Hi Rolando, It is indeed quite possible to build a CASPER correlator or other instrument using a combination of high-speed ADCs, FPGAs and CPUs (+ GPUs). In fact, many of the latest generation of CASPER instruments use this so-called ‘heterogeneous’ approach, using ADC/FPGAs to digitize, channelize and packetize data, and CPU/GPUs to perform correlation, additional channelization or other DSP. I suggest taking a look at some of the CASPER workshop talks for an introduction to these techniques (2012 talks at https://science.nrao.edu/science/meetings/casper2012)
If your goal is to eventually build a large correlator or multi-beam spectrometer with many features, a FPGA/CPU/GPU system might be a very good architecture and developing a IBOB-based 8-input heterogeneous system would be worthwhile. However, if your ultimate goal is a working 2048 channel 8-input FX correlator, it would be far easier to just use a newer-generation FPGA board with more resources (e.g. ROACH I/II). Cheers, Andrew On Jan 6, 2014, at 12:52 PM, Rolando Paz <[email protected]> wrote: > Hi CASPER > > I have a question about correlators, hope you can guide me ... > > I have a IBOB and two QUADC, and I want to develop a 8 inputs correlator, > implementing the correction of geometric delay. > > I succeeded compile a 1024-channels spectrometer, however when I try to > compile for 2048 channels arose several errors. I understand that the reduced > capacity of the FPGA IBOB not allow me more ... > > In order to reduce the work of the FPGA (using IBOB), is it possible to > obtain data (quadc-IBOB) and transfer them by the CX4 port, or 10/100MB port, > to a PC and then perform the correlation? > > In other words I want to make the correlation of the signals on the PC and > not on the IBOB. Is this possible? > > From this, new questions arise: > What would be the data format sent to the PC?, and > Which software use to process the data? > > :-) > > Best Regards > > Rolando Paz > >
