I am glad that I am still working on the project.
We had great discoveries in week 10 that the noise at certain frequencies I mentioned in last report came from the power supply of FPGA. We tried to use a paper box to attenuate the noise from the air, but only to find that the intensity of that certain noise didn’t fall down proportionally as we expected, which means the noise was not from the air. Then we probed the upper board where the microphones are thoroughly to look for the source of this noise. The result shocked us because the noise was from the power. Consider the lower FPGA board provided the power of the upper board, we believed the noise was from FPGA. The further experiments proved our suspect. Actually, the noise was from the power supply of FPGA.
After this important discovery, Mr. Richter Ed decided to replace the FPGA board with the more advanced NI Elvis II platform to provide the power of the upper microphone board. And it works! Though we only replace for one out of four boards, the noise was gone and the SNR reached almost the estimated value.
With the excitement and the encouragement our new discovery brought to us, we entered week 11. Since all the simulations work as we wish, Mr. Richter asked us to proceed with real world tests. To make the result more accurate, we firstly make a record with a single sound source. The purpose of doing this is to calculate the current lagging timetable, i.e., the current relative positions of 16 microphones. Then we played 2 audios via cellphones from different direction and make records for before-process effect and after-process effect. The comparison results are really encouraging. Though only 16 microphones were involved, we could apparently tell the differences between two records. The beam formed sound from the presumed facing direction was much clearer than the not-beam-formed sound from the same direction under an environment where there was another interfering sound source at a different direction.
By now, I could draw a line and encapsulate my project!
We had great discoveries in week 10 that the noise at certain frequencies I mentioned in last report came from the power supply of FPGA. We tried to use a paper box to attenuate the noise from the air, but only to find that the intensity of that certain noise didn’t fall down proportionally as we expected, which means the noise was not from the air. Then we probed the upper board where the microphones are thoroughly to look for the source of this noise. The result shocked us because the noise was from the power. Consider the lower FPGA board provided the power of the upper board, we believed the noise was from FPGA. The further experiments proved our suspect. Actually, the noise was from the power supply of FPGA.
After this important discovery, Mr. Richter Ed decided to replace the FPGA board with the more advanced NI Elvis II platform to provide the power of the upper microphone board. And it works! Though we only replace for one out of four boards, the noise was gone and the SNR reached almost the estimated value.
With the excitement and the encouragement our new discovery brought to us, we entered week 11. Since all the simulations work as we wish, Mr. Richter asked us to proceed with real world tests. To make the result more accurate, we firstly make a record with a single sound source. The purpose of doing this is to calculate the current lagging timetable, i.e., the current relative positions of 16 microphones. Then we played 2 audios via cellphones from different direction and make records for before-process effect and after-process effect. The comparison results are really encouraging. Though only 16 microphones were involved, we could apparently tell the differences between two records. The beam formed sound from the presumed facing direction was much clearer than the not-beam-formed sound from the same direction under an environment where there was another interfering sound source at a different direction.
By now, I could draw a line and encapsulate my project!