Forums

Interface Xilinx KC705 to BeagleBone?

Started by pfraser July 22, 2012
I'm playing with the idea of interfacing a BeagleBone
(cheap dual ARM Cortex A8 board) to a Xilinx KC705
Kintex development board. This will give me much more
CPU processing power than a microblaze could.

I thought I could probably do it with a passive interface
because the Kintex can deal with 3.3 Volt I/O.

I'd probably use a Xilinx 105 debug board on the FMC
HPC connector, and hand build an interface board between
the debug board and the BeagleBone.

That would leave the LPC connector free for an Avnet
HDMI input board (I'm playing around with some video
processing / measurement ideas).

I would then develop a Angstrom Linux driver for
the TI GPMC interface to the Kintex.

Anybody see any flaws in this plan? Any advice?
Anybody done some / all of this already, and prepared
to share so that I don't need to re-invent the wheel?

Thanks

Pete

pfraser <pete_fraser@comcast.net> wrote:
> I'm playing with the idea of interfacing a BeagleBone > (cheap dual ARM Cortex A8 board) to a Xilinx KC705 > Kintex development board. This will give me much more > CPU processing power than a microblaze could.
> I thought I could probably do it with a passive interface > because the Kintex can deal with 3.3 Volt I/O.
> I'd probably use a Xilinx 105 debug board on the FMC > HPC connector, and hand build an interface board between > the debug board and the BeagleBone.
> That would leave the LPC connector free for an Avnet > HDMI input board (I'm playing around with some video > processing / measurement ideas).
> I would then develop a Angstrom Linux driver for > the TI GPMC interface to the Kintex.
> Anybody see any flaws in this plan? Any advice? > Anybody done some / all of this already, and prepared > to share so that I don't need to re-invent the wheel?
Did you think about the ZYNC and the ZED board (http://www.zedboard.org)? If availability really starts August, you will have much less hassle to start... Bye -- Uwe Bonnes bon@elektron.ikp.physik.tu-darmstadt.de Institut fuer Kernphysik Schlossgartenstrasse 9 64289 Darmstadt --------- Tel. 06151 162516 -------- Fax. 06151 164321 ----------
I don't know the status of it, but this is listed in the "Cape Board 
Registry": http://specialcomp.com/beaglebone/BeagleBone_FPGA.html

If you dig into this, please post the status here.

BobH


On 7/22/2012 1:27 PM, pfraser wrote:
> I'm playing with the idea of interfacing a BeagleBone > (cheap dual ARM Cortex A8 board) to a Xilinx KC705 > Kintex development board. This will give me much more > CPU processing power than a microblaze could. > > I thought I could probably do it with a passive interface > because the Kintex can deal with 3.3 Volt I/O. > > I'd probably use a Xilinx 105 debug board on the FMC > HPC connector, and hand build an interface board between > the debug board and the BeagleBone. > > That would leave the LPC connector free for an Avnet > HDMI input board (I'm playing around with some video > processing / measurement ideas). > > I would then develop a Angstrom Linux driver for > the TI GPMC interface to the Kintex. > > Anybody see any flaws in this plan? Any advice? > Anybody done some / all of this already, and prepared > to share so that I don't need to re-invent the wheel? > > Thanks > > Pete >
On Sunday, July 22, 2012 4:27:45 PM UTC-4, pfraser wrote:
> I&#39;m playing with the idea of interfacing a BeagleBone > (cheap dual ARM Cortex A8 board) to a Xilinx KC705 > Kintex development board. This will give me much more > CPU processing power than a microblaze could. >=20 > I thought I could probably do it with a passive interface > because the Kintex can deal with 3.3 Volt I/O. >=20 > I&#39;d probably use a Xilinx 105 debug board on the FMC > HPC connector, and hand build an interface board between > the debug board and the BeagleBone. >=20 > That would leave the LPC connector free for an Avnet > HDMI input board (I&#39;m playing around with some video > processing / measurement ideas). >=20 > I would then develop a Angstrom Linux driver for > the TI GPMC interface to the Kintex. >=20 > Anybody see any flaws in this plan? Any advice? > Anybody done some / all of this already, and prepared > to share so that I don&#39;t need to re-invent the wheel? >=20 > Thanks >=20 > Pete
It seems the BB has a peripheral mux on all the pins so you can wire the FP= GA to the BB in pretty much any pinout and get to any of the peripherals (I= 'm not sure if there are limitations). But I would suggest that you duplic= ate the pinout of the LCD7 which seems to use a parallel interface and you = might be able to reuse some of the driver code. I'm thinking of doing some= thing similar. You might even be able to multiplex the interface and use i= t for both the FPGA board and the LCD7 display. =20 They are doing a 3.5" display, but I don't know if the specs are available = for that yet. They weren't the last time I looked.=20 Rick
Uwe Bonnes wrote:

> Did you think about the ZYNC and the ZED board (http://www.zedboard.org)? > If availability really starts August, you will have much less hassle to > start...
I wasn't familiar with that board. It looks interesting, and the price is right. The FPGA is way underpowered for the video processing I would want to do, but perhaps I could use this board instead of the BeagleBone, and couple it to the KC705 with an FMC to FMC cable. That would be a lot simpler mechanically than using the BeagleBone. I had originally intended to use perf board with connectors for the FMC-105-Debug, and headers for the BeagleBone but unfortunately the 105 (which I own) and the similar TED parts (which I don't) have connectors which are not set on 0.1" centers. That means an even worse mechanical kludge. So, in favor of the ZED board is mechanical neatness, and less work hacking up sections of perf board and gluing them together (though I would need to check that I could present the appropriate signals to the ZED board's LPC connector). Also, it's A9 rather than A8. In favor of the BeagleBoard is a reasonably mature Angstom distribution, and a full user community. Who knows what distribution will be available for the ZED board initially? Thanks Pete
rickman wrote:
> But I would suggest that you duplicate the pinout of the LCD7 > which seems to use a parallel interface and you might be able > to reuse some of the driver code. I'm thinking of doing something > similar. You might even be able to multiplex the interface and > use it for both the FPGA board and the LCD7 display.
Unfortunately the LCD7 seems to use the TI chip's dedicated 16-bit LCD interface. It looks like it probably uses mux mode 0, so I suspect the GPMC is still available, but they don't seem to use the GPMC. Thanks Pete
pfraser wrote:

> I'm playing with the idea of interfacing a BeagleBone > (cheap dual ARM Cortex A8 board) to a Xilinx KC705 > Kintex development board. This will give me much more > CPU processing power than a microblaze could. > > I thought I could probably do it with a passive interface > because the Kintex can deal with 3.3 Volt I/O.
I did this with the older Beagle Board, where the I/O is 1.8 V. I set up a Spartan 3AN FPGA with one bank at 1.8 V to interface to the Beagle, and the other 3 banks at 3.3 V for outside I/O. It is working quite nicely. If you have not dealt with user level to GPIO on the ARM processors, it is a bit complex, as there are multiplexers to swap various I/O devices onto the limited package pins. Then you set up memory mapping pointers to the GPIO registers, and you have to carefully mask off the pins that are in use by other parts of the system. Jon
BobH wrote:

> I don't know the status of it, but this is listed in the "Cape Board > Registry": http://specialcomp.com/beaglebone/BeagleBone_FPGA.html
But, does it really exist? Looks like a rendered drawing (although it sure fooled ME!) and no update since November, "PCB design almost ready". Jon
rickman wrote:


> > It seems the BB has a peripheral mux on all the pins so you can wire the > FPGA to the BB in pretty much any pinout and get to any of the peripherals > (I'm not sure if there are limitations).
There are MANY limitations. Each peripheral device can be mapped to up to 4 package pads, but not "any pinout". So, many peripherals are not available because the pads you need are not wired out, or are being used by something necessary. Jon
On Monday, July 23, 2012 2:47:57 PM UTC-4, pfraser wrote:
> rickman wrote: > &gt; But I would suggest that you duplicate the pinout of the LCD7 > &gt; which seems to use a parallel interface and you might be able > &gt; to reuse some of the driver code. I&#39;m thinking of doing somethi=
ng
> &gt; similar. You might even be able to multiplex the interface and > &gt; use it for both the FPGA board and the LCD7 display. >=20 > Unfortunately the LCD7 seems to use the TI chip&#39;s dedicated > 16-bit LCD interface. It looks like it probably uses mux mode 0, > so I suspect the GPMC is still available, but they don&#39;t seem > to use the GPMC. >=20 > Thanks >=20 > Pete
Funny, I get more info on the Beagle boards here than I ever have in the Be= agle google group. The guy running the thing just says to read the documen= tation and not many others respond. I have found the BB to be a bit diffic= ult to even find info on other than the chip data sheet. =20 Rick