<electronic_eel>
whitequark: what do you plan to do with the AUI board?
<electronic_eel>
do you have some device that needs AUI?
<whitequark>
i have a MAU
<whitequark>
for 10BASE-T unfortunately
<whitequark>
10BASE5 would be more fun
<whitequark>
thiccnet
<whitequark>
well strictly speaking i also have an absolutely *massive* ISA card that can use a MAU and also some other interfaces
<whitequark>
but i suspect that even had i had any device with ISA i would not be able to get it to work
<electronic_eel>
do you just want to use a MAU for kicks or is there a deeper reason behind?
<whitequark>
i made Glasgow for kicks
<whitequark>
and because i have ADHD
<electronic_eel>
I think I somewhere still have a 10Base2 MAU
<electronic_eel>
and a hub with 10Base2 and -T
<whitequark>
nice
<electronic_eel>
i actually used the hub sometimes, to show my apprentice at work how sniffing packets with a hub worked
Getorix has joined #glasgow
Getorix_ has quit [Ping timeout: 265 seconds]
Getorix_ has joined #glasgow
Getorix has quit [Ping timeout: 265 seconds]
ali_as has joined #glasgow
<Stormwind_mobile>
"I made Glasgow because I have ADHD" is kind of a deep statement.
Getorix_ has quit [Ping timeout: 265 seconds]
<Stormwind_mobile>
Speaking of review...
<Stormwind_mobile>
If anyone of you designs a PCB, do you do it all on your own? Do you send it out to someone for peer review? For professional work, I would guess this would be the norm, that either a colleague look at it, or you pay someone to review it.
<Stormwind_mobile>
That's not so practical for hobby work.
<tnt>
I usually just do it all on my own.
<tnt>
The few I did for work as well because ... we were not an EE firm so noone else could have reviewed it anyway.
<whitequark>
there are pcb checklists
<Stormwind_mobile>
I guess that works if you are experienced enough. My last go at PCB design has been more that ten years ago, and I have never sent in files to a fab before. I'm kind of worried messing something up due to rookie mistakes.
<Stormwind_mobile>
Not so much because of the money, but because of turnaround times...
<whitequark>
oh, don't worry
<whitequark>
even very experienced designers often have a bodge wire or two on revA PCB
<whitequark>
it used to be that people would actually do *production runs* with bodge wires
<Stormwind_mobile>
Or 600...:-D
<whitequark>
well sure, it depends on your experience level and carefulness and such
<whitequark>
but it's normal
Getorix has joined #glasgow
<whitequark>
glasgow revA had a place where I accidentally inverted polarity of a pin
<Stormwind_mobile>
Was looking at Dave Davill's D20 bodgefest. That looked like some serious punishment.
<whitequark>
D20?
<Stormwind_mobile>
The LED icosahedron
<Stormwind_mobile>
RPG nerds often times refer to that as a 20 sided die, a D20
<Stormwind_mobile>
Regular die is a D6
<whitequark>
oh, right
<whitequark>
i know what a d20 is
<whitequark>
didn't know the context
<Stormwind_mobile>
Oh, my bad.
<Stormwind_mobile>
Whatever, those photo stories made me itch trying some versions of blinky LED stuff on my own.
<tnt>
Stormwind_mobile: you can also just post your files somewhere (twiter / relevant irc / discord / whatever ...) and see if anyone can have a quick look.
<tnt>
(also, it's Greg Davill, not Dave :p)
<_whitenotifier-3>
[glasgow] PTS93 opened pull request #178: Fix formatting issue on BOM revC1 line 21 - https://git.io/Jvk7D
<Stormwind_mobile>
Oops 😅
<JJJollyjim>
Does Dave Davill do D20 development? Definitely!
<awygle>
Stormwind_mobile: everybody feels that way. the best advice is just to do it, but that's easier said than done.
<awygle>
i usually either get someone to review it or take a couple days between 'this is done' and 'time to fab', then look at it again with fresh eyes.
<Stormwind_mobile>
If it was just dependent on fresh eyes, and not on what to look for in the first place 😅
<Stormwind_mobile>
Well, I'll figure something out. I have a list of five or six projects of varying complexity that I can grind my teeth on.
<Stormwind_mobile>
I don't know if the world needs yet another Arduino programmable platform, but I think I'll try anyway.
ExeciN has joined #glasgow
ZirconiumY has joined #glasgow
ZirconiumX has quit [Ping timeout: 252 seconds]
ZirconiumY is now known as ZirconiumX
ExeciN has quit [Ping timeout: 272 seconds]
parataxis has joined #glasgow
Stormwind_mobile has quit [Ping timeout: 260 seconds]
ali_as has quit [Remote host closed the connection]
ali_as has joined #glasgow
<kc8apf>
First generation PowerMac G5 shipped with a carefully calibrated bodge wire to work around a timing issue in the U3 northbridge. You do your best and ship what works.
Stormwind_mobile has joined #glasgow
<whitequark>
nice
<kc8apf>
not sure my friend ever really got compensated enough for the months of long days he did to figure out the exact length required
<tnt>
I'm trying to understand some part of glagow platform code. In the resource of revAB, there is Resource("port_a", 0, Subsignal("io", Pins("45"))). What's the Subsignal for ? What's the difference with just Resource("port_a", 0, Pins("45")) ?
<sorear>
it was the same length on every board?
<whitequark>
tnt: look at revC
<whitequark>
that also has oe
<tnt>
Yeah, I've seen that. So the 'Resource' line defined the platform IO port, then you bind "sub signals" of that defined IO port to different pins, and 'io' is the full tristate instance and 'oe' is just the output enable. But then in revAB that subsignal is redudant and could be omitted and is just there for consistency with revC ?
<tnt>
Ok, no nevermind, I'm completely wrong.
<tnt>
I see what it does, it create a hierachy of signals in the resource.
<tnt>
I was wondering how just specifying the resource like that dealts with the revAB / revC difference ... and the answer is it doesn't.
<tnt>
The magic dealing with that is in build_pin_tristate() with "if hasattr(pin_parts, "oe"):"
<whitequark>
yeah
<whitequark>
btw this part of glasgow logic is actually kinda really bad
<whitequark>
i suggest not emulating it if possible
<whitequark>
it's what's inspired many of nmigen platform changes wrt migen
<whitequark>
except now it's stuck in this weird limbo because fixing this properly requires also fixing every applet
<whitequark>
eventually
<awygle>
Not quite a bodge wire, but I had a design where a transistor would fail short and kill the whole thing if connected to 18" or 24" of coax. We ordered a custom 21" cable and it worked fine.
<sorear>
Standing waves?
plaes_ has quit [Quit: Reconnecting]
plaes has joined #glasgow
plaes has joined #glasgow
<tnt>
whitequark: I'm trying to see how easy/hard it is to get it to build a bitstream for a given applet on a different platform so I kind of have to emulate what applets expect to find.
<awygle>
Yes. Had to locate the amp at a node instead of a peak.
ExeciN has joined #glasgow
<tnt>
ATM I'm a bit lost trying to figure out how a CompatModule can define which clock signal to use for itself and all its submodule (when that clock comes from an Instance inside that module).
<Stormwind_mobile>
awygle: that's kind of what my dad warned me about when he showed me CB radio stuff as a kid. 8-year-old me didn't understand why the power amplifier should burn out when I was messing about with the standing wave meter or not connecting the antenna properly 😅
ExeciN has quit [Ping timeout: 268 seconds]
<awygle>
oh god switching from solvespace "move me around the workspace" conventions to kicad conventions is like having a stroke
<awygle>
whitequark: reviewing now. anything in particular you want checked?
<electronic_eel>
whitequark: was also having a look at the aui board. unfortunately only schematics, layout seems to require a prerelase-5.99-kicad which I don't have
<electronic_eel>
I suggest to use caps in 0805 for the stepup (C7 and C9), because especially the output cap will have a much lower capacitance than the nominal 1µF if biased to 12V
<electronic_eel>
larger footprints aren't that much affected by this
<electronic_eel>
and maybe a dip jumper instead of solderjumper for JP1? you have enough space for that.
<awygle>
i agree with that. i'm a bit worried about the LDO with ceramic caps on the input and output, it's not mentioned in the datasheet whether that's OK.
<awygle>
no discussion of stability at all afaict
<electronic_eel>
ah, 1117, yeah, these older models can be affected by that. what is the expected current draw?
<electronic_eel>
oh, and the 5v are taken from glasgow, right? so they are not proper 5v, but usb vbus, which can be substantially lower. the dropout of the 1117 may be too high for that
<awygle>
the rest of it looks fine. i didn't double check your impedances or anything since i don't have a stackup, , but the traces are short enough that's unlikely be a problem. maybe don't run CO-A through DO-A so close together on the far right side, since you have room to spread them out a bit for less coupling.
<electronic_eel>
how about MCP1755S-3302E/DB instead of the 1117? they have nice foldback characteristics for overload, are stable with ceramics and just need 0.3v dropout
<awygle>
even in worse case probably not more than 400 mA total draw on that regulator
<Stormwind_mobile>
electronic_eel: what's that comment on 0805 and biasing?
<Stormwind_mobile>
Current footprint is smaller or larger?
<electronic_eel>
are the 400mA in short circuit limiting or in regular operation?
<awygle>
that would be a pretty exceptional condition
<awygle>
nominal might get up to 200 mA or so
<awygle>
based on a cursory analysis
<electronic_eel>
Stormwind_mobile: current footprint ist 0603 1µF, in my experience there is not much left of the 1µF at 12V
<awygle>
i'd probably want something beefier than an MCP1755 but it'd almost certainly be fine
<awygle>
people over-worry about capacitor DC bias effects, but as this particular capacitor participates directly in a sensitive analog control loop, this is a good time to worry about it :p
<electronic_eel>
awygle: the MCP1755 has nice overload characteristics with foldback, in case of short circuit this even helps. In regular use with up to 200mA it will perform fine
<awygle>
the short circuit current will be limited to 137 mA by the series resistors anyway, so i spose it's fine
<awygle>
(per driver, so 2x that total)
<electronic_eel>
right, that is below the nominal 300mA of the MCP1755, so it's fine
<awygle>
mhm
<Stormwind_mobile>
Interesting, thanks! One assumption we make is that capacity is a constant and don't worry about it. Didn't know that!
<whitequark>
MCP1755 is sadly not sold in russia
<whitequark>
and everything else is
<whitequark>
re CO-A/DO-A: ack
<awygle>
if you want to keep the LD1117, slap a parallel resistor footprint across C5
<Stormwind_mobile>
So do people abuse that for frequency modulation? That behaviour reminds me of a varicap diode...
<electronic_eel>
why do they think russians don't want nice regulators?
<awygle>
DNP it and then you'll have it if you need more ESR
<awygle>
err not parallel
<awygle>
series, and pop it as 0ohm
<awygle>
(it's in the goddamn name awygle, sheesh :p)
<whitequark>
the total current is limited by Glasgow
<whitequark>
the dropout vs Vbus is a valid point
<whitequark>
re footprint: ack
<whitequark>
thanks everyone!
<electronic_eel>
if you get stability problems with the regulator, they may only show up at certain loads. So bringup requires a bit of work with different loads on 3V3.
<awygle>
Stormwind_mobile: not usually, it's not typically very well controlled. i suppose you could if you put feedback around it. usually sensitive applications try to use C0G caps if possible, which are better on all axes.
<electronic_eel>
I suggest a proper testpoint for 3V3
<whitequark>
hm, alright
<awygle>
(except capacity per volume)
<awygle>
or electrolytics which have different problems but not DC bias
Stormwind_mobile has quit [Ping timeout: 265 seconds]
<electronic_eel>
whitequark: does the stuff need to be availably locally in russia or is china ok too? lcsc has mcp1755s in stock.
Stormwind_mobile has joined #glasgow
<whitequark>
former
<whitequark>
I don't have a good way to handle logistics like this and don't want to do it for this board
<whitequark>
ordering anything from anywhere takes fucking forever
<whitequark>
pcbway sent me some PCBs with EMS a month ago, still not arrived
<electronic_eel>
stuck in customs?
<electronic_eel>
whitequark: maybe add power leds for vp and 3v3, so you can directly see if there are any issues
<whitequark>
electronic_eel: no, this is normal
<whitequark>
it's slow.
<electronic_eel>
oh, another point: you plan to set vio of glasgow to 5v to power the regulator. but the output of the transceivers will be 3v3
<whitequark>
sure
<electronic_eel>
the level shifters on glasgow have their reference set to 5, so giving them 3v3 as high is not best practice
<whitequark>
hm. crap.
<whitequark>
maybe i should just drop the boost
<electronic_eel>
the 5v line on the planned addon interface on revD has it's uses...
<whitequark>
won't help this one
<whitequark>
(since it's not on ribbon
<electronic_eel>
I don't know the aui specs, is 12v required?
<whitequark>
yes, it needs at least 12V
<whitequark>
up to 15 or so
<electronic_eel>
so no boost and you need external power?
<whitequark>
yes
<electronic_eel>
another idea would be transceivers for 5v and keeping the boost
<electronic_eel>
would also get rid of the 1117
<whitequark>
not many options of available transceivers
<whitequark>
that can do >10 Mbps
<whitequark>
and if I have to figure out how to get parts from Mouser and wait a month I'd rather just not do this board at all
<electronic_eel>
doesn't ti have some?
<whitequark>
not available
<awygle>
can you take in 3.3 from Glasgow and boost to 12v from there?
<whitequark>
AUI requires 6 W, I can already do at most 1 W in ideal conditions
<whitequark>
dissipating a third of that on the glasgow LDO seems dumb
<whitequark>
might as well just drop the boost
<electronic_eel>
yeah, dropping the boost seems to be the way to go with these kind of restrictions
<whitequark>
should probably do a screw terminal instead or something
<electronic_eel>
what is wrong with the dc jack?
<whitequark>
hard to chain
<whitequark>
there are multiple incompatible definitions
<whitequark>
12V can be 5525 or 5521
<electronic_eel>
screw terminals allow even more kind of wrong connections...
<whitequark>
sigh. forget it. i'll just not do this board.
Stormwind_mobile has quit [Ping timeout: 268 seconds]
<whitequark>
this was supposed to be quick and i already spent 10 hours on it, not counting what i'll need to spend after this discussion
<electronic_eel>
don't ask for review then ;)
<whitequark>
i won't
Getorix_ has joined #glasgow
Getorix has quit [Ping timeout: 240 seconds]
<awygle>
sorry :/
<whitequark>
hardware is hard enough to make without having to satisfy my perfectionism and also everyone else's
<awygle>
that's certainly relatable
Stormwind_mobile has joined #glasgow
Stormwind_mobile has quit [Ping timeout: 265 seconds]
Stormwind_mobile has joined #glasgow
katharina has joined #glasgow
<katharina>
Hi, I am planning to assemble the Glasgow board on my own. Does anybody have any positive experiences with board manufacturing services in the context of the Glasgow? :)
<sorear>
I don’t have relevant experience, but
<sorear>
most of the problems I’ve heard of have been shipping related and so location would matter
<katharina>
Germany/Europe in my case,
<tnt>
katharina: just the PCB or assembly ?
<Twix>
katharina, i ordered the boards at jlcpcb. The quality is quite good
<katharina>
JLCPCB sounds good. Are there also providers that assemble it for you?
<electronic_eel>
assembly just for one board or a small batch (like 10 pcs)? just one board will be quite expensive
<katharina>
I was planning to build two, so yeah, I kinda figured it would be expensive if done is such small quantities.
<electronic_eel>
I remember someone here on irc had a quote from eurocircuits, IIRC it was about 200 EUR for a batch of 10. I think this is quite good
<electronic_eel>
200 EUR per board of course
<tnt>
yeah, that's probably good price for such low qty.
<tnt>
(not that it was probably 200 EUR excluding vat)