MIMO System Operation Comparison

Therefore, in MIMO system, processor and ASIC own half a symbol duration to com-plete operations. It can be expected that processor’s operations per second will be doubled of scheduleⅡ in SISO system as shown in Fig. 3.22 ~ 3.26 based on time schedule diagram

shown in Fig. 3.9.

2 4 8 16 32 64

0 100 200 300 400 500 600 700 800

compare 64-points MIMO

FFT Length of ASIC

MOPS

11n 64-points FFT(50Mhz)with radix-8 11n 64-points FFT(50Mhz)with radix-4 11n 64-points FFT(50Mhz)with radix-2

Fig. 3.22 64-points FFT operation comparison of MIMO time schedule

In Fig. 3.22, it shows the 64-points FFT MOPS of radix-2, radix-4 and radix-8 based on IEEE 802.11n standard, x-axis is what kind of branch ASIC FFT we will decide, if the processor provide us some restricted MOPS to use shown on y-axis. This figure is designed according to time schedule of MIMO system. The performance is twice than SISOⅡ system of Fig. 3.17, but the usage of time schedule in a symbol is improved. In a symbol duration, we can execute FFT two times and just only use one N-points branch FFT of ASIC well.

2 4 8 16 32 64 128 256 512 0

200 400 600 800 1000 1200 1400 1600 1800 2000

compare 128-points MIMO

FFT Length of ASIC

MOPS

11n 128-points FFT(50Mhz)with radix-8 11n 128-points FFT(50Mhz)with radix-4 11n 128-points FFT(50Mhz)with radix-2

Fig. 3.23 128-points FFT operation comparison of MIMO time schedule

In Fig. 3.23, it shows the 128-points FFT MOPS of radix-2, radix-4 and radix-8 based on IEEE 802.11n standard, x-axis is what kind of branch ASIC FFT we will decide, if the processor provide us some restricted MOPS to use shown on y-axis. This figure is designed according to time schedule of MIMO system. The performance is twice than SISOⅡ system of Fig. 3.18, but the usage of time schedule in a symbol is improved. In a symbol duration time unit, we can execute FFT two times and just only use one N-points branch FFT of ASIC well.

2 4 8 16 32 64 128 256 512 0

50 100 150 200 250 300 350

compare 512-points MIMO

FFT Length of ASIC

MOPS

16e 512-points FFT(50Mhz)with radix-8 16e 512-points FFT(50Mhz) with radix-4 16e 512-points FFT(50Mhz) with radix-2

Fig. 3.24 512-points FFT operation comparison of MIMO time schedule

In Fig. 3.24, it shows the 128-points FFT MOPS of radix-2, radix-4 and radix-8 based on IEEE 802.16e standard, x-axis is what kind of branch ASIC FFT we will decide, if the processor provide us some restricted MOPS to use shown on y-axis. This figure is designed according to time schedule of MIMO system. The performance is twice than SISOⅡ system of Fig. 3.19, but the usage of time schedule in a symbol is improved. In a symbol duration time unit, we can execute FFT two times and just only use one N-points branch FFT of ASIC well.

2 4 8 16 32 64 128 256 512 1024 2048 0

100 200 300 400 500 600 700 800

compare 1024-points MIMO

FFT Length of ASIC

MOPS

16e 1024-points FFT(50Mhz)with radix-8 16e 1024-points FFT(50Mhz)with radix-4 16e 1024-points FFT(50Mhz)with radix-2

Fig. 3.25 1024-points FFT operation comparison of MIMO time schedule

In Fig. 3.25, it shows the 1024-points FFT MOPS of radix-2, radix-4 and radix-8 based on IEEE 802.16e standard, x-axis is what kind of branch ASIC FFT we will decide, if the processor provide us some restricted MOPS to use shown on y-axis. This figure is de-signed according to time schedule of MIMO system. The performance is twice than SISOⅡ system of Fig. 3.20, but the usage of time schedule in a symbol is improved. In a symbol du-ration time unit of 2x2 antennas, we can execute FFT two times and just only use one N-points branch FFT of ASIC well.

2 4 8 16 32 64 128 256 512 1024 2048 0

200 400 600 800 1000 1200 1400 1600 1800

compare 2048-points MIMO

FFT Length of ASIC

MOPS

16e 2048-points FFT(50Mhz)with radix-8 16e 2048-points FFT(50Mhz) with radix-4 16e 2048-points FFT(50Mhz) with radix-2

Fig. 3.26 2048-points FFT operation comparison of MIMO time schedule

In Fig. 3.26, it shows the 2048-points FFT MOPS of radix-2, radix-4 and radix-8 based on IEEE 802.16e standard, x-axis is what kind of branch ASIC FFT we will decide, if the processor provide us some restricted MOPS to use shown on y-axis. This figure is de-signed according to time schedule of MIMO system. The performance is twice than SISOⅡ system of Fig. 3.21, but the usage of time schedule in a symbol is improved.

In the method of Fig. 3.26: 2048-points FFT operation comparison of MIMO time schedule. IEEE 802.16e standard, if we chose 64-points ASIC FFT, the processor operations will be used about 900 MOPS. In the other word, the MOPS of MIMO schedule will be doubled than the time scheduleⅡof SISOⅡ.

In this section, we introduce scheduleⅠ、Ⅱ in SISO system, and a schedule in MIMO system. According to time schedule system based on time scheduleⅠof SISOⅠand time scheduleⅡ of SISOⅡ, the cost of hardware in scheduleⅠis less than scheduleⅡ, but the uti-lization of scheduleⅠis less than scheduleⅡ. In MIMO system, bad utiuti-lization can be im-proved by changing ASIC and processor order. In 2x2 MIMO systems, it only needs a

pro-cessor and a branch FFT of ASIC. This schedule not only lower hardware cost, but also in-crease the utilization of module.

Therefore, we use the equation (2) to calculate the “Processor Operations” of 64-points, 128-points, 512-points,1024-points and 2048-points FFT and according to IEEE 802.11n/16e standards, we can predict the time of these three schedules (SISOI,SISOII and MIMO) in a symbol duration. Finally, MOPS (Million Operations per Second) has been eva-luated by our estimation. In next chapter, we want to implement the processor’s architecture of SISOⅠ in Fig. 3.6, SISOⅡ in Fig. 3.8 and MIMO in Fig. 3.10 with “Micro Blaze Proces-sor”, which is a embedded system implemented by FPGA tools

Chapter 4

Implementation of the Structure

with MicroBlaze Processor