#avr Logs
Feb 08 2022
#avr Calendar
01:35 AM specing_ is now known as specing
05:20 AM rue_bed: heh, boost converter
07:46 AM twnqx: turbo boost!
07:46 AM twnqx: i had to drop a texas instruments boost converter.. we were paying 19€ each for the last batch, time to say goodbye
07:47 AM twnqx: well, boost regulator
07:47 AM twnqx: and now i have a a 5 digits stash of Microchip boost regulators :) should last a year or two
08:20 AM exp: twnqx: which Ti one did you drop?
08:21 AM twnqx: TPS63020QDSJRQ1
08:21 AM twnqx: we paid 19€ each instead of the 0.8 or whatever TI asks...
08:21 AM exp: have not used that particular one but we had to drop some Ti LDOs and DCDCs for similar reasons
08:22 AM exp: also i have found the LDO specs to be a bit dubious
08:30 AM exp: currently using microchip integrated buck and searching for an alternative buck-boost :(
08:30 AM twnqx: i need to replace a dual-buck-dual-ldo TI chip
08:30 AM twnqx: >_>
08:30 AM twnqx: there is one (1) alternative chip apparently in existence
08:30 AM exp: if it's any consolation, i think they are doing ok with their lead times
08:30 AM twnqx: microchip?
08:30 AM exp: i've had a lot of parts moved up
08:30 AM exp: nah Ti
08:30 AM twnqx: yeah, 52 weeks...
08:30 AM twnqx: i can't *buy* TI
08:30 AM twnqx: whatsoever
08:30 AM exp: nah some parts which were 2023 moved to 2022 for me and i just got them delivered
08:30 AM twnqx: i can't even *order*
08:30 AM twnqx: how do you order them?
08:30 AM exp: these were mouser + digikey backorders
08:30 AM twnqx: digikey doesn't accept backorders
08:30 AM twnqx: mouser says 80 weeks lead time
08:30 AM twnqx: lm26480-*-AA
08:30 AM exp: digikey does backorder stuff if they can't supply it
08:30 AM exp: mouser we ordered regardless of lead times
08:30 AM exp: mostly LM34936 / TPS27S100 / TPS244160 i think
08:30 AM exp: uh TPS274160*
08:30 AM exp: we have some TCA4307s, TCA9548AMRGRs and others on order too, so hoping to see those move forward in a week or two
08:32 AM twnqx: hm, 500F capactors
08:35 AM twnqx: mouser says "6k on order, 2k 10.05.2023, 3k 26.05.2023, 2k TBD" :(
08:36 AM twnqx: actually that's 7k...
08:36 AM cehteh: link? what voltage?
08:36 AM twnqx: the caps? 2.1V
08:37 AM cehteh: i guess they go with a loud bang when you overvoltage them
08:37 AM twnqx: https://www.mouser.de/new/kyocera-avx/avx-prizmacap/ just noticed it on the frontpage
08:37 AM cehteh: lots energy stored
08:37 AM twnqx: there's a way to find out :P
08:38 AM cehteh: :)
08:38 AM exp: twnqx: if i come across a reel i'll let you know ;)
08:38 AM exp: do me a favour and find me some more LM5050-mk2s please :D
08:38 AM exp: those are still 2023
08:39 AM twnqx: i can get you 22 eval boards, and you can desolder them :P
08:40 AM exp: we may or may not have already taken that approach with some other bits
08:40 AM * exp whistles
08:40 AM exp: we have a year's worth of stock or so but i would kill for a reel of lm5050s too
08:40 AM exp: i should be able to find alternatives, but that takes time and redesigns
08:41 AM twnqx: haha
08:41 AM exp: also got a reel of parts from win-source electronics, who are surprisingly real
08:41 AM twnqx: we have ADP5034 scheduled as a replacement for the lm26480
08:42 AM twnqx: win-.source really seems hit&miss
08:42 AM twnqx: some say "great"
08:42 AM twnqx: some say "0% success rate"
08:42 AM exp: yeah sounds probable
08:42 AM exp: twnqx: why so tightly integrated? you supremely short on space?
08:43 AM exp: i've seen a few 'system power controllers' i think that are similar in design, but i'm not sure i've ever found them ideal
08:43 AM exp: only made one handheld thing so far though
08:44 AM twnqx: sadly this isn't even one
08:44 AM twnqx: still need a separate supervisor as sequencer
08:45 AM twnqx: https://i.imgur.com/N4rFpJV.jpg the part on the top-side is main power regulator, topside right bottom is the step-up
08:45 AM twnqx: there's... not much space :P
08:46 AM exp: what sort of bizarre expensive pcb process is that?
08:46 AM exp: filled and capped vias, but no floods?
08:46 AM twnqx: 6 layer 0.8mm thickness
08:47 AM twnqx: via in pad, so yeah, filled and capped
08:47 AM exp: i am surprised you pushed them so close to the edge connector too
08:47 AM exp: i guess when capped it's irrelevant
08:48 AM exp: but i get complaints about that, presumably as they don't want to plate vias with thick layers of gold
08:48 AM twnqx: we could have gotten away without via in pad, if it wasn't for the eMMC...
08:48 AM twnqx: 0.5mm ball spacing, no chance
08:48 AM twnqx: pitch*
08:49 AM exp: you can often just leave the vias open if they're small enough
08:49 AM exp: HD boards i'm using .15 final diameters
08:49 AM exp: and you do get some wicking but it's not crazy
08:49 AM exp: i'm guessing your customer doesn't mind the premium
08:49 AM exp: also i love the fiducials the size of a fat finger, you place in-house?
08:50 AM twnqx: btw
08:50 AM twnqx: https://www.youtube.com/watch?v=rUQYPRixwdo don't do mil-spec vibrations tests with such PCBs :P
08:50 AM twnqx: nah, there was space for the fiducials, so why not
08:51 AM exp: lol jesus that's some vibration
08:51 AM exp: is that hitting a unique mode or is it just that aggressive?
08:51 AM twnqx: milspec
08:51 AM exp: this is in my future so i'm not looking forward to it
08:51 AM twnqx: i think 810h
08:51 AM exp: so a couple more questions if you don't mind?
08:52 AM twnqx: we didn't go for it in the end :D
08:52 AM exp: this is obviously an rf board, but there's no ground flood and i don't see any maskless microstrips
08:52 AM exp: i see some differential tracks but they are masked still
08:52 AM twnqx: RF is only from modem to the antenna connector
08:52 AM exp: everything inner layer GCPWG?
08:52 AM exp: oh no impedance matching on any other traces?
08:53 AM twnqx: masking doesn't matter that much if you calculate t in
08:53 AM twnqx: mh, there's impedance matching on USB, and on DRAM at least
08:53 AM exp: funnily enough our expensive manufacturer doesn't give permittivity of their mask
08:53 AM exp: but the cheap one i use occasionally gives full specs lol
08:53 AM twnqx: but IMHO it's pretty much overrated
08:53 AM twnqx: the boards are like... 3€ each
08:54 AM twnqx: if you order batches of a few thousand
08:54 AM exp: that's not bad really
08:54 AM exp: i normally order a few hundred and via plugging adds a fortune
08:58 AM twnqx: in the end it's up to the manufacturer, no?
08:58 AM twnqx: you say "this track. 50ohm single ended. make it so."
08:58 AM twnqx: and they fix the geometry to match it
08:59 AM exp: twnqx: huh?
08:59 AM exp: i wish they bloody did, i'm the one who has to do that
08:59 AM exp: or did you design the schematic and have a company lay it out etc?
09:00 AM twnqx: i have someone in the US do the layout, and he precalculates the impedance of course
09:00 AM * exp is jealous
09:00 AM twnqx: but.. in the end the baord manufacturers i work with tend to adjust if you specify
09:00 AM exp: i do everything start to finish, mostly manufactured with eurocircuits
09:01 AM exp: who i don't think care in the slightest :D
09:01 AM twnqx: well
09:01 AM twnqx: i had to compain to them once
09:01 AM twnqx: because they put stopmask on the pads of a connector for the above module...
09:02 AM twnqx: german support "but... that can't work this way!" "what do you think why i call?"
09:02 AM twnqx: >_>
09:02 AM exp: i've never worked in such a scenario, but i've talked to a couple of people here and ##electronics that do something similar
09:02 AM exp: it doesn't seem efficient to me for complex boards
09:03 AM twnqx: what? stopmask OVER solder pads?
09:03 AM exp: for example i have spent a day carefully swapping GPIO pins so that i can route signals pretty much entirely on one layer
09:03 AM exp: no, using a 'layout department'
09:03 AM twnqx: well, i didn't do the schematic either
09:03 AM exp: layout is a solid part of design for me, and routing part of that to check feasibiliy
09:03 AM exp: it's just that i don't understand how the workflow happens
09:03 AM twnqx: "these are the parts, these are the specs, this is the pinout of the connector"
09:04 AM twnqx: "you are free to swap these as you need it"
09:04 AM twnqx: admittedly, it's mainly because i would have to swap and learn a new pcb layout software
09:05 AM exp: trying to think how many lines i'd need to give someone to lay out this board
09:05 AM twnqx: i'd never do DDR3 dram with eagle...
09:05 AM exp: i'm gonna try it with kicad :D
09:05 AM twnqx: this design was made with kicad
09:05 AM exp: excellent
09:05 AM twnqx: but every time i open kicad
09:05 AM twnqx: i have to throw up
09:05 AM twnqx: and close it ASAP
09:05 AM exp: i'm sorry you feel like that
09:05 AM twnqx: forward annotation? lolno
09:05 AM twnqx: backward annotation? lolno
09:05 AM exp: perhaps you should open since 6 ;)
09:06 AM twnqx: change a value in schematic? don't expect the BOM generated from the board to change
09:06 AM exp: that's still true because they're different parts
09:06 AM exp: you have to press the 'update board' button
09:06 AM twnqx: want an SMD part? sucks to be you, have to add it for EVERY SMT COMPONENT ON THE BOARD, not put it on the component
09:06 AM exp: not sure what that means
09:07 AM twnqx: export BOM -> 90 of 250 parts show up
09:07 AM twnqx: ah sorry, not bopm
09:07 AM twnqx: pick and place
09:07 AM exp: twnqx: most of these sound like bugs that have been resolved
09:07 AM exp: save for the schematic -> pcb change which is intentional
09:07 AM twnqx: assembly house: "uh, your BOM has 250 components, but the PNP only lists 90, are you sure that's correct?"
09:07 AM exp: you can certainly annotate backwards and forwards, as well as geographically reannotate and re-assign based on ID
09:07 AM exp: i have my own pick and place and pos files have been comprehensive
09:08 AM twnqx: yeah, there's also the "export everything button"
09:08 AM twnqx: but the whole stupidity of this sucks
09:08 AM exp: i'm not sure i understand the 'SMD part' criticism though
09:08 AM exp: could you explain that?
09:08 AM twnqx: why to i have to define a part, say, 0805 resistor
09:08 AM twnqx: and everytime i place it on the board
09:08 AM rue_mohr: who uses BOMs?
09:08 AM twnqx: THEN have to annotate every single fucking one as "yes, this is SMT, put it into the pick and place data"
09:09 AM exp: twnqx: wow were you using like kicad 4 or something?
09:09 AM rue_mohr: have you used eagle$ latley?
09:09 AM exp: whether something is smt or TH is part of the footprint data
09:09 AM exp: and was in 5 too
09:09 AM twnqx: this was 5.4
09:09 AM twnqx: or whatever was the last before 6
09:09 AM exp: something wacky is going on then, because that should be part of the footprint definition
09:09 AM twnqx: rue_mohr, i am stuck on eagle 7
09:09 AM exp: and you certainly should never be asked for some 0805 resistor
09:09 AM exp: all other bugs you listed are resolved
09:09 AM rue_mohr: did they delete any of your files yet?
09:09 AM twnqx: there is no fucking way i will ever in my life pay for time limited software
09:09 AM exp: (except sch -> pcb without clicky button)
09:10 AM twnqx: "they"
09:10 AM twnqx: ?
09:10 AM exp: twnqx: have you used the 'interactive bom' plugin?
09:10 AM rue_mohr: if your project files are gone, does it matter?
09:10 AM exp: it is one of my favourite things ever
09:10 AM twnqx: why would they be gone?
09:10 AM rue_mohr: oooh see if you get to find out!
09:11 AM exp: https://openscopeproject.org/InteractiveHtmlBomDemo/html/motherboard.html
09:11 AM exp: ↑↑ check this out
09:11 AM exp: for mass manufactured boards obviously not very useful
09:11 AM twnqx: rue_mohr, like i said, i refuse to upgrade to autocad shit
09:11 AM twnqx: eagle is at a dead end for me
09:11 AM exp: but for those of us dealing in 2-3 figure quantities...
09:11 AM rue_mohr: ah
09:11 AM rue_mohr: so your not using the latest version
09:11 AM twnqx: no
09:11 AM rue_mohr: wise
09:11 AM exp: my boss keeps pushing eagle
09:12 AM exp: i feel similarly to you twnqx
09:12 AM twnqx: i have an ultimate license on the small one, privately
09:12 AM twnqx: and i want to be able to work an hour here, an hour a month later
09:12 AM exp: i need a new warning in kicad that isn't in the core
09:12 AM twnqx: without paying full recurring price as a hobbyist for it
09:12 AM exp: because i need it to catch mistakes
09:12 AM exp: it'll take me an hour over the weekend to patch it in
09:12 AM exp: wonder how long it'd take autocad to implement
09:12 AM exp: they're at ~4 years on a major fusion feature currently
09:13 AM twnqx: it's still the cadsoft team doing the development
09:13 AM twnqx: and they have always been slow
09:13 AM exp: worse, they're bad at their job
09:13 AM exp: fusion 360 is really a bane
09:13 AM exp: twnqx: the feature by the way is a usable hole table
09:13 AM exp: that's all
09:14 AM exp: they have added one, but it only works on certain models and that's it, their advice to fix the problem is "don't use certain fusion features"
09:15 AM twnqx: as much as i'd like to switch to kicad
09:15 AM twnqx: it's not yet usable for me
09:15 AM twnqx: or "to" me
09:15 AM exp: twnqx: if you do try again, please file bugs for what you hate, or at least give me a list :)
09:16 AM twnqx: the absolute minimum i'd expect is full integration of schematic/bom editor, with instant both-side changes
09:16 AM exp: i doubt that's ever going to happen
09:16 AM twnqx: s/bom/pcb/
09:16 AM exp: there may be some 'auto sync' option
09:16 AM twnqx: well, that it won't ever be kicad
09:16 AM twnqx: it's pointless to have that without
09:16 AM exp: i haven't found it a particularly large impediment
09:17 AM exp: it's easier for me to work in 'phases'
09:17 AM twnqx: maybe your development style is fundamentally different from mine then
09:17 AM exp: last thing i want is to mess about with a schematic for 10 mins and find i've screwed up half of my existing board
09:17 AM twnqx: i am bidirectionally
09:17 AM twnqx: find something that's bad on board, edit schematic and see changes
09:17 AM exp: that is supported, but you have to press a button :)
09:18 AM twnqx: and of course, if i replace something i want to see how it looks on the board instantly
09:18 AM twnqx: and if it's bad press ctrl-z
09:18 AM exp: let's face it, someone will have made a plugin that syncs on focus
09:18 AM exp: ctrl+z works fine
09:18 AM exp: just not auto synced
09:18 AM twnqx: on both sides?
09:18 AM twnqx: yeah
09:18 AM twnqx: that is the issue
09:18 AM twnqx: to i look like i want to do the same single action more than once?
09:18 AM exp: i think it's a pretty minor issue all things considered
09:18 AM twnqx: it is ONE change, and it can't take more than ONE action
09:19 AM twnqx: i had boards blow up because of it
09:19 AM exp: well DRC does warn you about schematic parity
09:19 AM exp: so that won't go missed in current versions
09:19 AM twnqx: "oops, resistor didn't show up in pick and place because it wasn't set to SMT"
09:19 AM exp: yeah that also isn't a thing
09:19 AM exp: the footprints define the type now
09:20 AM twnqx: "BOOM" 5V on 1.2V Vcore
09:20 AM twnqx: sadly it was the feedback divider.
09:21 AM twnqx: great the prototype boards were only what, 500€ each?
09:21 AM twnqx: i think i paid 1k for the first 10 PCBs *alone*
09:21 AM exp: yeah that's feasible pricing
09:21 AM exp: but as i say that's now solved and guarded against in several ways
09:21 AM exp: footprints themselves define their category
09:22 AM exp: DRC checks schematic parity
09:22 AM exp: you can do forward or reverse syncs
09:22 AM exp: the only thing missing you've listed is the sync being automated
09:22 AM exp: which i'm betting you could script but i haven't seen it done yet
09:22 AM twnqx: also, full properties on both sides
09:22 AM twnqx: haven't tested in 6
09:23 AM exp: how do you mean?
09:23 AM twnqx: but highlighting something in schematic and now having highlighted in board is so fuckign annoying
09:23 AM twnqx: like selecting a trace in schematic (because you can see it)
09:23 AM twnqx: and then... nothing in board
09:23 AM twnqx: "hmmm where could it be..."
09:24 AM twnqx: "oh it has no custom name, let me check in schematic what the name is so i can use search in pcb"
09:24 AM twnqx: hello, the 1970s called and want their usability back
09:25 AM twnqx: yes, i know the history of eecad and eeschema
09:25 AM twnqx: but if you want a usefull suite those need to be merged in a sensible way
09:25 AM twnqx: and the "keeping it separate" is not that
09:26 AM exp: fwiw it won't currently highlight a trace when you click a schematic track, as there's no 1:1 mapping
09:26 AM exp: it will however do that in reverse, and works for all pins and components
09:26 AM exp: but that's a fair bug to point out, i'll have a look at that one
09:26 AM exp: there's no fundamental issue with separate designs though
09:26 AM exp: i expect to be able to work on a schematic for a few days, committing changes
09:26 AM exp: without overwriting half of an existing PCB
09:27 AM exp: it would however be nice if there was some 'auto-sync' for people who do want that workflow
09:27 AM twnqx: why would that overwrite anything?
09:27 AM twnqx: i do that in eagle day in and day out
09:27 AM exp: you just described it doing?
09:27 AM exp: you said a component change must be visible immediately
09:27 AM exp: so if i change an 0402 to 1210, that would definitely break a pcb
09:27 AM twnqx: edit schematic, place on board, look at ratsnest, gateswap if needed
09:27 AM twnqx: yeah sure
09:28 AM twnqx: but i only do the change BECAUSE of the baord
09:28 AM exp: sure, i just want it to be an explicit button press
09:28 AM twnqx: (well, or because i realized i need more power dissipation)
09:28 AM exp: i don't mind that at all, i do that in vim too for example
09:28 AM twnqx: anyway, need to put car to workshop, bbl
09:29 AM * exp has 71 gpios still to route
10:13 AM twnqx: that's... a lot of gpios
11:23 AM exp: i'm down to in the 40s :)
11:27 AM twnqx: through-hole and multilayer poses nteresting questions
11:27 AM twnqx: if you want to bypass some IC... do you isolate that IC's VCC pin from the supply plane, so you can explicitly route it from its bypass capacitor?
11:37 AM exp: twnqx: no i just place them manually
11:37 AM exp: using net ties is so ugly
11:38 AM twnqx: i'm not sure if i'll have a vcc plane anyway, or if i'll use two GND planes (so i don't have to care about referencing data lines)
11:39 AM twnqx: things to keep in mind when your 74xx design closes in on 100mhz operating speed :P
11:40 AM exp: if you have acceptable decoupling there's no significant advantage to a vcc plane IMO
11:40 AM exp: you do get better coupling so reduced EMI etc
11:40 AM exp: but that's better controlled locally
11:40 AM exp: my current board is P/S, Gnd, S, G, S, P/S/G
11:40 AM exp: cause that's just how it works out
11:40 AM twnqx: this is some hobby stuff, so 4 layers tops :P
11:41 AM exp: normally my stuff is 4 layer but on this one i want the extra isolation
11:41 AM exp: some 800khz gate driving going on
11:41 AM exp: plus quite a bit of power
11:41 AM twnqx: heh
11:41 AM twnqx: my carrier board for the thing i showed earlier... broke residential emission limits 1at 250mhz :(
11:42 AM twnqx: because i used a vcc and a gnd plane
11:42 AM twnqx: and a 50mhz oscillator had to change layersfor a brief time (18mm or so)
11:43 AM exp: that should still be fine, 50mhz isn't very high
11:43 AM exp: unmatched vias?
11:43 AM twnqx: it multiplied with the dram's 125mhz+ddr
11:43 AM exp: unterminated line?
11:43 AM exp: you should be able to achieve near 100% power matching
11:43 AM twnqx: building an antenna through bad return flow
11:44 AM twnqx: because mostly referenced to gnd, but those 18mm to +3v3
11:44 AM twnqx: and no caps to cross for the reverse current
11:44 AM exp: ah right, so just trace inductance
11:44 AM exp: that's way too short for any wavelength effects
11:44 AM exp: you could have sympathetic resonance or some unintentional coupling
11:45 AM twnqx: removing the oscillator made us pass the test
11:45 AM twnqx: so.. i blame it :P
11:45 AM twnqx: though we had ethernet no more.
11:45 AM exp: on tracks like that i often try and split them in half with a 0 ohm jumper
11:45 AM exp: 50mhz is starting to get into dubious ranges
11:45 AM exp: but you'll push the resonance of each end up and a bit of attenuation in the middle significantly affects VSWR
11:45 AM twnqx: 5th was the only problem, the 250mhz on top of the base noise of ram
11:46 AM twnqx: 50mhz was a no-issue
11:46 AM exp: interesting though, would have expected it quite far down
11:46 AM exp: i don't have intuition for dBc figures yet
11:46 AM exp: got some 3ghz tracks to route soon though so that'll be fun
11:46 AM twnqx: it wasn't much, and the peak was inside industrial limits
11:46 AM exp: two whole pages of the schematic are dedicated to decoupling
11:47 AM exp: just capacitor after capacitor after capacitor
11:47 AM twnqx: ouch
11:47 AM exp: and they laid them out as a grid, then populated only a few
11:47 AM twnqx: O_o
11:47 AM exp: now that's design skill
11:47 AM twnqx: there's some software that calculates for a given board how to kill emi by strategically placing caps
11:48 AM twnqx: but i forgot the name and can't find it any more
11:48 AM exp: in this case they've got a nice fat processor next to a high end (unlabelled) synth
11:48 AM exp: so i presume they just put some 10ghz DSO on it and decoupled everything until it was quiet
11:48 AM twnqx: lol
11:49 AM twnqx: that's one way to do it :D
11:49 AM exp: it looks exactly like that, there's a huge grid of decoupling caps on the other side of the board
11:49 AM exp: all specified in schematic
11:49 AM exp: then they're fit almost at random
12:24 PM twnqx: hm. trying to find some things to stabilize two PCBs mounted at 90° to each other.. it's hard if you don't know the english term :S
12:24 PM exp: 'edge connector'
12:24 PM qu1j0t3: bracket?
12:24 PM exp: 'straddle mount'
12:24 PM exp: twnqx: how do you want to connect them?
12:25 PM twnqx: thought of standard 2.54mm angled connector
12:25 PM twnqx: just want to give them some stability
12:26 PM exp: they make them with a rectangular outer shroud
12:26 PM exp: and a detent
12:26 PM exp: that will somewhat increase its stability
12:26 PM exp: twnqx: how many cols/rows?
12:27 PM twnqx: ~ 1x20
12:27 PM exp: https://www.digikey.co.uk/en/products/detail/te-connectivity-amp-connectors/6-103639-9/1122521
12:28 PM exp: that's the sort of thing i'd consider, maybe shorter if you need it
12:28 PM twnqx: i have about the bottom... well... 5mm if i stretch it of the angeled PCB :P
12:28 PM twnqx: the less the better
12:29 PM exp: https://www.digikey.co.uk/en/products/detail/hirose-electric-co-ltd/DF1BZ-20P-2-5DSA/1033128
12:29 PM twnqx: soldered down, no connectors... i have 15mm to put a 7 segment display in
12:29 PM exp: huh, soldered down?
12:29 PM twnqx: yeah
12:29 PM exp: just use a 1x10 from either side of the board
12:29 PM twnqx: it's just 2 7-segment-display, and 3 buttons, at an angle
12:30 PM twnqx: angle bracket seems to be what i am looking for indieed qu1j0t3
12:30 PM exp: guess i got the wrong end of the bracket ;)
12:31 PM qu1j0t3: o/
12:32 PM qu1j0t3: exp: Well our theories bracketed the target, at least
12:32 PM twnqx: lol
12:32 PM exp: :D
12:32 PM exp: i found most pcb fixings to be gigantic compared to what i needed
12:33 PM exp: the only acceptable ones i found were axial threaded mounts
12:33 PM exp: ie that mount to a NPTH
12:33 PM exp: use those a lot now
12:38 PM twnqx: https://www.mouser.de/datasheet/2/215/567-1571-743481.pdf this, in higher, narrower, and M2 threaded would be perfect :P
12:41 PM Rab: I've been hating on KiCad for 20 years, but I just looked at v6 last week and it's noticeably more polished than v5.
12:41 PM exp: yeah lots still needs to be done
12:41 PM exp: but i rarely fight with it more than my board
12:42 PM twnqx: i still don't get how you can do pinswaps without the instant visual feedback of the ratsnest change
12:42 PM Rab: Auto forward/back annotation is a killer feature in Eagle (not sure about Altium or other pro packages). I would definitely like to see that implemented in KiCad.
12:45 PM Rab: Eagle's scriptable console is also underrated. E.g. typing 'sho vcc' and having the net light up in schem and pcb. Or 'mo r20' and have R20 snap to your pointer, ready to place. Etc etc.
12:45 PM twnqx: also i massively dislike nearly every schematic i've ever seen in kicad, but i guess that's more about me being a visual person and the creators of the schematics :P
12:45 PM twnqx: rab: i was expecting kicad would have a hotkey for that
12:45 PM exp: Rab: the forward/back does work just fine but yeah it requires a button press, i think it's just a different work style
12:45 PM twnqx: but yes, the textcommand console is a bliss :D
12:46 PM exp: also kicad does have a python console but no shortcuts like that i'm aware of
12:46 PM exp: that would be very nice, as a vim user
12:46 PM Rab: Well, Eagle schematics have their own distinctive ugliness.
12:46 PM twnqx: i massively dislike the tendency of keycad users to just drop a chip, put a named line on each pin
12:47 PM twnqx: and then have the other end, potentially multiple times, somewhere else on the sheet
12:47 PM twnqx: kicad*
12:47 PM exp: that's normal in all EDAs i think
12:47 PM exp: you might not like some of my schematics ;)
12:47 PM Rab: I think that's a Make:r thing, not a KiCad thing.
12:47 PM twnqx: that's only ok for vcc and gnd in my books
12:47 PM exp: i do try and draw actual traces, but everything is broken up into functional blocks
12:47 PM exp: otherwise it's impossible
12:47 PM twnqx: i want to grasp a schematic at a glance
12:48 PM twnqx: with a single view
12:48 PM exp: twnqx: subsheets 100% unacceptable then?
12:48 PM Rab: I agree that overdependence on net flags isn't proper documentation.
12:48 PM twnqx: nah, that's what youo have busses for (in eagle)
12:48 PM exp: kicad has buses but that doesn't always work
12:48 PM exp: only for simple designs
12:49 PM twnqx: i personally rather go for 20 page schematics than make them less readable
12:50 PM exp: how do you connect between pages?
12:50 PM twnqx: as few labels as possible (and busses do that automatically in eagle)
12:50 PM exp: well that's reasonable
12:50 PM exp: each of my little blocks could also be a page
12:50 PM Rab: I have a paid license for pre-Autodesk Eagle v5, which is quite elderly now. I've been rooting for KiCad, 'cause I don't see moving to a subscription product.
12:51 PM exp: we have some eagle cloud license, but i am an open source zealot
12:51 PM exp: if it can be done with open source i strongly prefer to do it that way
12:51 PM twnqx: i started with a 5 ultimate, bought 6, and then 7... and that's where eagle ends
12:52 PM Rab: I don't think the functional block approach is a problem per se. I've seen schematics where it's just a bunch of chips dumped onto the page with no connections, and poorly-labeled nets.
12:52 PM twnqx: until autodesk sells non-time-limited, cloud- and internetless licenses again
12:52 PM exp: Rab: yeah it's a definite trade-off
12:54 PM Rab: I kicked some money to LibrePCB, which seems like a very clean free software reimplimentation of the Eagle UI. I think development has slowed lately though.
12:54 PM Rab: https://librepcb.org/
12:54 PM exp: looks derived from freecad too
12:55 PM exp: as soon as i can sort out a department budget i think i will try and justify some donations
12:56 PM Rab: One advantage of these giant, ancient projects with lots of cruft like KiCad and FreeCAD is that people have pitched in to address all kinds of edge-case functionality over the years...hard for a one-person project to compete. (Of course it's a big disadvantage too, from a maintenance standpoint.)
12:57 PM twnqx: https://docs.google.com/viewer?a=v&pid=sites&srcid=bWFsaW5vdi5jb218d3d3fGd4OjIyYWViMDc3YmU1MmY0Yg take a look at this if you want to understand what i absolutely can't stand
12:57 PM twnqx: just try to understand this schematic
12:58 PM exp: mine looks much like that at the moment really
12:58 PM exp: there's not much you can do when you get to that scale
12:58 PM twnqx: of course you can
12:59 PM exp: adding the column/row to net names automatically is a really nice thing
12:59 PM twnqx: you can connect the 8088, 8087, its address registers and data buffers
12:59 PM twnqx: and call the result "system bus"
12:59 PM twnqx: then you can connect whatever else to the system bus
12:59 PM exp: yes this particular schematic is well suited to a few buses
12:59 PM twnqx: and declare what the hell the "x-bus" is
01:00 PM exp: mine is not however as it's just 70 different control lines or analog etc
01:00 PM exp: several sub-pcbs connected to one big 'distribution' pcb
01:00 PM exp: you're not wrong in your criticism
01:00 PM twnqx: (it's connecting the onboard components to the ISA bus, but still)
01:00 PM exp: i just think that there are practical limits a lot of the time
01:01 PM twnqx: i also don't really like the way kicad handles sub schematics, but eh
01:01 PM twnqx: it is acceptable i guess
01:02 PM twnqx: i mean the enforced parent-child hierarchy
01:02 PM exp: it's not actually enforced afaik
01:02 PM twnqx: my pcb layouter drives me mad, too
01:02 PM exp: you can link arbitrarily
01:02 PM exp: it's just you have to specify an existing sheet and it'll ask you if you want it linked
01:02 PM twnqx: randomly split the voltage supervisor and the supply chips
01:02 PM exp: maybe there's some topology you've tried i haven't
01:02 PM twnqx: random filter beeds, not connected to their uplink nets
01:03 PM twnqx: not *visually* connected
01:04 PM twnqx: the resistor dividers for the voltage supervisor floating in free space as two resistors, two pins to the right and one to the left
01:13 PM twnqx: unrelated: DIL ICs are... huge
01:20 PM exp: haha
01:20 PM exp: when i first started in this industry it was very hard to gauge how big things actually were
01:20 PM exp: we mostly do cad driven design now
01:20 PM exp: so it's not so bad
01:21 PM twnqx: i just had to make room, and now my good looking PCB is fugly, and possibly unroutable :(
01:22 PM exp: don't forget to put a big socket on there too
01:22 PM exp: to really bulk up the size
01:22 PM twnqx: one? :D
01:22 PM exp: i think you can get smt dil sockets if you want to really make people puzzled
01:24 PM twnqx: https://i.imgur.com/GitdvDL.jpg
01:24 PM twnqx: it was completely reasonable until i added the bottom ICs :P
01:26 PM exp: two layer?
01:27 PM twnqx: ground planes inside
01:27 PM twnqx: so yeah, two signal layers
01:27 PM exp: remove the top annular rings on the pads
01:27 PM exp: don't need them for soldering, lets you squeeze some more space
01:27 PM exp: coupling might be a concern but down to your application :)
01:28 PM twnqx: that's just a potentially 50mhz data bus, nothing of concern :P
01:28 PM exp: i really prefer running everything high speed on inner layers
01:28 PM exp: it makes routing really tricky too
01:29 PM twnqx: yeah, i call these "inverse designs" :P
01:29 PM twnqx: ground outside, data inside
01:29 PM twnqx: but it makes coupling apparent, at least in 4 layer designs
01:29 PM exp: tbqh most of my stackups vary over the board surface
01:29 PM exp: almost everything i do has decent power sections on it, 10A+
01:29 PM twnqx: make sense
01:30 PM exp: so inevitably i'm running power+ground on the outside to take advantage of cooling+copper
01:30 PM twnqx: just do 2oz inside
01:30 PM twnqx: though it will mess with your impedance :P
01:30 PM exp: everywhere else i try to have contiguous ground layers as much as possible
01:30 PM exp: £££
01:30 PM exp: gotta be poolable really
01:30 PM twnqx: yes, that too
01:30 PM twnqx: mh
01:30 PM exp: we're hoping to do some runs soon where it'll be multiple full panels of just our boards
01:30 PM twnqx: poolable only matters for hobby stuff for me
01:31 PM exp: well that's what makes it cheap
01:31 PM exp: and by cheap i mean < £200/board lol
01:31 PM twnqx: money doesn't matter as much for prototyping, and mass production is outside poolable anyway
01:31 PM exp: at the moment we sit in the unfortunate segment of 'just bigger than prototyping' and 'too small to do 4 figure runs'
01:32 PM twnqx: hell i just have a customer that nees to replace the eSIM on our designs with a sim card holder
01:32 PM twnqx: 400€ surcharge. per PCB. for a single flex PCB with the card holder
01:33 PM exp: yes we are doing some hard gold plating on another run
01:33 PM exp: and that adds basically almost the entire PCB price in low volume
01:34 PM twnqx: yeah... our connector edge is hard gold, too
01:34 PM twnqx: guess what it wasn't on the prototypes for that exact reason :P
01:35 PM exp: yep ours neither, that's why i mentioned the vias close to it
01:35 PM exp: i'm trying to take care to ensure they're well away, as we're not filling/capping, just basic mostly-tented
01:37 PM exp: twnqx: today's crappy news for me: a connector supplied by manufacturer does not meet their own measurements
01:37 PM exp: so we are having to drill out a TH
01:37 PM twnqx: O_o
01:38 PM exp: their CAD is correct
01:38 PM exp: their drawing from their website is not
01:38 PM exp: i am sure that i've test fitted it previously and it was fine
01:38 PM exp: so i'm putting £10 on: pandemic has led to them slightly revising the design and didn't bother updating drawing properly
01:38 PM exp: i have no other explanation
01:39 PM twnqx: platis shrinkage.
01:39 PM twnqx: plastic*
01:39 PM exp: in this case it's a pin diameter
01:39 PM exp: on their drawing it explicitly calls out one as being smaller, and gives tolerances
01:39 PM twnqx: hm ok
01:39 PM twnqx: metal expansion? :P
01:39 PM exp: on their CAD it's identical to all other pins
01:40 PM twnqx: hm.
01:40 PM twnqx: did you contact them?
01:40 PM exp: not yet, only discovered a couple hours ago
01:40 PM exp: i have some old stock from a year or so ago, so i'm going to measure mine
01:40 PM exp: there are two different order #s that we use but that's because one is T+R and one is tray
01:41 PM exp: they don't make any other variations of this product either
01:41 PM exp: they might have made a production fuckup too i suppose
01:41 PM exp: not sure anyone would notice a pin being 0.2mm too wide
01:41 PM twnqx: production oopsie :P
01:41 PM twnqx: is it made in china?
01:42 PM twnqx: might just be their QA staff can't enter the country atm
01:43 PM exp: not sure, there'll be a label photo, one sec
01:43 PM twnqx: iirc someone from Würth told me they started to have issues due to that
01:44 PM exp: nope, made in Germany
01:44 PM twnqx: :S
01:44 PM twnqx: still not that surprised.
01:44 PM exp: anyway the pin does not carry any significant current
01:45 PM exp: and the pad annular ring is more than sufficient
01:46 PM twnqx: but it'll break the plating :(
01:46 PM twnqx: reminds me of an eagle issue...
01:46 PM twnqx: my first edition of the carrier board... did short out gnd and +3v3
01:47 PM twnqx: only chirping noises from the SMPS
01:47 PM twnqx: because eagle... does not isolate milling
01:47 PM twnqx: drill, yes. mills, no.
01:47 PM exp: i'm not sure what you mean exactly
01:47 PM twnqx: and a micro USB connector i used uses an oval cutout for stability... that requires milling
01:47 PM exp: it doesn't add clearance around them?
01:47 PM twnqx: yes
01:48 PM twnqx: so.. the inner layers went straight to it
01:48 PM exp: interesting, in kicad they're all considered part of the Edge Cuts layer by default
01:48 PM twnqx: and were plated through even
01:48 PM exp: oh wow, that's a pretty bad fuckup
01:48 PM exp: the fab should have caught that too really
01:48 PM twnqx: eurocircuits :P
01:49 PM twnqx: "we no care"
01:49 PM exp: nah they definitely do
01:49 PM exp: it's literally an extra option
01:49 PM exp: "Copper up to the Board Edge"
01:49 PM twnqx: luckily it was nothing a dremel couldn't fix
01:49 PM exp: they will call you out on it if you do pre-prod approval
01:49 PM exp: you know how eurocircuits analysis works?
01:49 PM twnqx: just cut through three layers from the bottom
01:50 PM exp: it's basically just a hundred randomly trained people sat monitoring a queue
01:50 PM exp: so sometimes they go to absurd time and care and find really obscure problems that might actually affect manufacturing
01:50 PM exp: and sometimes they barely read anything and just send it back to you
01:50 PM exp: it's quite the fun lottery!
01:51 PM twnqx: wow
01:51 PM exp: anyway their old pcb checker was a bit rough in a few areas
01:51 PM qu1j0t3: exp: Heh, that makes sense :<
01:52 PM exp: they updated that a year or so ago
01:52 PM exp: and the new one is pretty solid really
01:52 PM exp: it would absolutely have caught that
01:52 PM twnqx: that was late 2019
01:52 PM twnqx: i think
01:53 PM twnqx: i should still have that pcb stashed away somewhere
01:53 PM exp: yeah 2020 when they updated it
02:20 PM twnqx: https://www.ebay.co.uk/itm/184812839119
02:20 PM * twnqx goes into a corner to cry
02:29 PM exp: ouch
02:29 PM * exp goes for coffee