<rqou>
azonenberg: um, not to be rude, but there doesn't seem to actually be that many parts here?
<rqou>
er, looking more closely it doesn't have a lot of the parts that i would use
<rqou>
e.g. weird SO* in all the silly pitches, MOSFET footprints
<rqou>
azonenberg: do you not make "random grab bag of sensors" embedded stuff very often?
<awygle>
I don't understand why footprint libraries aren't 99% "auto-generate from IPC spec under <constraints to match company style guide>"
<rqou>
among other things, the IPC spec isn't free
<awygle>
ew really?
<awygle>
i guess i shouldn't be surprised
<rqou>
azonenberg: what kicad version do you use? latest stable, nightly, something else?
<azonenberg>
rqou: No, i do not :p
<azonenberg>
let me push my latest
<azonenberg>
sec
<azonenberg>
rqou: the majority of my designs are fast digital maybe with some RF analog
<azonenberg>
And i generally go leadless as much as i can
<azonenberg>
so you'll see lots of BGA and WLSCP and QFN
<azonenberg>
but i avoid TQFP and SO* unless i absolutely need a weird niche part i can't get any other awy
<azonenberg>
Every footprint i've used on a board in the last couple years is there
<rqou>
would you be willing to accept them if i added some?
<rqou>
i'll make them fit your style of course
<azonenberg>
I won't accept PRs for anything that i haven't personally board proven
<rqou>
uh...
<rqou>
hrm
<awygle>
azonenberg hates QFPs
<azonenberg>
This is my personal library and if something screws up, it could cost me $1K or more
<rqou>
awygle: btw ^ is the reason everybody has their own libs :P
<awygle>
yeah i mean if you have a 1K part, sure :P
<azonenberg>
I'm talking about the PCB
<azonenberg>
before components
<awygle>
i never build anything without doing a pcb-pooling prototype of some kind
<rqou>
btw i got bit by kicad's default 0402 footprint being really screwy and undersized, so it's not unwarranted
<azonenberg>
awygle: when there's batch fab that has 8 layers with ViP available, let me kno
<azonenberg>
know*
<azonenberg>
The kinds of designs i'm talkign about have half a dozen BGAs and one of them is a large xilinx part
<rqou>
azonenberg: if i board-prove a footprint would you be willing to add it?
<rqou>
or only if you personally tested it and did characterization?
<azonenberg>
rqou: maaaaybe but honestly, the amount of verification i'd do before merging a footprint would be almost as much as if i had made it myself
<rqou>
hrm, so maybe i'll just make my personal lib overlay yours
<azonenberg>
yeah a fork is probably better
<azonenberg>
i check the part against the datasheet a minimum of once as i create the footprint
<azonenberg>
once again after everything is set up and i think it's good
<azonenberg>
and again during the signoff review prior to tapeout
<azonenberg>
When you're a hobbyist planning designs with $300-$1K FPGAs, there is no respin
<azonenberg>
The first one has to work
<rqou>
sure
<azonenberg>
so the level of verification i do on these designs is pretty extreme
<rqou>
but i don't usually use parts like that
<azonenberg>
It also means that i almost never tape out a board with a show-stopper bug
<rqou>
i was thinking things like "hurr durr, this is an sd card slot"
<rqou>
which can be completely dead-bugged and still worjk
<azonenberg>
i literally cannot remember the last time i made a board that i had to throw away due to an un-reworkable issue
<azonenberg>
this is both due to heavy verification keeping critical bugs from making it to tapeout, and me being pretty decent at rework :p
<azonenberg>
I mean i've had boards i did several revs of, but it was either to improve performance in some way on the original
<azonenberg>
or because it was a planned iterative design where i did small boards to prototype parts of the AFE then another board to integrate
<rqou>
i don't think i've made a totally unsalvagable board either
<rqou>
but a board i did years and years ago with swapped S-D on a power mosfet was pretty close
<awygle>
I can't remember the last board I had to rework in a non-trivial way (more than "change resistors or caps")
<awygle>
I need to do more complicated boards, clearly :-P
<azonenberg>
Lol
<rqou>
azonenberg: btw are you using kicad nightly or last stable?
<azonenberg>
I'm using a nightly from a few weeks ago right now, it had some bug fixes i needed past stable
<azonenberg>
And this is why we have design review checklists
<rqou>
still works (most of the time at least)
<azonenberg>
Consider EMI
<azonenberg>
in close proximity to an RF front end
<rqou>
also ime it'll start failing signal integrity if the wires get a tiny bit longer
<rqou>
what i have in that photo worksforme(tm)
<azonenberg>
The list went on and on
<rqou>
ah lol
<rqou>
babby's first production design out of school? :P
<azonenberg>
tl;dr company took a PCB design firm they had used in the past for low speed embedded control projects
<azonenberg>
and gave them a microwave RF project
<rqou>
lolol
<azonenberg>
it went about as well as you'd expect
<rqou>
i'd be amazed if it worked
<azonenberg>
Oh, it worked when i was done with it :p
<rqou>
really?
<azonenberg>
I charged them $100/hr for a pretty significant amount of debugging and rework
<rqou>
so they didn't mess up the RF parts too badly?
<azonenberg>
when i was done parts of the design were barely recognizable
<azonenberg>
But i salvaged it
<rqou>
did you fab new distributed element filters?
<azonenberg>
i'm not quite that 1337, this was a direct digitization design that had a pretty simple AFE
<azonenberg>
but there was at least one bug in every major subsystem requiring rework
<rqou>
ah, so it didn't have a pile of rf voodoo
<azonenberg>
oh and they had a PLL chip that required an external loop filter to work
<rqou>
wait, direct digitization?
<azonenberg>
they had all the passives for that DNP'd :D
<rqou>
doesn't that mean fast adc and very $$$?
<azonenberg>
they were on the board
<azonenberg>
Just not loaded
<rqou>
at least that's an easy fix
<azonenberg>
Yeah, but it shows you the level of care that went into the design
<rqou>
i've done that by accident by yolo-ing the parts ordering
<azonenberg>
And you see why they were willing to pay me so much to rework it now? :p
<azonenberg>
They were actually marked as DNP on the schematic
<rqou>
"eh, i need to compute values for this thing, i'll do the math later"
<rqou>
oh
<rqou>
wtf
<rqou>
i expected it to be more like "Cxx = TBD uF"
<azonenberg>
Nope
<azonenberg>
When i was done and had shipped the board back to the client their engineers were all @_@ at the stuff i had pulled off, once i sent them the bug list they were amazed it was salvageable
<azonenberg>
I had a pretty long phone call with the manager afterwards
<azonenberg>
telling him just how much money they had wasted by not doing a proper design review prior to tapeout
<azonenberg>
I gave them the link to my design review checklist, and i suspect nobody is ever going to forget a review in that team going forward :p
<azonenberg>
apparently they were not doing a full sign review at all
<azonenberg>
signoff*
<rqou>
does your checklist include things like "make sure nothing is labeled TBD uF"? :P
<azonenberg>
"power / voltage / tolerance ratings specified if important" i think encompasses that
<rqou>
on this topic, i just remembered a fun story from my grandfather
<rqou>
at one point my grandfather did QC in a machine shop making custom orders
<sorear>
what was on that board that was worth $thousands of your time to salvage?
<rqou>
one day he walked onto the machining floor and looked at the part being machined, and then looked at the drawing
<rqou>
it was a shared drawing for something like three different variant parts
<azonenberg>
sorear: lots of SiGe RF stuff, big FPGA
<azonenberg>
plus lead time for a respin, PCB NRE
<azonenberg>
(this was a prototype, firmware dev was halted until they got a working unit)
<rqou>
one set of dimensions had a very faint checkmark next to it, but it wasn't definitive that that was the right one
<rqou>
turns out it was wrong
<rqou>
:P
<rqou>
somehow the "which set of dimensions to use" information got lost during the calling/faxing/etc. between the shop and the customer
<rqou>
and they had already made thousands of parts
<azonenberg>
lool
<rqou>
at least they hadn't finished the _entire_ order yet
<azonenberg>
Oh, another fun story
<azonenberg>
Customer shipped us a product for pentesting that had moving parts
<azonenberg>
Used in a life-safety application
<azonenberg>
The limit switches on the axes were in software
<azonenberg>
There was no watchdog timer
<rqou>
uh
<azonenberg>
And the firmware would lock up if you pressed UI buttons at a certain time
<rqou>
were they toyota? :P
<azonenberg>
You can probably guess how that ended
<rqou>
you guys mechanically broke it?
<azonenberg>
The system had a UI consisting of a non-touchscreen LCD
<rqou>
was it a car? :P
<azonenberg>
and one push button
<azonenberg>
We bricked it with that one button
<azonenberg>
Mechanically
<rqou>
> was it a car? :P
<azonenberg>
To the point that we had to end the pentest early
<azonenberg>
because there was nothing to test anymore
<azonenberg>
and i cant say more about what the thing was or you'd know the exact product
<azonenberg>
yay NDAs
<rqou>
can you say what industry?
<rqou>
e.g. automotive, medical?
<azonenberg>
Probably shouldn't
<azonenberg>
other than, if it broke somebody could die
<sorear>
"1 button" is an interesting UI choice
<azonenberg>
sorear: there are occasionally products i test where i really wish i could write the report differently
<azonenberg>
I have to say "these are the bugs i found, here's what we recommend you do about them
<azonenberg>
But i wish i could say "this entire product is fundamentally a horrible idea and should never have been created"
<azonenberg>
This was one of them
<rqou>
azonenberg: still no "toys" in the IOA lab? :P
<azonenberg>
rqou: I have not tested any teledildonics hardware, no
<azonenberg>
still waiting, it's going to happen
<azonenberg>
the industry is growing and security risks are becoming more known
<rqou>
O_o i just thought of a "fun" device that fits your description of "only one button but can kill someone"
<rqou>
but i know it's not it
<rqou>
anyways, azonenberg what do you think about an IoT connected "TENS" device? :P
<azonenberg>
tens?
<azonenberg>
And the closest i came was when i told a customer that the whole session handling / login subsystem of their product was fundamentally broken
<rqou>
not a real TENS device, a "TENS" device
<azonenberg>
and needed to be reimplemented from scratch
<sorear>
I was thinking "One button -> awfully slow UI -> probably for an intrinsically slow, hence large, device with multiple axes -> maybe some kind of crane?"
<rqou>
i really need to get myself some soldering tweezers
<sorear>
wondering if socketed large FPGAs are a thing / if CPU sockets have survived into 2018 for any reason other than "direct to consumer marketing is more profitable"
<azonenberg>
sorear: BGA sockets exist but are super expensive
<azonenberg>
normally only used for test/burning
<azonenberg>
And because there's so many CPU SKUs
<azonenberg>
doing a production run of a board with each possible CPU isnt economical
<azonenberg>
also most cpus are LGA
<azonenberg>
(i think the socket itself is a bga)
<azonenberg>
but i dont know any fpgas available in lga
<awygle>
woo comcast outage
<rqou>
hmm that woman doing phone repair actually uses leaded solder despite the potential for intermetallics or whatever
<rqou>
also rosin flux. totally not the techniques i expected
<azonenberg>
Do you know how long her stuff lasts? :p
<azonenberg>
if it only has to last the rest of a typical phone's lifespan
<azonenberg>
it may not matter
<rqou>
i just hate rosin flux because it's sticky and messy
<azonenberg>
i use sticky flux on purpose
<azonenberg>
it helps keep things from moving around
<rqou>
i usually use a tacky chemical flux
<azonenberg>
and it doesnt drip all over the board
<azonenberg>
i use tacky flux but not sure what its made of
<azonenberg>
chipquik smd291nl
<rqou>
i seem to find rosin is much harder to clean
<azonenberg>
its rebranded amtech iirc
<azonenberg>
but not sure what it actually IS
<awygle>
catching up on longs, i have firsthand knowledge that BGA sockets a) exist and b) are super expensive :P
<awygle>
rqou: what's your preferred tacky chem flux?
<rqou>
i have some mg chemicals flux that i bought ages ago
mumptai has quit [Remote host closed the connection]
<awygle>
i wonder how hard it would be to set up a home PCB washing station
<rqou>
ok, i just checked and it's number 8341
<rqou>
thanks for asking, i actually finally need to buy more
<awygle>
rqou: is this stuff reasonably easy to clean with IPA?
<rqou>
er, kinda?
<awygle>
(yes i know it's "no clean")
<rqou>
i find my boards don't clean very well
<rqou>
but it might be because i've just been using drugstore IPA
<awygle>
neither do mine (unless i spend tons of time on it)
<rqou>
azonenberg: thoughts on cleaning pcbs by just dunking them in dichloromethane? :P
<azonenberg>
Sounds like a terrible idea
<awygle>
which is a huge issue when doing >5GHz RF
<rqou>
will it affect the board?
<rqou>
i wouldn't expect it to
<azonenberg>
Dont know, i'd be more worried about connectors dissolving
<azonenberg>
and i use ACS reagent grade IPA for cleaning
<rqou>
although afaik anything else plastic will totally get wrecked
<awygle>
i don't remember what exactly they did, but one of our MEs once tried to clean a PCB in something that ate the (iirc) Rogers material
<rqou>
lolol
<rqou>
maybe DCM?
<awygle>
maybe RO4003?
<awygle>
something expensiveish anyway
<awygle>
but probably not one of the Dk ~= 10 materials
<rqou>
hmm i wouldn't actually expect that to dissolve in anything
<awygle>
it didn't really *dissolve*, primarily just delaminate
<sorear>
fiberglass = glass + glue so it seems plausible
<rqou>
i thought the epoxies used are supposed to cure into a covalently-bonded matrix that you can't dissolve?
<sorear>
hmm, maybe
<rqou>
hrm, that woman's soldering technique isn't particularly impressive
<rqou>
it seems like it's mostly about just buying a small enough tool
pie_ has joined ##openfpga
<awygle>
i want a soldering microscope, a vacuum pickup tool with a foot switch, and a real reflow oven
<awygle>
and a tweezer tip for my iron
<rqou>
not a chinese special reflow oven?
<awygle>
and for none of that to cost any money
<awygle>
chinese special might be better than my current toaster-based solution
<awygle>
an unmodified toaster would probably also be better than my current hack job :P
<sorear>
i assume reflow works better with the board horizontal
<awygle>
well, it's a toaster _oven_
<Ultrasauce>
azonenberg: ya rope rescue is quite the endeavor isnt it
<awygle>
my next design, if i find myself building one, will use two PRTs for temperature sensing and have at least one fan
<Ultrasauce>
a friend of mine took a bad fall in a freak climbing accident a few months ago, the rescue took 3 hours...with the unit already in the area doing exercises
<awygle>
and a real board with a real AFE instead of an Arduino hack-job
<rqou>
awygle: instead of that, want to work on an open-source DIY hotplate? (chemistry lab hotplate)
<awygle>
rqou: i'm probably oversubscribed but i'll do board review for you :P
<rqou>
the mechanical design is probably the bulk of the work
<awygle>
yeah i'm a terrible ME
<awygle>
i wish i was better at it
<rqou>
same
<rqou>
did you at least watch the first ~minute of that video? :P
<awygle>
i have a minor in ME but it was a waste of time
<rqou>
oh you do?
<awygle>
rqou: i was on the phone with a customer so... i half-watched it on mute :P
<rqou>
you need to have it not on mute :P
<awygle>
cracked three-lead package
<rqou>
also imho berkeley's ME curriculum isn't as good as EECS
<awygle>
oh no argument
<rqou>
awygle: you really need to watch it with sound
<awygle>
"this cube thingy" lol
<awygle>
i'm watching it now
<rqou>
"A lot of complex scientific terminology is used in this video..."
<awygle>
triac failed short i guess? i feel like that's sort of uncommon
<rqou>
you need to go to 0:29 :P
<awygle>
what about it?
<rqou>
"complex scientific terminology" :P
<awygle>
ah lol
<awygle>
so that's basically just a BLDC and some heaters?
<rqou>
yeah
<rqou>
they can't even design that reliably
* awygle
was just talking about doing a BLDC controller earlier today...
<rqou>
whee, same here!
<awygle>
or technically yesterday i guess
<awygle>
okay! i get to go to bed now
<rqou>
somehow people who make "instruments" (whatever that means) really don't understand how useful it is for things to be computer-controllable and networked
<awygle>
i wonder if you could just buy nichrome in sheets and route out a heating element with a cheap mill
<rqou>
maybe?
<rqou>
nurdrage found a chinese seller that sells the heating elements
<awygle>
yeah i just finihsed that video
<rqou>
but min order is 30 for like $500 or so
<awygle>
but that's unsatisfying
<Ultrasauce>
I assume you're familiar with the VESC project?
<rqou>
no?
<Ultrasauce>
opensauce bldc driver
<awygle>
yeah i'm familiar
<awygle>
huh, online metals doesn't seem to sell nichrome, and you can only get mesh or wire from mcmaster
<mithro>
daveshah: I'm hoping to merge that giant pull request today
<mithro>
daveshah: Shouldn't have let it get so large
<daveshah>
mithro: Awesome. Will be good to have master up to date
Zorix has quit [Ping timeout: 245 seconds]
digshadow has joined ##openfpga
<mithro>
daveshah: Where are we at with the verilog generation stuff?
rohitksingh has quit [Quit: Leaving.]
Zorix has joined ##openfpga
<daveshah>
mithro: I have two major TODOs. Fixing the multiple mode syntax and adding support for autogenerating model.xml includes
<daveshah>
mithro: both should be finished in the next few days
<mithro>
daveshah: Okay cool
<daveshah>
mithro: then in the longer term we need a better approach to specify timing, and support for the num_pbs concept in some way (I.e verilog for generate)
user10032 has joined ##openfpga
<awygle>
how are you extracting timing information?
grantsmith has quit [Ping timeout: 240 seconds]
<mithro>
awygle: That is a question for digshadow -- but basically linear equations
<digshadow>
awygle: I'm cleaning that up right now to ask for help on it
<digshadow>
if you are interested to take a look
<digshadow>
as part of cleanup I found a potentially small issue, so I want to run a few checks before sending things out
grantsmith has joined ##openfpga
m4gul0 has quit [Quit: Connection closed for inactivity]
elms has quit [Quit: Connection closed for inactivity]
<digshadow>
Information on the timing fuzzer if anyone wants to play with it
<mithro>
digshadow: timfuz :-P
<digshadow>
mithro: yes, it psychologically probes tim, learning how he works. Don't tell him though
<digshadow>
mithro: I also sent the link to clifford, asking for feedback
indy has quit [Ping timeout: 265 seconds]
pie__ has joined ##openfpga
pie_ has quit [Ping timeout: 252 seconds]
indy has joined ##openfpga
digshadow has quit [Read error: Connection reset by peer]
digshadow has joined ##openfpga
<digshadow>
I recall that a lot of the xilinx partial reconfiguration technology was an acquisition of a product called recobits, recologic, or something like that. But I'm looking now and I don't really see anything about that. Does anyone recall something of this sort?
<mithro>
daveshah: Merged now!
<daveshah>
mithro: Awesome. Will see what I can do on the XML stuff tomorrow
<awygle>
rqou: just found my diploma at the very bottom of a box
<rqou>
lol
<rqou>
not even framed?
<awygle>
it was next to an empty frame
<rqou>
lool
<awygle>
i suppose i should hang it but i'm not sure where it's appropriate
<rqou>
mine is just on a desk in my parents' house
* awygle
hangs it in the kitchen, directly above the fridge
<rqou>
lol
<awygle>
it has that instant-devaluation factor that anything i actually achieve gets - as soon as i have it it becomes no big deal
<awygle>
poll: what should i do with all these boxes full of the extras left over from boards i've kitted and assembled/had assembled before?
<awygle>
it seems like a waste of money to get rid of them but a waste of time to do an extensive inventory...
<rqou>
dump them on azonenberg? :P
<awygle>
they are largely uninteresting - interesting parts tend to be expensive and thus i don't buy overages
<awygle>
i think maybe i'll dump all the passives and keep the ICs and connectors. that shouldn't take too long to go through.
digshadow has quit [Read error: Connection reset by peer]