<cr1901_modern>
I find it more than a bit weird that nobody came up with an Ethernet wing for Pipistrello. Hell, even a custom PMOD should be fine if you can live with 10BASET...
<cr1901_modern>
Not that programming for the ENC* PHYs is pleasant for any definition of the word
<mumptai>
nobody cared enough, or they didn't share it with the world
<mumptai>
digilent has a enc28xx on pmod iirc
<cr1901_modern>
so scrap the ppp stuff and require a PMOD if you want to use ARTIQ w/ Pipistrello. Problem solved?
<mumptai>
hmmm
<sb0>
that only addresses one of my concerns, and makes the pipistrello setup more difficult to assemble
<mumptai>
i'm unsure about the application, but if ppp is to annoying, there is always slip ... which is sa simple as it gets (and it is annoying for that reason too)
<cr1901_modern>
It appears in order to use Kasli you also need to buy a Eurocard back plane?
<mumptai>
ohh, neither h/w nor software problems
<mumptai>
but i remember discussions about similar situations, in the context of embedded/electronics stuff in robots, and it kinda hinges around the mindset of the user.
<cr1901_modern>
mumptai: I think smoltcp throughput is considered a s/w problem
<cr1901_modern>
Spartan 6 is too (I think everyone there uses Vivado anyway)
<mumptai>
personally, i have to keep up with ise14.7 for some more years ...
<cr1901_modern>
Same, unfortunately
<cr1901_modern>
Doesn't help that a lot of dev boards meant for consumer use are still S6 or even S3
<mumptai>
and regarding the (new) users, talk to them about their motivation and environment (this is really hard if they don't know they are future users ;))
<cr1901_modern>
(Well at least ISE support in Migen is pretty solid, so I wouldn't have any concerns of ISE support rotting if I left it)
<mumptai>
with ISE, as long as i can plug it into a VM and get the output i consider it done
<mumptai>
i still have some s3 & s6 custom boards i have to support
<mumptai>
the low-end fpga board with tqfp will probably remain s6 for some time :-/
<cr1901_modern>
I'm a little pissed that s7 doesn't have tqfp. Yes, I really do need to do a BGA board, but it annoys me all the same.
<mumptai>
same here
<mumptai>
i guess it can be doen cheaply, but it could be so much more predictable if the solder joints were visually inspectable
<cr1901_modern>
$$$$$
<cr1901_modern>
Well, $ if economies of scale applies, which costs $*50
<mumptai>
the reflow is okay, placment tricky, but the process control stuff for small runs is annoying
<cr1901_modern>
process control?
<cr1901_modern>
and use a stencil for placement?
<mumptai>
placement is basically a calibrated jig with a vacuum tips
<mumptai>
the balls are self-centering to some degree
<mumptai>
but in the end you do stuff, and then it has do be correct, and diagonstic is hard
<mumptai>
and the whole process has to be reproducible
<cr1901_modern>
Hrm, the ppl in ##openfpga don't seem to have those issues making one-offs or low volume
<cr1901_modern>
Are you doing stuff that's not on the "hobbyist" end of BGA?
<mumptai>
for non-hobbyist stuff, i usually pay people to assembly the boards, they are effecticly a black box, parts in & assembled boards out
<cr1901_modern>
So why such an elaborate setup for one-offs?
<cr1901_modern>
(Or am I misunderstanding why you're mentioning a calibrated jig?)
<mumptai>
its not that elaborate, its just slow to use
<mumptai>
its a z-axis that you use to drop it on the board under a stereo microscope
<mumptai>
and allignment marks in the copper
<mumptai>
but pcb holder, under-side heater, microscope are already there
<mumptai>
and temperature sensors are more or less required if one uses the heater
<sb0>
cr1901_modern, the backplane is optional.
<sb0>
cr1901_modern, and what is your artiq use case?
<GitHub78>
[artiq] sbourdeauducq commented on issue #685: @cjbe How was the KC705 connected to the PC when you made the dump in #731? Did it go through some network router that can fragment a packet that is too large for MTU of the outbound port? https://github.com/m-labs/artiq/issues/685#issuecomment-301310588