<d1b2>
<esden> the cap will be really hard to replace... it has an annoying list of requirements...
<d1b2>
<timonsku> tell me, I'd rather work that out than find a compromise to make both connector and hiding cap work well 😄
<whitequark>
what's the problem with the sync connector?
feldim2425_ has quit [Remote host closed the connection]
feldim2425 has joined #glasgow
<d1b2>
<timonsku> if I increase the thickness you can't actuate the latch anymore with your fingers
<whitequark>
oh god
<whitequark>
i... misunderstood how these connectors work my whole life
Nephirus has quit [Ping timeout: 260 seconds]
* whitequark
facepalm
<d1b2>
<timonsku> 😄
<whitequark>
well that explains why everything related to the sync connector is so damn awkward all the time lmao
<d1b2>
<timonsku> hehe
<whitequark>
i just plopped it there without using my brain
<whitequark>
"how hard can it be? it's just a 2 pin molex"
Nephirus has joined #glasgow
<d1b2>
<timonsku> I mean you do get two variants, one with a bump and one without. With the bump you definitely need to be able to get your finger in there. With the flat "latch" or key rather you could wiggle it out without not too much issues
<electronic_eel>
esden: the CDSOD323-T36S has 75 pF. I'd say it should be below something like 150 pF. often the ones with large capacitance don't have the capacitance specified at all
<whitequark>
timonsku: yeah... i didn't realize there were 2 different kinds
<whitequark>
even though i even unlatched some
<whitequark>
i think we can just use the keyed connector?
<whitequark>
it's not meant to withstand force or vibration
<electronic_eel>
esden: CDSOT23-T36 might work. it is SOT-23 though, so we'd need a bit more space
<d1b2>
<timonsku> whitequark: yea not sure what esdens last state was, last time he wanted to try the latched ones again I think
<electronic_eel>
esden: the ones with large footprint (SMA, SMB, SMF) are the ones that often also have large capacitance. they are designed more for power rails than for data lines
<electronic_eel>
esden: we don't want those
<whitequark>
the one i have on revC1 is latched, on revC2 is keyed
<d1b2>
<timonsku> in the prev. rev I had the top at the same level as the top of the sync connector, that would be enough to cover up the cap without making the spot too thin
<sorear>
pure mercury. get someone with high cold tolerance to demo it at -50C, then it disappears if anyone else tries to inspect it
<d1b2>
<Attie> "high cold tolerance"...
<DX-MON>
LOL
<d1b2>
<OmniTechnoMancer> Indeed, clearly need to properly lacquer the case
<DX-MON>
bonus points with DU.. explodes on sufficiently heavy impact into a cloud of toxic dust
<DX-MON>
on a more serious note.. whitequark, do you think Glasgow can handle at least shiming and logging JTAG communication between, eg, an Atmel JTAGICE3 and an ATXMega to help me with trying to build a FOSS (probably BMP-based) interface to my keyboard so I can debug stuff?
<DX-MON>
I don't recall what JTAG applets already exist so I figure asking would be the fastest way to know
<whitequark>
iirc i looked into implementing ATmega debug via JTAG and it turned out the debug core is 100% undocumented
<whitequark>
what sort of shimming do you need?
<DX-MON>
ATMega might be, ATXMega seems.. at least a spot less
<d1b2>
<OmniTechnoMancer> Can you not just put a logic analyzer on the JTAG lines?
<DX-MON>
record all traffic in both directions, pass the lines from the JTAGICE through unchanged
<whitequark>
what OmniTechnoMancersaid
<DX-MON>
Omni: I don't have one atm, but I do have a Glasgow now.. additionally, I want to be able to try stuff out with the link once I get more of a handle on what's going on with it
<whitequark>
Glasgow has an integrated LA
<DX-MON>
ah, that works
<whitequark>
$ glasgow run analyzer
* DX-MON
plomps dunce hat on her own head
<whitequark>
:D
<whitequark>
you can also do $ glasgow repl jtag-probe
<whitequark>
to try things out in the console
<DX-MON>
pokie
<DX-MON>
*okie
<DX-MON>
perfect
<DX-MON>
I know it's a 4-bit instruction interface and I know some of the layout but that's about all I've been able to confirm so far while getting very frustrated at things like OpenOCD..
<DX-MON>
and before anyone says.. no, OpenOCD does not have AVR support.. what it has is a SIGSEGV'ing stub
<d1b2>
<OmniTechnoMancer> Helpful segfault
<whitequark>
ouch
<DX-MON>
yeah.. I'm trying to fix the craptacular state of AVR debug tooling as it's hurting my ability to figure out what is going on inside my keyboard's firmware
<DX-MON>
contributing where I can on that because nobody should have to go through this misery
<whitequark>
mood :3
<whitequark>
if you ever write a glasgow applet i'm definitely going to help out
<whitequark>
i don't have any xmega hardware, but i'll grab some and set up a testbed
<DX-MON>
I definitely want to :)
<d1b2>
<esden> @timonsku the ones I got last are not really latched... they have a friction bump, so you can push pull them without having to access the tongue itself. But they have a positive retention. Also they will be sitting closer to the pcb as far as I can tell. (probably all part of the stream tomorrow, where I will be testing the USB-C connector, including destructive testing)
<DX-MON>
JLC doing that rather nice deal over christmas means I might as well throw two JTAG headers on a PCB and taps to a header that fits port A so I can keep the wiring to my keyboard nice and clean
<d1b2>
<esden> (for context I am talking about the Sync connector issue timon brought up earlier)
<DX-MON>
(keyboard uses 1.27mm headers same as the JTAGICE)
<d1b2>
<timonsku> hm in my experience the molex kk with the firction bump are pretty hard to undo without touch the latch at all
<d1b2>
<timonsku> but I only ever tried 4 pin or more
<d1b2>
<esden> @timonsku ahh interesting, I will definitely have to test that
<d1b2>
<timonsku> oh yea has to be torx or at least hex socket
<whitequark>
i *way* prefer torx(+) or hex to philips
<d1b2>
<timonsku> the plan was no so far because its quite tight in that spot and the piece would be very thin but I did increase the opening recently to at least give it a try
<whitequark>
it would be unfortunate to not have Vio LEDs since those are partly a safety feature
<whitequark>
less relevant now with the reset button
<whitequark>
maybe it's possible to do something at an angle? utilizing internal reflection
<d1b2>
<timonsku> they are visible, just not at shallow angle if I don't put a light pipe there
<whitequark>
yeah, i understand
<whitequark>
a lightpipe would be ideal but i understand if it's mechanically unefasible
<d1b2>
<timonsku> yea it was hard to work with some features, never easy to do a case after the fact :) but I do want to try
<d1b2>
<esden> maybe they can mill the anodization or mask it to create a reflective lightpipe?....
<d1b2>
<timonsku> previously I had some other features be in the way of that
<whitequark>
excellent :D
<d1b2>
<esden> milling post processing is definitely not ideal at all...
<d1b2>
<timonsku> possible but probably not cheap if you want it to look nice
<d1b2>
<esden> right :/
<d1b2>
<timonsku> I will try a light pipe first
<d1b2>
<timonsku> I increase the "hole" from 1.5mm to 3mm which should be ok I feel, just not sure how well it would stay in place
<d1b2>
<timonsku> but I do feel it could be ok, will see. I will order a piece of the LED acrylic that worked best and try it all
<sorear>
curious how this will inform future revs
<whitequark>
back when this all started i had no expectation that there will ever be a milled case
<whitequark>
heck, even by revC1 i didn't
<whitequark>
revD should be co-developed with a case, certainly
<d1b2>
<timonsku> I did design it in a way that could also use the case with just the bottom in a meaningful way with pan head screws. So if only the top parts change you might still be able to use at least the protective bottom plate with a new rev. without needing to buy the whole case or just buy a new top for cheaper.
<whitequark>
revD will likely be larger
<whitequark>
it does not seem feasible to put 2 shrouded connectors side to side without making the PCB bigger
<whitequark>
oh ys and then do $ glasgow run selftest
<whitequark>
yeah the LEDs definitely need work
pie_ has quit [Ping timeout: 240 seconds]
<d1b2>
<Hardkrash> For the "light-pipe" in the case, fill with clear epoxy. Try various tapes to get a texture on the "lens" surface.
<d1b2>
<Hardkrash> It works quite well and is a pretty simple to do.
pie_ has joined #glasgow
<whitequark>
I'd think that is prohibitively labor intensive
<whitequark>
but maybe i'm wrong
<d1b2>
<Hardkrash> I might have shipped a lot of things with that, and the glue was not labor intensive part...
<whitequark>
interesting
<d1b2>
<OmniTechnoMancer> Pot entire Glasgow in clear epoxy :P
<d1b2>
<timonsku> that is definitely very labour intensive, not really doable within the planned price point 🙂
<d1b2>
<timonsku> thanks whitequark, I will give it a try tomorrow!
<d1b2>
<Hardkrash> @OmniTechnoMancer only practical if you get Evan and Katlin to do it.... https://evanandkatelyn.com
<d1b2>
<TiltMeSenpai> rofl
<d1b2>
<Hardkrash> lthough i'm not sure if they will do clear....
<d1b2>
<OmniTechnoMancer> I am unsure if potting the whole thing would even give you the light pipe effect
<d1b2>
<Hardkrash> OMG sed s expressions edit discord's last message!
<d1b2>
<TiltMeSenpai> s/ expressions do
<d1b2>
<TiltMeSenpai> it's not really sed
<d1b2>
<bob_twinkles> note that those don't properly mirror over the IRC bridge, so be mindful when using them in bridged channels
<d1b2>
<TiltMeSenpai> and I don't even know if it's a setting somewhere, so I'm not sure if you can disable it or not
<d1b2>
<OmniTechnoMancer> Pushing the up arrow on desktop also edits the last message in the message box
<d1b2>
<Hardkrash> yea, i had been avoiding edits, buy trying to use s /
_whitelogger has joined #glasgow
<d1b2>
<Greg> Then*
<d1b2>
<Greg> Injection molded light-pipes would be ideal. But probably not achievable with the margins/qtys.
<whitequark>
yeap
<whitequark>
i think if you wanted something like that you could do silicone castings
<whitequark>
silicone mold castings*
<whitequark>
it's a lot less labor intensive than resin windows, and a lot cheaper than injection molding
<d1b2>
<Hardkrash> @Greg we have very different experience with glued windows, shipped over 10e6.
<whitequark>
wow
<whitequark>
1 million!!
<d1b2>
<Hardkrash> 10
<whitequark>
TEN million
<whitequark>
holy shit
<whitequark>
i can't entirely conceptualize shipping 10 million of anything, much less something this finicky
electronic_eel has quit [Ping timeout: 240 seconds]
electronic_eel has joined #glasgow
<d1b2>
<esden> I guess almost any technical problem can be mass produced with the right process and dedication...
<d1b2>
<Greg> Cool! Did you have feature designed in to make this step less error prone?
<d1b2>
<esden> If not for 2020 being the S***show that it is I would have gone to China this year and figured out a bunch of logistics stuff ...
<d1b2>
<Hardkrash> Tell me about it, I REALLY wish I could have gone.
<d1b2>
<Greg> My experience was with an off the shelf case with CNC'd cutouts, The cutouts were at awkward angles which required holding the parts during curing.
<d1b2>
<Hardkrash> @Greg, I can't go into much more details, but managing the flow with gravity and a good form will go much further than any amount of masking and odd fixtures.
<d1b2>
<Greg> It was also winter while I was making these. and the resin would not cure without the assistance of an oven.
<d1b2>
<Hardkrash> Oh that's going to suck, also I would put doing anything less than 50 would be painful.
<d1b2>
<esden> I have seen a lot of the casters pay a lot of attention to resin temperature.
<d1b2>
<esden> It seems to be a very important parameter...
<d1b2>
<Hardkrash> The other option is to use a 1 part glue and give it plenty of time.
<d1b2>
<Greg> Knowing I was only making 10 units, I was not really able to experiment too much with different resin products. So I just created a process that worked. Just not very time efficient.
PyroPeter_ has joined #glasgow
PyroPeter has quit [Ping timeout: 256 seconds]
PyroPeter_ is now known as PyroPeter
Stephie has joined #glasgow
Stephanie has quit [Ping timeout: 240 seconds]
<_whitenotifier-f>
[GlasgowEmbedded/glasgow] whitequark pushed 1 commit to master [+0/-0/±1] https://git.io/JkPHS
Stormwind_mobile has quit [Remote host closed the connection]
midnight has quit [Ping timeout: 244 seconds]
midnight has joined #glasgow
pie_ has quit [Ping timeout: 240 seconds]
pie_ has joined #glasgow
oeuf has joined #glasgow
samlittlewood has quit [Read error: Connection reset by peer]
samlittlewood has joined #glasgow
ExeciN has joined #glasgow
sven- has joined #glasgow
sven has quit [Read error: Connection reset by peer]
sven- is now known as sven
<whitequark>
electronic_eel: there is a nasty problem with the idea of tying ICE LED to CDONE
<whitequark>
the CDONE pin *must* be pulled up for FPGA to finish configuration (the configuration FSM won't release IO otherwise, i think this might be used for ganged configuration?)
<whitequark>
which means it must be above Vih for VCCIO_2, which is 2 V
<whitequark>
this is not obvious from the ice40 manuals... you need to read them already knowing about this problem
<whitequark>
hm, actually, i'm wrong, we already need the resistor divider to exceed the FX2 Vih, so if that input works reliably, so will the FPGA input...
Stormwind_mobile has joined #glasgow
pie_ has quit [Ping timeout: 240 seconds]
pie_ has joined #glasgow
d1b2 has quit [Remote host closed the connection]
d1b2 has joined #glasgow
<_whitenotifier-f>
[GlasgowEmbedded/glasgow] whitequark pushed 2 commits to master [+0/-1/±10] https://git.io/Jk1dC
<_whitenotifier-f>
[GlasgowEmbedded/glasgow] whitequark 23aacd7 - firmware: move ICE LED handling to FPGA code. NFCI.
<_whitenotifier-f>
[GlasgowEmbedded/glasgow] whitequark cf871da - firmware: remove LED abstraction. NFCI.
bvernoux has joined #glasgow
ali_as has quit [Ping timeout: 264 seconds]
d1b22 has joined #glasgow
d1b2 has quit [Read error: Connection reset by peer]
<electronic_eel>
whitequark: does the fx2 have internal pullups? i don't think so. so there must be a pullup integrated in the ice40 on the CDONE, because with the current revs it works somehow
<whitequark>
there is a pullup in ice40
<whitequark>
"Weak" pullup
<electronic_eel>
and it is acitve low?
<whitequark>
uh, define "active"
<whitequark>
CDONE is pulled low when the FPGA is not configured, and it also inhibits release of I/O from tristate
<electronic_eel>
so it is pulled low when not configured, and when configured it switches to input and must then be pulled high for the config to become active?
<whitequark>
basically
<electronic_eel>
the weak pullup won't be enough for driving an led. but we could use a simple little logic buffer to do that
<electronic_eel>
or a small nfet
<whitequark>
or a pair of resistors (ugh)
<whitequark>
we can probably just wire one of the level shifters.
<electronic_eel>
yeah, whatever kind of parts we also need on the pcb for other features
<electronic_eel>
timonsku: yesterday you wrote that putting lightpipes for the vio port leds in the case is difficult. there is some space around those leds and some resistors that could easily be moved around. would it make it easier for you if the vio leds were moved a few millimeters?
bvernoux has quit [Read error: Connection reset by peer]
d1b23 has joined #glasgow
d1b2 has quit [Read error: Connection reset by peer]