FPGA based cnc controller / routing pcb

[guru_florida]

I am just finishing an FPGA design. It's a basic board with sram, flash rom, sd card, sound codec, +more. It will run a version on ucLinux. It is primarily intended for something else, but I was thinking of using it later as a reference design for cnc control.

If you were to design a dream board for cnc that includes the G-code interpreter and all...would it be with an FPGA (perhaps running linux) or would a cheaper solutions such as an Atmel AVR be more suiting. Maybe a blackfin? The nice thing about the FPGA over AVR is being able to design the stepper or servo PWM in hardware. This is purely opinion but I am interested in a good brainstorming discussion on this.

Initially my cnc machine is controlled by laptop and a amplifier board i bought off ebay. But eventually I would like my machine to be more self contained. I would like to be able to send files to it over the network or usb, and be able to recall programs saved to its memory. I also want a joystick type remote that has digital rotary dials for manually moving the machine and digital read out. All possible with the board I am making.

Second, for those routing pros: In my current FPGA design I have a pcb routing conundrum(??). It's a 4 layer board and I have 3 power supplies. 3v, 2.5 and 1.2. The 3v is is the primarily used supply so it takes the power plane. 1.2v is FPGA core supply only. 2.5v is Vaux provided by a linear reg and will probably be disabled in most of my applications. Good practise tell me not to split the ground and power planes with traces. But I also want to keep my 1.2v (and my Vaux if possible) lines short. So do I route on the power plane my other supplies or route these supplies on the signal plane? I am leaning towards routing them on the bottom layer even though I have them routed on the power planes in the pics. In any case, I will put bypass caps on each 1.2v pin so this should supply the fpga core well with a low impedance source. (bypass caps are not all there in the pics.)

In my pic I dont like how the power lines split the power plan down the middle...if it happened to be at the edge of the board I wouldnt think much of it...but I am thinking of current loop distances.

I have read a really enlightening book on high speed pcb design but I guess I am still a little on the fence.

Colin

[Xerxes]

Layout looks quite pro to me. Some good design rules:
http://circuitcalculator.com/wordpre...ed-pcb-design/

Are you planning to have these machine assembled? If yes, then it is probably cheaper to use only single side SMD's despite of larger PCB area.

[guru_florida]

Thanks for the feedback. This is more of a reference design then any particular use, so I am going low quantity and self assembly. When I go larger qty I would probably be using smaller area. I also really wanted to use the BGA FPGA so that I can go up to 1M gate count if I ever needed it, but the risk of having a problem under the BGA and not knowing if it is design flaw or solder flaw didnt appeal to me. I may still go BGA later when I know the design is good.

In my design my expansion bus is really the Flash ROM signals plus a few others like system clock, more CS lines, etc. but I have all the signals routed to the expansion header then to the flash rom to eliminate making any unintentional antennas or requiring termination.

So you dont think those two thick lines splitting my power plane nearly in half will cause any problems?

C

[guru_florida]

Oh yes, also, what pcb house do you use?

Thanks,
C

[Xerxes]

Connect power layer by few traces across the split on other layers. Traces don't have to be thick because only high frequencies will pass thru the smallest inductance path. DC will go around the split on the smallest resistance path.

In the last order I used Pcbcart.

[guru_florida]

I did some rearranging last night because I couldnt sleep. I managed to get those power rails as poly pours on the signal layers. I think this looks the best upon reflection. It provides a low resistive path to the fpga, I will have decouping caps also.

Thanks,
C

[Episs]

Reading your tread it look like you want to make similar functionality cnc controller board, have a look at my tread.

building FPGA motion controller
past two months there are no new info, because my new Cyclone III in 256BGA package board is not ready yet.

What fpga processor core will you use ?? is it MicroBlaze™ ?
and did you thought about software, what features will it have and so on.. ???

I counted that your board has 60pin connector, could be enough for controller, but you never know until you start to build it and then you will see how much you really need.

[guru_florida]

I may turn this board into a cnc controller board, but I am not sure yet. I have been progressing a lot with emc2 (linux) and I now have an all solid state pentium with a 2GB CompactFlash card and Wireless Ethernet, Media Card Reader, etc running EMC2. Though it is much larger than my 3"x3" fpga board, my need for a cnc controller has somewhat diminished!

After making my first cuts and evaluating the whole cnc'ing process. I realize there is much yet to do. I want to remake my Z-axis, stiffen up my machine more, add digital read out and manual input controls. A lot of this I can do with off the shelf components and since I only want one cnc machine this is the cheaper solution for sure! (My fpga baord was intended for other projects anyway, so its not wasted.)

Also, I have been learning advanced g-code, including flow control such as loops/if's, etc, and getting just to where emc2 has me now, would require a lot of programming and effort. So I am left with asking myself "Why go there?" when I am so close now to start cutting parts and making money! ...and robots for my plan of global domination!

If I do use my fpga board I will simply wire it for usb or network connectivity to a remote emc2 computer. The board would then just house the real-time portions and I would write hal (emc hal!) modules to support it. The board would also be the IO for DRO and manual input controls. I think the EMC2 project could use such a "plug-and-play" board that controls steppers/servos, DRO and controls. The Mesa board can do it too, but there would still be some more circuitry required for it and its not so plug-and-play.

[Xerxes]

Some "hardware accelerators" exist for EMC. See:

http://www.atelierrobin.net/p41.htm
http://www.pico-systems.com/univstep.html
http://axis.unpythonic.net/01183862490