[rescue] Loading FCode via RARPD/TFTP

[Malte Dehling]

On Thu, Jun 24, 2021 at 06:47:20PM +0200, Romain Dolbeau wrote:
> Le jeu. 24 juin 2021 C  18:11, Malte Dehling <mdehling at gmail.com> a C)crit
:
> > I'm pretty happy with the results :-)
>
> As you should be! Nice results, nice write-up and nice setup.
> I never made it past the 'hack it into working for myself' stage; I
> didn't really expect anyone else to be interested...

Thanks.  I wrote a post about this on some forum and a few people were
interested so I decided to give some more details.

> > What makes it expensive?
>
> Low volume for the carrier... Once you go above 5 or 10 the per-unit
> cost drops drastically for such a small, not-too-complex board.
> Also I was never skilled at soldering and failing eyesight made it
> worse, so I had the board fully populated by the manufacturer
> (SeeedStudio) as well, adding to the costs.
> I ended up paying ~300b, for 5 PCBs with two fully populated. I didn't
> need more, and wasn't sure anything would work (never designed a
> meaningful PCB before), so I minimized the total cost rather than the
> per-unit cost.
> Of course, as everything seems fine except perhaps the micro-sd
> connector, I now regret I didn't get all 5 populated...
>
> > I've never done any FPGA work so this is probably a naive question: why
> > did you choose this particular board if it's so expensive?
>
> It's not super-expensive, far from it (150b, with VAT), but there's
> cheaper boards with bigger FPGAs of the same family. And usually with
> tons of extra features that made them unsuitable for my use case...
>
> Basically, it's the only one I found that ticked all the boxes:
> a) fitting in the SBus form factor (easy)
> b) expansion connector that can be easily used to plug on the carrier
> (harder than it sounds, high-density connector are common and I wasn't
> sure I could deal with those, 2.54mm/0.1" headers are easy)
> b) modern FPGA (trivial), I preferred Xilinx because there's more online
help
> c) enough user pins routed to the connectors (SBus has 82 signals to
> deal with, and nearly all of them are needed; you can only skip parity
> and some of the interrupts) (not-so-easy as many dev boards adds extra
> peripherals that consume pins)
> d) onboard flash for self-configuration at power-up that could be
> configured with an external power supply (not always obvious)
> e) the ability to be powered up by a 5V source on some of the
> expansion pins (hard, as the documentation doesn't always make it
> obvious; it's usually from a specific connector; and 5V isn't very
> high, e.g. the 2.13 I use is fine with it but the 2.14 requires
> 6V...).
> f) some memory I could use (easy to find but somewhat optional as DRAM
> is not a trivial thing, though my recent move to Migen/Litex makes it
> more likely I could do something with that).
>
> As a bonus, the one that I chose
> (<https://www.ztex.de/usb-fpga-2/usb-fpga-2.13.e.html>) has multiple
> variants, so I could buy the small FPGA one for testing and move to a
> bigger FPGA one if my project didn't fit. Turns out even the small one
> is largely sufficient for now.

Very interesting, thanks a lot for taking the time to explain!

Cheers,
Malte

--
Malte Dehling
<mdehling at gmail.com>