[J-core] Turtle Board Documentation

[Rob Landley]

On 3/8/21 4:17 PM, John Paul Adrian Glaubitz wrote:
> Hi Patrick!
> 
> On 3/8/21 11:08 PM, Patrick Oppenlander wrote:
>> I've received my Turtle Board. Thanks for sending that to the opposite
>> side of the planet!
> 
> Interesting. Are these available for purchase now or are those still prototypes?

Still prototypes. We'd very much LIKE to make them available for purchase but it
turns out that "low to mid volume retail sales" requires completely different
infrastructure from "high volume B2B" and "let's have crowdsupply handle
fulfillment" was not the complete solution we'd hoped for. (Little unaddressed
issues like "who does hardware testing", for example. Once you scale past "our
engineer tested four boards this afternoon by hand on a bench in our lab" our
existing solutions jump straight to "create a $50k dedicated/automated test
harness" which is not a cost easily amortized over an initial production run of
a couple thousand units. There ARE solutions in between, but none we've used
before in-house.) Nothing unsolvable, but fairly small roadblocks put it back on
the todo list where it gets buried by everything else we have to do...

As for the prototypes, ages ago we made a dozen of the 1v0 boards
https://github.com/j-core/jcore-soc/tree/master/targets/boards/turtle_1v0
(LX25, didn't quite fit in all PI cases) which turned out to be a bit fragile
and are all dead now (the power/serial USB port would come off after a couple
hundred plug/unplug cycles because it was HELD ON BY THE SOLDER, the new ones
have a proper clip). I have a 1v0 we cut up a USB cable and soldered the wires
to so technically it is still usable, but...

Then we did a run (I think 50?) of the 1v1 design
https://github.com/j-core/jcore-soc/tree/master/targets/boards/turtle_1v1
(LX45, lots of small fixes, more blinky lights, WAY more durable; we've only
killed one so far and it's probably fixable it's just sitting on a desk in
Asakusa because we still have spares in tokyo and canada).

We still have plenty of the 1v1 prototypes, and I grabbed a handful of those out
of the bucket on my way back from the Tokyo office to Austin in December so I'd
have them if we needed them (and then I got stuck here by the shifting covid
travel restrictions again, at this point it doesn't look like I can return to
tokyo until I can get vaccinated...).

This left me with extra boards, and rather than (eventually) take them BACK to
Japan I got permission to mail 3 of them out so far to people who've expressed
interest. No promises but I might have one more (there's pile of anti-static
bags in the suitcase with various boards in them but that includes things like
GPS hats, dunno what's left) if you'd like to try to run Debian on it?

I would very much like to get these available retail, but at this point I
suspect we need to partner with somebody who already does this to actually make
it work. (It's not my area of expertise, everybody is always busy, and the
boards we have work great for us so it's kind of solved already from our
perspective? We want to get this out into the world but it's unlikely to be a
significant revenue center any time soon, just a thing we'd like to do for the
open source community, but also a surprising amount of work even _after_ we have
enough boards for our own use...)

> Adrian

Rob

P.S. We sometimes talk about doing a 2.0 design someday with an ICE40 J1 instead
of Atmel, maybe Artix-7 and DDR3, possibly USB-C and an SDIO wifi chip... dunno
when Pi 3 form factor stops being useful and we'd need pi 4 to get cases? Except
what we REALLY need to make is a Lattice ECP5 board that can run J2 (since that
can use the open GHDL toolchain, although the result's probably like 12mhz
because ECP5 is _slow_). But again... that doesn't solve the problem that we're
not set up for retail sales. We're pretty good at designing new boards in house
and doing a run of a few dozen prototypes, we've already made multiple "hat"
boards that plug into the 1v1 Turtles for various development projects. And we
can set up high volume manufacturing to hand off to a B2B hardware partner that
can deploy a zillion boards into some supply chain. It's the "sell small
quantities retail" part in between that's... not what we do. And things like
kickstarter or "amazon fulfillment" turn out not to actually do the bits that
are missing.