<awygle>
q3k: if you can share your symbol, my intent is to clean one up and submit it upstream
<awygle>
(by "clean up" i mean "comply with the KLC even when they're dumb)
<q3k>
what part do you want?
<awygle>
i did not
<awygle>
any of the 381s
<q3k>
after applying this it worked OOTB, just that the resulting symbols are very big
<awygle>
okay i'll try with this
<q3k>
as it generates one per bank and then one for everything else
<awygle>
okay, that worked. i see what you mean, the symbols are not ideal, but they're definitely to a clean-up-able state. thanks a ton!
<q3k>
y/w
lexano has joined ##openfpga
_whitelogger has joined ##openfpga
_whitelogger has joined ##openfpga
unixb0y has quit [Ping timeout: 252 seconds]
unixb0y has joined ##openfpga
<reportingsjr>
whitequark: awygle: Are you two going to a do a run of glasgows at some point?
pie_ has quit [Remote host closed the connection]
<whitequark>
reportingsjr: yes, once we debug them
<whitequark>
after revC basically
<whitequark>
probably either a run of revC or revD
<whitequark>
might still need revD with DFM fixes even if revC works well, not sure yet
<reportingsjr>
excellent
pie_ has joined ##openfpga
<rqou>
huh friggin let's encrypt has managed to flake out the past two times it tried to automatically run
<rqou>
always seems to be an error on their side
<rqou>
hmm i suppose running at 00:00 gmt on the first of every month might trigger flakiness
<whitequark>
unsurprisingly
<sorear>
don't they rather specifically tell you not to do that
<rqou>
where?
<sorear>
I found «If you offer client software that automatically configures a periodic batch job, please make sure to run at a randomized hour and minute during the day, rather than always running at a specific time» in https://letsencrypt.org/docs/integration-guide/ , not quite what you're after
<rqou>
also kinda hard to do that with cron
<feuerrot>
rqou: sleep $RANDOM? (works at least in zsh&bash)
<cr1901_modern>
I was gonna suggest the same thing, but it seems so immoral
<awygle>
systemd has this built in </troll>
pie_ has quit [Ping timeout: 240 seconds]
<awygle>
wow, kicad's unchangeable schematic stroke font isn't even fixed width
pie_ has joined ##openfpga
_whitelogger has joined ##openfpga
azonenberg_work has quit [Ping timeout: 244 seconds]
azonenberg_work has joined ##openfpga
<azonenberg_work>
awygle: yeah that is a bit annoying
<azonenberg_work>
i think freely licensed fonts are hard to come by
<azonenberg_work>
stroke fonts that is
<awygle>
i mean it's schematic
<awygle>
they should support whatever font imo
<azonenberg_work>
In the new opengl renderer they probably will
<azonenberg_work>
There is active work towards porting the schematic tool to the new graphics backend used in the layout tool
<azonenberg_work>
i see chatter about it on the mailing list on a daily basis
<azonenberg_work>
but its far from ready for prime time
<awygle>
yeah i know
rohitksingh has joined ##openfpga
_whitelogger has joined ##openfpga
X-Scale has quit [Ping timeout: 252 seconds]
<whitequark>
azonenberg_work: poke?
<azonenberg_work>
ack
<azonenberg_work>
whitequark: ?
<cr1901_modern>
^need to resend UDP packet
tinyfpga has joined ##openfpga
* awygle
goes to get a SNACK
digshadow has quit [Remote host closed the connection]
Miyu has joined ##openfpga
* azonenberg_work
sends awygle a SNRST
Miyu has quit [Ping timeout: 252 seconds]
* awygle
FIN
<rqou>
interesting, seems like a lot of people like electroluminescense but very few "maker" type people have ever actually experimented with it
<azonenberg_work>
Cable tray (at least the one for power) is re-hung in the lab after sheetrock and paint are up on this half
<rqou>
wtf your walls aren't white?
<azonenberg_work>
There are three home-run wires in the tray (only one in this picture), but so far only one is wired back to the breaker panel
<azonenberg_work>
The storage room upstairs now is fully wired all the way, we turned on briefly to test each outlet then flipped the breaker off until we're ready for inspection
<azonenberg_work>
There's actually six different coating systems used in the lab :p
<azonenberg_work>
Sorry, seven
<azonenberg_work>
The ceiling, except the last ~2 feet at this edge of the room, is matte white like a normal ceiling
<azonenberg_work>
This edge is recessed slightly above the remainder of the ceiling because the geniuses who built the garage didn't frame it with the intent of it ever being rocked
<azonenberg_work>
So the garage framing doesn't quite line up with the house framing and this is the offset
<azonenberg_work>
We decided to paint that a different color (matte navy blue) to make it look intentional :P
<azonenberg_work>
Then the upper halves of the walls (except for the one facing the house) are a very pale high-gloss gray, almost white
<azonenberg_work>
The lower halves are a slightly darker high-gloss gray
<azonenberg_work>
The floor will be a dark gray ESD epoxy with a high-gloss finish
<azonenberg_work>
The "shelf" of the exposed foundation is high-gloss white
<azonenberg_work>
I wanted this wall to be a darker, more contrasting color to draw attention toward the Antikernel Labs logo that will eventually be hanging on it
<azonenberg_work>
We had some blue paint left over from the upstairs and decided to use it
<azonenberg_work>
The reason for the two-tone wall is similar to that of the ceiling, portions of the wall stick out due to structural necessity and rather than trying to hide it i decided to add a little bit of contrast
<azonenberg_work>
here's a view looking toward the street, looks a bit more "typical lab"
<rqou>
red outlet covers?
<rqou>
UPS-backed?
<azonenberg_work>
Eventually, yes
<azonenberg_work>
The "door knocker" shape of flex conduit is a temporary jumper bridging UPS input to output until it's installed
<azonenberg_work>
just so i have power in all of the outlets
<rqou>
never would have expected to see red outlets in a home :P
<azonenberg_work>
Each wall of the lab has at least one 20A GFCI'd and UPS-backed circuit
<azonenberg_work>
(including the center wall)
<azonenberg_work>
Most walls also have a second, non-UPS'd circuit for running power tools, soldering irons, etc
<azonenberg_work>
this particular one does not because it's just going to be server racks and storage cabinets
<azonenberg_work>
There's not really going to be any reason to run heavy equipment there because of the traffic patterns around the door, i can't fit any kind of workbench against it
<azonenberg_work>
And i actually *wanted* red receptacles
<azonenberg_work>
but those are basically impossible to find spec-grade
<rqou>
spec-grade?
<azonenberg_work>
You can get el cheapo consumer ones that fail after a few dozen mating cycles
<azonenberg_work>
Or you can find hospital grade ones that are $$$$
<rqou>
ah lol
<azonenberg_work>
But regular white spec-grade 20A receptacles plus red cover plates get the message across too
<azonenberg_work>
if you look closely the cover plates also have circuit ID numbers on them
<rqou>
waiting for you to eventually decide to install a third color for balanced 60v line-ground outlets :P
<azonenberg_work>
And "spec grade" in the electrical world is an abbreviation for "meets federal specification W-C-596" aka the GSA procurement standard for electrical fixtures
<rqou>
lol of course the government has a spec for that
<azonenberg_work>
UL listing for an outlet only requires basic safety standards, this standard also mandates durability and longevity requirements
<rqou>
inb4 old political meme of mil-spec ashtrays
<azonenberg_work>
There is actually a substantial difference
<azonenberg_work>
The contacts are physically larger and mate across the whole surface of the blade
<azonenberg_work>
The shutter assembly is a lot more "clean" feeling than the cheap tamper resistant outlets (buying high grade is extra important with TR because cheap shutters are awful)
<azonenberg_work>
They have a very firm grip on the plug, you cant easily pull it out by accident
<rqou>
wait so is this why the TR outlets at home despot basically all don't work? :P
<azonenberg_work>
Yes :p
<azonenberg_work>
They do actually sell this receptacle at the HD, individually boxed, for like $12 each
<azonenberg_work>
But if you go to Platt and tell them you want to order by the case, they'll cut it down to $8 and change
<azonenberg_work>
The $1 TR outlets are garbage
<azonenberg_work>
I actually have some of those that i bought early on in the project, when i was thinking of using them for non-lab circuits
<azonenberg_work>
I installed them in the "temporary construction use" outlet box next to the breaker panel
<azonenberg_work>
it's sooo easy to knock things out of them, they have no grip
<azonenberg_work>
plugging things into them is difficult
<azonenberg_work>
These receptacles also have a solid metal bar on the back all the way from top to bottom
<azonenberg_work>
continuously
<azonenberg_work>
rather than being super thin injection molded plastic with metal "ears" at top and bottom for mounting
<azonenberg_work>
So you have a much more solid backing to plug things into
<azonenberg_work>
althouhg some of the benefit of this is negated with the electrical boxes i used... the 2-gang boxes do wobble a little bit when you push on the "outboard" side
<TD-Linux>
what's the point of a tr receptacle in a lab? dust ingress?
<azonenberg_work>
TD-Linux: mandated by code basically anywhere these days
<TD-Linux>
ah, didn't realize that. tfw still in apartment without grounded outlets
<azonenberg_work>
the only 120V receptacles that don't have to be TR are ones that are >5 ft above floor level, dedicated to a specific appliance and physically behind it so you cant reach it wihtout moving the e.g .refrigerator
<azonenberg_work>
or one or two other weird niche cases
<azonenberg_work>
i.e. "physically impossible for a toddler to access"
<azonenberg_work>
I think it's silly, if a toddler was running around my lab unsupervised
<azonenberg_work>
the uncovered receptacles are about the least likely way they'd kill themself
<azonenberg_work>
Maybe the cabinet with a big bottle of HCl? Or the toolbox full of sharp objects? :p
<azonenberg_work>
I'd rather simply say "this is not a safe place for a kid", trying to make it one would render it useless for the intended purpose
<azonenberg_work>
But the NEC doesn't distinguish between an extra bedroom and a lab
<azonenberg_work>
TD-Linux: you also need AFCIs on basically all circuits, and GFCIs on a lot more places
<TD-Linux>
I don't either, my lab is right next to my bed
<azonenberg_work>
Pure GFCI is almost not a thing anymore
<azonenberg_work>
it's either AFCI, or AFCI+GFCI
<TD-Linux>
I was super impressed, my grandparents are in a new apartment building and when a self-healing christmas light burned out, the afci tripped
<azonenberg_work>
Reminds me i need to fire up my ham radio at some point
<TD-Linux>
the antifuse arcing was enough to be detected
<azonenberg_work>
See if eaton solved the nuisance tripping problems
<azonenberg_work>
(I think my VHF radio is far enough away from arc frequencies that i'll be OK but apparently they've had issues in the past)
<rqou>
i was just about to ask about that lol
<azonenberg_work>
TD-Linux: anyway, i am going all out with the cable plant here
<azonenberg_work>
All metal-clad wiring, no romex anywhere except in legacy portions of rooms i haven't yet renovated
<azonenberg_work>
all 20A circuits except when bridging to legacy 14-gauge wiring
<azonenberg_work>
Each room has a minimum of one dedicated 20A circuit for wall receptacles, each floor has one circuit for LED lighting
<azonenberg_work>
the lab and office have one circuit per wall since i expect to have lots of heavy stuff in them
<TD-Linux>
nice. if I owned a house it would also be hard to resist the temptation to go all out
<azonenberg_work>
I wasnt originally planning to go quite this nuts but the previous owners of the place smoked
<TD-Linux>
is there mcmaster but for home depot?
<azonenberg_work>
So all the old walls had to come down
<azonenberg_work>
the old cable plant was awful, the breaker panel had its UL listing *revoked ex post facto*
<azonenberg_work>
rqou: yes it was a federal pacific
<azonenberg_work>
some rooms had no working lights even when we changed bulbs
<azonenberg_work>
the basement had flooded because the sump pump had no power to that circuit
<TD-Linux>
that's why you get ETL certificated kit, no chance of that happening :^)
<azonenberg_work>
The roof was leaking, one room had no grounds in the ouelts
<azonenberg_work>
Insulation wasnt up to modern standards either
<azonenberg_work>
So we gutted it to bare studs, then removed the studs in one room where the framing was atrocious too
<azonenberg_work>
and are building it back up
<azonenberg_work>
In-wall cable plant and insulation are done, sheetrock is up in a couple of rooms but we're far from done
<azonenberg_work>
$wife's dad is coming down here to visit soon and depending on how much work he gets done we might accelerate our timetable
<azonenberg_work>
otherwise, interpolating from our progress the past month, i estimate all walls up circa mid december
<azonenberg_work>
Part of that depends on how much we want to finish
<azonenberg_work>
It's easier to do finish electrical after painting, rather than trying to mask off light fixtures so you dont get paint on the globes etc
<azonenberg_work>
But taping, mudding, and sanding is a ton of work
<azonenberg_work>
And if i really wanted to i could hang all the rock and do the minimum finish work to pass fire code pretty quickly
<azonenberg_work>
But then i'd have to remove half of the electrical fixtures to paint
<azonenberg_work>
So i'm still undecided as to whether i want to do that
<azonenberg_work>
Especially since we just found an apartment probably a 2 minute walk from the house that costs half as much as the hotel we're living in now, and has a month to month lease
<azonenberg_work>
We're planning on using it as a bedroom and maybe place to cook and otherwise spend all our time at the new house, but this means that getting it finished NOW is slightly less urgent
<azonenberg_work>
And it may be worth taking our time to get it a little more complete before actually taking up residence
<gruetzkopf>
the "red/green/blue/yellow/orange" socket thing is easier in europe, as outlets are usually 3-part assemblies (rarely two)
<TAL>
but is there a VDE for it when to use what color? I doubt it
<TAL>
also the color choice is kinda bad for people with a deficiency in color vision; there should be other ways to differentiate between different circuits
<gruetzkopf>
it's usually also written on the sockets
wpwrak has quit [Read error: Connection reset by peer]
wpwrak has joined ##openfpga
<felix_>
rqou: that intel ballout you linked does make sense; the two outer rows can have a smaller pitch, since they can be routed on the top layer; the inner rows have a bigger pitch so that you can put vias between the balls. so you get a smaller package that can still be routed without needing fancy hdi stuff
<felix_>
whitequark: i have an xps 13 9360 and a tb16 dock, but no device to sniff the communication on the cc lines
<felix_>
yeah, professional grad digital video stuff sadly is quite a trashfire :/ from the stuff with real-world usage displayport is by far the most sane protocol
<whitequark>
felix_: that would be very interesting to sniff to me
GuzTech has joined ##openfpga
rohitksingh has quit [Quit: Leaving.]
<felix_>
what do i need to sniff the cc lines? if it's less than 50 euros and takes me less than 2-3 hours, i'd try that. from tuesday on i'll be on conferences for a 9 days, so not at my officce
<whitequark>
felix_: can you disassemble the dock?
<whitequark>
(or the laptop, though I imagine the dock would be easier)
<felix_>
i can try to disassemble the dock. i\m not too happy about trying to solder things to the dock, but if it's easy enough, i'd do that
<whitequark>
felix_: it's exactly one wire
<whitequark>
unfortunately all other methods of sniffing it are more invasive
<whitequark>
i wasn't happy about disassembling my laptop and soldering to the flash to dump the pd firmware either...
<felix_>
ok, that doesn't sound too bad
<felix_>
the soldering one wire to the dock's pcb
<whitequark>
do you have a scope? la? which?
<whitequark>
you'll need to poke it a bit with the scope to see which cc line has communications
<whitequark>
there aren't any pictures of tb16 so i can't direct you to any specific connector pin yet
<felix_>
i have access to some rigol scope and different logic analyzers (mostly a bunch of the cheap fx2 boards)
<whitequark>
there's a few different footprints of usbc connectors
<whitequark>
ok, rigol works, the comms are slow
<whitequark>
you'll need an LA that has Vih of 0.7V
<whitequark>
or a comparator/opamp and a few resistors
<whitequark>
USB PD communications use Vh=1.15V
<felix_>
i can probably bodge that together
<whitequark>
sweet
<whitequark>
that's all, i just need a capture of what happens exactly when you plug it in
<whitequark>
on the cc pin
<whitequark>
then i can reproduce that and reflash my adapter to do the same thing and hopefully it'll do something useful
<felix_>
sounds nice way to procrastinate working on the slides for my talk at the osfc... ;P /o\
<whitequark>
i'd be really happy if you can do that for me.
<felix_>
at least i got 50 of these boards assembled yeasterday and the day before https://imgur.com/a/D5m9vvi
<whitequark>
nice
<felix_>
*yesterday
<felix_>
those are for easier automatic testing of spi flash firmware images on real hardware
<awygle>
real castellation or vias sawed in half? Lol
<awygle>
that's really cool
<felix_>
i checked the box "castellated holes" at pcbway; 350 of those small board adapters cost around 74$
rohitksingh has joined ##openfpga
<felix_>
the yield rate of those boards wasn't that good; they needed to make another batch to have enough good boards...
<awygle>
ah, bummer
<felix_>
yeah, i was quite worried that the boards won't arrive in time or that there was still some bug in the design and i won't have time for another revision
_whitelogger has joined ##openfpga
<azonenberg_work>
TAL: pretty sure that red vs white is possible to distinguish even for folks who are color blind, just in terms of overall reflectance
<azonenberg_work>
But i also have circuit IDs on all the cover plates in lab/office spaces
<azonenberg_work>
$wife wont let me do it in the rest of the house, sadly... :(
<whitequark>
yeah color blindness isn't absolute
<TAL>
yep, red vs white; but there is also green, orange and sometimes even yellow in the mix ;)
<whitequark>
yo uhave red-green
<qu1j0t3>
azonenberg_work: Like i said, use uv sensitive marker :)
rohitksingh has quit [Quit: Leaving.]
<TAL>
usually not a big deal bc people might even see different shades but i had this case twice that someone asked me which outlet he should use
<qu1j0t3>
or rfid
<sorear>
most “color blind” people are dichromats or anomalous trichromats
rohitksingh has joined ##openfpga
<awygle>
I have a hard time with blue and green but I've never been sure if that "counts"
<azonenberg_work>
TAL: yeah i'm just thinking, in the case of "red cover/white receptacle" vs "white cover/white receptacle"
<azonenberg_work>
even a true achromat should be able to tell one is a slightly darker gray than the other
<azonenberg_work>
unless they are unable to perceive green light whatsoever. vs just being unable to distinguish it from other wavelengths
<TAL>
yup, but the colors in europe are not really standarized
<azonenberg_work>
That and, this is my own lab and i have normal color vision
<azonenberg_work>
And i'm not expcting anyone else to be plugging stuff into it
<sorear>
Well, you can’t easily perceive absolute reflectance in inconsistent lighting
<azonenberg_work>
no but you can notice receptacles and plate are different
<azonenberg_work>
in a well-lit laboratory space
<TAL>
ya they're usually able to do that, but it can be pretty hard with a floor outlet box near a desk
digshadow has joined ##openfpga
rohitksingh has quit [Quit: Leaving.]
<azonenberg_work>
TAL: well the other thing that helps is my convention
<azonenberg_work>
in this lab, all receptacles are either 1ft above floor level (UPS backed) or 4ft (non UPS backed)
azonenberg_work has quit [Ping timeout: 252 seconds]
GenTooMan has joined ##openfpga
azonenberg_work has joined ##openfpga
<reportingsjr>
whitequark: I did the same thing you did a few days ago.. Just spent about an hour trying to figure out what was wrong with pulseaudio. Finally got frustrated and unplugged my speakers and plugged them back in. It was a bad connection and now everything is happy.
<whitequark>
lol
<whitequark>
rqou: poke?
<felix_>
displayport alternate mode of the usb type c on the xps13 works; will disassemble the tb16 later
<whitequark>
felix_: right, displayport is kind of less interesting than thunderbolt
<whitequark>
i have that working
<sorear>
we still want dp>1.2 re tho
prpplague has joined ##openfpga
<rqou>
whitequark: currently participating in a social function lol
<prpplague>
call social($me)
<gruetzkopf>
DP alt is also well understood, found some ST document showing the bitfield
pie__ has joined ##openfpga
pie_ has quit [Ping timeout: 252 seconds]
<awygle>
azonenberg_work: do you have a LATENTBLUE booked yet? :p
<awygle>
goddammit, there are plenty of available TLAs
<awygle>
why not Optical Monitoring Diagnostics
<azonenberg_work>
Lol
<prpplague>
azonenberg_work: i was also surprised how much difference between a 20% 100nF and a 5% 100nF made in reliability
<awygle>
like, this isn't hard
<azonenberg_work>
in addition to the ID EEPROM, SFPs can (at least at 1G, they're not required to) have an i2c interface to query parameters of the interface
<prpplague>
azonenberg_work: network stuff isn't my area
<azonenberg_work>
prpplague: you... didn't use Y5V, did you?
<azonenberg_work>
For 10G stuff honestly i'd use NP0
<prpplague>
azonenberg_work: hehe no
<awygle>
NP0, X7R, X5R in a pinch
<azonenberg_work>
I normally use X*R for general purpose decoupling but for any signal coupling >1 GHz i'd use NP0
<prpplague>
yea X7R is my go to
<azonenberg_work>
I also generally spec Samsung MLCCs
<azonenberg_work>
because they have very nice datasheets
<gruetzkopf>
i do have a working-ish cat5-o-nine-tails
<gruetzkopf>
using a "normal" realtek 8port switch with an extra phy on its management port
<azonenberg_work>
lol nice
<gruetzkopf>
the magnetics are the biggest problem
<azonenberg_work>
That's why i wanted to go capacitive coupling
<azonenberg_work>
you would probably have to force 100base-T mode as 10baseT doesn't cap couple too well
<azonenberg_work>
There's a TI appnote on this
<gruetzkopf>
no problem on that part
<awygle>
everybody has an app note on that
<prpplague>
i have an app note on how to use app notes
digshadow has quit [Quit: Leaving.]
<awygle>
my favorites are from microsemi iirc? microchip? one of the micros
digshadow has joined ##openfpga
<gruetzkopf>
(pinning it)
<sorear>
luckily for you they’re the same company now
<whitequark>
wtf
<whitequark>
stm32 uses parity bits in the bootloader
<whitequark>
this is the first time i ever see those anywhere
<rqou>
it does?
<whitequark>
yes
<whitequark>
and my UART misreads what it sends, naturally
<whitequark>
because it thinks that parity 0 is a start bit
<awygle>
lol
<awygle>
idk why more uarts don't use parity
<whitequark>
christ i thought i could get away without a parity
<whitequark>
but NO
<rqou>
oh the _uart_ bootloader
<rqou>
not the usb dfu bootloader
<rqou>
never used the uart
<whitequark>
yeah the maple usb dfu bootloader is so fucked i can't get it to work
<whitequark>
so i'm gonna flash a black magic bootloader into it
<whitequark>
so i'm adding parity to glasow uart
<sorear>
can’t you like. crc8. parity is a terrible solution in 2k18
<whitequark>
so i can flash a black magic into this maple mini
<whitequark>
so i can make it talk swd to the thunderbolt adapter
<whitequark>
so i can fix swd in glasgow
<whitequark>
so azonenberg can make jtaghal work with glasgow
<whitequark>
so i can debug the thunderbolt adapter
<whitequark>
so i can get it to work under linux
<whitequark>
so i can use an eGPU
<whitequark>
my fucking life
<azonenberg_work>
lol
_whitelogger has joined ##openfpga
<jn__>
that's a deep yak stack
<whitequark>
jn__: this is an average yak stack for me
GuzTech has quit [Ping timeout: 245 seconds]
<whitequark>
i should tell you about the thing where i want a/c to work next summer so i'm getting a three-phase driver for a frequency inverter now
<rqou>
wait what
<whitequark>
i bought a 8 kW multihead outdoor unit and two commercial Carrier fancoils...
<uovo>
Ꙩ_ꙩ
uovo is now known as egg|egg
<sorear>
Ok but do you want the eGPU for its own sake or
<whitequark>
gonna make a chiller from the outdoor unit by gutting it and replacing the brains with a three phase inverter and an expansion valve controller
<felix_>
hmm, the plug of the tb16 is quite big; i wonder if the power delivery device controller is in the base unit or in that plug
<whitequark>
since i'm too cheap to get a new inverter i bought a dead inverter off craigslist
<whitequark>
felix_: it is definitely inside tb16
<felix_>
ok
<rqou>
have you considered just buying a normal window ac unit or something?
<whitequark>
it will be a TI TPS65983 next to the USB C connector
<whitequark>
rqou: no
<whitequark>
first, RU doesn't really have window AC units
<whitequark>
it's all split systems
<whitequark>
i haven't ever seen a window AC unit actually before HK
<whitequark>
well ok i've seen some in the US
<whitequark>
rqou: second, those are weak shit
<whitequark>
this apt needs 5-6 kW to be cooled in entirety during the hottest part of summer
<whitequark>
i added some safety margin
<whitequark>
it's also a 9.5 kW heater
<rqou>
the HK solution is to have multiple window units :P
<whitequark>
yeah fuck window units
<whitequark>
they vibrate
<whitequark>
i have sensory processing issues that make sounds like the ones that window units make hell for me
<whitequark>
the carrier fancoils, on the other hand, have a rotor that was manually balanced on the factory
<whitequark>
and they sound really nice
<whitequark>
in fact i'm running them right now just for the sound
<whitequark>
it helps a lot
<gruetzkopf>
i have ~22kW of cooling installed at $location
<whitequark>
nice
<gruetzkopf>
all split
<whitequark>
i went for 8 kW because that's the largest unit that's still single phase
<whitequark>
and this apt doesn't have 3ph
<gruetzkopf>
yep
<whitequark>
in fact I doubt the wirnig in it can even tolerate 10 A of 1ph
<whitequark>
I had to redo all electrical wiring in it in preparation
<whitequark>
and the wiring that goes into it from the communal space is uhhhhhh
<rqou>
not going to use those hilarious-looking rotary phase converters to get 3 phase? :P
<gruetzkopf>
i have two 7kW units
<whitequark>
well it's an aluminium wire (return appears to be via some sort of building framing structure)
<whitequark>
and it's twisted together with another aluminium wire
<whitequark>
then it's fixed to a wooden board.
<rqou>
and then mix in some copper, right? :P
<whitequark>
when i opened the similar twists in the conduit boxes inside the apartment the insulation has all charred
<whitequark>
i thought it was just old first, no, it's literally charred black and falls apart
<whitequark>
rqou: yeah but that was inside the apt
<whitequark>
not outside
<rqou>
wait seriously?
<whitequark>
yes
<rqou>
goddammit
<rqou>
that happens outside the US too?
<gruetzkopf>
only thing we paid money for was some copper tubing
<whitequark>
rqou: also fuck rotary converters
<whitequark>
the 2.2 kW 3ph inverter I got can fit in my hand
<whitequark>
and it's barely a kilo
<whitequark>
it had dead FETs and FET drivers tho and when I replaced those it made a nice explodey
<rqou>
yeah solid state inverters are nice
<zkms>
as opposed to what
<zkms>
mercury arc inverters?
<whitequark>
zkms: rotary converters
<rqou>
which is basically an optimized version of "what if we hooked a motor back into a generator?"
<whitequark>
optimized?
<whitequark>
the ones i've seen in ru are literally just a motor and a generator
<implr>
huh, interesting, I guess we need to look around - we could use some for the hackerspace
<implr>
the coolant is most painful because of regs
<rqou>
do these also predate solid-state power electronics?
<rqou>
implr: just ask whitequark to get you some from russia :P
<implr>
the eu won't let you buy any unless you're certified
<zkms>
rqou: yes.
<whitequark>
you can use butane for refrigerant
<implr>
rqou: illegally imported r143a from the ukraine is very much a thing, yes
<whitequark>
(not coolant)
<implr>
but we'd rather not do that
<whitequark>
implr: or do that
<rqou>
lool
<whitequark>
why not
<implr>
yeah, refrigerant, derp
<rqou>
wait aren't you in PL?
<whitequark>
fuck the system etc
<implr>
rqou: yes
<whitequark>
RU still has R22 sold
<rqou>
does the government actually care over there? :P
<rqou>
as opposed to "meh, eu said to ban this"
<whitequark>
even though it's in theory signed the relevant bans
<whitequark>
and R141b
<whitequark>
in fact I had to specifically search for an R404 unit
<whitequark>
which was a pain
<whitequark>
sorry, R410a
<implr>
rqou: kiiinda, yeah
<zkms>
rqou: as a saving grace they usually run at 400Hz and not 60/50Hz
<implr>
whitequark: we did consider butane, but that's a bit too explosive
<whitequark>
have you done the LEL/HEL calculations?
<whitequark>
I was concerned about R600 fridges at one point but not anymore
<whitequark>
though, it is true that a residential split system uses more refrigerant than a small fridge
<rqou>
why are there so many random combinations of fluorocarbons used as refrigerants?
<whitequark>
rqou: different boiling points and working pressures
<whitequark>
then you have things where uh
<whitequark>
R134a is a replacement for R22 with an oil change
<whitequark>
HFO1234yf is a drop-in replacement for R134a
<whitequark>
sorry, HFO1234ze, not yf
<whitequark>
HFO1234yf is close but not an exact replacement for R134a
<whitequark>
HFO1234ze is excellent actually, it has a zero ODP and less GWP than CO2
<whitequark>
so it actually even rules over R744
<felix_>
whitequark: https://imgur.com/a/c0yHr6e taking that thing apart wasn't the best idea though; to reassemble that, i'll need to buy new thermal pads, since the one on the main chip broke into pieces :/
<whitequark>
not to mention far nicer working pressures
<whitequark>
felix_: damn, sorry :/
<rqou>
wait what is this device?
<rqou>
interesting connector
<whitequark>
felix_: ok what the hell
<whitequark>
that is a thunderbolt controller
<whitequark>
so it terminates thunderbolt *inside the connector*?!
<felix_>
i stared taking apart the dell tb16 device. the last two photos are the inside of the big connector from the first photo
<felix_>
inside the connector that is usually inside the tb16 dock
<whitequark>
felix_: hmmmm weird
<whitequark>
felix_: ok can you please do two things
<whitequark>
i see how it's laid out now
<whitequark>
the uC with the green dot is an SPI flash 25C series
<whitequark>
please hook up an LA to pins 1, 2, 4 (ground), 5, 6 and capture a trace of it booting after you plug it in
<felix_>
U5 is a winbond spi flash
<felix_>
or just desoldering it and using a raspberry with flashrom?
<whitequark>
I mean if you want to bother with that, sure
<whitequark>
but I'll be fine with an LA trace as well
<whitequark>
whatever is easier for you
<rqou>
wtf that big inductor is the exact same one used on the sonnet cards
<whitequark>
the flash is probably 1 MB in size
<whitequark>
rqou: intel reference design.
<rqou>
but this isn't the same board layout so they did change something
<whitequark>
felix_: now. one of those 6 wires that aren't under any shield has the CC pin
<felix_>
i have a raspberry with soic8 clip and falshrom installed in the drawer with my coreboot stuff, so just pulling the chip might be easier
<whitequark>
felix_: one of those wires should have long-ish (dozens of bits) bursts of manchester encoded data at 1.15V
<whitequark>
when you plug it in
<whitequark>
that should be easy to recognize with a scope
<felix_>
i don't know if it's a bad idea to just connect the soic8 clip to the flash without desolddering
<rqou>
wait are you sure CC runs to the TB controller?
<whitequark>
felix_: it is fine to do that
<whitequark>
the TB and PD controllers tristate SPI bus after boot
<felix_>
above u5 is some chip in either csp package or directly wirebinded to the board. i could try to find out the mubers written on it
<whitequark>
rqou: it could in theory be contained entirely within the plug on the other end, but then there will be an UART running to the TB controller
<felix_>
ah, then it might
<whitequark>
because they negotiate via UART
<whitequark>
that seems unlikely to me
<rqou>
so where's the TI chip? hidden in the plug?
<whitequark>
not a clue
<whitequark>
it might even be inside the dock for all I know
<whitequark>
rqou: btw intel reference designs are schematic
<whitequark>
layout is up to customer
<whitequark>
rqou: I think the TI chip is in a COB package here for some reason
<rqou>
do they have an example layout?
<whitequark>
see those top five testpoints running to the COB package?
<whitequark>
that's probably SWD
<rqou>
yeah
<whitequark>
or maybe its UART
<rqou>
sonnet has a full 10-pin header though
<whitequark>
there's a few options
<whitequark>
the bottom 5 probably run to the flash
<whitequark>
they need to preseed it at the factory
<rqou>
preseed what? i thought that's what the green dot is for
<whitequark>
firmware
<whitequark>
dunno what the green dot is
<rqou>
usually dots indicate "preprogrammed"
<whitequark>
hm ok
<rqou>
sonnet cards clearly have swd for the ti parts
<rqou>
unsure about alpine ridge jtag
<rqou>
might be on the edge connector
<whitequark>
rqou: the ti part can expose SWD on SBU pins in USB
<whitequark>
so you don't strictly need SWD on board
<rqou>
what
<rqou>
everything so overcomplicated
<felix_>
seems to be a cypress part and not a ti one
<whitequark>
it has a huuuuge digital/analog crossbar
<whitequark>
felix_: ohhhh
<whitequark>
yeah cypress makes PD controllers too
<whitequark>
EZ-PD lol
<felix_>
cypd1105-3 maybe
<felix_>
yeah, that seems to be plausible
<rqou>
wait, is that a different reference design then? :P
<whitequark>
rqou: intel has a number of designs you can choose from
<whitequark>
with PD controllers from different vendors
<rqou>
how the heck are the other vendors cooperating with this ecosystem?
<rqou>
just because apple?
<whitequark>
because intel
<rqou>
how does anybody debug this?
<whitequark>
they suffer.
<whitequark>
well
<whitequark>
lecroy sells an incredibly expensive (call for quote) pd debugger
<whitequark>
they all use it in the docs
<felix_>
flashrom says "No EEPROM/flash device found.". so it's hot air time i guess
<whitequark>
hm that's weird
<felix_>
it's a W25Q80DV
<whitequark>
yep, intel reference design
<whitequark>
it's gonna have like 100 kB used out of all that