Dear member of the Ettus mailing list, particularly those of you actually expert in RFNOC.
We have started recently to try to use the rfnoc_devel branch of UHD and of the FPGA source code in order to implement a project where we want to transfer part of the processing on the FPGA. In a preliminary stage we are using gnuradio and the uhd bindings to simulate the producers, but we plan to move to C++ code later on. We have however found several road blocks. I would therefore like to explain what we want to achieve, what our problems are, and see whether you have any idea how to get past them. Application: We have several processes (say 2 to 10) producing streams of complex numbers to be interpreted as I/Q symbols in time domain at a low rate (say 12.5MHz per stream, which is 200MHz/16). We would like every stream to be up-sampled (to 200MHz), shaped with an FIR filter, and up-converted to a user-defined and stream-specific frequency. Finally, we would like to sum up all digital samples together, and forward them to a radio block. Problems: 1) We would like to send the stream values to the USRP through uhd::tx_streamer::send() as separate buffers, but in an FPGA image there seems to be only one DMA_fifo block with only two inputs. Is there a way to increase the number of input ports to this block, and/or to have more such blocks? (without compromising the available data bandwidth) 2) We were not able to find documentation about RFNOC DUC blocks. In particular we don't know how to modify the shaping filter and what is the range of the allowed up-conversion frequencies. 3) It seems that the add/sub block only has two inputs, and that these two inputs share the crossbar bandwidth, which makes it impossible to join several streams already at 200MHz. Is there a way out of this (maybe with a custom block)? 4) More in general, which bandwidth or functional limitations are implicit in the crossbar implementation provided by RFNOC? Best regards/Mit freundlichen Gruessen, Stefano Bettelli Senior research engineer Optical & Quantum Laboratory Huawei Technologies Duesseldorf GmbH _______________________________________________ USRP-users mailing list USRP-users@lists.ettus.com http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
