2009-7-20 22:44
A and B
LTE uplink adopts BPSK, looks like a Pi / 2-BPSK, i.e. Pi / 4-QPSK promote natural, even symbol constellation of the normal use, the odd symbol constellation using the offset, the amount of this phase jump can be reduced, aims to improve the spectrum utilization, due to the reduced amplitude variation caused by the filter
This would have on the old BPSK modulation scheme A no interest, but the BPSK mapping mode in LTE 36.211-7.1 A attracted attention, instead of mapping to (+1 -1), but is mapped to 1 / sqrt (2 ) * (1 + j, -1-j).
A first glance, the thought and mapping it to (+1 -1) is the same, as long as the minimum Euclidean distance between the symbols on the same line.
But B suddenly raised a problem as long as the first mapping method to extract the in-phase component, while the second has to extract the in-phase and quadrature components, that is a first embodiment of the quadrature component of noise can be suppressed, and the first two ways but then the noise completely and theoretical performance with phase and quadrature components of the difference should be 3dB
A rip off immediately, because very mention acetate corroboration lethal, SSB and DSB, known, DSB demodulation gain 2, gain 1 demodulating SSB, DSB reasons only the quadrature component, while the SSB quadrature-phase are used, (in the case of completion of the synchronization, in fact, the DSB and SSB system with the same performance with respect to the base, because only half the bandwidth of SSB, bandpass filtered input SNR already high 3dB)
A reset of the brain, the implementation process both in the brain to reproduce again, both the performance is certainly the same, because the difference is, vine leaves curve (to borrow the term) a difference of pi / 4. A confidence, start counter, consider DSB and SSB, SSB demodulator gain is a cause of a signal which is the quadrature component are also discarded the (quadrature component signal is Hilbert transformed, can be utilized right? looks like all the book's approach is to directly abandoned, so the easiest receiver), which is different and BPSK said earlier, the evidence does not hold.
A then B basic argument to refute, but knowing that it could not tell unreasonable grounds, (in fact, I understand why, no matter which way the mapping, complex Gaussian noise is only one degree of freedom to work, but can not make farmers understand the reasons are not a good reason), methyl continue to crash in. . .
A model of this proposition, expressed as a problem, the channel statistical properties of two identical, a signal is transmitted from a good channel, or sub-sub-signal power, with two good channels pass, (farmer uncle laugh, of course, Like, the kind of food I was a paying taxes to the government or to pay twice, is not the same! cut, engage in communications really are an idiot). The generalized problem, so to express, more than the same statistical properties of the N channel, assigning a power signal to each channel, its performance and whether the distribution, intuition tell us, of course, is irrelevant.
An example of two channels, A demonstrated as follows:
Signals S, noise N, the signal distribution coefficient k1, k2, of course, k1 ^ 2 + k2 ^ 2 = 1
Not fool a receiving end, the signal power is large the channel course it more reliable, have a large coefficient of a point, then the received signal R is
R = k1R1 + k2R2, R1, R2 respectively, the output signals of two channels, (MMSE laugh)
R=k1(k1S+N1)+k2(k2S+N2)
=(k1^2+k2^2)S+(k1N1+k2N2)
=S+(k1N1+k2N2)
Reply 2 F:
1), for BPSK, the two signal points regardless of whether the rotation without any change to the quadrature receiver performance, the variance is assumed that N1 and N2 are both 1, R = S + (N1 ) and R = S + (1 / sqrt (2 ) N1 + 1 / sqrt (2) N2) are also a noise variance, the noise power is the same.
2), BPSK due to the phase jump will cause the envelope fluctuation and spectrum spreading, so will adopt the phase rotation, i.e. the odd and even symbols of different symbol constellations. For BPSK, can be employed odd symbols (+ 1, -1), the even symbol using (+ j, -j), i.e., even-numbered symbol constellation is rotated Pi / 2, thus called the Pi / 2-BPSK, and that Pi / 4-QPSK principle is the same. The advantage of this is that the phase of two symbols before and after a maximum of 90 mutations (Pi / 2-BPSK) and 45 degrees (Pi / 4-QPSK).
In addition to quadrature receiver, BPSK efficiency is very low, therefore less used, LTE is the BPSK except that a small amount of control signaling transmitted, even symbol plus pi / 2 phase rotation is to reduce abrupt phase changes.
ps: log I constellation symbol is not even wondering why phase rotation must be pi / 2 in (for pi / 2-BPSK, the normal rotation even-numbered symbols), but why all symbols have the entire 45 degrees rotation of a phase and whether the impact on performance. I think the reason for the entire 45-degree rotation is to reduce the quantization error, without rotation, concentrated in the I signal, the quantization error is large, as performance, no effect