Honored Contributor
Honored ContributorOk, I assume after 1 second sampling all stop there so data chunk is not so large and 64MB of ram are useful as buffer to store on main memory.
Final data is large 50Ms * 4 Byte -> about 200MB I don't see simple solution other than use Linux to manage data stream and also apply some software, 4096 point can be stored on fast FPGA memory as dual port, then changed to new sample by DMA, output buffer must be transferred away before restart a new FFT computation, this on Cyclon has no trouble and you can allocate multiple buffer then use main memory as storage. I don't see simple solution only using IP cores and Linux you state is difficult help a lot, program to transfer streaming to network or USB is not so complicated and can be debugged without touching hardware on FPGA fabric. I am using Linux on PC as main development and is faster on compilation, before Intel acquisition was more than 10 times faster, I am using Linux on SOC due I developed all network communication and software emulation on this platforn, I don't wish use windows is slow and cumbersome to have working software in it. Again you can simply allocate buffers then transfer by ftp or storage sharing or some other simple way ffrom Linux domain. If you don't have a support managing TCPIP or USB I suspect generate a big nightmare. You can find a lot of help on Linux, if your data can be transferred with some time try also evaluate interface Raspberry PI to FPGA and avoid use the complexity of DE1-SOC environment. De1-SOC has 1GB of ram not 512K organized as two devices 16x512 so you can feed both devices with the stream from both channel in parallel as 32 bit. http://www.terasic.com.tw/cgi-bin/page/archive.pl?language=english&categoryno=205&no=836&partno=2 Regards
