<pie_>
does anyone know how to do a drag selection using the kezboard in kicad?
<Zorix>
only ever used the mouse
rohitksingh has joined ##openfpga
<tnt>
I know there is 'move exact' but it's not exactly dragging ...
rohitksingh has quit [Ping timeout: 240 seconds]
<RaYmAn>
'?' key pops up a more or less complete list of special keys
<RaYmAn>
or not
<RaYmAn>
it used to
<RaYmAn>
Ctrl+F1 apparently now
jevinskie has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<tnt>
whitequark: wrt to glasgow doc ... not sure about others, but I figured since it's a "WIP" project that you'd get to writing doc later and so not bothering you with that now :p
<whitequark>
well kinda, i'm trying to come up with interfaces that are actually nice to use
<whitequark>
so i only spend time writing up on things that i know will stay the same
<_whitenotifier>
[Glasgow] kbeckmann opened issue #78: async has become a reserved keyword in Python 3.7 - https://git.io/fpBNE
m4ssi has quit [Remote host closed the connection]
jevinskie has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
jevinskie has joined ##openfpga
rk[ghost] has quit [Ping timeout: 240 seconds]
<pie_>
ive done a bit of googling but havent found anything for certain yet...i have 10 subsheets in eeschema to duplicate something 10 times, is there a way to not have to lay out each of them repeatedly in pcbnew?
<whitequark>
nope
<pie_>
well shit?
<tnt>
I haven't replicated it in kicad, but in eagle I used to use a custom script ... you could do the same with python bindings I guess.
rohitksingh has quit [Ping timeout: 244 seconds]
<tnt>
Mmm, next-pnr doesn't seem to know about IOB registers in timing analysis.
rohitksingh has joined ##openfpga
<pie_>
i know it doesnt work in probably a lot of cases but the falstad circuit simulator is so good...
<pie_>
i havent seen anything else thats that interactive
<pie_>
though i havent looked very hard
<pie_>
ok so that was meant to be a question...do you guys know anything more professional thats perhaps still free and good and such
<tnt>
qucs maybe ?
<pie_>
thanks ill check that out. now that you mention it it sounds familiar, i may have run into it at some point
<pie_>
looks like there might be some qucs -> kicad conversion stuff
<pie_>
donts see anything in the other direction sadly
<pie_>
kicad ui is nicer than qucs :c
<_whitenotifier>
[whitequark/Glasgow] whitequark tagged ef4e170 as revB https://git.io/fpRvf
<TavyCats>
pie_: I love how interactive the falstad circit sim is
<TavyCats>
makes it very easy to fuck around and explore ideas
<pie_>
TavyCats: yeah
<pie_>
thats like the whole point :P
<TavyCats>
like breadboarding but its easier to know pinouts and its generally less of a mess
<pie_>
do any professional tools let you do that
<TavyCats>
like, watch scopes/etc as you're altering the circuit?
<pie_>
yup
<TavyCats>
not sure, but I wonder how close to realtime ngspice is
<TavyCats>
and also other things, because it'd be cool to pair kicad with continually running simulation and visual scope outputs
<pie_>
apparently kicad has WIP (?) ngspice integration
<pie_>
which is why is probably why you brought it up
<TavyCats>
it has functional-ish ngspice integration
<TavyCats>
sadly my distro is being silly wrt even building libngpsice
<pie_>
the wonders of linux :p
<TavyCats>
so I guess, time to patch the ngspice.spec and kicad.spec and have a local repo
<TavyCats>
hrmph, copr is like the thing to use for this I should figure out what my FAS account was
<pie_>
whats copr, whats fas
<TavyCats>
copr is like ppa but fedora, fas = fedora account sysem which is fedora's sso thingy
<pie_>
ah
<pie_>
dumb question...i have this falstad circuit: http://tinyurl.com/ycjzhchs , theres a switch at the bottom left. why is the current through my red LED lower than the green one?
<pie_>
do p mosfets have a bigger voltage drop or what?
<pie_>
sidenote: i only have a very vague idea what im doing
<qu1j0t3>
pie_: different colour LEDs have different drops
<pie_>
qu1j0t3: i dont think the simulator accounts for that
* qu1j0t3
didn't look at the serkis
<qu1j0t3>
pie_: It would, if you set the colour
<qu1j0t3>
pie_: the differences are pretty significant
<pie_>
qu1j0t3: it takes floating point rgb values
<qu1j0t3>
meh
<pie_>
to set the color
<qu1j0t3>
ic
<pie_>
i find it more likely im missing some mosfet fundamental
<pie_>
or just derping something else in the circuit :p
<pie_>
hmmm...actually the voltages across them look a bit different, i think they may not be hooked up "the same" ...
<pie_>
though im not sure if this can actually be done the way i want
<pie_>
basically i want to switch between the leds depending on the logic signal...
<qu1j0t3>
this would actually do the job, i think)
<daveshah>
Indicator LEDs don't need 24mA unless you want them to be visible from space
<qu1j0t3>
indeed!
<tnt>
pie_: just connect your wire to one inverter, then the second inverter at the output of the firstinverter and connect your led across the second inverter with a current limiting resistor.
<pie_>
ah. lol.
indy has quit [Read error: Connection reset by peer]
<pie_>
that seems kind of dumb but i see how it works :P
<qu1j0t3>
dumb can be smart!
<tnt>
when you think about the push-pull driver output stage of those inverter, well ... you made a h bridge :p
<pie_>
tnt: huh.
futarisIRCcloud has joined ##openfpga
<pie_>
tnt: um, do i use the same voltage divider circuit?
<tnt>
that chip can take the 5v both at the input and as vcc.
<pie_>
ok but what about the non-chip end...
<tnt>
what ?
<pie_>
i mean the led has two ends, the dual inverter goes on one end, what goes on the other end?
<tnt>
I said, "place the led across the second inverter"
<tnt>
_across_
<pie_>
ahaaa
<pie_>
ok thats messed up, i love it :D
<pie_>
tnt, daveshah: thanks
indy has joined ##openfpga
<pie_>
time to learn how to add devices to kicad
<tnt>
I'm pretty sure that device is in kicad already
<pie_>
hm, i didnt find it but ill take another look
<tnt>
it has implicit vcc/gnd pins though IIRC which is annoying.
<pie_>
or im missing something
<pie_>
is there something other than the standard library?
<pie_>
tnt: yeah about that, do they show up in the pcb editor or do they need to be explicitly added or something?
<tnt>
ther eis a 74xGxx library
<tnt>
which has the 74x2G04 in there.
<tnt>
you need nets names vcc and gnd in the schemtic and it will use that.
<tnt>
daveshah: is there anyway to use/extract the 'typical' timing from icecube instead of the 'worst cast' ones ?
<daveshah>
tnt: you can change the voltage and temperature for a start
<tnt>
daveshah: you can do that in nextpnr ?
<daveshah>
No
<daveshah>
We only have the 85 degrees and 1.14V timing data in icecube
<daveshah>
*from icecube
<daveshah>
But it would be theoretically possible to get the 25 degree and tighter voltage ones too for example
<tnt>
yeah, that was my question, is there anyway to extract other timing data ? I've been running my up5k at 1.35v and I'd like timing data matching that more closely :p
<daveshah>
These settings are definitely configurable
<daveshah>
But certainly not all the way to 1.35V
<daveshah>
Basically it would be a case of modifying the project files icefuzz creates and rerunning icefuzz and the timing extraction stuff
<tnt>
yeah, that's fine. It's mostly so I can safely use the nominal 1.2v one and be sure the rail isn't going to ever dip below that.
<pie_>
tnt: figures, i was looking for dual inverter
<daveshah>
The other option is to do it more experimentally
<daveshah>
Using a ring oscillator to characterise the change in delay vs vcc
<daveshah>
azonenberg_work has done stuff like this before
<tnt>
daveshah: yeah, that's something I wanted to do as well to see how much exactly it varies.
<qu1j0t3>
ha.
<daveshah>
If you were really pushing things you could build a rig to bin devices too
<sorear>
characterise individual LUTs and timing paths, then do timing-directed placement to put the most critical paths on the best LUTs in a given device :3
azonenberg_work has quit [Ping timeout: 268 seconds]
<tnt>
daveshah: huh yeah, not that much :p I had just noticed that the vcc dependency was _really_ strong when testing with icecube. I guess to be ultra low power they operate right at the edge. And that's why i pushed vcc a bit just so I could use the 'typical' timing and still be safe because it's easily 50% faster which is not insignificant.
<daveshah>
Would be quite interesting to see the static power vs Vcc curve too, for that matter
<daveshah>
I know the 1.2V ECP5-5Gs are 30% faster fabric wise than the 1.1V ECP5s
<daveshah>
So there is certainly a strong voltage dependency there too
<tnt>
daveshah: yeah. Although ...yesterday, I filled 3000 FF, toggling at every cycles at 82.5 MHz and couldn't get the damn part to heat more than 1C above ambiant. I don't have current sense on the vcore but given the 5V usage, it couldn't have been more than 45 mA ...
<tnt>
Toggling half the IOs at 170 MHz did raise the power and temp a bit. But not even that much.
<sorear>
which part?
<tnt>
UP5k
<tnt>
(ice 40)
renze has quit [Quit: Spaceserver reboot?!]
renze has joined ##openfpga
<tnt>
Mmm, I think I have all the gear to actually make automated curve tracing for speed and power (freq counter, SMU, thermocouple and power resistor for temp testing) ...
Miyu has quit [Ping timeout: 260 seconds]
<SolraBizna>
I can't even seem to do something simple like read an SPI memory.
<SolraBizna>
Without an oscilloscope, I can't even tell if I'm affecting the SPI bus at all.
<tnt>
SolraBizna: beware that the ice40 put the flash to sleep after config. It needs to be woken up with 0xAB command.
<SolraBizna>
I'm doing that.
<tnt>
Ah well, that was what got me :p
<SolraBizna>
I reset the FT232, put it into MPSSE mode, send FFFFFFFF... to clear mode bits (to be safe), send AB to get out of power down, wait 10ms to be *very* sure it's out of power down, and then try 0B
<SolraBizna>
(with appropriate booping of the CS pin, and CRESET on the iCE40 being held low)
<SolraBizna>
... and now it works.
Miyu has joined ##openfpga
<SolraBizna>
waiting 700ms, attempting to read from the FT232 (and getting nothing), and *then* sending Fast Read works
<SolraBizna>
waiting 500ms flatly doesn't work, and waiting 600ms gets corrupted data most of the way through the read
azonenberg_work has joined ##openfpga
<pie_>
i might as well ask, is there maybe something where i can combine a mux and individual flip-flops? right now i have a 4bit -> 16 line mux (one for data, one for clk) setting data in 10 individual 1bit flip-flops
<pie_>
i.e. i want to be able to set individual "bits" without resetting the rest of them