ICACT20210153 Question.1
Questioner: gueye.ibrahima1@esp.sn 2021-02-09 오전 1:15:24 |
ICACT20210153 Answer.1
Answer by Auhor mihai.sima@ieee.org 2021-02-09 오전 1:15:24
Chrome  Click!! |
| Hello, thank you for the presentation, I have a question about the massive mimo system, I ask what are the problems related to the massive mimo transmission |
In Massive MIMO, many antennas are used to provide higher throughput and increased spectrum efficiency. Two technical problems need to be addressed in implementing a Massive MIMO system: (1) the build of a base station with a very large number of receive antennas, and (2) the amount of computation required to estimate the transmitted signals.
|
ICACT20210153 Question.2
Questioner: lijingwei10@huawei.com 2021-02-08 오후 4:00:07 |
ICACT20210153 Answer.2
Answer by Auhor mihai.sima@ieee.org 2021-02-08 오후 4:00:07
Chrome Click!! |
| In this paper, R is 128, T is 16, if we increase T, can the Bit-Error Rate keep the similar perfermance using coarse Analog-to-Digital Converters? |
This is a good question. A larger T (e.g., 32 or 64) means a larger number of users (terminals). The answer to your question is Yes, but the computation becomes increasingly difficult, since larger matrices (32-by-32 or 64-by-64 for the assumed examples) will need to be inverted. |
ICACT20210153 Question.3
Questioner: lijingwei10@huawei.com 2021-02-08 오후 4:09:18 |
ICACT20210153 Answer.3
Answer by Auhor mihai.sima@ieee.org 2021-02-08 오후 4:09:18
Chrome Click!! |
| Have you evaluated or predicted the profit quantitatively by using the coarse ADCs? |
According to our calculations (to be presented in a separate paper), the noise power is significantly greater than the information signal power on each individual antenna. This means that if a high resolution ADC is employed most of its output bits are nearly independent of the information signal. This provides the motivation for the use of coarse quantization, as for high-resolution quantizers the bits of lower significance are mostly independent of the data signal. |
ICACT20210153 Question.4
Questioner: kzyano@atr.jp 2021-02-10 오후 4:23:23 |
ICACT20210153 Answer.4
Answer by Auhor mihai.sima@ieee.org 2021-02-10 오후 4:23:23
Chrome Click!! |
| Thank you for your presentation. If higher-level modulation such as 16QAM or 64QAM is used, do we need more bits than for QPSK to achieve the same BER? |
Thank you for your question. For higher-level modulation we do need more bits if we keep the Eb/N0 the same. |
ICACT20210153 Question.5
Questioner: jmr77@mju.ac.kr 2021-02-10 오후 4:48:09 |
ICACT20210153 Answer.5
Answer by Auhor mihai.sima@ieee.org 2021-02-10 오후 4:48:09
Chrome Click!! |
| Usually, before using fixed point arithmetic, it is necessary to first examine the dynamic range of each signal. Through this, the optimal bit-width and the location of the fixed point can be found. Has this process been preceded? |
The dynamic range of the signal is adjusted by an AGC, so that the signal is Gaussian with average 0 and standard deviation 1 for the presented simulations. Please notice that for 1-bit quantization, the dynamic range has no meaning (negative signal gives logic 1, positive signal gives logic 0). In this heavily coarsed quantization situation there is a loss of 3dB, but at least we get something out of this MMIMO communication system (maybe we should not use this with 5G, but with IoT it would not be a bad idea). |
ICACT20210153 Question.6
Questioner: rusly.eic04g@nctu.edu.tw 2021-02-11 오전 12:37:13 |
ICACT20210153 Answer.6
Answer by Auhor mihai.sima@ieee.org 2021-02-11 오전 12:37:13
Chrome Click!! |
| Thank you for your presentation. My question is, how would fixed arithmetics affect the computational time for channel estimation? |
Fixed-point addition is faster than floating-point addition (where normalization is required). Multiplication latency is about the same in both fixed-point and floating-point arithmetic. As a result, fixed-point arithmetic leads to a shorter computing time. Thank you for your question. |
