 1 and 4, burst errors are more readily corrected.

While this type of interleaving according to the present invention can be provided on the dowinlc it is not feasible for the uplink (mobile-to-base direction) due to the impossibility of tirmng the transmissions of nongeographically co-located transntters associated with the various mobile stations. Despite the fact Cat the uplink 25 is not conducive to this type of interleaving, mobile stations should be provided with some type of information regarding upon which uplink time slot they should transmit their data. Consider that conventional half rate communications provided one downlink time slot for each mobile station and, therefore, each mobile station simply transmitted its uplink burst in a corresponding time slot. However, since according to the present 30 invention each mobile station may now receive half rate information on two time slots

in each frame, the mobile stations must implicitly or explicitly recognize their assigned uplink time slot. An example of an explicit uplink time slot assigmnent is provided below m a discussion of mode signalling There are various ways in which the information associated with users A and B 5 can be interleaved within each time slot. For example, interleaving can be performed on a symbol-by-symbol basis, which type of interleaving is referred to herein as "symbol multiplexingn. Symbol multiplexing can take many forms, however generally bits associated with user A's information stream are allocated to be carried by one or more symbols, followed by bits associated with user B's information stream being 10 allocated to one or more symbols, in some repeating pattern. Symbol multiplexing could take the form of every other symbol having bits therein solely associated with a user's information stream, two symbols at a time having information bits solely associated with a user's bit stream (as shown in Figure 8) or any desired pattem.

Symbol multipleung could have varying pattern of repetition, e.g. AAl3ABBAABABB, 15 etc., as long as the pattern was known apnor' to all remote units.

Interleaving of data from two different users in each time slot can also be performed on a bit-by-bit basis rather than a symbol-by-symbol basis. This can be achieved according to exemplary embodiments of the present invention by allocating one or more bits associated with user A's information stream to a symbol and filling 20 the remaining positions in the symbol with one or more bits from user B's information stream. An example of this type of inter-symbol multiplexing is illustrated in Figure 9.

Therein a portion of a time slot is shown wherein three bit symbols are used to transmit information. Thus, the first symbol has two bits of information from user A's information stream and one bit of information from user B's information stream, while 25 the second symbol has two bits of information stream from user B's information stream and one bit of information from user A. Like the previous embodiment which used symbol multiplexing to provide a mechanism for interleaving the data from two different users in two different time slots in one frame, inter-symbol multiplexing can also be used to distribute data associated with each user across spaced apart tune slots 30 in one frame to protect against burst errors in half rate communications.

-1 Those skilled in the art will appreciate that multiplexing bits from different users in one time slot effectively creates two sub-channels within each time slot. In addition to assigning these sub-channels to different users, it should be appreciated that these sub-channels could also be assigned to different connections associated with be 5 user. For example, information from both a voice connection and a data connectic associated with one user could be transmitted at half rate by interleaving information from each connection over multiple time slots in a frame using either of the symbol multiplexing or inter-symbol multiplexing techniques described above. Likewise two voice or two data connections from one user could be transmitted in a similar fashion.

to Since the preceding exemplary embodiments rely on information associated with two data sources, e.g., users, to be multiplexed over each time slot, a question arises as to how to handle the case where only one active user is being transmitted to by a base station. Various solutions are possible. First, this one active user could be switched to a full rate format (including use of a full rate vocoder). Of course, the 15 mobile station would have to be informed of the switch, which is essentially equivalent to a handoff from half rate to full rate coTmaunicatiors. This information can be passed to the mobile station using the mode signalling described below. Secondly, the two subchannel concept described above could be replaced in this case by using the same vocoder as for the two sub-channel approach but applying more channel coding to fin 20 in the bits normally supplied by the source of the second channel.

A third solution is to rnaintam the two suchaTnels and send arbitrary or predeffned data as the non-existent Uuser B's" information. In this way, the predefined, half rate channel coding structure and existing vocoder can be used even when only one active user is receiving information on a half rate downlink channel. Another 25 possibility is to keep the channel coding structure, but eliminate the vocoder and subsitute predefined data where the speech coded bits would normally occur. Still another option is to eliminate the channel coding and the vocoder and simply fill user B's Formation stream with predefined bits. Lastly, the second sub-channel could be filled by repeating each information frame transmitted by the one active user, i.e., 30 copying the active user's speech frames into the second sub-channel.

- 1 1 Multi-user multiplexing of information described in the foregoing exemplary embodiments can be used to provide interleaving and reduce the effects of burst errors in half rate radiocormnunications. However, multiuser multiplexing also provides the opportunity for multi-user detection, i.e., using the decoded information from one sum 5 channel to aid in the decoding of information in another sub-channel. This aspect of the present invention will now be described.

The symbol multiplexing and inter-symbol multiplexing described above may also be used to provide additional gain to the decoding/demodulation process.

Generally, given that two subhannels are multiplexed together in one time slot and a 10 radio channel that is subject to fading and interference, it is likely that at least one of the two sub-channels can be decoded correctly. For example, user A's mobile station will receive, and attempt to decode/denodulate the information in its assigned time slots within each frame. User A may be able to successfully decode its own information, after which no additional processing would be needed. if, however, user 15 A is unable to decode its own informadon, e.g., due to a CRC failure, it may then attempt to decode user B's data which has been multiplexed with user A's information.

If user B's can be successfully decoded/demodulated, then the entire burst can be re-

encoded and redemodulated and user A's equipment can take a second pass at attempting to decode its information using the known symbols and/or bits of user B's 20 information.

A process for using known information symbols within an information stream to aid in the decoding/demodulation process of unknown symbols per se has been described in earlier literature and, therefore, is not described in detail herein. The interested reader is, however, directed to U.S. Patent No. 5,673,291, entitled 25 Simultaneous Demodulation and Decoding of a Digitally Modulated Radio Signal Using Known Symbols" and issued on September 30, 1997 and to International Patent Publication No. WO 98/04047 entitled Method and Apparatus for Detecting Communication Signals Having Unequal Error Protections, and published on January 29, 1998, the disclosure of each of which is incorporated by reference.

- -: -: - -

-12 In the context of symbol multiplexing, e.g., as shown in Figure 8, this process of decoding/demodulating one sub-channel and then reencoding/remodulating and decoding/demodulating the other suchannel is straightforward. User A's equipment will first de-interleave and attempt to decode/demodulate the 'A' symbols. If 5 successful, the process ends. Otherwise, user A's equipment will attempt to decode the de-interleaved sub-channel including the 'B' symbols. If successful, the symbols are reencoded and demodulated, and user A's equipment attempts to decode/demodulate the 'A' symbols using knowledge of the 'B' symbols.

In the context of inter-symbol multiplexing, e.g., as shown in Figure 9, 10 determining the CB' bits in user A's equipment before determining the 'A' bits may increase the certainty associated with demodulating the 'A' bits. The exemplary inter symbol multiplexing illustrated in Figure 9 portrays three bits per symbol. One exemplary modulation using three bits per symbol is 8-PSK, which has eight constellation points as seen in Figure 10. The transmitter sets the in-phase and 15 quadrature phase signals corresponding to one of the constellation points. flue radius of the circle represents the signal amplitude. At the receiver, noise and other signal impairments will, at the lime of symbol decision, generate a received signal which is different than any one of the eight specified constellation points. The more noise associated with the radio channel, the more distance will separate the actual 20 (transmitted) constellation point from the received point on the circle. The received point is interpreted as representing the closest one of the eight points on the constellation. Thus, the most likely error-event is that a signal point is interpreted at the receiver as a constellation point which is adjacent to that which was actually transmitted. 25 Because of this characteristic impact of noise on the modulation, a common mapping of bits to the constellation uses Grey coding in order to minimize the number of bit errors for the most likely error event. Figure 11 illustrates one such mapping.

For each constellation point, the To nearest neighbors on the circle are assigned bit combinations that differ in value in only one place. Thus, moving from signal point to

-13 signal point along the circle only one bit changes at a time. This type of mapping minimizes the number of bits in error when an error occurs.

A Grey coded mapping may be appropriate when symbol multiplexing embodiments of the present invention are employed. However, when intersymbol 5 multiplexing embodiments are employed other mappings according to the present invention may produce better demodulation results and add processing gain.

For example consider the exemplary bit mapping for an 8-PSK constellation illustrated in Figure 12(a) which can be used in conjunction with the inter-symbol multiplexing techniques described above. As mentioned above, by decoding the sum 10 channel for the other user first, a known bit or bits may be identified for a symbol involving the sub-channel of interest. If the most significant bit (MSB), i.e., the leftmost bit assigned to each point in the constellation of Figure 12(a) is known, then the demodulation problem reduces to QPSK. For example, if the MSB is known have a value of '1', then the constellation reduces to that shown in Figure 12(b). If the MSB 15 is known to have a value of '0', then the constellation reduces to that shown in Figure 12(c). Using the mapping illustrated in Figure 12(a) for 8-PSK modulation and the multi-user detection techniques describedabove, the modulation may be more tolerant to noise and other signal impairments than the Grey coded mapping illustrated in 20 Figure 11. This can be readily observed by noting that the minimum distance between any of the running points in Figures 12(b) and 12(c) are 90 degrees apart, while the distance between the re: g points in Figure 11 is 45 degrees. However, this does not necessarily mean that the mapping in Figure 12(a) is optimal for all applications.

For example, the mapping in Figure 12(a) is particularly poor when the MSB is not 25 known, since every point has as its neighbor an MSB of the opposite value which maximizes a possibility that the wrong constellation point will be selected when the MSB is not [mown. By way of contrast, the mapping of Figure 11 is better than that of Figure 12(a) when the MSB is not known Other mappings should also be considered. For example, Figure 13(a) 30 illustrates an optional mapping for the MSB when the two least significant bits (LSBs)

-14 are known. For each case, the remaining points are separated by 180 degrees. For example, as seen in Figure 13(b), when the two LSBs are both zero, the remaining constellation effectively reduces to BPSK. This mapping is also optimal for the case I where the MSB is detected without knowing the two LSBs as can be seen in Figure 5 13(c). Therein, it will be appreciated by those skilled in the art that since four of the signal points have only one neighbor with a different MSB value and for the other four of the signal points the neighbors have the same assigned MSB values, this represents an optimal mapping in terms of MSB detection.

However, a weakness of the mapping of depicted in Figure 13(a) becomes 10 apparent when trying to determine the 2 LSBs when the MSB is known. Consider Figure 13(d) wherein it is assumed that the MSB has a value of '1'. Therein, it can be seen that the two signal points in the middle of the four remaining points both have neighbors with different values spaced apart by 45 degrees, which is one of the worst possible results for combatting noise (although they are, however, Grey encoded).

15 Yet another constellation mapping is illustrated in Figure 14(a). This mapping may present a slight improvement over the exemplary mapping of Figure 13(a) for the case where the MSB is known and the 2 LSBs are being detected. This can be seen in Figure 14(b), where the MSB has a value of '0', wherein it is apparent that the two f middle remaining signal points have the same LSB value.

20 Thus, it will be apparent that changing the constellation mapping used in the modulation process can provide different noise immunity properties given knowledge of one or more bits in each received symbol. Since the inter-symbol multiplexing technique provided above provides bits associated with two different sub-channels to be disposed in orate symbol, the opportunity to select a preferred mapping arises. The 25 selection of a particular mapping, however, depends upon the channel coding involved in each suchannel, the bit pattern used in selecting bits for the inter-symbol multiplexing, the system design and Me desired result.

For example, the most heavily protected bits (e.g., class 1 bits in the DAMPS example described above) from each user or connection could be mapped to an MSB of 30 each symbol while using a mapping that maximizes the detection of these bits. Then,

-15 the re-encoding/remodulation procedure described above will make it easier to decode the remaining 2 LSBs. Another alternative is to increase the channel coding for he already most highly protected bits from the source and use the mapping of Figure 12(a). Since more channel coding is used, the performance associated with detecting 5 these MSBs may be better than that for the mapping of Figure 11 even though the mapping is less favorable for MSBs per se. Then, since the remaining signal points in the constellation of Figures 12(b) or 12(c) are optimally place, less channel coding can be used to protect these bits.

Yet another alternative is to allocate the most channel coding to a few LSBs 10 using the mapping in Figure 12(a) and detect these bin first, which mapping is optimal for this form of detection. The MSBs, which have the least favorable mapping in this constellation, may be allocated to the least important bits in a source.

Another alternative is to use the mapping illustrated in Figure 13(a) and allocate bits in an information stream to a few LSBs. These bits are again detected first, but the 15 MSBs are now better positioned than in the preceding example. If there is an uncoded class in the bit streams that are being multiplexed together, then the latter alternative provides a better quality frame of bits when the condition of the channel is fairly good.

The fanner method increases the likelihood that the most important bits can be correctly recovered while the least important bits may more frequently have bit errors.

20 Yet another strategy involves first detecting the MSBs using the mapping in either Figure 13(a) or Figure 14(a) then, after re-encoding, detecting the LSB. In this detection two of the four signal points have no close neighbors of different value.

Then, after another re-encoding, potentially from a different channel, the remaining bit is detected. I he receiver may first try with one channel. If it cannot successfully 25 decode (e.g., by checking a CRC) the other channel (other bit) is decoded. Since the channel coding spans several bits and, therefore, several symbols, another channel may be successfully decoded even though the first channel decoding failed. The subsequent re-encoding may reduce the number of errors in the symbol-to- bit detection, which then can accommodate the ret= g errors in the channel which, at the first attempt, could 30 not be channel decoded.

-16 Those skilled in the art will appreciate that for a particular strategy, especially if the three bits are detected at three different instances with intermediate re-en' oding, better mappings than illustrated herein may be possible. Moreover, for each suF-

channel different classes of bits may be transrrutted. Each class of bits comprises those 5 bits subject to the same channel coding protection. For example, one suchannel may have class one and class 2 bits as described above, while another sub-channel carries data which involves a third and, possibly, a fourth class of bits. Those skilled in the art will appreciate that these classes of bits may be mapped using more than one bit-to-

symbol mapping, i.e., in one symbol this class is mapped to a first subset of bits, while 10 for another symbol this class is mapped to a second subset of bits.

For each particular source, modulation method (e.g., 8-PSK, 16QAM, etc.), air interface (e.g., IS-136, GSM, IS-95, PDC, etc.), extensive computer simulations can be used to determine an optimal allocation of channel coding and multi-user multiplexing method as described herein. For example, some symbols may be 15 allocated to a single channel. If these symbols can be detected correctly, the next channel can use these symbols tO estimate the state (phase) of the channel, i.e., using these symbols as reference or pilot symbols.

In addition to their effects on modulation and demodulation techniques, i overhead signalling and channel coding should also be considered. For the two voice 20 suchannels which are multiplexed onto each time slot, common channel coding can be used. However, in D-AMPS, the traffic channel frequency carries multiplexed therewith a fast associated control channel (FACCH) which provides more urgent, overhead information to the mobile stations. Two possibilities exist for handling the PACCH in the context of multi-user multiplexing. First, as illustrated, for example in 25 Figure 15, speech and FACCH can be separately channel coded. In this example, the FACCH words are shortened to fit into the half rate portion of each time slot.

Alternatively, each FACCH can steal 2 speech frames or the amount of channel coding on FACCH information can be reduced to fit into the half rate scheme.

A second possibility is that the signalling FACCH can be common 30 ('CFACCH') for both users multiplexed onto a half rate channel. In this solution, the

WO 99/59269 - - - PCI.ISE99/00798

-17 FACCH word length can remain the same as the full rate FACCH. However, thTs solution affects both user A and user B (user B's speech frame will also be blanked out), even though the message is intended for only one user. Thus the use of CFACCH will cause shorter, but more frequent interruptions in the transmission of 5 speech than the use of distinctly coded FACCH's for each user. A discriminator inside the CFACCH word can indicate the intended recipient, i.e., A, B or both. The CFAACH can be divided into two sub-fields, one for each user.

Other overhead signalling may also require consideration when two users are multiplexed into a single downlink time slot. Consider the exemplary downlink slot 10 formats of Figures 16(a) (8-PSK) and 16(b) (8-DPSK) which are designed for full rate communication. In Figure 16(a), the SYNC field provides synchronization bits for

acquiring timing alignmcut to the slot. The REF field provides reference symbols

which may be used as a reference to assist in begin coherent demodulation of information. Several PLT fields are interspersed to provide pilot symbols. The pilot

15 symbols provide phase reference information so that the receiver can track the tane-

varying effects of the radio channel on the transmitted information. Various payload DATA fields are also provided in this exemplary downlink timeslot format. The

ending ramp field provides a period during which the transmitter can ramp down its

output power to reduce adjacent channel interference. Figure 16(b) illustrates an 20 exemplary downlink slot format for differentially encoded, modulated information.

Since differential encoding is used, pilot symbols are not necessary.

The structure of these exemplary full rate dowrdink time slots may require certain operational adjustments in order to operate with the multi-user multiplexing described above. For example, with two users per time slot a Power Control (PC) 25 function has to accommodate controlling the output power of the transmitters of each user. These exemplary downlirk slot formats, however, provide for only one power control bit. Two alternatives for controlling the output power of each user can be provided. Fast, an additional bit could be added to the power control field, with some

implicit understanding between the users as to which bit controls which user's output 30 power. Second, the usage of this bit could be time multiplexed, e.g., user A could use

-18 the PC bit in time slot 1 while user B uses the PC bit in time slot 4. The first solution reduces the payload data by one bit, while the second solution reduces the speed of the power control loop by a factor of two.

For both full rate and half rate communications (one or two users per slot), the 5 PC bit may be grouped into a multi-bit value before an interpretation of the PC data is made. For example, after receiving two time slots of data, the value 00 may mean no change, 01 increase with dB, 10 increase with y dB and 11 can mean no change or reserved or increase with z dB. These changes (x,y,z) can be further filtered over time and, when the output of that filter makes the desired power level differ from the actual 10 power level by w dB, a change can be made.

Of course, those skilled in the art will appreciate that it is not possible to use the re-encoding techniques described above if encryption is applied after channel codung as, for example, in the D-AMPS example provided above in Figure 4. If encryption is applied after channel coding, then user A would not be able to decrypt user B's sum 15 channel information, check the CRC and then re-encode/remodulate the information for the second pass. Instead, exemplary embodiments of the present invention provide for encryption (if any) to be applied before channel encoding, e.g., on the speech frame.

Lastly, mode signalling can be provided to inform mobile stations of various effects created by the foregoing exemplary embodiments of the present invention. For 20 example, the following signalling can occur at the start of a call and during a call, including as part of a handoff command. With reference to the above-described embodiment wherein only one active user is currently transmitting on an otherwise half rate time slot, the mobile station may be informed of the format (channel coding, other sub-channel's data, etc.) of the second sub-channel. Moreover, a flag can be sent to 25 the mobile station informing it as to whether the second sub-channel which is multiplexed into its designated time slots can be used for re-encoding as described above. Other mode signalling may include the usage of the PC bit(s) and whether C:FACCH is enabled. For example a single bit flag could be sent to a mobile station which indicates a modality package, e.g., MMFLAG = 1, might mean that the 30 recipient mobile station should transmit on the uplink in the time slot corresponding to

-19 its second downlink time slot m a frame, that it should consider the odd bits in each of its two assigned downlinlc time slots to comprise its sub-channel and that it should read the PC bit in its second downlink time slot in each frame to adjust its power.

Alternatively, the explicit information can be reduced by defining each of these characteristics for each sub-channel and then simply informing the mobile station to which sub-channel it is assigned. That is, for a given multiplexing scheme according to the present invention, e.g., symbol multiplexing, inter-symbol multiplexing, or a combination thereof, two (or more) suchannels are created. The mode signalling bit can indicate to the mobile on which sub-channel it will receive information, in response 10 to which it will i nplicitly (through pre-programming) know which bits belong to its channel, which power control bit is associated with its transmit power, etc. Additional exemplary embodunents of the present invention will now be discussed to further illustrate how the re-encoding techniques described above can be applied to improve the demodulation/detection of multiple information streams. As 15 will be appreciated by those skilled in the art, demodulation refers to the process of channel estimation and extracting soft received samples. Typically, both the phase and amplitude of the channel are esli'T ated for use in the demodulation process. However, for some modulations or when equalization is not performed (e.g., no time dispersion), only the phase may be tracked. Synchronization words and pilots may be used in 20 estimating the channel.

When the modulation is contained in the absolute value of the symbol, for example the phase in 8PSK, it is important to undo any phase changes that the fading of the transmitted signal has introduced before coherently demodulating the received signal. For differentially encoded signals, where the modulation is contained in a 25 phase shift from one symbol to another, it may be less important to accurately track the phase changes of the channel prior to demodulation. However, differentially encoded signals can first be coherently demodulated and, subsequently differential decoding can take place to improve performance as compared with a receiver which only loolcs for the phase difference between symbols. Thus, for modulation based on either an 30 absolute or relative symbol value there can be an interest to track the channel.

-20 Moreover, when symbol interference due to timedispersion is present or when the modulation format contains information in the amplitude of the sigral, e.g. 16 QAM, it is also of importance to track the amplitude of the channel which, due to the Rayleigh fading, can be subject to rapid changes.

5 The accuracy of channel estimation will improve if re-encoding is performed as bits are decoded. Thus, the hvo-pass demodulation process described above can be generalized to periodically re-estimate the channel and use the re-estimated channel estimates in subsequent demodulations. The following provides an example of how such a process can be performed.

10 Initially, define different classes of the information streams based on their varying chaMel coding Tom each source frame as 1,2,...n. For the purposes of this example, assume that the interleaving is two slot diagonal interleaving (as for IS-136 systems). In this example, each user U1, U2,...Uw receives and processes slots...,m 15 2,m-l,m,m+ l,m+2,., etc. These slots are transmitted, using the IS-136 example, on time slot pairs 1&4 or 2&5 or 3&6. When processing slot m in the receiver in order to generate speech frame j, half of the bits in time slot m are related to another speech frame (i.e. speech frame j t 1) which other speech frame includes bits which are also sent in time slot m+1 (i.e., due to the two slot interleaving used in IS-136, see 20 Figure 7). The demodulation of time slot m according to the present invention could be performed as follows: Receive slot m for the final processing of source frame j Demodulate Decode class 1 bits of user U1 25 Re-encode, re- interleave using the decoded data

-21 Demodulate Decode class I bits,of user U2 given the result from previous decoding step Re-encode, re-interleave using the decoded data Decode class 1 bits of user Uw given the result from previous decoding steps Re-encode, re-interleave using the decoded data Demodulate 10 Decode class 2 bits of user U1 given the result from previous decoding steps Demodulate Decode class 2 bits of user Uw given the result from previous decoding 15 steps Decode class n of user Uw given the result from previous decoding steps Demodulate 20 For each user U1.,.Uw, source frame j is now recovered. The process can continue with time slot m+1 to recover frame j+l.

Half of the bits of time slot m will be used in the processing of frame j +l which will happen during processing of time slot m+ 1. In order to have the best quality samples from time slot m when processing time slot m+ 1, a last demodulation of slot 25 m is performed after all the decoding for slot m is finished. Furthermore, the least protected class (n) in the set of classes is often not protected at all by error correction coding. A final demodulation after all the protected classes are processed will enhance demodulation of the unprotected bits.

l; - -22 As described in the above-incorporated International Patent Publication No. WO 98104647 entitled "Method and Apparatus for Detecting Communication Signals Having Unequal Error Protection", the information fed back for use in the re-encoding, re-

interleaving and demodulation process may be soft (i.e., likelihood) information t 5 provide improved performance. For example, the modulation constellation in Figure 12(a) may not collapse to distinctly either that shown in Figures 12(b) or 12(c).

Instead, the fed back soft information indicates a likelihood that the constellation of Figure 12(b) is relevant and another likelihood that the constellation of Figure 12(c) is relevant. 10 Of course, the algorithm described above is just one of numerous variations on the re-encoding theme which can be employed in conjunction with multi-user detection according to the present invention. For example, the estimation of the radio channel need not be performed after every single class has been decoded for every user. It may be sufficient to only re-estirnate the radio channel after the most protected class of bits 15 is decoded. Half of the bits in any given time slot will be used when processing a subsequent tune slot. Hence, it may advantageous to perform a final channel estimation and demodulation when all channel decoding is completed for a burst in the interest of getting as high a data quality as possible when starting the decoding of the first selected class of bits and user in the subsequent burst. Moreover, if there is an 20 indication, e.g. by a CRC check or He soft information from the previously decoded bits, that the information is not reliable, the receiver may skip to another user's data or to another class before continuing in order not to introduce further degradation of the received sandpiles.

U.S. Patent No. 5,673,291 to Paul Dent (mentioned above) teaches further 25 refinements which may be employed in the context of the present invention. For example, the channel estimation and the channel decoding described above need not be separated into two distinct processes. For example, when using a Viterbi channel decoder, each state can have a separate channel estimator, which state is updated for each decoded bit. Another refinement is to decode the information from the multiple

-23 users and multiple classes in parallel. For each step in the Viterbi decoder, modified input data to all other Viterbi decoders may be produced.

For speech transmission, the delay when processing the data should generally be mirumized. However, if more delay can be tolerated than has already been introduced 5 by the speech coder and the two slot interleaving, further performance improvement can be gained. For example, when frame j is decoded (which happens after processing of slot m has finished), the data from slot m-1 that was used in the decoding of frame} during slot m can be reintroduced in slot m-l. Decoding frame j-1 can now be further improved. Having now improved the data in slot m-l and since half of those bits are 10 used together with data in slot m to form frame j, the slot m can now be reprocessed.

Thus, the performance can be improved by iterative processing. However, since frame j-l can not be released to the speech decoder until one slot later, this iterative processing introduces delay. A tradeoff between delay and performance improvement would be to release the frame j-l to the speech decoder after slot m-1 is processed, so 15 that the iterative processing is used to enhance the bits of slot m-1 that affect frame j.

In this process a stored copy of frame j-1 is used. If bits in frame j-1 were changed, this will have no effect on the frame j-1 delivered to the speech decoder. Armorer, somewhat more elaborate example is to reintroduce the now improved bits in slot m-1 into slot m-2 and then work forward in time to first further improve slot m- I and 20 finally slot m. It should be apparent to the person skilled in the art that many variations are possible.

A simple but practical example is now presented with limited complexity to ease understanding and which does not perform any iterative inter-slot processing. The source is a speech coder win three classes. Class la which is protected with a CRC 25 and encoded with a first coding rate, class lb has no CRC and is encoded with a second coding rate and class 2 has no CRC or channel coding. The algorithm is designed based on the assumption that class la is more protected (i.e., has a lower coding rate and thus more redundancy) than class lb and that there is no FACCH type of interruption of the speech service (as described above with respect to Figure 15. The

-24 following steps can be performed in this exemplary algorithm (italicized expressions defined below): Receive slot m for the final processing of source frame j. The assigned channel is A. [1] Demure. Setq=0 [2] Decode class la of user A If CRC=OK THEN update received date, Set q=a Decode class la of user B If CRC = OK THEN updated received dme, Set q=q+a 10 [3] Case 1: If user A CRC=Not OK AND user B CRC-Not OK THEN GOTO [ERR]

Case 2: If user A CRC=OK AND user B CRC-Not OK THEN GOTO t4] Case 3: If user A CRC=Not OK AND user B CRC=OK 15 Decode class la of user A user A CRC=OK, THEN update received date, Set q=qa and GOTO [4], OTHERWISE GOTO tERR Case 4: If user A CRC=OK AND user B CRC=OK THEN GOTO[41

20 14] IF q>-a THEN re phase adds and Set q=0 Decode class lb of user A If metric=OK THEN update received dote and Set q=q+b Decode class lb of user B If metnc=OK THEN update received date and Set q=q+b 25 [5] IF q> =b THEN re phase adjust Decode class 2 for user A Extract class 2 bits for user A and all bits related to frame j + 1 for both users [6] Send frame j to speech decoder. Store bits associated with frame 30 j+l. GOTO [END]

-25 [ERR] Declare fraTne i as undecodable. Store bits associated with frame j+l [END] Proceed decoding frame j +1 In the foregoing exemplary algorithm, the term update received date means that 5 in further processing of the received data the result of the decoding shall be reflected (soft or hard information) in the further processing of the received data. The actual formatting (channel re-encode, interleaving, inter-symbol multiplexing) is implicitly understood in the algorithm.

The tenn demote means to extract the soft values from the received burst.

10 This includes the aspect of first estimating and then compensating for the phase shift introduced by Rayleigh fading. Prior to demodulating, time synchronization has been performed in which a symbol spaced set of samples has been extract from an over-

sampled received signal.

The term re-phase adjust means to improve the correction of the phase errors 15 introduced by the Rayleigh fading using the decoded data.

The value of the consents a and b in the foregoing example are dependent on the relative protection level of class la and class lb. The intent is to set these constants to values such that re phase adjustment is performed only if the decoding has been successful enough to warrant a re phase adjustment. Note that the next frame (j+l) 20 may benefit from the re phase adjustment even if class la can not be recovered correctly. In the foregoing example, a relatively simple approach is taken to decide whether the received samples should be re phase adjusted, i.e., after having decoded class la from both users, perform re phase adjustment if at least one of the decoding passes was successful. Phase re adjust is performed aged if at least one of the class lb 25 information streams was successfully decoded.

A more general example for determining whether re adjustments should take place could be to use a and b as variables which are set equal to quality metrics received from the channel decoder and then to examine at each iteration described above whether q =c. In this more general example, die value of c is a threshold

-26 value which could be determined based on simulations involving many values including, for example, the channel coding, and could vary depending on soft output from the decoding processes.

When evaluating the Quality of the decoded class la bits, the CRC can be used 5 as a quality indicator. For cams lb, an absolute or relative threshold of the Viterbi decoder's cost function selected path can be used. A relative threshold can be to compare the metric of the selected path with the selected path of the class la bits, scaled with an appropriate value depending on e.g. the number of gross bits in class la and class lb respectively.

10 It will be seen from the foregoing example that the frame j is discarded if neither user A nor user B's class la bits provides a positive CRC check. This provision was described in the foregoing example merely to limut complexity for illustration purposes. In most cases, further processing of the class lb bits will not permit correct recovery of the class la bits. However, depending upon the amount of 15 channel coding used in each class, discarding the frame when the CRC fails for the class la bits may be undesirable. Thus, the class lb bits of user A could be decoded at step [3] instead of giving up. If successful, according to a metric, decoding of class la bits could be retried However, this latter approach increases the decoding complexity.

According to yet another exemplary embodiment of the present invention, noisy 20 radio channel conditions on the uplinlc can be addressed by switching between several c of me transmission modes described above. Oneexemplary context in which noisy radio channel conditions may arise is in range-limited systems wherein the distance between a base station and a mobile station may be such that the noise in the radio channel may reduce the received signal quality at either the base station or the mobile 25 station to unacceptable levels. This problem can be more readily addressed for the downlink by increasing the antenna height at the base station andJor increasing the transmit power. For example, the base station may be able to transmit at 50 Wlcarrier or more. However, the uplink is snore problematic since the antenna size of the mobile cannot be easily increased nor are the power amplifiers used in mobile stations

# - -27 designed to support significant increases in transmit power, e.g., mobile stations typically can transmit at levels on the order of 0.6 W. Thus, when noisy channel conditions are experiP-^-d in the uplink, e.g., as detected based upon the received signal quality at the base station, exemplary 5 embodiments of the present invention provide for a user to switch from, for example, the half rate transmission described above to using full rate bandwidth (e.g., bandwidth equivalent.to nvo full tune slots per frame per user). The additional bandwidth can be used in several different ways to add additional bits that provide for better received signal quality at the base station. For example, the full rate vocoder 10 and channel coding can be used which may provide additional robustness in the context of noise resistance. Alternatively, the half rate vocoder can still be used, but additional channel coding ran be provided, the increased redundancy of which can be used tO correct more errors. Yet another alternative is to copy the bits from each time slot into the new (full rate) time slots. This provides a form of diversity which information can 15 be combined or selected at the base station to Improve the received signal quality. If the signal quality received at the base station improves to an acceptable extent, the system may recognize the reduction in noise effects and return the mobile station to half rate transmission mode. Of course, as described above, changes in the transmission mode may be accompanied by suitable mode signalling to the mobile 20 station to indicate the change.

Although the invention has been described in detail with reference only to a few exemplary embodiments, those skilled in the art will appreciate that various modifications can be made without departing from the inversion. For example, although the present invention has been described in conjunction with half rate 25 communications those skilled in the art will appreciate Mat the concepts set forth herein can be extended to one third rate, one quarter rate, etc. with additional users or sources being multiplexed into the same bandwidth. Accordingly, the invention is defined only by the following claims which are intended to embrace all equivalents thereof

Claims (5)

( CLAIMS

1. A method for transmitting information from a first source and a second source in 5 a radiocommunication system comprising the steps of: interleaving bits from said first source and said second source together into a single symbol by: assigning at least one bit from said first source to at least one most significant bit (MSB) of said single symbol; I 10 assigning at least one bit from said second source to at least one least significant bit (LSB) of said single symbol; and transmitting said single symbol, among other symbols.

2. The method of claim 1, wherein said symbols are 8-PSK symbols and said at 15 least one MSB includes one bit from said first source and said at least one LSB I includes two bits from said second source. I

3. The method of claim 1 further comprising the steps of: assigning, as said at least one MSB. highly protected bits from said first source; 20 and employing a bit-tsymbol mapping that increases a probability of detection of said at least one MSB.

4. The method of claim 3, wherein said highly protected bits are class 1 bits.

25 '

5. The method of claim 1, wherein said first and second sources comprise different: information streams forwarded to a base station in a radiocommunication system.