<Darius>
esden[m]: very soothing apart from the alarm :D
omnitechnomancer has joined #glasgow
<omnitechnomancer>
It does appear quie perplexed when it goes to put down the part and doesn't have it
<esden[m]>
Yeah sorry I will try to make an alarm less video tomorrow. I might have figured out the right frequency and amplitude settings for the connectors 🤞
<esden[m]>
Yeah it is odd that it goes to the placement position after it already fails the alignment check…
<esden[m]>
(Organically grown software is fun 😉 )
<Darius>
esden[m]: it was interesting to see what it did!
josHua[m] has joined #glasgow
<josHua[m]>
"has anyone seen my, uh"
<josHua[m]>
at some point when I was an undergrad in Pittsburgh, PA, there was a tale of some friends of mine who were walking along one evening, when a car stopped at a stop sign next to them, the passenger rolled down the window and asked them "DOES ANYBODY KNOW WHERE THE, UH" and then the driver drove away
<josHua[m]>
which is sort of like the PnP machine going to place an object that it did not pick
notgull has quit [Ping timeout: 264 seconds]
notgull has joined #glasgow
redstarcomrade has quit [Read error: Connection reset by peer]
bvernoux has joined #glasgow
ar-jan has joined #glasgow
jstein has joined #glasgow
<omnitechnomancer>
Does it do alignment checks on things like resistors?
<esden[m]>
You mean if the parts are being checked before they are placed? If so. Yes of course they are. When you look carefully you can see that with each cycle a mirror is flipped down to look at the underside of all 6 nozzles. Each nozzle has a camera on the back side oriented at 90 degrees to the nozzle. Here is a clip showing what the in flight vision is doing.
<esden[m]>
* omnitechnomancer (@_discord_157742445801504768:catircservices.org) You mean if the parts are being checked before they are placed? If so. Yes of course they are. When you look carefully you can see that with each cycle a mirror is flipped down to look at the underside of all 6 nozzles. Each nozzle has a camera on the back side oriented at 90 degrees to the nozzle. Here is a clip showing what the in flight vision is doing.
<russss>
lol, "where did I leave that connector?"
<russss>
does it work out the optimal location of the feeders for you?
<esden[m]>
There is an optimizer built into the machine software. It does an ok job. I think it could be better. And there are very expensive offline tools to do a better job. I sometimes manually massage the feeder placement. But it does try to place the feeders so it can pick up multiple parts in one go as often as possible. The heads are exactly two 8mm tape feeder slot far apart.
<esden[m]>
I also “gave” the project 4 x 100nF feeders. It speeds up the assembly by almost 30s per panel. Also I don’t have to swap the reels as often.
galibert[m] has joined #glasgow
<galibert[m]>
nF?
<esden[m]>
0.1uF …
<esden[m]>
The most common capacitor there is? 😉
<galibert[m]>
ahhhhhhhhhhh
<galibert[m]>
I thought it was hardware you added, not just component lines
<galibert[m]>
well, it's hardware, but I think you get what I mean
<esden[m]>
Usually you only set up one feeder per BOM line. But for some really high count parts it makes sense to set up multiples.
<galibert[m]>
Yeah, it sure makes sense
thechickenundert has joined #glasgow
<thechickenundert>
Any updates now that feb 29th passed?
<josHua[m]>
I wonder if one could reasonably encode that optimization problem into z3
<whitequark[cis]>
which
<galibert[m]>
Pick&place I think. Scroll up a little
<whitequark[cis]>
ahhh
<tpw_rules>
isn't it almost literally traveling salesman
<duskwuff[m]>
traveling salesman but even more complicated because you're picking up parts on multiple heads, some of which may be interchangeable with each other
<esden[m]>
Oddly the first thing i want to address is reverse engineering the file format that the "programs" are stored in. As there is no way to manipulate and write part descriptions, feeder locations and parameters otherwise. The import file format only supports part positions, fiducials and panel arrangement.
<esden[m]>
When that is done I can consider playing with a better step and feeder optimizer.
<esden[m]>
But it would indeed be nice to improve on what the machine comes with.
<josHua[m]>
reverse engineering file formats and writing optimizers definitely is one of those things where you go 'wait, I thought I was buying a tool, not a project'
<josHua[m]>
but I guess you knew you were buying a project as soon as you saw the words 'pick n place'
<esden[m]>
sigh ... yeah exactly
<esden[m]>
I did ask them for a spec, but they: "haha, it is proprietary, nope"
<josHua[m]>
'not for long'
<esden[m]>
me: "but this vendor you offer software for can write it" ... they: crickets
<esden[m]>
s/for/from/
<esden[m]>
The files seem to be some derived thing from the access database file format. The part databases are definitely access, and I can read them with the mdb tools.
<esden[m]>
In any case we should move the further discussion about this to an appropriate channel. 😄
<esden[m]>
There are a ton of different optimization dimensions. You have to think which nozzles you have loaded onto the heads, and when. Which order you place the parts. Where you put the feeders, and and and and.
<esden[m]>
I have to improve the back side stencil and move the apertures of the two caps further apart. These two live to hug each other and create a short. :/
<galibert[m]>
If you need help with RE, I do have some experience on the topic
<esden[m]>
@galibert thanks very much appreciated. I will gather some materials and information and make it available so we can collaborate on figuring it out.
<galibert[m]>
Cool, let's do that :-)
<esden[m]>
But first, more glasgow assembly 🙂
<galibert[m]>
Yay!
<galibert[m]>
One day I'll have mine mine mine!
<Wonka>
yeah, still patiently waiting for mine too :)
<esden[m]>
yesterday I managed to assemble 56 top sides. Let's see how many I will manage today. Considering that I am reworking a bunch of boards today it will likely be less units. I am trying to get into a routine. We have about 100 units SMD assembled and 20 fully assembled so far. Still figuring out some inefficiencies and ramping up production speed.
<esden[m]>
So far it seems like I am able to do ~200 a week? ... I hope we can do better than that.
<esden[m]>
I want to ship a few boxes of glasgows to Mouser this week, and write a way overdue backer update. Wish me luck.
<galibert[m]>
there something that put solder paste before the P&P?
<esden[m]>
No, not on this machine. There are machines that have jet printers or other funky stuff. Some have dispensers for glue. But a dedicated stencil printer is a much more common solution.
<galibert[m]>
that's what you have, a stencil printer?
<esden[m]>
I don't have a fully automated one. Even though it would be nice to have on. I used just a desk and a stencil for ages. It works much better than you would expect. Definitely better than the cheap stencil printers that you can find dime a dozen on ebay. Recently I got a nice Essemtec manual stencil printer donated by tomkeddie (@_discord_574338939620556811:catircservices.org). It is indeed an improvement in repeatability, but I would not be
<esden[m]>
able to justify the price.
<esden[m]>
s/on/one/
<galibert[m]>
Nice. And after the P&P it goes into an oven I guess?
<galibert[m]>
It's very interesting to see how semi-industrial production happens
vegard_e[m] has joined #glasgow
<vegard_e[m]>
individual solder paste dispensing is likely too slow for any sort of volume, and I don't have the impression it's very widespread in prototyping either
<tpw_rules>
stencil printers are not a good deal?
<esden[m]>
They only make sense for really small batches on a limited selection of footprints. Screaming circuits have a jet printer, and they told me they rarely use it, and it does not work for BGA at all.
<tpw_rules>
oh, it cannot print arbitrary shapes?
<esden[m]>
tpw_rules (@_discord_285821454715715585:catircservices.org) you are better served with some scrap pcb, a table and a stencil taped down to the desk. The cheapo stencil printers are rickety, don't have good back side support for the pcb, paste squeezes everywhere under the stencil. It is just YUCK.
<tpw_rules>
do you order the stencils out?
<tpw_rules>
i've done that approach before but the stencil making step is kind of annoying
<esden[m]>
Yeah, I just order them from the PCB manufacturer or from OSH Stencils.
<vegard_e[m]>
I remember a university I visited once had a manual pneumatic solder paste dispenser, but that was a decade and a half ago, doubt it sees much use nowadays
<esden[m]>
I definitely don't make my own stencils. Not worth the time and effort.
<tpw_rules>
i've gotten stung with ridiculous shipping costs before
<esden[m]>
I really value the cheap metal stencils one can get these days. Definitely not worth "saving" on that.
<vegard_e[m]>
I figure advances in fiber laser tech has made stainless steel stencils a lot more affordable since then
<tpw_rules>
maybe i should look into that...
<vegard_e[m]>
I remember using kapton stencils from some place…
<esden[m]>
tpw_rules (@_discord_285821454715715585:catircservices.org) if you make the mistake not to adjust the stencil size when ordering ... 😉
<tpw_rules>
i've gotten kind of a ridiculous budget for spending on pcb manufacturing equipment nobody is gonna use...
<tpw_rules>
so far it's all going to a vapor phase soldering system
<whitequark[cis]>
i once built a vapor phase soldering system using a 50 HKD hot pot and a PID controller
<whitequark[cis]>
worked just fine, as long as you didn't lift the lid of the pot until it was done cooling
<tpw_rules>
that sounds dangerous to be in the same room as. not that the existing system wouldn't be
<esden[m]>
I still use this piece of junk to do Glasgow back side. But would not recommend buying one.
<whitequark[cis]>
naaaah it's fine
<tpw_rules>
what i really want is a manual pick and place
<whitequark[cis]>
don't breathe so much of the galden 230 that it condenses in your lungs and you'll be ok
<tpw_rules>
hook up a mouse and a hi res camera to a 3d printer or somthing and just click and drag
<whitequark[cis]>
tpw_rules: these exist
<whitequark[cis]>
oh, like that
<whitequark[cis]>
i guess these exist too but the utility seems lower since it's kind of expensive
<tpw_rules>
i mean if it has manual lead screws or something then that could work too
<whitequark[cis]>
why not just an XY gantry?
ar-jan has quit [Ping timeout: 272 seconds]
<tpw_rules>
idk what you mean by that
<tpw_rules>
manual or motorized?
<whitequark[cis]>
like a carriage on two guiderails that lets you maneuver a part without dropping it
<tpw_rules>
(also i think the real answer is the nozzles)
<tpw_rules>
can i buy a good set of those
<russss>
I find the laser cutter/transparency film trick is pretty decent for stencils if you're just doing a handful of boards (and have access to a laser cutter) but most of the time these days I just get China to do the assembly
<esden[m]>
I am not sure how much it actually helps for the few prototype assemblies.
<tpw_rules>
it has some smarts?
<vegard_e[m]>
no, it just stabilizes
<tpw_rules>
well it can apparently lock or something
<tpw_rules>
what's the price?
<tpw_rules>
oh, found it
<tpw_rules>
(1995 eur)
notgull has quit [Ping timeout: 264 seconds]
<vegard_e[m]>
I'm also not sure how well it works, the one at uni was just sitting in a storage room and we didn't bother grabbing it and setting it up when we were assembling some boards
<tpw_rules>
ah
notgull has joined #glasgow
<omnitechnomancer>
Is a stencil printer here a device used to apply solder paste through a stencil, or a device used to make a stencil?
<vegard_e[m]>
seems to me it'd only be worthwhile if you'd be hand assembling a lot, and I'm not sure why you'd want to be doing that…
<esden[m]>
I feel it is more hassle than it is worth. Unless you have major tremors that you need to work around.
<vegard_e[m]>
the former
<tpw_rules>
mine are usually solved by lunch
<tpw_rules>
but can get annoying :P
<whitequark[cis]>
it would've really helped me back when i was on TCAs and couldn't solder a single 19 AWG cable in half a hour of work
<whitequark[cis]>
(don't do antidepressants, kids)
<omnitechnomancer>
Thanks, the name is somwhat ambiguous :/
<tpw_rules>
i was hoping a stencil printer was like an inkjet thing which just squirted out paste in the right place
josuaH[m] has joined #glasgow
<josuaH[m]>
Interesting how it all scales down to a manual tool!
<tpw_rules>
but i guess the name does not make sense for that
<vegard_e[m]>
that would be the jet printer esden (@_discord_269693955338141697:catircservices.org) mentioned earlier
<galibert[m]>
a 3d-printer-like would make more sense than an inkjet mechanism though, no?
<russss>
yeah "jet printer" but they sound like a lot of hassle
<esden[m]>
any regular production line just uses automated stencil printers. They work, the stencils are not that expensive in the scope of things. And they are faster than a jet printer. Unless you try to make an all in one prototype building machine. (and yes Essemtec seems to be the main one doing this)
<russss>
or you can always bolt an air-driven syringe onto a 3D printer, people were keen on doing that back when they were convinced they could make the reprap print itself
<vegard_e[m]>
probably need to heat it to get the viscosity right too
<omnitechnomancer>
might be helpful for aligning BGAs?
<esden[m]>
People underestimate how much effort it takes to do pcb accuracy stuff. A 3D printer gantry is not enough.
<esden[m]>
Listen to the liteplacer folks... they finally realized that the hard part is not the xy movement... but the vision...
<galibert[m]>
vision is hard?
<SnoopJ>
it's always harder than you think it is
<russss>
yeah and the cost of Chinese PCBA is getting so cheap that it feels like not worth even bothering now
<russss>
but stencils are always the best solution for paste
<esden[m]>
Yeah, calibrating everything out and making vision reliable is quite hard.
<vegard_e[m]>
we've got a local company setting up an assembly shop nearby, my work used it to run a prototype series recently
<esden[m]>
Plus, in the current machines the speed you need to do vision is pretty stunning too. Did you see my earlier video I posted?
<galibert[m]>
yeah, the speed is impressive
<vegard_e[m]>
board has a nrf52 in one of those awful aqfn packages, and when I ran a verification on the board I noted one of the core decoupling caps didn't have a voltage, so I noted that as a likely bad solder joint
<esden[m]>
Each vision alignment per part has a dozen filters and vision steps it does. (there is a debug mode to enable capturing all the picture steps, I should get them one day and post it...)
<galibert[m]>
cute
<esden[m]>
Another issue with 3d printers is how flimsy they are.
<vegard_e[m]>
so the guy went and got the rest of the series xrayed, found that there were solder issues on 9 of 10 boards or something like that, shit's hard to get right
<galibert[m]>
But you just need to use AI! ;-)
<vegard_e[m]>
then he sent me the one board that didn't show issues on xray and that still showed no voltage on the same decoupling caps, turns out that in the last chip revision that pin is now internally disconnected and no longer used…
<esden[m]>
I need an xray 😭
<vegard_e[m]>
I've considered buying one of those dentist xray things you can find on aliexpress
<vegard_e[m]>
looks like you can get an xray source and digital sensor for $1k or so
<russss>
not really something I'd DIY
<galibert[m]>
Try not to xray yourself too much
<russss>
occasionally you can find a bargain on a Faxitron or similar enclosed machine
<esden[m]>
fixaitrons are not powerful enough unfortunately
<russss>
we have a 40kV one, it does struggle a bit with thicker boards
<esden[m]>
ok, a friend of mine was recommending to get something stronger than that to get good results
<russss>
yeah, we mostly have this because it was too cheap to turn down
<russss>
but it is fun
<russss>
the medical/parcel stuff tends to top out around 40kV and so anything more powerful is much harder to find
<galibert[m]>
Catherine: So *that* is why you glow in the dark
<whitequark[cis]>
i know many people who did this and only one (who i went to high school with) ended up in the hospital after xraying his hand wayyyy too many times
<tpw_rules>
what could the hospital do? xray it again to check for damage?
<russss>
lol
<whitequark[cis]>
it was a closed specialized hospital that was mainly serving local nuclear industry workers
<whitequark[cis]>
so they had experience with radiation damage
<vegard_e[m]>
the thing I'm most worried about with that xray thing is how horrible the software for the digital sensor is going to be…
<whitequark[cis]>
RE the hardware :3
<whitequark[cis]>
plenty of ppl doing that too
<tpw_rules>
mikeselectricstuff did that long ago but did not get too far
<tpw_rules>
but it was half-dead when he arrived
<tpw_rules>
s/he/it/
<russss>
last I heard mike's machine is dead now but it is similar to our faxitron and we sent him the disk image
<tpw_rules>
i mean the xray cassettes
<russss>
oh yeah he has those too
<omnitechnomancer>
gotta compute some tomography!
<galibert[m]>
Heh, one of my colleages did his PhD on tomography image reconstruction
<sys64738_2574[m]>
re: xray: isn't there this one hackaday project that's basically DIYing an xray scanner?
<josHua[m]>
do you have one of those process calibration thermocouple things for your oven that you can run through on the conveyor and get a log?
<josHua[m]>
if so, you're better than former-client's CM...
<esden[m]>
josHua (@_discord_256468461818085377:catircservices.org) ugh not at the moment… there is a long story attached to that
<esden[m]>
I hope soon to have one … I just need to find some time 😭
<josHua[m]>
we were having a bunch of failures in the field that we diagnosed to be cracks in BGA pads, from dye and pry testing, and we were like 'what is the solder profile' and they gave us a nice graph
<josHua[m]>
and we were like 'no, what does it measure at', and they were like 'oh. idk, our box went off for calibration like a year ago and we haven't heard back, and the other one is broken'