--- Quote Start ---
Hi, 50MHz on two stream of 14 bit sound as 1.4 Gbps, so it is also difficult to sustained stream data to pc. not in some way compressed GigaLan has not enough bandwidth and also USB3 get saturated too.
If you accept sample some data then do FFT and Wavelet on chunk then you can transfer data to pc to do some postprocessing.
So how long is sample to process from FFT, how many FFT point and how is wavelet applied to?
RAW data has to be transferred to pc also?
Cyclon V can manage this but the best is SOC with dual core ARM, if you wish to manage real time data processing forget Windows ( I HATE due is not real multitask nor SMP) and use task spreading from Linux platform or other embedded OS.
ALso to do multiple transfer you need split process on thread and use one processor core, Nios can be used to manage task on low level as register programming and DMA coordination on FPGA fabric side.
NOt sure this is the best but again too many details are missing.
Regards
Roberto
--- Quote End ---
Only 1 sec RAW DATA is enough
I don't want to work with Linux bcoz I'm not expert in Linux.. I think it takes long time to learn it
some experts in Xilinx SOC have told me you can manage the RAM and USB with SOC easily without Linux. also told me SOC is better than Microblaze (is similar to Altera Nios)
I'm a master student, All of my time is 7 month.
DE-1 SOC board has only 512MB DDR3 RAM on the HPS and 64MB on the FPGA side
Arria V GX Starter board has 256MB DDR3 RAM but has lots of resources.
Cyclone V GX Starter board has 512MB LPDDR2
my main problem:
which board is better?.. I need an "USB device port" or "Ethernet port" for transferring data to my laptop
I know transferring the data over them needs IP core or writing some codes in FSM
which one is better and easier? Nios or SOC (without Linux)
ps: if I want to process on the FPGA I have to process 4096 point FFT or Continues Wavelet on the sampled data. Continues Wavelet needs Huge Convolution process.
or I can transfer the RAW DATA to my laptop (2x50Mx2Byte=200Mbyte) and then apply the FFT or wavelet on them by Matlab.