Erik, You can use the generic IIO Device blocks to talk to the ADRV9008/9. The PHY Device is "adrv9009-phy", Device is "axi-adrv9009-rx-obs-hpc" to get the observation path, and the channels are voltage0_i and voltage0_q.
In regards to processing data, even on the FPGA fabric 450 MHz is challenging. I would recommend looking at captured buffers in isolation and prove that your signal processing is working. Then migrate things to the FPGA fabric. If you have more questions please post them on the ADI support forums here: https://ez.analog.com/ -Travis On Thu, Jun 27, 2019 at 10:36 AM Müller, Marcus (CEL) <muel...@kit.edu> wrote: > Hi Erik, > > On Thu, 2019-06-27 at 09:19 +0000, Erik Heinz wrote: > > Hi Marcus, > > > > thank you for clarifying. I already guessed that signal processing in > the FPGA is required, but I had the hope that we can start with a CPU-only > setup with limited performance and improve it by FPGA implementations later > on. > > > > Do you have a rough guess what bandwidth a Cortex-A53 ARM on an ZCU102 > board would be able to handle, considered some basic data processing within > gnuradio? > > Puh, that really depends on the kind of processing your basic > processing is! > Also, you'd be surprised how much CPU you can spend on getting data > from devices to userland, so how you'd go about that can have a heavy > influence. > > But: Really no big deal: Simply prototype your whole signal processing > on a capable PC, identify the bottlenecks, port and test them on the > FPGA, and test whether what you consider "lightweight" stuff works in > isolation on a comparable ARM at sufficient rates. > > > > This approach, however, would only work if there were some support and > some hooks within the Linux IIO kernel driver and/or libiio for including > customer FPGA designs. > > I'll go ahead and admit that I don't know IIO well enough to comment on > this. But: there's certainly folks at analog devices that know! Travis > would be my go-to person here. > > Best regards, > Marcus > > > So far I have not found any hint that such support exists. In this case > running gnuradio on the Zync board would be a dead end solution that > cannot be accelerated easily, and no appropriate approach for the given > task. > > > > Best regards, > > Erik > > > > ________________________________________ > > Von: Müller, Marcus (CEL) <muel...@kit.edu> > > Gesendet: Dienstag, 25. Juni 2019 13:05 > > An: discuss-gnuradio@gnu.org; Erik Heinz > > Betreff: Re: [Discuss-gnuradio] ADRV9008-2, Zynq FPGA boards, and > gnuradio > > > > Hi Erik, > > > > so, GNU Radio *itself* is purely host-based software; in other words, > > you'd be processing 100s of MHz on the poor Zynq's ARM (will not even > > remotely happen). > > > > What you'd need is to do the signal processing in the FPGA, for the > > most significant part. After you've selected your channel and reduced > > your sampling rate to that, yes, GNU Radio could very well deal with > > that data (if the ARM CPU is up for that). > > > > There's an accelerator framework, that Ettus developed for their Zynq- > > based products, called RFNoC. That works relatively nicely with GNU > > Radio. Pretty sure Analog Devices themselves also have libiio- > > compatible ways of signal (pre)processing on Zynqs, but I'm not > > knowledgeable about that at all. > > > > Which FPGA would fit your bill depends on the kind of signal processing > > you'd need to do. I'd assume the 7000 series Zynq would be a bit on the > > limits of its bandwidths dealing with 450 MS/s. > > > > Best regards, > > Marcus > > > > On Tue, 2019-06-25 at 10:04 +0000, Erik Heinz wrote: > > > Hello everyone, > > > > > > we are developing an application where a large bandwidth (some 100 MHz > at 5–6 GHz mid frequency) needs to be processed to extract a small > bandwidth > > > payload signal. > > > > > > For rapid prototyping and testing algorithms I am thinking about using > an ADRV9008-2 evaluation board, a Xilinx Zynq FPGA board, and gnuradio. > > > > > > I understand that there do exist ready-made Linux distributions that > run on boards like the ZC702 (Zynq-7000 based) or ZCU102 (Ultrascale+ > based) and it is possible to run gnuradio on top of it. > > > I also found some hint that there does exist software to access an > EVAL-ADRV9008 board connected to the Zync board from within gnuradio. > > > So I assume in principle it should be possible to process the 450 MHz > bandwidth from the ADRV9008-2 directly on the Zync board using gnuradio in > real time. > > > > > > Can anyone confirm that this assumption is correct and that such a > setup could work? Has this be done before? > > > Would both the ZC702 and the ZCU102 be suitable? > > > Where should I start reading? > > > > > > Lots of questions. Thank you for some answers. > > > Erik > > > > > > _______________________________________________ > > > Discuss-gnuradio mailing list > > > Discuss-gnuradio@gnu.org > > > https://lists.gnu.org/mailman/listinfo/discuss-gnuradio >
_______________________________________________ Discuss-gnuradio mailing list Discuss-gnuradio@gnu.org https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
