<awygle>
apparently you can do x/y tolerances of 5 mil which isn't bad
tinyfpga has joined ##openfpga
<awygle>
qu1j0t3: i don't know anything about machining really. why don't you need "actual flat surfaces and good tolerances" for moulds?
<qu1j0t3>
you don't.
<qu1j0t3>
you specifically don't.
<awygle>
so i inferred, but why?
<qu1j0t3>
that's why i'm saying 3d printing for moulds is neat.
<qu1j0t3>
it's a good fit for the tech
<qu1j0t3>
have you seen the YT videos of people doing backyard castings like that? really fun
pie___ has quit [Ping timeout: 244 seconds]
<awygle>
my first-order intuition (clearly wrong) is that the mould would need to be _more_ precise than the final thing, because you'll never get a perfect cast so the casting process will add some slop
<mithro>
awygle: You added support for the lm4k into icestorm?
<awygle>
mithro: yes
<awygle>
it... probably works
<awygle>
i got a bitstream
<mithro>
awygle: Why?
<awygle>
but i didn't get a working board (because i screwed up my dev board)
ZombieChicken has joined ##openfpga
<awygle>
mithro: i have an application that wants a very small fpga with >32 I/O pins
<mithro>
awygle: Hrm? I mean what made you add support for the lm series rather then using an existing hx/lp or up version?
<awygle>
the lm was the only option that worked out for that
<mithro>
awygle: Ahh okay
<awygle>
also, it aesthetically offends me that we only support 3/6 ice40s
bitd has quit [Ping timeout: 265 seconds]
<mithro>
awygle: Ha, kinda does me too :-P
<awygle>
if you find somebody who wants to do Ultra or UltraLight, i will sponsor a dev board for up to 2 people for each family, and help out (as much as i can) with the actual work.
<awygle>
(cc cr1901_modern1 maybe?)
<daveshah>
yeap happy to help too
<daveshah>
the bulk of the work for both of those should be done with the UltraPlus already
<qu1j0t3>
awygle: you machine it.
<qu1j0t3>
awygle: and drill it and plate it and whatever else needs doing to it.
<mithro>
awygle: I think the issue is finding people with time rather than dev board?
bitd has joined ##openfpga
<qu1j0t3>
awygle: i'm actually more interested in 3d cnc, last weekend saw a really nice $500 (CAD) kit, designed locally
<qu1j0t3>
awygle: thinking of all the things you could use as a tool, incl cutting laser
<qu1j0t3>
awygle: can do pcbs
<awygle>
mithro: my experience has been that people with time tend to lack money, and people with money tend to lack time. i don't have any time to donate, but i do have money. thus...
<awygle>
qu1j0t3: so you machine and drill and plate the mould? i see how that cuts down the "# of things machined" metric but if you're machining it anyway i don't understand why 3d printing adds value
<awygle>
and that looks really cool, i've been thinking about pcb mills lately for various reasons
<mithro>
awygle: Same! Reason I send people dev boards and stuff for free :-)
<awygle>
oh daveshah, congratulations on trellis blinky
<daveshah>
thanks!
<daveshah>
I think we should have something at icestorm levels of stability/usability by say October/November
<awygle>
how did you P&R it? and did you add yosys support?
<daveshah>
the P&R is a new tool that needs a bit more work but will be released around the end of the month
<daveshah>
right now I'm using a really simple techmap for Yosys that just maps LUTs or single DFFs onto their own slice
<awygle>
oh awesome
<daveshah>
next step is to add proper Yosys support, and a packer for the P&R that packs LUTs and DFFs into slices
<awygle>
so i should start my next ecp5 project with the expectation that i don't need diamond? :p
<daveshah>
yeah :p
<daveshah>
beta testers will definitely be needed
* awygle
signs on the dotted line
<awygle>
sorry i never managed to like... actually _contribute_ lol. this summer has been unexpectedly busy so far and i've been dedicating all "project" time to glasgow to try to get that finished.
<kc8apf>
qu1j0t3: I'd only plan to use that mill on wood. Runout on those palm routers is terrible
<daveshah>
no worries, I would really like to see Glasgow too
<whitequark>
yeah me too
<kc8apf>
awygle: you expect the cast object to be imprecise so the mould doesn't need to be very precise. You machine, drill, and plate the cast part to make it precise where needed
<whitequark>
UP5K might even be usable with a decent pnr
<awygle>
i wonder how intense you have to get process-wise before homebrew pcb etching gives better feature sizes than mills
<awygle>
kc8apf: oh, i had that totally backwards. that seems to me like it loses you a lot of the mass production advantages of casting, but my cost model is probably totally wrong
<whitequark>
awygle: already
<whitequark>
milling PCBs is shit
<whitequark>
I mean
<whitequark>
already on something like 12 mil
<whitequark>
I tried isolation milling and it's an atrocious process
<awygle>
whitequark: for "me, as a hobbyist, trying to etch stuff, and having no idea what i'm doing" vs "me, as a hobbyist, trying to mill stuff, and having no idea what i'm doing", my intuition is that i am more likely to screw up the etching than the milling, because there are more computer controls on the milling lol
<awygle>
sounds like that intuition is wrong?
<kc8apf>
awygle: with an adjustable wrench, you only need a few precision surfaces. Mill time is expensive. Casting the rough form means fewer mill passes.
<whitequark>
awygle: you need to fix the PCB so it's flat and exactly where it should be
<whitequark>
that is not rivial
<whitequark>
trivial
<whitequark>
if you use photoresist, there are some variables that require experimentation, but not so much skill
<awygle>
kc8apf: yeah that's what i meant about cost modeling wrong. i assumed process steps had a big constant cost but a small per-part or per-minute cost.
<gruetzkopf>
getting a useable result from home-etching is quite easy
<whitequark>
you can get PCBs precoated with photoresist
<whitequark>
laser printer transparencies UV leds
<whitequark>
all you need
<whitequark>
I got down to 50 micron resolution, limited mostly by laser printer and etching itself
<gruetzkopf>
over here in germany using laser printer toner as resist is quite common
<whitequark>
if you can etch with HCl+H2O2 mixture you generally getbetter results
<whitequark>
gruetzkopf: I hate toner transfer
<awygle>
what about process control on the etch itself? etchant concentrations, exposure time, etc?
<whitequark>
shit process with unreliable results
<awygle>
not so important?
<gruetzkopf>
photoprocess is much more reliable, yet
<gruetzkopf>
yes
<whitequark>
awygle: exposure time is calibrated, takes a few hours
<whitequark>
if you're slow, anyway
<whitequark>
etchant contentrations are looked up
<whitequark>
I even tried doing plated holes
<gruetzkopf>
the calibration?
<whitequark>
yes
<whitequark>
you need to spin a few test boards and that's it
<awygle>
my concern is more "how do i know i did the concentration right". but i guess "don't be an idiot" is probably sufficient.
<whitequark>
you'll know it when you see it
<awygle>
i should really try this someday just for the experience
<awygle>
probably not in my poorly ventilated apartment though
<whitequark>
build a ghetto fume extractor
<gruetzkopf>
yeah, once you know your exposure times it's extremely repeatable
<whitequark>
I actually had some serious trouble with ferric chloride etching
<whitequark>
underetch overetch...
<awygle>
my lab currently does not border an exterior wall. i'd need to swap lab and bedroom (in which case i'd have carpet in the lab) or put it in the living room
<whitequark>
HCl+H2O2 is reportedly much better
<azonenberg>
yes
<gruetzkopf>
yes it is
<azonenberg>
I've seen some people who use HCl+H2O2 to make CuCl and then etch with that
<whitequark>
I should try that now that I have both
<azonenberg>
i dont like that, i had problems with repeatability
<whitequark>
conc HCl and 30% H2O2 is sufficient right
<azonenberg>
So i made fresh HCl : H2O2
<azonenberg>
Loool
<awygle>
yeah i was thinking about the CuCl thing. doesn't it "regenerate" somehow?
<azonenberg>
whitequark: I used RCA-2 for copper etching
<azonenberg>
as in the HCl:H2O2 mix from the RCA clean
<awygle>
it seems the most economical of materials which appeals to me despite its frequent irrelevance to total costs
<whitequark>
ha nice
<azonenberg>
except instead of using conc HCl, conc H2O2, and water
<azonenberg>
i replaced the water and conc H2O2 with 3% H2O2
<whitequark>
I have a bottle of 30% H2O2 in my fridge
<whitequark>
that shit is -nasty-
<azonenberg>
it came out to about the same amount of peroxide but no handling of 30% required
<azonenberg>
slightly less i think?
<whitequark>
I mean, I could probably etch with conc nitric too, except it'd eat the resist as well
<azonenberg>
i used one part conc HCl to 7 parts 3% H2O2
<azonenberg>
by volume
<azonenberg>
The official RCA recipe comes out to 7 parts of ~4% peroxide after you mix the 30% and water, to 1 part acid
<gruetzkopf>
the same german hobbyist forum that likes deep-fry soldering is also a fan of CuCl
<azonenberg>
so cutting it to 3% didnt hurt the etch too badly
<awygle>
helllll yeah ice40up symbol merged _finally_
<whitequark>
hell yeah
<cpresser>
commit-id or it didn't happen :P
* awygle
submits ice40hx symbol, lays in supplies for another long winter
<awygle>
i drafted the I/O cells for C, but i haven't started replacing the fpga yet
<whitequark>
nice!
<whitequark>
I'm kinda fixing solvespace in the meantime
<whitequark>
it's been without any attention for a literal year
<whitequark>
so many pull requests...
<awygle>
i continue to be impressed by your willingness to maintain literally anything
<awygle>
seems very much like a thankless job, so - thanks :p
mumptai has joined ##openfpga
<rqou>
mithro: I'm "available" now?
<rqou>
(at work though)
bitd has quit [Ping timeout: 240 seconds]
bitd has joined ##openfpga
pie___ has joined ##openfpga
pie___ has quit [Ping timeout: 240 seconds]
<felix_>
for me buying the more expensive brand photoresist-covered pcbs instead of the cheap ones made the process of etching pcbs much more bearable; the expensive ones didn't require very precisely timed exposure. i think i used sodium persulfate as etchant. with todays inexpensive pcb facorites, i don't really bother etching my own pcbs any more. oh and i tried that toner transfer stuff once; it was awful ;P
<felix_>
and i even used pages from and old german shop for electronics stuff that were said to be good for toner transfer
<azonenberg>
lol yes toner transfer is awful
<azonenberg>
i never got anywhere with it
<azonenberg>
i got ok-ish results for lithography
<azonenberg>
But oshpark is cheap enough now its not worth DIYing it anymore
<azonenberg>
not unless you need like same-day turn time
<azonenberg>
but honestly, my boards are complicated enough thats rarely an option anyway
<felix_>
for the last few pcbs i didn't order, i used some magic spray that made the toner more solid and less light-permeable, which slightly improved my results with litography
<felix_>
yeah, even 2 layer is quite a pain to make at home
<qu1j0t3>
anybody here do cnc routed pcbs?
<azonenberg>
basically, the way i see it
<azonenberg>
look at how much time it'd take to DIY the board
<azonenberg>
multiply by $150/hr
<azonenberg>
if that's more than the oshpark price, send it out
<azonenberg>
(they always win)
<qu1j0t3>
>>> $150/hr <<< yeah they would
<felix_>
yep. but the geometry usually has to be bigger than on litography process, but at least doesn't require chemicals
<azonenberg>
felix_: did you run through my list before sending it out?
<azonenberg>
Also thoughts on moving the signoff list to a github repo full of markdown files?
<azonenberg>
google doc doesnt strike me as the best platform
<felix_>
i'll run through the list before sending the board to the manufacturer
<felix_>
putting that as markdown files in a git sounds like a good idea to me
<azonenberg>
Let me do that now before i forget
<azonenberg>
it's lunchtime, i have a few mins
<felix_>
haven't finished the board yet; need to fix some footprints and route the few missing parts; got a wisdom tooth removed yesterday though, so i didn't make any progess yesterday and today
<felix_>
and i should send some bug reports for the kicad libraries...
<azonenberg>
fuuun
<azonenberg>
i'm getting my last one pulled in a month or so
<whitequark>
your last kicad bug report?
<azonenberg>
my last wisdom tooth, lol
<whitequark>
h
<whitequark>
oh
<felix_>
even though the tooth wasn't easy to remove it went surprisingly well; didn't need much painkillers
<azonenberg>
felix_: i metabolize lidocaine very quickly
<azonenberg>
my problem wasnt the removal, it was after
<azonenberg>
they had to give me so many injections during the procedure that my jaw muscles got all shredded on that side
<azonenberg>
i couldnt open my mouth more than 1/3 of the way for at least two weeks
<azonenberg>
i didn't have much pain from the actual procedure though, they prescribed me 5 mg hydrocodone and 800mg ibuprofen
<felix_>
uh oh, that sounds rather bad
<azonenberg>
i took like 2 or 3 of the ibuprofens and that was it
<azonenberg>
So for the last tooth they're going with titrated IV sedation rather than a local nerve block, since jamming that much lidocaine into your jaw isnt exactly good for you
<whitequark>
wow
<whitequark>
I just found a place that does N2O and that was mostly for anxiety and not so much actual tooth pain
<felix_>
i took 400mg ibuprofen maybe an hour after the removal and then one before sleep. maybe another one tonight and then it should be good
<azonenberg>
Propofol, midazolam, and maybe a touch of ketamine
<whitequark>
they yanked three of them real quick I felt almost nothing
<azonenberg>
Not thrilled about it, but the alternative isnt good either
<whitequark>
for that matter I found that drilling teeth is far worse than extracting
<azonenberg>
my first two teeth weren't a huge deal
<azonenberg>
The third one was a giant pain, they planned to do two in the same visit and only got to one
<azonenberg>
apparently i have good bone structure
<azonenberg>
they had to rip it to pieces to get it out, and use a lot more force than usual
<felix_>
yep, didn't enjoy that procedure, but much better than the tooth getting infected when i'm nowhere near a known-good dentist...
<azonenberg>
its funny that i only ended up taking a couple of ibuprofens after that
<azonenberg>
in general i have a pretty high pain tolerance
<azonenberg>
But i guess i have a high lidocaine tolerance too :p
<azonenberg>
working on moving it over but that will be the new home
<felix_>
ok
<awygle>
yeah my dentist was like "oh we did it this way because we thought you only had two roots but turns out you had three, sorry"
<awygle>
idk what they gave me but i don't think it was lido, as i tend to react poorly to it. but i was definitely awake and had my dentist kneeling on my chest for the duration.
<felix_>
huh, didn't the do an x-ray before the extraction? o_O
<whitequark>
wow
<awygle>
they did but apparently misread it
<felix_>
ouch
<awygle>
it turned out to be fine. like azonenberg, i only had jaw pain and only took ibuprofen.
<awygle>
the worst part was the dressing was disgusting and very uncomfortable
<azonenberg>
felix_: this surgeon is doing a CT scan of the area before the procedure
<azonenberg>
apparently i have a nerve very close to the tooth in question and they want to make sure they avoid it
<azonenberg>
it wasnt visible in the normal xray
<felix_>
oh. but better be a bit too careful than damaging a nerve
<azonenberg>
yeah
<rqou>
heh fiducials
<rqou>
I've actually never put fiducials on my boards because i never get them mass manufactured
m_w has joined ##openfpga
<mithro>
daveshah: Is the 25F the smallest ECP5?
<daveshah>
mithro: it's the smallest ECP5 *die*
<daveshah>
there are smaller ECP5 chips
<mithro>
daveshah: Ahh
<daveshah>
the 12F
<mithro>
daveshah: The ECP5 25F doesn't fit on my screen :-)
<daveshah>
mithro: in reality the CrossLink is a 6k ECP5 with a few other simplifications too
<daveshah>
that also has hard MIPI, so might be of interest to you in the future
<mithro>
daveshah: Oh, I'm actually also super interested in the CrossLink :-P
<daveshah>
mithro: It will probably be what I look at once the ECP5 is close to done
<daveshah>
it's quite an interesting part. I think it is Lattice's first attempt to crossbreed pure Lattice with iCE40; the oscillators, pad part of the IO (pullup resistor options), NVCM, I2C and possibly other hard/AMS parts are seemingly similar to the UltraPlus
<mithro>
daveshah: Interesting
<daveshah>
I'm not sure whether it uses the ECP5 40nm process or the iCE40 40nm process
<mithro>
daveshah: I tend to find hard logic things tend to be bad as they are always broken in weird ways
<daveshah>
mithro: I suspect the latter, given the identical choice of 3.3k, 6.8k, 10k
<daveshah>
mithro: yeah, the iCE40 has lots of problems with it
<mithro>
daveshah: It's not unique with Lattice, seen similar problems with Xilinx and Altera stuff
<daveshah>
mithro: And likewise in the MCU world errata are all to common
<mithro>
Yeah
<mithro>
I wish the FPGA manufactures would do the *minimum* required to get things into the FPGA (IE SERDES and such) and then release open source FPGA implementations
cr1901_modern1 has quit [Quit: Leaving.]
cr1901_modern has joined ##openfpga
<daveshah>
mithro: yeah, to be fair the ECP5 are quite nice in being free of too much crud (other than the messy DSPs)
<daveshah>
e.g. the PLLs are not complicated at all, almost too simple - no dynamic reconfig for them for example
<awygle>
daveshah: have you taken a look at the serdes yet?
<daveshah>
awygle: no, no license
<daveshah>
planning just to buy the cheapest devkit with a license voucher
<awygle>
yeah the versa's only 200$
<daveshah>
I actually have a Versa already, a donation by q3k, although no voucher, so I'll probably get the embedded vision one for balance
<q3k>
i found the voucher recently
<q3k>
at least one of them
<q3k>
want it? :D
<awygle>
oh wow that's only 200$? thought that was much more for some reason
<q3k>
i got it for $100, they had a sale for a while
<q3k>
even at $200 it's excellent bang for buck (if you're talking about the ECP5-5G)
<daveshah>
q3k: yes please
<awygle>
no i was talking about the embedded vision board
<awygle>
i have the versa-5g already for 250$ iirc
<daveshah>
lol the embedded vision is always on promotion, I always see ads for it and the promotion end date always seems to be about ceil(month+3)
<Ultrasauce>
now there's a rabbit hole to fall into when I have time again
<cr1901_modern>
https://ycnrg.org/lattice-diamond-on-ubuntu-16-04/ Very curious- these are the install instructions that worked for me, but Googling "install lattice diamond ubuntu 16.04" (without quotes) doesn't even put this on the first page
<azonenberg>
Not familiar with it
<azonenberg>
awygle:
<cr1901_modern>
second page, last result ._.
<awygle>
huh really? i ran into that on a similar search, i think just "lattice diamond ubuntu" or the like
<awygle>
i was able to make that work and also make alien work
<awygle>
azonenberg: ah well, thought i'd ask
<cr1901_modern>
awygle: did you use these exact insns/page?
<awygle>
yup
<awygle>
i remember the dig at alien at the top lol
<cr1901_modern>
Hmmm, wonder what changed in the past few months
<awygle>
you'll also need to use bash instead of dash as /bin/sh
<cr1901_modern>
I remember the crescent moon favicon
<cr1901_modern>
b/c that's important
<cr1901_modern>
it looks like the 3-up moon from SMW
<awygle>
it appears to be intended to, based on the signature
<awygle>
criminally underutilized item btw
<cr1901_modern>
Lol indeed
<cr1901_modern>
There's only 7 in the game (I just looked it up) ._. Thought there were a few more
<cr1901_modern>
(and based on what signature?)
<openfpga-bot>
[jtaghal] azonenberg pushed 2 new commits to master: https://git.io/fNtwr
<openfpga-bot>
jtaghal/master ddd1b71 Andrew Zonenberg: Initial version of LockableDevice. Not fully implemented for STM32 yet.