<whitequark[cis]>
it's a shame the design files for the case aren't available; I think we should add the drawings to the repository at least
<whitequark[cis]>
it's unclear how big the "ridge" on the case bottom is, but that seems to be 3.0 mm or so?
<whitequark[cis]>
in that case I think it might be insufficient to have this much keepout, since the board play is more than .4mm for sure
<whitequark[cis]>
and the only thing stopping SDA from being shorted to 5V is the anodizing
<whitequark[cis]>
for the main connector, most of those pins are GND, so we'll never see it, except for VIO, which will safely short and isn't an issue (other than a functional problem; not a safety issue)
<whitequark[cis]>
but the prospect of getting 5V on the SDA pin is really concerning if that was really what was happening
K900 has joined #glasgow
* K900
is suddenly very happy he 3D printed a case
<electronic_eel>
not good
esden[m] has joined #glasgow
<esden[m]>
the movement tolerance is about .2mm, as we hame a combined tolerance of .1 all around the board paremeter.
<esden[m]>
hitting that test pad is very unlikely
<electronic_eel>
i measured resistance on my early prototype case - the anodizing was quite good, except where the lasered text was on the top
<esden[m]>
ohh well I guess there is also the lip
<electronic_eel>
you mean the edge of the lip could be conductive?
<esden[m]>
unlikely... just thinknig of the metal overlap in general
<esden[m]>
if there is some conductive debree coaught between the board and the pcb it could pierce through the anodization
<whitequark[cis]>
yeah, that is what I think might have happened?
<esden[m]>
s/coaught/caught/
<whitequark[cis]>
it was really odd, the symptom is that all i2c register accesses fail with "no such register"
<whitequark[cis]>
with known good gateware that worked a minute ago
<whitequark[cis]>
and then just the uart applet failed too, after multiple tries with selftest leds
<electronic_eel>
whitequark[cis]: do you see any scratch marks or similar around the area where the test point is?
<whitequark[cis]>
no
<electronic_eel>
i mean in the case
<whitequark[cis]>
and we weren't able to reproduce it
<electronic_eel>
perfect, i love bugs like this
<whitequark[cis]>
even by screwing things quite hard, manipulating the board while in the case, etc
<whitequark[cis]>
but the failure was very real and it happened during normal use
<whitequark[cis]>
in fact it interrupted me writing a GPIB applet at last
<whitequark[cis]>
electronic_eel: hehe :)
<electronic_eel>
so it could also be something else related to the case or mounting
<whitequark[cis]>
but there is a 5V testpoint inline with the SDA testpoint
<esden[m]>
I will try to find some time in near future and export the case outlines in a way that we can import it into the KiCad files.
<whitequark[cis]>
so if anodization fails under those two but not elsewhere in the case, the case will short the two TPs
<esden[m]>
So that we can see what the issue is.
<whitequark[cis]>
of course it's much more likely the case will be grounded, but that's not completely reliable either, I think
<electronic_eel>
whitequark[cis]: did you do some rework or similar to your glasgow pcb? it could also be some bad solder joint or similar introduced by that and mechanically moving it out of the case "fixed" it
<whitequark[cis]>
nope, as shipped by esden
<esden[m]>
We might need to ship some stickers on the test pads in the future.
<esden[m]>
or some masking kapton tape to prevent this short
<whitequark[cis]>
yeah
<esden[m]>
for now I will add to my todo list to figure out where the possible overlap is
<electronic_eel>
yeah, i think kapton tape would be the easiest solution i think
<whitequark[cis]>
thank you!
<esden[m]>
sure, I hope we can find a workable solution. And I will adjust what I can for the next case production batch.
<electronic_eel>
i did some testing regarding the case and esd when i got it. like zapping it with a piezo from a ligther and similar things. but i couldn't trigger any crashes or data errors this way
<electronic_eel>
but i didn't take a look at the testpoints and their clearance
<whitequark[cis]>
we didn't until just now either
<electronic_eel>
hmm, the testpoints aren't the only way this could happen. there are also a bunch of components which are connected to the relevant nets. so there could also have been some conductive debris (like a stray blob of solder) between a component and the case
<whitequark[cis]>
yes, potentially
<whitequark[cis]>
but testpoint clearance is one obvious option (though made less likely by our inability to reproduce it)
<electronic_eel>
yeah, especially if you try to wiggle the pcb in the right direction or apply pressure at the correct spot and it still isn't reproducible
<whitequark[cis]>
yep
<electronic_eel>
you could also put a bit of solder onto the testpoint, like half a millimeter or something and then try again
<electronic_eel>
if all this doesn't do it, then i'm a bit in doubt if it is really the testpoint and more think it was another spot
<_whitenotifier-f>
[GlasgowEmbedded/archive] whitequark 6d45daa - G00085: IEEE488.1™ IEEE Standard for Higher Performance Protocol for the Standard Digital Interface for Programmable Instrumentation