hi karl, to add to what jack emailed about using a snap board for your non-astronomical array:
snap has 12 ADC's on board. if you want more analog inputs, you could plug in an ADC16 board; then you'd have 12 + 16 = 28 inputs, up to 250 Msps each. as jack pointed out, snap only has a pair of 10Gbe ports (20Gbit/sec in, 20Gbit out), so assuming 100 MHz bandwidth, and 4 bit real, 4 bit imaginary data, (we typically use 4+4 for correlators and beamformers), you would be limited to 24 analog inputs: 24 inputs x 100 MHz x 8 bits = 19.2 Gbit/sec snap uses the new jasper tool flow, is easier to program, and is cheaper than roach2. on the other hand, you already have a lot of roach2 experience, and if you have a spare roach2 board, then that's the cheapest option, as you only need to procure an adc16 board..... best wishes, dan On Thu, Mar 15, 2018 at 10:05 PM, Jack Hickish <[email protected]> wrote: > Hi Karl, > > On Thu, 15 Mar 2018 at 21:59 Karl Warnick <[email protected]> wrote: > >> Hi all, >> >> I have a non-astronomical comms antenna array project that does not have >> detailed specs for bandwidth and number of antenna elements. I need to >> build a programmable platform that I can use for multichannel sampling >> and real time DSP. >> >> Our group has considerable experience with ROACH based systems over the >> years and hybrid FPGA - GPU architectures, but the students with much of >> the expertise have graduated and moved on, as students tend to do. We >> have an x64 board and ROACH that on paper could do the job, but the >> hardware seems old enough now that I wonder if it might be wise to move >> to a new platform for my new project. We also like the easier >> programmability of GPUs for matrix based array signal processing >> algorithms. We also have a system with digitizers, polyphase filterbanks >> running on ROACH boards, ethernet switch, and HPCs with GPUs, but that >> system is in operation at GBO and is probably overkill for the current >> project. >> >> This leads to my question. To realize a system with 16 analog channels >> and analog bandwidth ranging between a few MHz up to 100 MHz (I realize >> that this is a rather ill defined range, but I feel fortunate to have >> the flexibility), that can do digitization, filterbank to separate into >> frequency channels, and enough processor power for real time XB >> (correlator/beamformer) type signal processing, with current CASPER >> hardware, what would be the ideal, recommended hardware setup? >> > > I think your options here are either a/some SNAP board(s) -- 12 ADC > channels at up to 250 MSample/s sampling rate, or a ROACH2 + adc16 card -- > 16 channels at up to 250MSample/s. Either would give you Ethernet output on > SFP+ connectors. SNAP can do up to 20Gb/s output, ROACH2 up to 80Gb/s. > In either case, there should be some simulink designs you could use as a > starting point, from PAPER/HERA or other projects. > > Cheers > Jack > > >> >> The relatively modest bandwidth requirement may also point to a >> commercial FPGA/ARM core board with an ADC expansion board, and I'm >> pursuing that path as well. From another project, I have an expansion >> board of our own design with eight ADC channels that plugs into a >> microZED board, which actually comes somewhat close to meeting the >> current requirements. Expanding the ADC board and moving to a bigger >> commercial digital board is an option. >> >> Thanks to all for any feedback! >> >> Best, >> Karl >> >> -- >> Karl F. Warnick >> Department of Electrical and Computer Engineering >> Brigham Young University >> 459 Clyde Building >> Provo, UT 84602 >> (801) 422-1732 >> >> >> >> >> >> -- >> You received this message because you are subscribed to the Google Groups >> "[email protected]" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> To post to this group, send email to [email protected]. >> > -- > You received this message because you are subscribed to the Google Groups " > [email protected]" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To post to this group, send email to [email protected]. > -- You received this message because you are subscribed to the Google Groups "[email protected]" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected].
