SPREAD SPECTRUM MODULATION SYSTEM
This invention relates to modulation systems for the transmission of data by- radio. It is particularly suited for use on high capacity channels such as satellite transponders.
The cost of leasing a satellite transponder forms a substantial part of the total cost of the link of which it forms a part. It is therefore desirable to maximise the data rate which can be carried by the transponder.
The factor which limits the data rate which can be handled by a transponder is the frequency response of the amplifier. Those used in satellite transponders typically have a 72MHz bandwidth, as shown in Figure 4. The Nyquist transmission rate, (maximum symbol rate in baud, equal to twice the bandwidth) puts a practical limit on satellite transmissions of approximately the bit rate used by the smallest synchronous transport module (STM-1 ), 1 55Mbit/s. (In comparision, ransatlantic submarine cables operate at STM 64, that is 10 Gbit/sec) .
A greater bit rate can be obtained for the same symbol rate (baud) by the use of phase shift keying (PSK) or quadrature/amplitude modulation (QAM): schemes in which phase as well as amplitude are used to define symbols each representing two or more binary digits (bits). For example, two different amplitudes and four different phases can be used to define eight symbols, each representing a different three-bit character, allowing a three-fold increase in speed over a simple binary system. More generally a scheme with 2" symbols allows the representation of "n"-bit characters, and a consequent n-fold increase in speed. However, such a system requires a highly linear amplifier response to ensure that the amplitude levels are distinguishable. This limits the available bandwidth to that region of the amplifier's nominal bandwidth over which the frequency response is approximately linear. Using the same 72 MHz channel referred to above, and assuming a usable bandwidth of .37.5 MHz, (see Figure 4) leads to a Nyquist rate of 75Mbaud. An eight bits/symbol coding scheme (256 QAM), e.g. using sixteen amplitude levels and sixteen phases, could therefore allow 600Mbit/s to be carried.
However, the requirement for a highly linear amplifier response would require that the amplifiers are not operated near their saturation levels, as they become non¬ linear in such circumstances. This therefore requires that the antennas in the
receiving apparatus have a larger gain factor than would otherwise be necessary. The use of high bit/symbol QAM schemes is thus inefficient.
Low capacity systems such as digital cellular radio systems use a variety of other systems to optimise the use of bandwidth. The modulation technique used in the "GSM" digital cellular radio standard is known as Gaussian Minimum Shift Keying (GMSK). This is a modified form of minimum shift keying (in which digital bits are represented by half-cycle sinusoidal pulses, the frequency representing a 1 differing from that representing a zero by half the bit rate). MSK has high sidebands relative to the main lobes in the frequency domain because of the discontinuity in the MSK waveform at the edge of the binary interval, caused by the rapid change between the two carriers - these large sidebands can lead to interference with adjacent signals. In GMSK, the pulses are filtered using a Gaussian-shaped impulse response filter - the sidelobes of the resulting waveform are much lower compared to those in MSK, allowing the transmission of data over a narrower bandwidth while maintaining the same capacity to reduce interference levels as frequency modulation, and ensuring a constant-envelope transmission.
An alternative coding system, to be used in the proposed "UMTS" or "3rd Generation" digital cellular telephone system, is known as spread spectrum. In this arrangement a binary "chipping" code is modulated onto the information, at a chip rate very much faster than the information rate. The resulting encoded signal requires a relatively high bandwidth for its transmission. However, other transmitters, using different chipping codes, can share this bandwidth. This is known as code-division multiple access (CDMA). The resulting composite signal includes all the encoded signals. This composite signal is decoded at the receiver, using a chipping code congruent to that of the transmitter whose signal is to be extracted. The resulting sequence comprises a a non-random element corresponding to the required signal, and a superimposed quasi-random element, which will appear as noise, caused by the non-congruity of the decoding chipping code with the chipping codes of the other signals. Provided transmitter strengths are controlled such that all channels are received at a similar amplitude, the signal/noise ratio should always be adequate to avoid corruption of the original signal by the noise from the other signals.
CDMA and GMSK have both been used, as alternative approaches, for single-user point to point transmissions. A proposal by Ogawa et al, described in two
papers published by the IEEE in 1 994 (Third International Symposium on Spread Spectrum Techniques, pages 399-403) and in 1 996 (Wakaki, Ogawa et al, Fourth International Symposium on Spread Spectrum Techniques, pages 564-568) uses the two systems together, by generating a CDMA signal and applying a GMSK modulation to the result. (The CDMA chipping code is actually applied as a pair of interleaved codes). At the receiver, the demodulated signal is split into two channels in quadrature, alternate chips being handled by each channel (each quadrature channel therefore using one of the pair of chipping codes) . This permits the signal to be recovered from a noisy background by differential detection techniques, without carrier recovery.
The present invention relates to a novel method of using CDMA in high capacity links, in a different manner to Ogawa's proposal, and for a different purpose. (
According to the present invention, there is provided CLAIM 1 The modulation process may be performed separately on each member of the plurality of signals, which are then combined for transmission. Alternatively, the modulation may be applied to the combined signal. The separation of the original signal into the plurality of signals may be done by any suitable time division demultiplexing or packetising process. It should be noted that this demultiplexing or packetising need take no account of the nature of the original signal, which may itself carry a multiplex. Indeed, a decoupling of the extraction process from the original multiplex can be advantageous, as will be described.
Also according to the invention, there is provided CLAIM 1 1 .
By transmitting a high bit rate signal as a plurality of lower bit rate signals, the bandwidth limitations of the transponders in the system can be avoided, even though the total number of bits (chips) is much greater. Since all the spread spectrum transmissions come from the same source, their amplitudes can be controlled to be of the same level, ensuring that no one member of the plurality of signals predominates. By using GMSK modulation, which has a constant envelope over a minimum bandwidth, the multiple spread spectrum carriers can be carried by amplifiers operating at or near saturation without the production of spectral side lobes.
In contrast with the Ogawa system, the secondary spread spectrum signals generated from one original signal are transmitted simultaneously, rather than being
interleaved. Also unlike Ogawa's proposal, the secondary signals are not required to be detected by being operated in quadrature, so there may be three or more such secondary signals, and in practice there would be many more than this. The number of secondary signals is limited only by the number of available distinguishable chipping codes: in the decribed embodiment two hundred chipping codes are used. As an example, to transmit a 600Mbit/s signal over a 75MHz bandwidth transponder using a QAM scheme would require a symbol rate of 75 Msymbols/second, necessitating 8 bits per symbol: that is a 256 level QAM scheme. This would require an amplifier with a very linear response curve, in order to distinguish so many amplitude levels. Using the present invention, the signal can be divided into (say) two hundred 3Mbit/sec signals, each of which is then encoded using a 25 chip/bit encoding scheme to generate a 75Mchip/sec signal. These signals can then all. be carried simultaneously over one 75MHz bandwidth transponder.
An embodiment of the invention will now be described, by way of example, with reference to the drawings, in which:
Figure 1 illustrates schematically the general arrangement of a first embodiment of a transmitter system according to the invention
Figure 2 illustrates schematically the general arrangement of a second embodiment of a transmitter system according to the invention Figure 3 illustrates schematically the general arrangement of a first embodiment of a receiver system according to the invention
Figure 4, which has already been discussed, illustrates a typical transponder frequency response.
It should be noted that the receiver of Figure 3 may be used with either embodiment of transmitter
The transmitter system in Figure 1 comprises an input 10, feeding a distributor or splitter 1 1 , which has a plurality of output channels 1 1 a, 1 1 b, etc. These output channels each comprise a spread spectrum coder 1 2a, 1 2b etc, feeding a respective GMSK modulator 13a, 13b, etc, each comprising a Gaussian filter 131 a, etc, and a voltage controlled oscillator 132a etc. The outputs from the modulators are then recombined in a addition unit 14 (operating in the frequency domain) for transmission at an output 1 5.
The transmitter system in Figure 2 is similar to that of Figure 1 , and corresponding units are indicated by the same reference numerals. However, in this embodiment the outputs from the spread spectrum coders 12a, 12b etc are fed to an addition unit 24 operating in the voltage domain. The single output from this addition unit is fed to a GMSK modulator 23, (comprising a Gaussian filter 231 and a voltage controlled oscillator 232) for transmission at an output 15.
The receiver system of Figure 3 operates to reconstitute the signal generated by the transmitter 1 5 of Figure 1 or Figure 2. In the arrangement of Figure 3, an input 35 is fed to a demodulator 36, whose output feeds a plurality of spread spectrum receivers 37a, 37b etc, each comprising a decoder 371 a, etc and an assembler 372a, etc. The outputs from these receivers 37a, 37b are then combined in a multiplexer 38 to genarate an output 39.
The operation of these embodiments will now be described. For illustrative purposes, the input of Figures 1 and 2 is assumed to be. a 600Mbit/s signal, corresponding to STM-4.
The 600 Mbit/s signal at input 10 is extracted by the splitter 1 1 and converted into two hundred individual 3Mbit/s secondary signals, each fed to a respective channel 1 1 a, 1 1 b, etc. The distribution of the digits between the channels may follow any suitable pattern, but typically the incoming bit stream is split into groups each comprising a predetermined number of bits (one or more), the groups being fed to the outputs 1 1 a, 1 1 b etc of the splitter 1 1 in rotation. Preferably the groups each comprise a small number of digits, so that in the event of failure of one of the channels, only a small number of adjacent bits are lost, allowing error correction facilities to recover the signal from the remainder of the final output. It is preferable for the pattern of distribution to be such that it does not correspond with commonly used time frames, so that any control channel information (such as error correction digits, address digits, synchronisation digits etc) in the input signal 10 is carried by different channels on different cycles, reducing the possibility that the control information might be compromised by a single failed channel. It will be understood that all the channels 1 1 a, 1 1 b, etc are processed in the same way, so it is only necessary to describe the process as it is applied to one channel 1 1 a. Firstly, the 3Mbit/s secondary signal is processed by a spread spectrum encoder 12a. This applies, to each binary digit it receives, a pseudo-random chipping,
unique to that channel. The chipping rate is 25 times the original bit rate. This chip rate gives 225 possible chipping codes - over 33 y2 million. However, not all these codes provide clear distinctions between the signals. The usable limit is determined by the spreading ratio of the code and the requirements for sharp auto-correlation and low cross-correlation - characteristics that are displayed, for example, in the "Kasami large set", as discussed by John G Proakis: Digital Communications (McGraw-Hill, 1 983: ISBN 0-07-050927-1 ) pages 545 - 569, which gives a treatment of spread spectrum, CDMA and code properties.
The chipping codes selected for the two hundred individual channels are selected to allow clear distinction between them, as is well known in CDMA spread spectrum systems. The individual pseudo-random chipping codes may repeat over a greater period than 25 bits, to provide even greater distinction between the individual bitstreams.
The bit-rate of the output from each pseudo-random chipping coder 1 2a is therefore 25x3Mbit/s, or 75Mbit/s. If transmitted in this form this output would require highly linear amplifiers with correspondingly large earth station antennas, because of the relatively large sidebands and the envelope shape illustaretd in Figure 4. To avoid these problems, the ouput is fed to a GMSK modulator 1 3a, which provides a constant envelope while minimising the side-bands. The GMSK modulator 1 3a comprises a Gaussian filter 131 a, which smooths the square-wave pattern, thereby eliminating the high-frequency components in the signal caused by the abrupt changes in level. The output is then fed to a voltage-controlled oscillator 1 32a, which generates a frequency response dependent on the signal level (voltage) input to it, thereby generating a frequency modulated signal. The frequency-modulated signals generated by each GMSK modulator 13a,
1 3b, etc are then combined in a frequency-domain addition unit 14 to generate an output signal 1 5 for transmission.
In the variant of Figure 2, the outputs from the chipping coders 1 2a, 1 2b etc are fed to an addition unit 24. It should be noted that the result of this addition is a 75baud signal with 201 potential levels (ranging from all outputs 1 2a, 1 2b etc being zero to all outputs being 1 ). This signal is then smoothed by a Guassian filter 231 and then modulated by a voltage controlled oscillator 232, to generate a frequency- modulated output 1 5.
It will be seen that the embodiment of Figure 2 requires only one GMSK modulator 23. However, this single modulator 23 must be capable of accurately coding 201 levels. Conversely, the embodiment of Figure 1 requires two hundred modulators 13a, 13b, etc, but each one may be quite simple as it only needs to distinguish two levels. Moreover, in the first embodiment, failure of one modulator would only lose 0.5% of the data, (which may be recoverable- if the input signal 10 includes error correction digits), whereas such a failure in the second embodiment would lose all the data.
A problem which can occur with spread-spectrum systems is that weak channels can be drowned out by the the pseudo-random noise generated by other stronger channels. This problem does not occur with the present system, as all the channels have the same origin, and their amplitudes can thus be easily controlled to be similar.
The receiver system of Figure 3 operates to reconstitute the signal generated by the transmitter 1 5 of Figure 1 or Figure 2, and operates in the reverse sense to the transmit apparatus of Figures 1 and 2, to recover the original signal from the two hundred secondary signals received in the input signal 35. The input signal 35 received from the transmitter is fed to a demodulator 36, which comprises a frequency to amplitude converter 361 for converting the FM signal 35 into an amplitude modulated Gaussian signal. The resulting signal is then quantised by a sampler 362 to identify the individual amplitude level present at each sampling interval. This reproduces a 75Mbaud bit stream, having 201 amplitude levels, similar to that previously generated by the addition unit 24 in the transmitter of the embodiment of Figure 2. This bit stream is then fed to each of two hundred spread-spectrum receivers
37a, 37b etc. Each receiver 37a, 37b etc comprises the following elements. (Decoder 37a will be described, all others are similar). First a decoder 371 a applies the convolute pseudo-random chipping code, congruent with the chipping code applied by the corresponding coder 12a in the transmitter. Because there are 201 levels, and not just two, in the input signal, this process does not directly regenerate the original 3Mbit/s signal in the corresponding channel 1 1 a, 1 1 b etc. Instead, a new 75Mbaud, 201 -level signal is generated. This signal is then passed to an assembler 372a which groups the output chips into 25-chip words, and adds the sum of the 25 chips
together to produce an output at 3Mbaud. This result has two components: a decoded "1 " or "0" corresponding to the original input on the corresponding input channel 1 1 a, having been decoded correctly by the decoder for each of the 25 chips; and a pseudo-random component formed by the decoding process acting on the other channels. Averaged over 25 chips, the pseudo-random element is a small perturbation on the required signal, and the correct level, "1 " or "0" is therefore readily determined, and can be output from the decoder 37a as a recovered secondary signal.
The 3Mbit/s secondary signals regenerated by the decoders 37a, 37b, etc are then combined in a multiplexer 38, operating on the two hundred channels 37a, 37b etc, to reassemble the data in the same order originally extracted by the distributor 1 1 , to reconstitute the original 600Mbit/s signal 39.