<steakpizza>
nice! What are the undocumented parts which is required to get the partial reconfigurability working without vivado. I'm assuming using this vivado flow with prjxray, one can reverse engineer it?
steakpizza has quit [Quit: Page closed]
steakpizza has joined ##openfpga
X-Scale has quit [Ping timeout: 248 seconds]
cranc has quit [Quit: Page closed]
<kc8apf>
just generating partial bitstreams in general
<kc8apf>
currently, we can generate them from a very simple language (FASM) that sets individual configuration bits
<kc8apf>
see minitests/roi_harness
<kc8apf>
if you're willing to write out the whole config mem that way, you're good to go
<kc8apf>
right now, we have only documented the CLBs. Documenting IOBs would be a bit help. Otherwise, building something better than FASM to describe placed-and-routed components and convert them into configuration memory frames would be useful.
<kc8apf>
heading off for the night
<azonenberg>
awygle: So i managed to get this PHY to transmit 1000base-T in master mode with autonegotiation disabled
<azonenberg>
in theory this should mean TX only, no RX
<azonenberg>
and thus no echoes to worry about
<azonenberg>
but i'm having trouble getting a nice PAM5 eye out of it
<steakpizza>
kc8apf: cool cool. i'll bug you again once I do some digging on the things you mentioned
<azonenberg>
I'm falling back to 100baseTX to poke around more and found some bugs in my eye analysis tool
steakpizza has quit [Quit: Page closed]
<azonenberg>
awygle: ok something is wrong here
<azonenberg>
I'm not seeing differential signals, wtf
<awygle>
lol that's not ideal
<azonenberg>
I have probes on B+ and B- of my test rig
<azonenberg>
and i am seeing signals of equal polarity and similar amplitude
<azonenberg>
wtfff
<azonenberg>
gotta look into this more
<azonenberg>
let me sanity check on tragiclaser since i know that works
X-Scale has joined ##openfpga
<azonenberg>
Soooo my initial crosstalk measurements are probably invalid
<azonenberg>
I have to investigate this more
<shapr>
azonenberg: are you building your own gigE thing for work or fun?
<azonenberg>
shapr: I'm building a gig-e switch for fun
<azonenberg>
this is some signal integrity verification work for it
<shapr>
I sometimes work on writing my own SIP stack, but a 1000base-T switch is off into what I consider hardcored.
<azonenberg>
This is what i get with the KSZ9031 into my characterization board
<azonenberg>
something is seriously wrong
<jn__>
on the topic of names, tragiclaser could reasonably be the name of a destined-to-fail *optical* ethernet project :)
<balrog>
azonenberg: btw not that you’re interested but did you see my mention of intel’s 10GBaseT PHY? Probably the only documented one that exists and even it has a control MCU that requires external firmware
<azonenberg>
awygle: all measurements from before are invalid
<azonenberg>
tl;dr I put the magnetics in the schematic wrong (media side to the SMAs and chip side to the RJ45)
<azonenberg>
But i assembled it correctly
<awygle>
O no
<azonenberg>
Which means that the magnetic is in 180 deg off from how it should be
<azonenberg>
the center tap is connected to one of the data pairs
<azonenberg>
etc
<azonenberg>
everything is borked and i'm not actually doing truly differential measurements
<azonenberg>
Rework is possible, but easier to respin
<awygle>
Yeah that's not good
<awygle>
Whoops
<azonenberg>
actually
<azonenberg>
Hold on a bit
<azonenberg>
I'm not so sure
<azonenberg>
ok yeah
<azonenberg>
its wrong
<azonenberg>
Not only is it wrong, the line card i spent so much time on layout for is wrong
<azonenberg>
Welp
<gruetzkopf>
ouch. been there, done that., sadly
ZipCPU|Laptop has joined ##openfpga
steakpizza has joined ##openfpga
Lord_Nightmare has quit [Ping timeout: 248 seconds]
Lord_Nightmare has joined ##openfpga
steakpizza has quit [Remote host closed the connection]
steakpizza has joined ##openfpga
GenTooMan has quit [Quit: Leaving]
futarisIRCcloud has quit [Quit: Connection closed for inactivity]
cr1901_modern has quit [Read error: Connection reset by peer]
steakpizza has quit [Remote host closed the connection]
cr1901_modern has joined ##openfpga
steakpizza has joined ##openfpga
steakpizza has quit [Remote host closed the connection]
finsternis has quit [Ping timeout: 255 seconds]
finsternis has joined ##openfpga
mumptai has joined ##openfpga
finsternis has quit [Ping timeout: 240 seconds]
finsternis has joined ##openfpga
Bike has quit [Quit: Lost terminal]
finsternis has quit [Ping timeout: 276 seconds]
finsternis has joined ##openfpga
rohitksingh_work has joined ##openfpga
DocScrutinizer05 has quit [Disconnected by services]
DocScrutinizer05 has joined ##openfpga
finsternis has quit [Ping timeout: 265 seconds]
sgstair has quit [Remote host closed the connection]
finsternis has joined ##openfpga
finsternis has quit [Ping timeout: 240 seconds]
finsternis has joined ##openfpga
<azonenberg>
gruetzkopf: good news is the shift is only one pin i think
<azonenberg>
So layout shouldnt be hard to fix
<azonenberg>
And, most importantly
<azonenberg>
i found the bug before actually fabbing said line card :p
<azonenberg>
On the plus side both the magnetic and RJ45 footprints are now pcb proven
<awygle>
three categories: PCB Unproven, PCB Proven, and the rare PCB Disproven
<whitequark>
lol
<rqou>
whitequark how do you know so many people with bad relationships with their parents?
<whitequark>
rqou: that question doesn't make any sense to me.
<whitequark>
the question that does is "how come I know anyone at all whose parents weren't scumbags"
<whitequark>
(I'm not sure)
* whitequark
thinks
<rqou>
wtf why so negative?
<whitequark>
because lifie?
<whitequark>
*life
<whitequark>
just... realize that your own relationship with your parents is highly atypical
Ekho has quit [Quit: An alternate universe was just created where I didn't leave. But here, I left you. I'm sorry.]
<azonenberg>
awygle: ok so, fixing the sch
<azonenberg>
Now, labeling the schematic symbol more clearly :p
<rqou>
whitequark: um, IME many people in the Chinese/ABC community have ok/good relationships with their parents?
<rqou>
although unfortunately the behavior that you quote-tweeted from Fiora is also not uncommon
<whitequark>
rqou: then ABC community has its shit together more than usual I guess?
<whitequark>
although I wonder just how much of it is just people keeping things secret for as long as possible and killing themselves when it becomes too unbearable
<whitequark>
optics are misleading
<rqou>
hmm maybe
<rqou>
that is also not unheard of
<whitequark>
rqou: oh btw the failure crate is hitting 1.0 soon
<whitequark>
I'm not sure how I missed that
<whitequark>
there's already an 1.0 version in the repo, it just doesn't seem published yet
<rqou>
hmm nice
<azonenberg>
awygle: going to have to reroute most of the diffpairs
<azonenberg>
i think
Ekho has joined ##openfpga
sgstair has joined ##openfpga
<awygle>
rqou: the behavior quote tweeted from fiora is not indicative of a good relationship with one's parents
<rqou>
yeah, i didn't say everything was all perfect
<awygle>
didn't mean to imply you did. but a lot of people (hello, younger me) don't realize that's not ok
<azonenberg>
awygle: hmmmm
<azonenberg>
it looks like we arent going to be full clamshell
<azonenberg>
there's now a 2-pin offset
<azonenberg>
if we want to keep the passives lined up
<azonenberg>
That should be OK but it does mean a lot of layout has to be redone on the switch
<awygle>
azonenberg: less crosstalk? lol
<azonenberg>
lol maybe? but more to the point, we wont be turning our differential-mode signals into common-mode ones anymore :D
<awygle>
a pita for sure but hopefully you can drag a lot of stuff without redoing it entirely
<azonenberg>
Yes i think i can save the rj45 escape routing and all of the phy stuff
<azonenberg>
just a few things around the magnetics need redoing
<azonenberg>
the board as it stands hooks D+ and tap to the diffpair
<azonenberg>
and terminates D-
<azonenberg>
You can probably see how that ends badly
<awygle>
Indeed
soylentyellow_ has quit [Remote host closed the connection]
soylentyellow has joined ##openfpga
mumptai has quit [Remote host closed the connection]
<azonenberg>
awygle: also 8 in-lb is too much for SMT edge launch SMAs
<azonenberg>
i forgot to mention earlier
<azonenberg>
i had to reflow one of the connectors after i torqued it on :p
<azonenberg>
cracked the solder
<whitequark>
maybe just bad reflow?
<whitequark>
you've said you're on the bottom edge of the profile
<whitequark>
and connectors have high thermal mass
<azonenberg>
whitequark: no, i was turning *very* hard to torque it on
<whitequark>
hm ok
<azonenberg>
i actually hot-aired this board rather than using the oven
<whitequark>
ahh interesting.
<whitequark>
i wonder if it still cracked with leaded
<azonenberg>
Because of the awkward geometry
<whitequark>
oh
<azonenberg>
8 in-lb is a LOT
<azonenberg>
for something that tiny
<azonenberg>
apparently brass SMAs are recommended to be 3-5
<azonenberg>
So i'm gonna try and get a 3 in-lb torque wrench and try that
<whitequark>
azonenberg: random realization: with vapor phase reflow i can freely mix leaded and lead-free
<awygle>
hm i like the idea of just doing "take n" for training
<awygle>
this is sort of cool but it's all gradient descent, seems like it would have serious local-minima problems
<awygle>
then again, i also a) don't know haskell and b) don't know machine learning. so it's hard to see what this is being contrasted against, and how it's better.
<awygle>
would be interesting to see something like this in rust, which i could actually read effectively lol
<pie___>
awygle, well, i guess "usual caveats apply"
<pie___>
i cant read it too well either
<pie___>
well, mostly skimmed, on my <understand later pile>
<pie___>
i think the main thing might not be the actual optimization method but the method of building the model?
cr1901_modern has quit [Read error: Connection reset by peer]
<pie___>
idk what this looks like usually in any other framework
<awygle>
yeah, it talks at the end about how you can use many optimizers
bitd has quit [Quit: Leaving]
cr1901_modern has joined ##openfpga
<azonenberg>
awygle: pushing out the v0.3 respin to oshpark now
<azonenberg>
Fingers crossed this is the last bug :p
<awygle>
azonenberg: yeah i was gonna say lol
<azonenberg>
in retrospect, the attenuators acting funny should have tripped me up earlier
<azonenberg>
i thought it was just that the attenuators were differential mode and the bad magnetics were screwing me up
<azonenberg>
It's funny, this is the first time i've respun a board this much in years if not ever
<azonenberg>
i know i've done two revs of a few designs
<azonenberg>
Three is unheard of
<azonenberg>
and there's no active components or power supply on it
<azonenberg>
it's one of the simplest designs i've done in a long time
<pie___>
awygle, if i was going to quote a specific sentence i think this might be it, but idk "This is the entire crux of this approach, and lets us not only draw from mathematical models directly, but also combine and reshape models in arbitrary ways just by using normal function composition and application, instead of being forced into a rigid compositional model."
<awygle>
i'm sorta relieved that the attenuators aren't bad since i gave you the numbers for them :p
<awygle>
pie___: yeah i guess i'm missing the "rigid compositional model" side of that comparison
<awygle>
well, i clipped this, and five years from now when i have time to learn ML and get annoyed by "rigid compositional models" i'll... have forgotten it existed :p but maybe i'll remember it and discover an alternative
<pie___>
awygle, yeah i asked the author on irc, waiting for reply :D
<azonenberg>
awygle: also
<azonenberg>
i figured out how to make a KSZ9031 send 1000base-T in master mode with autonegotiation disabled
<azonenberg>
as in, single direction PAM5 data
<azonenberg>
Tried to plot a PAM5 eye on tragiclaser and it was pretty unreadable, but i dont know if that's just b/c i used a lower quality magnetic there (cant rmemeber if it was rated for gig)
<azonenberg>
I also dont know how well the clock recovery code in my eye analysis works for PAM
<azonenberg>
I've tested it on NRZ and MLT3
<azonenberg>
It's also possible my scope doesnt have the bandwidth :p
<azonenberg>
(they recommend 1+ GHz for characterizing gig-e)
<azonenberg>
Also TIL, gig-e over copper has FEC
<azonenberg>
The PAM5 is basically PAM4 with an extra symbol thrown in that used somehow for error correction
<awygle>
that's interesting but pretty unsurprising
<azonenberg>
well apparently its not actually a standalone fec layrer
<azonenberg>
the trellis code has some invalid symbols and automatically corrects them
<azonenberg>
but it's basically a fec-included line code
<azonenberg>
Higher-end baseT PHYs are black magic to me, but i can wrap my head around 100baseTX or the optical line codes just fine
<gruetzkopf>
i need to build a HDB3 line code/decoder
<awygle>
I need to build a cheeseburger. I haven't eaten lunch yet.
<gruetzkopf>
(help me, i'm deployinh a new ISDN network in 2018 ;)
<zkms>
what's ISDN?
<genii>
zkms: What there was before DSL
<gruetzkopf>
the integrated services digital network
<gruetzkopf>
before and alongside DSL
<gruetzkopf>
the first DSL technologies were used for ISDN access
<gruetzkopf>
later DSL modes are used for ATM, and current ones for plain ethernet
<awygle>
lol atm
<azonenberg>
my parents internet is 768 x 128 Kbps ADSL
<azonenberg>
running PPP over ATM
<azonenberg>
they may have gotten a speed boost since last time i checked, not sure
<pie___>
azonenberg, do your parents know this
<azonenberg>
i think it was 864 x 160 or something now?
<azonenberg>
pie___: they're just happy it works
<pie___>
how could you do this to them (jk(
<genii>
gruetzkopf: I have some old ISDN endpoint things here from when the call center in our building closed, if they are of any use to you
<azonenberg>
gruetzkopf: there is no tv cable in the house
<azonenberg>
(built in the 1930s)
<azonenberg>
And they're a very long distance from the nearest DSLAM
<genii>
Crap builds up around here faster than i can get rid of it
<shapr>
I know that feeling
<awygle>
mm fed
<pie___>
awygle, build a burger workshop
<pie___>
or, *build-a-burger apparently
<awygle>
bad idea: irc<->mailing list bridge
<pie___>
can we have mailing lists, except irc
<awygle>
just thinking about communication, and what methods i code as "serious"
<awygle>
(email and github)
<shapr>
I almost never use email
GenTooMan has joined ##openfpga
<rqou>
woo, i got to go to my fancy ceremony :P
<rqou>
now i just want my fancy piece of paper to go with it :P :P
<jn__>
rqou: the fancy square hat ceremony?
<rqou>
yes
<rqou>
i assume europeans don't do that?
<jn__>
nice
<rqou>
(iirc i read that germany does not do this?)
<jn__>
i'm not actually sure
<pointfree>
The mortarboard comes from the UK.
<rqou>
i'm aware
<rqou>
Cambridge iirc?
user10032 has quit [Quit: Leaving]
Zorix has quit [Ping timeout: 246 seconds]
<Bike>
doesn't cambridge give you a sword
<rqou>
i thought that was a myth?
<Bike>
it better not be a myth
<rqou>
apparently schools in finland give you a sword
<rqou>
for a doctorate
<Bike>
that's what i'm about
<awygle>
congrats rqou
<pointfree>
Students used to wear the cap and gown everyday. It indicated to law enforcement that they are students (university campuses are/were autonomous and not under the jurisdiction of the state). It has origins going back to ancient Greece, Rome, and also the Anglican Church.
<Bike>
what do you mean not under the jurisdiction
<pointfree>
Congrats rqou!
<Bike>
cap and gown is kind of a weird license to kill
<Bike>
also yes congratulations
<rqou>
thanks everyone
<pointfree>
Bike: In a lot of countries universities have their own police and have some sovereignty from the state, although this seems to have eroded over the ages. Still, Boston or Cambridge police is not allowed to police people on Harvard campus for instance. That's the responsibility of Harvard police.
<azonenberg>
pointfree: interesting, because at RPI the campus public safety officers were the *primary* LE
<azonenberg>
But city police still had jurisdiction
<azonenberg>
They just didn't routinely patrol there unless called
<awygle>
yeah berkeley has UCPD
<Bike>
i'm used to universities having their own police, but that seemed more like a pragmatic measure since a campus is a small city full of young adults
<pointfree>
I remember once, a Cambridge police officer was hassling people on Harvard campus. My father called the Harvard Police and they promptly got that officer off the campus.
<rqou>
lol
<awygle>
all right i have a very stupid question
<awygle>
other than "does my processor fit" and "do i have enough PCIe lanes, USB ports, etc", is there a substantive reason to pick one motherboard over another?
Bike has quit [Ping timeout: 260 seconds]
<rqou>
how many corners get cut? :P
<pie___>
as an EE you should know, as a customer, good fucking luck? xD
mumptai has quit [Remote host closed the connection]
<sorear>
pointfree: I can’t remember if I asked if you’re still in the area
<pointfree>
sorear: I live in San Mateo, CA
<rqou>
still in the expensive AF apartment?
cr1901_modern has quit [Ping timeout: 240 seconds]
<sorear>
ah.
Zorix has joined ##openfpga
<kc8apf>
awygle: they all mostly suck
<kc8apf>
the things OEMs do to AMI Aptio....
* awygle
nods and smiles
<rqou>
needs moar coreboot
<reportingsjr>
awygle: maybe if you're going to overclock stuff?
<pointfree>
rqou: Yeah I'm bleeding money but I'd don't really feel like moving. The problem is temporarily solved by not checking Mint.
<rqou>
##openfpga dorm? :P
<kc8apf>
rqou: I'll settle for Linuxboot
<awygle>
"supports RGB light strips in 7 colors" SOLD
<gruetzkopf>
genii: if that's G703 over HDSL it may indeed be useful
<gruetzkopf>
which continent are you on?
<genii>
North America... Toronto, Canada, more specifically
<gruetzkopf>
RWTH aachen has campuswache, which has police-like duties on campus
<pointfree>
I'm also renting 120sqft of storage for books.
<awygle>
i can't tell if i should be looking at JEDEC or "A-XMP OC MODE" RAM speeds...
<pointfree>
Is anyone here going to Bay Area Maker Faire this weekend?
<kc8apf>
I'll probably be there on saturday
<kc8apf>
tinyFPGA has a table
<rqou>
gruetzkopf: it's probably different in European universities because you don't have our drugs/alcohol/partying culture
<gruetzkopf>
wouldn't say that
<gruetzkopf>
but RWTH doesn't have a closed campus, it's all over the town
<jn__>
we have a dedicated street for students partying :)
<jn__>
kind of
<rqou>
we do too :P
<rqou>
it's called "frat row" (cc awygle)
<sorear>
I expect Harvard to be a bit unusual with the “predates the USA” thing
<pointfree>
kc8apf: Cool! I'll probably go on Saturday too. I'll look for the tinyFPGA table. I'm looking to pick the brains of Cypress engineers at their booth.
<awygle>
oh god the newegg shopping cart timed out, i'm gonna end up buying two motherboards...
<kc8apf>
*sigh* it's that a sign of the times. Cyress having a booth at Maker Faire
<awygle>
they had one at the seattle "mini maker faire" also
<awygle>
selling heavily-subsidized psoc dev boards
<rqou>
sorear: not Oxford with its "predates proper written records?"
<tinyfpga>
pointfree: stop by and checkout the demos
<tinyfpga>
pointfree: I’m finishing up the demos, handouts, and some giveaways now
<sorear>
rqou: did somebody mention Oxford upthread?
<kc8apf>
tinyfpga: I'm serious about the cargo van. It's been to Sac before and I have very little happening this week.
<pointfree>
awygle: The PSoC 5LP/4 boards were free/gratis at the Bay Area Maker Faire last time I was there.
<gruetzkopf>
genii: apparently they're T1 only, no e-carrier :(
<awygle>
pointfree: huh, they were like 10$ at the seattle one
<pointfree>
awygle: That's how much they cost online. Bay Area Maker Faire does not allow selling things, but you can give away freebies.
<rqou>
yeah I'm not going to maker faire
<awygle>
ahh that makes sense
<jn__>
unrelated electronics question: what's a good op-amp for low frequency (probably <1kHz most of the time) stuff in the 0V-15V range?
<genii>
gruetzkopf: Guess I should probably just send them to the e-waste then :(
<jn__>
TL084CN sounds alright from the datasheet and is cheap
<reportingsjr>
jn__: there are a million
<reportingsjr>
tl072 is also very popular
<pie___>
jn__, whatcha doin? :D
<jn__>
ok, i'll pick something
<awygle>
what's your figure of merit? do you care about power consumption for example?
<jn__>
pie___: i'm going to build a modular synth
<pie___>
oooh
<gruetzkopf>
on 100mmx160mm PCBs? ;)
<Ultrasauce>
cool i got one in the tubes also, which designs are you looking at?
<jn__>
i did some cheap, crappy casework last weekend
<Ultrasauce>
tl074 is perfectly sufficient for basically everything in that application
<gruetzkopf>
there's cheap-ish cases for eurocards
<jn__>
damn, 160mm won't fit
<pie___>
alternatively; "AAAH ANALOG MAGIC. WITCHCRAFT. *runs away screaming*"