The present invention relates to a method and a device for transmitting signals in a GSM system.
Despite the fact that Global System for Mobile Communication (GSM) networks have been commercially deployed for almost two decades, interest on the continued improvement of the GSM/Enhanced Data rates for GSM Evolution (EDGE) technology has not dwindled. Network equipment manufacturers, mobile equipment manufacturers and telecom operators continue to develop the GSM system further. Improvements to the hardware/spectral efficiencies for both voice and packet data services are being actively sought. To this end, precoded EGPRS/EGPRS phase 2 (EGPRS2) has been proposed. Precoding involves transformation of the symbol sequence using some suitable transform. Typically a Fourier Transformation is used.
These bits are mapped to symbols drawn from a Phase Shift Keying (PSK)/Quadrature Amplitude Modulation (QAM) symbol constellation. The letters s,x,t,g will be used to denote PSK/QAM symbols that carry training, payload, tail or guard bits respectively. Thus,
(b1, . . . , bα, bφ+1, . . . , bφ)→{right arrow over (g)}=(g1, . . . , gη) (guard)
(bα+1, . . . , bβ, bε+1, . . . , bφ)→{right arrow over (t)}=(t1, . . . , tν) (tail)
(bβ+1, . . . , bγ, bδ+1, . . . , bε)→{right arrow over (x)}=(x1, . . . , xD) (payload)
(bγ+1, . . . , bδ)→{right arrow over (s)}=(s1, . . . , sN
where η is the total number of guard symbols, v is the total number of tail symbols, D is the total number of payload symbols and Ntr is the number of training symbols. The total number of payload plus training symbols is N=D+Ntr and the total number of symbols in the burst is K=N+η+v.
The output of the symbol mapping block is the sequence of symbols
It is convenient to intercalate the training symbols and the payload symbols for synchronization and channel estimation purposes. A vector {right arrow over (z)} of length N is constructed from the payload {right arrow over (x)} and training symbols {right arrow over (s)} accordingly.
The location of the training symbols is given by the indices (n(m))m=1N
for 1≦m,i≦N. The pre-coding operation is {right arrow over (Z)}=WH·{right arrow over (z)}.
Multiplication by the matrix WH can be implemented efficiently using the fast Fourier transform. Next, an integer L≧0 is chosen and the last L terms in {right arrow over (Z)} are appended at the beginning of {right arrow over (Z)} to form a new vector {right arrow over (Z)}P. In other words, a cyclic prefix of length L is added. For example the values L=0 (no prefix) or L=5 (typical GSM channel length) may be used.
Using vector notation, the precoded symbols with the cyclic prefix added are
The output of the vector precoding (IDFT) block is the sequence of complex numbers
This sequence is pulse shaped to obtain the baseband signal, as follows.
Here ρ is the pulse shaping filter, T is the symbol period (in seconds), φ is the phase and τ is the duration of the burst (in seconds). Finally, the pulse shaped signal (5) is sent to the Radio frequency (RF) modulator.
In order to make precoded EGPRS backwards compatible with EGPRS/EGPRS2, the linearized GMSK pulse shaping filter is used, see 3GPP TS 45.004 Modulation. This is a partial response pulse, which introduces a significant amount of intersymbol interference (ISI) in the transmitted signal. Therefore, the length L of the cyclic prefix must be chosen large enough to cover not only the time dispersion in the radio channel and the receive filtering but also the large time dispersion introduced by the pulse shaping filter. This entails a larger overhead and therefore a loss of bandwidth. Furthermore, due to backward compatibility, the size N of the IDFT is not a highly composite number. Even though there are efficient Inverse Fast Fourier Transform (IFFT) algorithms for any size, the algorithms for highly composite lengths are much faster. Fast, highly efficient IFFT's are desirable for backward compatibility with legacy base station hardware, and reuse of existing mobile station (MS) platforms.
Moreover, highly efficient Fast Fourier Transforms (FFT's) also decrease the power consumption in the MS. For example, at the normal symbol rate N=142=116+26=2×71 (71 is a prime number) is not a highly composite number.
Hence there exists a need to reduce the above problems and to provide a more efficient method for transmitting signals in a GSM system. In particular when the signals are precoded.
It is an object of the present invention to provide an improved method and device for transmitting data in a GSM system and to address the problems as outlined above.
This object and others are obtained by the method and device as described herein.
Thus, as realized by the inventors, one of the first tasks performed in a typical receiver for precoded EGPRS is Cyclic Prefix (CP) removal. However, since a significant amount of intersymbol interference (ISI) is introduced already at the transmitter, it is possible to modify a part of the CP in the transmitter, after having performed pulse shaping, without incurring any loss of link performance. In accordance with embodiments described herein a shortened precoded EGPRS burst is provided. In accordance with one embodiment the burst shortening is performed while preserving the spectral characteristics of the signal.
In accordance with one embodiment a method of transmitting a burst of signals in a cellular radio system supporting data transmission using EGPRS/EGPRS2 is provided. The method comprises providing additional symbols in the EGPRS/EGPRS2 burst thereby forming a long burst, and pulse shaping the long burst to forming a long baseband signal whose duration exceeds the duration of one EGPRS/EGPRS2 time slot. The long baseband signal is then shortened to a shortened burst having the duration of an EGPRS/EGPRS2 time slot wherein the shortened burst fulfills the same spectrum mask requirement as the EGPRS/EGPRS2 burst, and which can be transmitted as a conventional EGPRS/EGPRS2 burst.
In accordance with one embodiment the burst length of the long burst is set to 144 symbols.
In accordance with one embodiment the transmitted symbols are precoded.
In accordance with one embodiment the additional symbols are payload and/or training symbols.
In accordance with one embodiment the shortened burst is generated by truncation of said long baseband signal followed by multiplication by a window function.
Thus, a long burst containing more payload or training symbols than an EGPRS/EGPRS2 burst is shortened to the same length as an EGPRS/EGPRS2 burst and fulfill the same spectrum mask requirements. The additional symbols can be used to increase throughput (more data symbols), improve link performance (more training symbols or lower code rates), or for other purposes such as Peak to Average Power Ratio (PAPR) reduction.
Moreover, longer bursts can be set to a highly composite sizes for the IFFT/FFT if precoding is used, which has benefits in terms of processing power and power consumption. For example the burst length can be set to 144 symbols.
In summary, burst shortening alleviates the problems with precoded EGPRS/EGPRS2 presented above while preserving backward compatibility with EGPRS/EGPRS2, and without any losses in link performance.
In accordance with one embodiment of transmitting signals in a GSM system with precoded EGPRS signals bursts, the precoded EGPRS signal bursts are shortened.
Embodiments herein also extend to a device adapted to transmit signals in accordance with the above. The device can typically be implemented in a module comprising a micro controller or a micro processor operating on a set of computer program instructions stored in a memory, which instructions when executed by the module causes the module to perform power control in accordance with the method as described above. In particular the module can comprise controller circuitry for performing the above methods. The controller(s) can be implemented using suitable hardware and or software. The hardware can comprise one or many processors that can be arranged to execute software stored in a readable storage media. The processor(s) can be implemented by a single dedicated processor, by a single shared processor, or by a plurality of individual processors, some of which may be shared or distributed. Moreover, a processor or may include, without limitation, digital signal processor (DSP) hardware, ASIC hardware, read only memory (ROM), random access memory (RAM), and/or other storage media.
The present invention will now be described in more detail by way of non-limiting examples and with reference to the accompanying drawings, in which:
For simplicity, the below examples consider burst formats of the form (4) without tail symbols. However, the invention is not limited to such an embodiment but is equally applicable to other formats.
In
A detailed implementation example will now be given, which is illustrated in
{right arrow over (z)}=[z1, . . . , zN+λ]T (6)
The new symbols need not be added at the end, they can be placed anywhere in the burst. Next, the vector {right arrow over (z)} is precoded, step 405.
{right arrow over (Z)}=(N+λW)H·{right arrow over (z)} (7)
Note that the new size of the Fourier transform is N+λ. After precoding, a cyclic prefix of length L is added, step 407 and step 409.
Guard symbols are added to the left and right of (7), step 415.
A good choice for the guard symbols can be made by picking them so that the signal generated after pulse shaping has approximately constant amplitude over the guard. For example, if the pulse shaping filter is the linearized Gaussian Minimum Shift Keying (GMSK) filter, then the guard symbols may be defined by, step 413.
The positive real numbers Al, Ar are the amplitudes of the left and right guard symbols step 411. These amplitudes can be adjusted to control the power of the signal during ramp up or ramp down.
The signal (9) is pulse shaped, step 417 using the filter ρ.
Note that the signal (12) is λT seconds longer than the signal (5).
Below a description, step 419, is given how to shorten (12) so that
1. It has the same length τ as (5),
2. It fulfils the same spectrum constraints as (5), and
3. The link performance is not degraded.
Firstly, the constant λ is chosen in such a way that the time dispersion induced by the multipath propagation radio channel (i.e. from transmit antenna to receive antenna) plus the time dispersion caused by the receive chain is less than (L−λ+1)·T seconds.
Secondly, the first λT seconds of (12) are erased.
y
λ(t)=y(t+λT), 0≦t≦τ (13)
Thirdly, two raised cosine ramp functions are defined, step 421.
The parameter β determines the duration of the ramping period.
Finally, a new shortened baseband signal is generated by the multiplication of (13), (14) and (15), step 423 and 425.
s(t)=yλ(t)·rβ1(t)·rβr(t), 0≦t≦τ (16)
Note that a portion of the signal S defined by (16) corresponding to λT seconds of the filtered cyclic prefix, is affected by the left ramp (14). This does not have any effect upon link performance since by design S is subject to a time dispersion of less than λT seconds of Hence, the received signal can still be accurately represented as a cyclic convolution of the sent symbols with a channel filter. This is a property that allows easy demodulation by means of the DFT, and it explains why the shortened signal (16) does not suffer from degraded performance. Moreover, the spectrum of the product of the ramps (14) and (15) is easily obtained from the usual raised cosine spectrum and its impulse response, see Digital Communications, J. Proakis, Mc Graw Hill 4th Ed. (2000), but interchanging the time and frequency domains. Indeed, the product rβl(t)·rβr(t) defines a raised cosine window in the time domain, while the usual definition gives the raised cosine window in the frequency domain, see Digital Communications, J. Proakis, Mc Graw Hill 4th Ed. (2000). Thus, the spectrum of rβl(t)·rβr(t) is a modified sinc function, see Digital Communications, J. Proakis, Mc Graw Hill 4th Ed. (2000) and the magnitude of its side lobes is controlled by the roll off β. Since the spectrum of s is given by the circular convolution of the frequency response of the 3 terms
yλ(t), rβl(t) and rβr(t ) (17)
will satisfy the same spectral mask requirements as (5) provided β is large enough. Observe that since the ramps have the value 0 for t<0 or t>τ+λT, the signal (12) can be extended to t<0 or t>τ+λT without affecting the burst shortening method. This can be useful in order to ensure that the spectrum of (12) is as narrow as possible and that the high frequency components decay as fast as possible. It is also clear from the description above that the raised cosine ramps (14) and (15) can be replaced by other ramping functions provided the spectrum of the shortened signal (16) satisfies the same spectrum mask requirements as (5).
An exemplary burst shortening method is depicted in
The size of the Fourier transform is N+λ=144=24×32, which is a highly composite number. An IFFT of size 144=24×32 is typically 3 to 10 times faster than an IFFT of N=142, depending on hardware capabilities and implementation details.
The amount of time dispersion due to multipath radio propagation and receive filtering that can be effectively compensated is 5T=18.5 μs. With these values of the parameters the spectrum mask requirement for normal symbol rate, see 3GPP TS 45.005 Radio Transmission and Reception is fulfilled. This is depicted in
The left and right portions of the power vs. time mask, see 3GPP TS 45.005 Radio Transmission and Reception, together with simulated shortened bursts, are shown in
Thus,
Further
This shows that the output power does not exceed the maximum allowed power during ramping up or ramping down according to the power vs. time mask as set out in 3GPP TS 45.005 Radio Transmission and Reception.
By using the methods and devices as described herein for shortening the burst applied to EGPRS/EGPRS2, in particular a precoded EGPRS/EGPRS2 burst a number of advantages compared to conventional transmission is achieved. Thus, either the payload or the training sequence (or both) can be increased, resulting in improved link performance, increased throughput or both. In addition it can be used to increase the length of the IDFT. This is beneficial because the prime factor decomposition of the length of the IDFT determines the computational efficiency of the IFFT. For example, an IFFT of length 142=2×71 is typically 3 to 10 times more complex than an IFFT of length 144=24×32. Thus, a tool for simplifying the computational burden on the precoder is provided. This is essential for backward compatibility with legacy base stations. Moreover a more effective FFT also saves battery and processing time in the mobile station. In addition it is possible to preserve the spectrum mask of EGPRS/EGPRS2.
This application claims priority to U.S. Provisional Patent Application No. 61/314,666, filed Mar. 17, 2010 and incorporated herein by reference in its entirety.
Number | Date | Country | |
---|---|---|---|
61314666 | Mar 2010 | US |