ColorCNC Colorlight 5A-75E/5A-75B as FPGA controller board

Yeah I know could sweep the millimeter probe, on a 0.1" pitch I'm okay with that but with a 50 thou pitch not so happy, that's just me & my eye sight. I need to learn some python someday I guess, so I might as well have a go. My son is pretty good at python, part of his physics course at uni, so I'll bend his ear. Plus I can do one buffer in one go.

What you are saying about pins 1,2,3,5,6,7 is a bit misleading (reading it would appear that the common 8 pins are connected to the FPGA without buffers). The A,B,C,D,E,OE,STB,CLK are also connected to buffers, but if they are similar in design to the other boards, 2 connectors share a buffer for those pins.

Considering there are 96 individual pins across all 16 connectors and 8 pins are common across the 16 connectors, giving a total of 104 pins it seems a waste just to use the board for step generation.
My plan is to use 74LVC245 for the inputs, these run on 3.3v but have 5v tolerant inputs, I've used these before on a BoB for my BeagleBone Black.
There is also the option of removing the buffers from the fpga board and bridge the pads with wires and have the buffers on a separate pcb connected via short cables.

Testing the 5A-75E V8.0 this weekend and have the board flashed with the built file, responding to pings. Linuxcnc loaded with the module successfully, however it seems like the board will stop responding to the etherbone frames after the first exchange. Acts as if it gets overloaded and wedged. At that point, board no longer will respond to pings and has to be reset in order to restore connectivity again. I've noticed romanetz has the etherbone buffer depth set to 1060 bits, however poking around i see  github.com/enjoy-digital/liteeth/blob/93...rlight_5a_75b.py#L45 has max set at 255 bits. Is there any implications in going beyond that value?


Frankly not sure what is going on and would appreciate any ideas!




 

Sorry not too sure about your issue.

I have started initial investigation into FPGA to Buffer connections for the 5A-75E v7.1. Only J1 to J16 have been done yet.

github.com/ozzyrob/chubby75/blob/master/...V7.1.md#connector-j1

Why not use a bi-directional level-shifter, for example:



In that case you don't have to use a full block for inputs or outputs. 

Also, I'm looking into the possibility to transform the code into a more configurable package based on Litex (so more boards will be supported, as long as they have Ethernet):

• The board should be configurable with a JSON-file where users can define inputs (max 32), outputs (max 32), stepgens and encoders;
• The board-file is build using the Litex recommended toolchain, which also gives the layout of the Etherbone package;
• The configuration-file should also be used by the LinuxCNC component to generate pin-definitions.

Any thoughts on this?

TOLP2 wrote:

Why not use a bi-directional level-shifter, for example:

You still have to remove the 74HCT245, where the chip was you would have bridge the pads and your circuit would have to be off board.

The 5A-75E V7.1 uses 48 inputs for the GPIOs and Encoders. So to work with that firmware it is far easier to replace a bank of 6 74HCT245 with 6 74LVC245 in a SOIC package. IMHO this seems a simpler solution.

I've used 74LVC245s with another cnc project and they worked very well.