ideas-collection-tread for a new rio board

Hi,
i would like to design a new fpga board and am still looking for ideas.
Or if I can find someone, my knowledge of KICAD is limited...

My idea:

using KiCAD
should fit into a DIN-Rail housing (3D-Printed)
IO's preferably with indicator LED's
Optocouplers at the inputs
uln2003 at the outputs or better also isolated !?!?
1 power input up to at least 24V
FPGA power via isolated DC->DC converter
modbus onboard
expansion port (isolated ?)
onboard W5500 Ethernet
Optional SPI (with Flash-programming support)
Plugable terminals

I am not sure about the FPGA:

TangNano9K (devboard):
    + cheap
    - cannot be programmed directly via SPI (flash pins)
    - 1.8V on some pins
    - nextpnr is buggy / gowin ide is commercial and does not run directly on the RPI4

ICE40UP5K (no devboard):
    + nextpnr runs perfectly and also on the RPI4
    + programmable via SPI from an RPI4
    + cheap
    - few IO's and logic cells

ECP5 (no devboard)
    + nextpnr runs perfectly and also on the RPI4
    + programmable via SPI from an RPI4
    + price is ok
    + very many IO's and logic cells
    - BallGridArray (I'm afraid of this and I can't really solder it by hand)
    - I am afraid that you need 4 layer PCB's

I would be delighted if we could do this as a joint project

greetings
 Olli

Hello,

My 2cents as follows.
I would not worry at all about 4 layer pcbs, there is very little price difference to 2 layers and it typically just makes things easier.
The BGA could be done example as a separate pcb that has the FPGA and decoupling caps and then you would add it to the "main" pcb, either by soldering, headers etc. If the FPGA is available example in LCSC then its quite cheap to have the pcbc assembled by example jlcpcb. As for the ECP5, some of them come in TQFP package.

I would say choose the best from a cost/performance point of view and dont worry about packages or layers.

I can help on the KiCad and 3D (case) stuff if you want.

EDIT: As for the isolation, I would minimize it as it always add quite a lot of cost. After you have a working board, isolation can always be put to the version 2 of the board if needed.

You can consider Tang Primer 25K

onceloved wrote:

You can consider Tang Primer 25K

I had already thought about it, I also have a board at home, but the pin spacing of the connector is even worse than with the BGA ;(

RDA wrote:

I would say choose the best from a cost/performance point of view and dont worry about packages or layers.

I can help on the KiCad and 3D (case) stuff if you want.

EDIT: As for the isolation, I would minimize it as it always add quite a lot of cost. After you have a working board, isolation can always be put to the version 2 of the board if needed.

I think you're right, it's better not to make too many compromises,
then it will probably be the ECP5.

about the isolation, I already have all the components, so that shouldn't be a problem.

3d: i already have openscad templates for the housing

I will still need a lot of help with KiCad

Hello, so I would first define if this something for you or something for a wider audience? I mean, making the pcb from parts you already have is probably good for you, but might not be "good" for everyone else. So while it might be tempting to base your design around the parts its not necessarily a good thing for the rest of the people.

After you have decided on this, then "map" out what you actually need in terms of #I/O, type etc. and decide your major parts based on price, availability and so on.

For the outputs I would use FET's instead of BJT's, but I assume the ULN2003 is also something you already have?

As for the KiCad maybe you can use one of these designs as a "base":
Orangecrab
Butterstick
Basic ECP5
 

if i say 'already have all the components', i mean, i already used it or tested it,
and i think it is save to use it again

in the end i would like to have it so that you can have it completely manufactured by jlcpcb.
so it is completely independent of what parts I have

IMO, make it 2 boards. breakout board, and the core board

I wish we had an updated breakout board standard, like the generic db25 5 axis cnc board, but updated. then instead of having everyone remake the same thing but slightly different, we can just make core boards, and there would be a larger pool of users using the breakout board across other linuxcnc platforms , and we could all benefit from the economy of scale

meister wrote:

if i say 'already have all the components', i mean, i already used it or tested it,
and i think it is save to use it again

in the end i would like to have it so that you can have it completely manufactured by jlcpcb.
so it is completely independent of what parts I have
 

Sorry I misunderstood this completely. Could you list all the parts you would like to use for the "bare" fpga part?

cakeslob wrote:

IMO, make it 2 boards. breakout board, and the core board

I wish we had an updated breakout board standard, like the generic db25 5 axis cnc board, but updated. then instead of having everyone remake the same thing but slightly different, we can just make core boards, and there would be a larger pool of users using the breakout board across other linuxcnc platforms , and we could all benefit from the economy of scale
 

I think this could be a good idea. Can we try and put together a list of what the core board should have? What connector or style of board should it be? I assume it should have at least:
Connector to a base board, what style? SO-DIMM like CM3+ and many other SOMs? I would say something relatively high pitch to make it small.
Power stage, what kind of input voltage range or we settle for 24VDC? OVP? Wrong polarity protection?
All decoupling caps,
Flash
Clock
Some leds for power and debug
...
As the FPGA will already be in a BGA package, I would not worry about example the connector being "solderable" as the board will already need a hot air station to be assembled but most likely its easier to let a fab house assemble it.