<awygle> hmm this is weird. the LMs are only supposed to have one PLL, but the floor planner shows two.
<awygle> i guess they must have two but only one is balled out
<awygle> daveshah: i'm confused about gbufin_db. i assumed it was information copied from the "global buffer" cells in the floor planner, but the third digit doesn't match (e.g., icebox.py has (6, 0, 6) but the floor planner says (6, 0, 2) for the up5k)
GenTooMan has quit [Quit: Leaving]
digshadow has joined ##openfpga
digshadow has quit [Client Quit]
rohitksingh_work has joined ##openfpga
knielsen has joined ##openfpga
Bike has quit [Quit: Lost terminal]
<awygle> arright that's enough of that. i feel like i made good progress today though. only a couple things left in icebox.py to modify. extra_bits_db and gbufin_db.
<daveshah> awygle: awesome
<daveshah> Extra_bits_db can be worked out by creating designs which use each global pad input (not internal global buffer) and seeing which extra bit appears in the asc
<daveshah> For gbufin_db, you will have to manually place a global buffer at each location and see which global is used in the output of icebox_explain
<daveshah> It doesn't line up the datasheet, that is correct, the datasheet numbers apply to the pad inputs only
<awygle> Hm, OK... Weird, but sounds pretty doable.
<daveshah> It shouldn't take too long to work out
<awygle> I'll hopefully get that done soon then
<daveshah> The PLL is strange, but a PLL without VccPLL bonded is useless
<daveshah> Maybe they were planning fancier packages
<daveshah> Two of the global buffer "pad inputs" are probably the oscillators btw
<awygle> On a closer look the upper PLL block is greyed out (though present) in the floor planner
<awygle> Thanks for all the help. This would take a lot longer without a guide. Icebox really is a huge hack lol
<awygle> And I don't feel bad for saying it because it's in the documentation
<daveshah> Yeah, I think Xray is a lot nicer and I'm certainly aiming for better with prjtrellis
<daveshah> icestorm has the problem of being the first
<awygle> yeah, definitely
<awygle> although historically isn't there a "second system effect" saying x-ray should end up way overdesigned? :p
<awygle> how is prjtrellis coming along? i haven't heard much since tinyfpga did his initial work
<daveshah> Really good
<daveshah> I have a C++ library for bitstream manipulation
<daveshah> That can be bound to Python scripts for fuzzing etc using Boost::Python
<rqou> goddammit my final is tomorrow, I'm not done studying, and I'm sick
<rqou> fuck my life
<daveshah> Luke and I have worked out a fast approach to fuzzing, we just need to do the actual fuzzing now
<awygle> awesome
<awygle> i'm at pretty much the limit of my capacity to absorb new projects, but i really want to write a fuzzer or two lol
<awygle> i have both personal and professional interest in the ecp5's
<daveshah> awygle: Some preliminary info on the architecture here (beware, bits incomplete and broken): http://prjtrellis.readthedocs.io/en/latest/
<daveshah> Sure, we have a lot more to fuzz than in Xray
<azonenberg_work> Yeah i am well over my limit for projects :p
<daveshah> Because we're aiming for the whole chip, not just an ROI
<awygle> right
<azonenberg_work> xray wants the whole chip too
<awygle> i particularly want to try and fuzz a serdes
<azonenberg_work> But they're starting with a manageable subset
<daveshah> Yes, but there's no partial reconfig in the ecp5
<azonenberg_work> daveshah: sure but you can fully reconfigure
<azonenberg_work> with a black box bitstream
<azonenberg_work> and say "patch your stuff in at this offset"
<azonenberg_work> That's how i did my initial coolrunner work
<azonenberg_work> i created a skeleton for io routing etc and patched in my own PLA config
<daveshah> Yeah, well we already have all the tile offsets for the ecp5 so it can even be frame x, bit y in tile z
<azonenberg_work> Even better
<daveshah> I'm sure we'll start with that
<daveshah> But it is more regular than the 7-series so the whole chip is very doable
<daveshah> awygle: SERDES fuzzing would be awesome
<rqou> what the fuck so many projects that I don't have time for
<azonenberg_work> daveshah: my experience is that 7 series is very regular and orthogonal
<azonenberg_work> But there's just so many tile types :p
<azonenberg_work> ultrascale seems to unify even more as far as slice structure
<daveshah> Similar on the ECP5, but with only one logic tile type
<daveshah> Quite a few IO tile types
<azonenberg_work> i love how many toolchain projects we have going on now though :)
<daveshah> Anyway the nice thing is that tinyfpga found a feature called NCL, that means we can generate bitstreams for our own designs created after PAR
<rqou> random question: do nsaids cause drowsiness?
<daveshah> So we can create a design with a single interconnect and find out just the bits for that interconnect
<azonenberg_work> rqou: not to my knowledge, although most 1st gen antihistamines
<azonenberg_work> do
<azonenberg_work> e.g. diphenhydramine (benadryl)
<azonenberg_work> rqou: the main important side effect of nsaids that i know of is that of an anticoagulant
<daveshah> Effectively using that and TCL commands, to fuzz interconnect we can enumerate all arcs I a tile
<rqou> yeah that's not a problem for me
<rqou> anyways, goodnight since I'm sick and my final is tomorrow
<daveshah> This should take less than an hour per interconnect tile with 8 threads
<awygle> good luck rqou
<daveshah> The hard manual work is the configuration bits
<daveshah> rqou: good luck!
<awygle> some nsaids have bad effects on your stomach or kidneys
<awygle> Is acetaminophen an nsaid? That's liver
<azonenberg_work> awygle: acetaminophen does cause liver issues especially with OD
<awygle> daveshah: what do you mean by configuration bits in this context?
<azonenberg_work> But it's not a NSAID
<azonenberg_work> It's a pain reliever but not an anti-inflammatory
<daveshah> awygle: everything that's not routing
<azonenberg_work> It's useful to have both it and a nsaid (e.g. ibuprofen) in a first aid kit since it's safer to use in the presence of bleeding or clotting problems
<awygle> Ah
<daveshah> We will need to create fuzzers for them, that iterate through all possibilities manually
<daveshah> Like for the SERDES for example
<awygle> daveshah: okay, just making sure I understood
<awygle> and that you weren't talking about bits to configure the routing or something
<azonenberg_work> some nsaids, like aspirin, are such strong anticoagulants that they are actually used medically for that purpose rather than as an anti-inflammatory
<daveshah> awygle: I'm not sure if there's a better term for them, I agree configuration is not ideal
<daveshah> But non-routing configuration is a bit long to type all the time
<awygle> I am directed to avoid acetaminophen due to a liver enzyme thing
<azonenberg_work> 81 mg aspirin is commonly used as a "maintenance" anticoagulant in cardiac patients that dont need something stronger like warfarin
<awygle> daveshah: yeah agreed. Tile config? Something like that. Not important really.
wpwrak has quit [Read error: Connection reset by peer]
<daveshah> awygle: in the moment in the code they are always qualified to "config words" which are either a single bit only, or multi bit and specified as a word (LUT init for example), and "config enum" which have several textual values (IO type for example)
wpwrak has joined ##openfpga
<daveshah> I'm not sure whether we will keep that difference or not though
<awygle> interesting
<awygle> Well I'll try to get LM off my plate as quickly as possible so I can switch over :-P
<rqou> i though a big (or at least very depressing) use of acetaminophen is to mix it into prescription opioids to "discourage abuse" (by killing you if you abuse it) (cc whitequark)
<daveshah> Sure it would be great to have you on board
<awygle> But for now, sleep. Thanks again for the help!
<azonenberg_work> rqou: apparently that isnt the only reason
<azonenberg_work> apparently it also has some kind of synergistic effect with the opioid
<azonenberg_work> such that you can use less of it
<rqou> huh
<azonenberg_work> But yes they are commonly combined, i was prescribed i think 300mg acetaminophen + 5 mg hydrocodone for my wisdom teeth extraction
<azonenberg_work> Didnt actually use it though
<rqou> so the killing people part only appears thanks to our totally fucked drug policy?
<awygle> I was prescribed an unconscionable amount of percocet (acetaminophen+oxy) for my sleep apnea surgery
<azonenberg_work> i have never seen any indication from a reputable source that this is the intent
<azonenberg_work> rqou: For me to consider an opiate i'd have to be in "uncontrollable screaming" level pain
<awygle> Something like 30 days with three refills iirc
<azonenberg_work> awygle: o_O they gave me five tablets lol
<awygle> I took like 4 total and did not enjoy it
<rqou> afaik it's not actually the intent, but the stupid drug restrictions end up causing people to die/end up with destroyed livers anyways
<awygle> It inexplicably just made me angry
<azonenberg_work> they also gave me 800mg ibuprofen "horse pills"
<azonenberg_work> Which i took... 3 of? lol
<awygle> I hate those lol. I have ibuprofen. I'll just take that.
<awygle> And not choke to death
<rqou> awygle: wtf. i guess you're not in Appalachia so you can actually get opioids? :P
<azonenberg_work> Yeah i'd rather take a couple of regular 200/500 mg ibuprofens to get 800ish
<azonenberg_work> than swallow one of those horse pills
<awygle> I am not gonna get into this one rqou, especially since I don't have a clearly formed position of my own anyway besides "I did not need 120 percocet for a septoplasty"
<azonenberg_work> Speaking of wisdom teeth i have a dentist appointment in a couple days
<azonenberg_work> At which point they're supposed to schedule me for pulling the last one (3 of 4 gone atm)
<azonenberg_work> Yaaaay
<rqou> heh i guess I've been influenced by *) whitequark *) being in Berkeley and would rather the government just decriminalized everything and worked on the real problems
<pie___> teeth for the tooth god
<azonenberg_work> You mean a blood sacrifice to the tooth fairy?
<azonenberg_work> :p
<azonenberg_work> awygle: yeah no matter what your thoughts on drug legalization are
<azonenberg_work> it's pretty undeniable that the us massively overprescribes opiates
digshadow has joined ##openfpga
<pie___> why solve problems when you can just hide the pain
<azonenberg> lol yeah
* pie___ starts looking at cat pictures on the internet
<whitequark> rqou: no, it's not "stupid drug restrictions"
<whitequark> opioid formulations are literally designed for murder in the US
<whitequark> it's very much conscious, you can read all the materials introducing them
* whitequark reads more backlog
<whitequark> oh you said that
<whitequark> yeah
<tnt> Just curious, anyone ever implemented USB FS on an ice40 ? (without phy, just the normal io pins)
<daveshah> Yes, have a look at tinyfpga's project: https://github.com/tinyfpga/TinyFPGA-Bootloader
<daveshah> He wrote a bootloader that implements the USB for programming on the ice40 itself
<daveshah> but it could be adapted for other uses too
<tnt> daveshah: oh, nice, thanks.
Patater has joined ##openfpga
<daveshah> tnt: He hangs out here too (but will be sleeping right now I think)
<tnt> even better :)
rohitksingh_wor1 has joined ##openfpga
pie___ has quit [Ping timeout: 256 seconds]
rohitksingh_work has quit [Ping timeout: 256 seconds]
bitd has joined ##openfpga
pie___ has joined ##openfpga
pie___ has quit [Ping timeout: 240 seconds]
rohitksingh_wor1 has quit [Read error: Connection reset by peer]
Bike has joined ##openfpga
soylentyellow_ has joined ##openfpga
gruetzko- has joined ##openfpga
benreynw1r has joined ##openfpga
soylentyellow has quit [*.net *.split]
benreynwar has quit [*.net *.split]
fouric has quit [*.net *.split]
gruetzkopf has quit [*.net *.split]
fouric has joined ##openfpga
rohitksingh has joined ##openfpga
wpwrak has quit [Ping timeout: 264 seconds]
rohitksingh has quit [Quit: Leaving.]
m_t has joined ##openfpga
<whitequark> oh
<whitequark> oh god I figured it out
<whitequark> the initialization code for paged RAM segment uses a movx @r0 instruction, where the actual offset into external RAM, which is same as code RAM on this chip, is given by XPAGE+r0
<whitequark> or rather XPAGE<<8+r0
<whitequark> on this chip, XPAGE is at 0x92 and is called MPAGE, but on the original 8051, it's at 0xa0 and is called p2, and MPAGE is 0 by default
<whitequark> so, what the CRT does is it clears first sizeof(PRAM) bytes of XRAM, which is the same as CRAM
<whitequark> the external interrupt 0 vector is at address 3.
<whitequark> so, once I exceed 3 bytes allocated in PRAM...
<whitequark> in theory, the XPAGE symbol should be weak
<whitequark> in practice, for some reason that's broken in sdcc
<whitequark> oh *facedesk*
<openfpga-github> [libfx2] whitequark pushed 1 new commit to master: https://github.com/whitequark/libfx2/commit/b860f9dfb5c5e670745626e929c7d5c679eefa3a
<openfpga-github> libfx2/master b860f9d whitequark: Declare _XPAGE as a *non*-static SFR so crtpagesfr.asm is overridden.
<awygle> damn lol. well done finding it
genii has joined ##openfpga
<daveshah> awygle: morning
<daveshah> would you be OK pushing your icebox work so far to the PR? It would be great to review it
<awygle> daveshah: well i would, but i'm at work now and don't have access to it lol
<daveshah> awygle: sure, no worries
<awygle> i'll either finish it this evening and then push it, or just push it this evening if i don't have time to work on it
<rqou> huh, adafruit is amazingly good at marketing
<rqou> (this is a comment on all of the posts that the algorithm(tm) has been putting on my timeline recently about the "feather" spec)
<rqou> (and in case anybody was wondering, the final is at 7pm pacific time so i still have some time to study)
<rqou> at least i'm feeling reasonably less sick today
<cr1901_modern> feathers are good. I have 2 of them
<cr1901_modern> One arm, one xtensa
<rqou> i dislike them
<cr1901_modern> You're allowed to be wrong :)
<rqou> random cramped pinout with too few pins and random features like battery charging that just feel hacked on
<daveshah> Would have used a few LoRa feathers in a commercial project, if they had proper antenna connectors
<daveshah> Even at 915MHz, using 0.1" pins just feels uck
<rqou> random missing features like that are my constant complaint with adafruit/sfe products
<rqou> "everything except this tiny bit"
<daveshah> Had my own design with an SMA that I could modify and did the job so went for that
<daveshah> Yeah, it's annoying because the LoRa ones could be really useful, if they thought for a minute what professional users might want
<whitequark> "If you are making a mainboard Feather you must have onboard LiPoly charging" mehhhh
<whitequark> that means I can't have LiFePO4?
<rqou> is LiFePO4 not compatible with the generic lithium battery chargers?
<gruetzko-> nope
<rqou> hmm
<whitequark> the voltage is different by 100mV and I think there's something else
<awygle> ew, 100mil header antenna connector?
<awygle> gross
<rqou> so yeah, yet another example of "random missing features"
<gruetzko-> voltage is far different
<gruetzko-> 3.2 nominal 3.65 max-charge
<awygle> i like the idea though. 100-mil breakout boards are good, generally.
<rqou> also, i personally would either not care about battery charging at all or would rather have a "fancy" BMS
<gruetzko-> instead of 3.7ish nominal 4.2 max-charge
<rqou> but adafruit/sfe seems to really like "half-solutions"
<rqou> i guess i do too in a different way :P
<gruetzko-> over here most lora stuff is 868MHz
<rqou> we all use half-solutions, but we all want different halves :P :P
<gruetzko-> 433 would be legal as well
<daveshah> gruetzko: yeah, we're testing at 868MHz right now (but the country its being depolyed in is 915)
<daveshah> awygle: I checked again and they do also have a spot to solder in a u.FL, apparently
<daveshah> still, not brilliant
<awygle> that's better at least
<daveshah> given that almost all 868/915 antennas are SMA or similar, so you need an adapter
<daveshah> also, the ufl connector is on the bottom
<gruetzko-> i'd rather have SMA
<gruetzko-> i'm in talks with the regulatory authority here to officially get rid of the modulation = FM requirement on my weird 2m license
<gruetzko-> because i've seen 140-150MHz lora modules with 10W output :D
<awygle> SMA is always better, especially than u.fl or the gpo family
<daveshah> gruetzko: oh nice
<rqou> gruetzko-: or you can get a real ham license? :P
<whitequark> awygle: what about smp :P
<whitequark> oh gpo
<awygle> whitequark: smp == gpo
<daveshah> in the UK I know there are no modulation restrictions, so long as you don't encrypt and send your callsign
<awygle> yup
<gruetzko-> pff
<gruetzko-> same here for HAM
<rqou> although iirc i was told that getting a ham license in germany is a pain in the ass
<daveshah> that applies to any license
<whitequark> daveshah: why can't you encrypt?
<whitequark> also is steganography fine
<daveshah> it's against the rules of amateur radio
<daveshah> the spirit is everything is public
<rqou> it's so much more convenient to operate with a US license under reciprocal agreements :P :P :P
<daveshah> rqou: yep, same with a UK license
<daveshah> whitequark: you are allowed to encrypt only if required to by a police officer or person of authority in an emergency
<gruetzko-> this is a commercial allocation in the "reserved for railway maintenance" section
<cr1901_modern> I've had a license since Dec 2013... I think I've made 4 contacts?
<rqou> iirc gruetzko- told me that the US has some of the lowest filing fees
<cr1901_modern> I keep wanting to get back into it
<daveshah> can't say that's used much at the moment, but once the UK switch their emergency services over to LTE perhaps they'll be relying on amateurs again...
<cr1901_modern> Esp CW, but... God, learning CW is a PITA
<rqou> lol i don't know CW
<awygle> i've had a license since 2012-ish but have never operated with voice, only APRS
<rqou> similar
<gruetzko-> emergency services here are on TETRA with 2m/4m FM fallback
<gruetzko-> railway stuff is on GSM-R
<daveshah> yep GSM-R was being rolled out in the UK
<cr1901_modern> awygle: All my existing contacts are voice
<rqou> actually i just remembered, the us ham bands _do_ have modulation restrictions on some of them
<rqou> probably for hysterical raisins
<gruetzko-> the local GSM-R base station is one of my timing references
<daveshah> a few in the UK do for the same reason (military shared I think)
<cr1901_modern> Naturally I don't like doing it :P
<gruetzko-> and what i calibrate my SDR freq lock agains
<rqou> a bunch of the US ones have a 100kHz-wide CW-only section
<cr1901_modern> CW bandwidth is 1/100 of that
<rqou> and then we have a weird channelized 60M band
<daveshah> I've been wanting to make a cheapo SDR with an iCE40, cheapo DAC, ADSL line driver to transmit and scanner AFE ADC
<daveshah> but ENOTIME
<cr1901_modern> that's 100 ppl that could use that band simultaneously
<rqou> and a bunch of our bands are split with data-only in the lower half and voice-only in the upper half
<whitequark> daveshah: ADSL line drivers are magic
<gruetzko-> i've been writing 68k code for a radio data terminal DB used
<whitequark> we use them for our mass spectrometer
<daveshah> whitequark: I really want to have fun with them
<rqou> O_o
<whitequark> to drive the rods
<gruetzko-> ADSL line drivers are extremely cool
<daveshah> I might just make a QRP transmitter with them one day
<rqou> whitequark: link to a part?
<daveshah> I hope they are linear enough you could transmit across all the HF bands at 1-2W without a LPF
<daveshah> but maybe that's being optimistic
<daveshah> never actually got round to making measurements
<whitequark> rqou: not offhand
<rqou> NDA?
<cr1901_modern> daveshah: Perhaps QRP is useful around your parts? There's nobody to talk to here
<rqou> cr1901_modern: get out of NJ? :P
<cr1901_modern> rqou: Not happening
<cr1901_modern> Although, tbf a Pixie clone isn't exactly a reliable QRP transceiver
<cr1901_modern> which is what I have
<gruetzko-> i think i do have a few ths6182 here
<whitequark> rqou: no I just don't have the PN
<rqou> so i'm looking at the datasheet for the THS6182, and maybe i just suck at analog but i don't see anything too magic?
<whitequark> daveshah: they are very linear
<whitequark> we tried using a fancy class A stage with beryllia substrate that cost like $600 and it *sucked*
<daveshah> whitequark: yeah, I suppose they have to be with the crazy modulation schemes ADSL use
<daveshah> particularly the newer VDSL ones, they go pretty high frequency too
<whitequark> yep that's why
<daveshah> it's a classic example of crazy technology being cheap because of a mass market use
<rqou> oh wow i just looked at the graphs again
<rqou> you can push it under just the right circumstances from dc to 100 MHz
<whitequark> yes
<daveshah> they can drive pretty low load impedances too
<rqou> alright, gotta add some of these to my next part order
<rqou> why are more people not using these?
<daveshah> think I'm going to make a PMOD with a DAC and one of these tonight
<rqou> ugh why does everybody like PMOD?
<awygle> no particularly good alternative is my guess
<daveshah> what do you suggest? I'm not fucking around with FMC for this
<rqou> signal integrity is really "meh" considering that it usually gets used with fancy programmable logic
<daveshah> only talking about 60MSPS here
<daveshah> ice40 stuf
<rqou> be azonenberg and use samtec Q-strips? :P
<awygle> i agree they do feel sort of fisher price but *shrug*
<whitequark> lol FMC
<whitequark> fuck FMC
<daveshah> seriously fuck it
<daveshah> as if things weren't bad enough, digikey had their fully populated and half populated FMC reels swapped at one point
<daveshah> so I ordered two of the wrong part for a time critical project
<daveshah> digikey offered to send two correct ones
<awygle> what about ~~SYZYGY™~~
<daveshah> of course they sent two more wrong ones
<whitequark> daveshah: LOL
<rqou> wait, what's wrong with the FMC connector?
<whitequark> rqou: overpriced piece of shit
<rqou> ah ok
<daveshah> most IO cards don't need several hundred IO
<rqou> i just asked samtec for a sample :P
<rqou> but they were only willing to send me one
<rqou> kinda silly tbh because samtec free samples are sent overnight, but they have silly restrictions on how many parts they will send you
<rqou> Q-strip parts are kinda pricey but not insanely ridiculous
gruetzko- is now known as gruetzkopf
<awygle> well, they restrict it because it's a, like, 60$ part
<whitequark> why not just put an SMA there?
<rqou> awygle: no, they wouldn't send me more than two QTH parts either :P
<awygle> well those are like 8-10 dollar parts iirc
<awygle> they scale with price
<rqou> yes
<awygle> (nonlinearly)
<rqou> but how much is overnight shipping?
<awygle> that's fair. i'm sure they get a good deal, but i've always been kind of blown away by that.
<rqou> yeah i didn't even ask for that
<rqou> and obviously don't need it because the parts are still sitting in a drawer ~2 years later :P
<awygle> does anyone actually _use_ this syzygy thing?
<awygle> i mean it's basically the same approach azonenberg and i independently came to because q-strips are just good connectors
<daveshah> awygle: the syzygy crowdsupply made a total of $3505
<daveshah> which given the price of the boards doesn't amount to much
<awygle> i didn't even know they did a crowdsupply
<daveshah> whitequark: looking at http://www.ti.com/lit/ds/symlink/ths6204.pdf I reckon I'll hit thermal limits before distortion, clipping or slew rate limits
<awygle> i just thought the "standard" was interesting, even if they went about it in a terrible way
<rqou> obviously they sucked at marketing because i've never even heard about it
<daveshah> also, they picked a crazy unpronounceable name, and yet it still doesn't feature when you google it
<daveshah> if you go for a stupid name, at least go for a unique one
<awygle> lol
<gruetzkopf> i have some ADSL DSLAMs which use these TI drivers
<rqou> tbh "feather" isn't a unique name either
<gruetzkopf> massive heatsink in them
<rqou> but ladyada is a mit meng grad who actually knows how to do teh marketing :P
<daveshah> gruetzkopf: I think in theory up to 16W CW output should be doable from an electronic point of view
<daveshah> but not sure if even fan cooling would get that far
<gruetzkopf> big copper heatsinks, good thermal compound, tornado-grade fans in this one
<gruetzkopf> the one FPGA IO i *always* want more of is transceivers
<awygle> well yeah cuz they're fukkin sweet
<whitequark> awygle: so I made the analog frontend work completely
<whitequark> I think
<whitequark> unless we want some feature like voltage-follower mode in hardware which I think we don't reallly
<whitequark> the board reacts to an alert within ~6ms, which seems like it should be enough for most although not all uses
<rqou> why does glasgow have fancy analog?
<awygle> it's not "fancy" it's literally an 8-bit I2C ADC
<rqou> why do you even need that?
<whitequark> rqou: we sense target voltage with an ADC, generate it with DAC+LDO, and have the ADC continuously monitor Vsense so that if something funky happens the LDO gets shut down
<whitequark> basic safety really
<rqou> i would have just put Vtg directly into the level shifter :P
<awygle> potentially too much load on the DUT
<rqou> meh
<whitequark> yeah the level shifter can eat up to 35mA or so
<rqou> lots of other programming cables do that and it kinda works
<whitequark> classic rqou
<awygle> "kinda" being the operative word
<whitequark> settles for nothing more than "kinda works"
<rqou> even the ones that just shove Vtg directly into an ftdi Vccio kinda work :P :P
<rqou> (yes, i have actually had problems with this)
<whitequark> does everyone from $FANCY_SCHOOL have the same level of standards? :P
<rqou> apparently not awygle :P :P :P
<awygle> rqou: now if you'd said "plug the AFE directly into the Vccio" i would have been more on baord
<awygle> but the AFE is nonoptional imo
<whitequark> yeah, we wouldn't need the level shifter if not for 5V support
<awygle> my standards are somewhere between rqou's and whitequark's. although i do feel like they should be more whitequarkian, i just don't have the energy
<whitequark> ha
<rqou> i like to think that i have standards too, but the "don't have the energy" part kicks in very frequently
<daveshah> you could just use a highish current op amp (or two bog standard ones in parallel) as a buffer
<whitequark> i don't have the energy for my bullshit either :P
<awygle> you're also in school still, which vigorously rewards halfassery
<whitequark> mwahaha
<awygle> daveshah: yeah but then you can't _not_ have a Vtg
<daveshah> yeah, this is true
<awygle> err no that's not true nvm
<awygle> you either need a Vtg or you have to do ~half the AFE
<daveshah> yeah, and I suppose then you may as well go the whole hog
<rqou> just do the pickit2 "possibly blows up your board" technique :P
<awygle> yup
<rqou> (that they actually fixed, and then eevblog complained about it)
<whitequark> lol eevblog
digshadow has quit [Ping timeout: 240 seconds]
rohitksingh has joined ##openfpga
<rqou> lol whitequark why do you seem to have even more issues with banks than normal?
<whitequark> dunno
<whitequark> part of it is being an expat
<whitequark> e.g. I am almost entirely sure that my nationality is the sole reason HSBC and Hang Seng refused to open me an account
<whitequark> at least Hang Seng had the courtesy to say that no, they will definitely never open an account for me
<whitequark> HSBC just kept asking for more bullshit documents that I cannot possibly have
<rqou> hmm interesting
rohitksingh has quit [Quit: Leaving.]
<rqou> shittybank just pestered me "do you also have a US passport?"
<rqou> and after i said "yes, fine, report me to the IRS, whatever" they opened an account for me
<rqou> but then i'm technically not an expat
<whitequark> clearly that's because US is a white people country and RU isn't
digshadow has joined ##openfpga
<rqou> well, i had already gotten my HKSAR passport at the time
<whitequark> ah, could be too
<rqou> this was also pre-FATCA, but it was imminent
<rqou> somebody probably noticed "oh, your HKID says '***AO', need to do extra checking"
<rqou> oh yeah, HKIDs hint at place of birth, to allow for extra <strike>discriminating</strike>vetting :P
<whitequark> I mean they demanded my passport anyway
<whitequark> HSBC also wanted a notarized translation of a water bill from my RU place of residence, which is insane
<rqou> oh wtf
<whitequark> for one, no such document with my name on it has ever existed.
<rqou> yeah, my parents did have to jump through some hoops to show an address of residence too
<q3k> i think demanding bills for everything is an (ex-)commonwealth thing
<whitequark> q3k: standard chartered wanted no such bullshit
<q3k> they also love their bills in ireland
<whitequark> they just took my passport, confirmation of employment, and *maybe* a HK utility bill?
<rqou> whitequark: obviously HSBC only launders money for some people, not you :P
<q3k> i have a registered ltd company here, and when i was opening a bank account for it they still wanted utility bills as proof of address of the office
<whitequark> rqou: look if I wanted to launder money I'd just use a Bitcoin exchange
<q3k> also the institution of letterheaded paper being somewhat official
<whitequark> well technically ANXPRO *did* require me to comply with AML
<rqou> can we please just declare the war on <foobar> a failure and kill KYC/AML?
<q3k> all large exchanges these days do the whole KYC/AML dance
<Bike> using a utility bill seems like a hack, but everyone seems to...
<azonenberg_work> rqou: what's AO mean?
<rqou> A = "has right of abode in HK" O = "born in other countries"
<rqou> as opposed to Z = "born in HK"
<rqou> or X = "born in mainland" W = "born in Macau"
<rqou> like i said, it's for extra <strike>discriminating</strike>vetting :P
<rqou> so you know how in the US the fight about national ID cards is always a huge political issue because of how it can be used to deny rights to undocumented people?
<rqou> HK more-or-less managed to make a national ID card specifically to deny rights to undocumented people
<whitequark> oh.
<whitequark> awygle: hang on
* awygle hangs
<q3k> well it's not like you can do much in the states without an SSN
<whitequark> do I understand it correctly that the Crystal_SMD_3225-4Pin_3.2x2.5mm was not in fact 3.2x2.5mm in size?
<rqou> q3k: yeah well in the states carrying ID in public is not compulsory (it is in HK)
<awygle> whitequark: i think it was more that the "hand soldering" footprint made the pads comedy-large
<whitequark> oh
<Bike> i used to wonder why us conservatives liked so many things about a nominally communist country and man, when it finally clicked for me
<awygle> i just switched to the non-hand-soldering, which uses the pad dimensions defined in the datasheet for that crystal (or, well, within 0.1mm)
<whitequark> ahhh that's fine
<awygle> btw for next time - on the crystal symbol, you want the drivers to be hooked up to the parallel lines next to the rectangle, not to the case
<whitequark> awygle: what
<awygle> yeah that was terrible, sec...
<whitequark> oh for fucks sake
<whitequark> THAT was the cause
<awygle> yup
* whitequark facepalms
<awygle> no shade or anything, just wanted to make sure that particular bit of implicit knowledge was transferred to you
<whitequark> I actually know this in theory
<whitequark> I have no idea why my brain decided that *those* particular lines on the symbol were for the electrodes
<whitequark> also, didn't you review the schematic? :P
<awygle> yes :p
<awygle> frankly, it looks better your way
<awygle> something to do with the aspect ratio of the symbol
<rqou> oh lolol i would have missed that too
<awygle> i fixed it and then was like "... but it's ugly now"
<rqou> i usually don't draw the symbol with the extra ground terminals
<whitequark> hahaha
<rqou> i hate how tightly coupled the symbol pins and footprint pins are in kicad
<whitequark> yeah eagle got that better
<whitequark> i don't think there is any reason this can't be fixed in eeschema-new
<awygle> i should use eagle at some point
<awygle> just for the experience
<whitequark> eagle isn't actually bad
<awygle> it's still free for shitty designs, right? or did they take that away when they went subscription?
<rqou> ime eagle was pretty dumb, but it was consistently pretty dumb
<whitequark> I think it's still free but you should just grab Eagle 7 and crack it
<whitequark> no one has the time to deal with subscription bullshit and it's a more authentic Eagle experience anyway
<rqou> yeah i was using a cracked 6
<whitequark> in more ways than one
<awygle> lol "consistently pretty dumb" is actually a decent bar
<whitequark> yeah I agree with rqou here for once
<awygle> better than "inconsistently pretty dumb"
<whitequark> shockingly, their built-in homegrown scripting language (the second one, ULP) is actually kind of well designed
<awygle> did i complain about placing pins in altium here or was that in #kicad
<rqou> hilariously nobody seems to have ever bothered to crack the linux build of eagle
<whitequark> pcb-gcode is written completely in ULP and it's fairly pleasant to contribute to
<whitequark> I think it was on rutracker
<rqou> eh whatever, just a quick ida exercise
<rqou> oh i was talking about 6
<rqou> which still had the custom license manager
<rqou> huh
<whitequark> awygle: you did complain
<whitequark> awygle: hm you didn't update the comment on the schematic with LED current
<rqou> so one "fun" thing i've done with eagle is automate generation of pdfs of the schematic/board assembly sheet
<rqou> unfortunately this required Xvfb
<whitequark> I thought Eagle has headless mode or something?
<awygle> whitequark: ah good point, I'll fix it
<whitequark> I vaguely recall it
<rqou> whitequark: afaik only for the CAM processor to generate gerbers
<rqou> iirc doesn't work for running ULPs
<whitequark> ahh
<rqou> heh whitequark i expected you to be more upset about the use of Xvfb :P
<rqou> at least it didn't involve xdotool as well :P :P
<whitequark> rqou: well, you don't really have a choice there, do you?
<whitequark> other than reimplementing eagle's rendering algorithm which is not a good idea
<whitequark> I would have used Xvfb too
<rqou> hey, some people have tried reimplementing eagle's renderer instead
<rqou> hmm this reminds me... does kicad actually work headlessly?
<gruetzkopf> today in gruetzkopfs bright ideas: nintendo-16k-AES-XTS-dmcrypt-loopdev-vfat
<whitequark> rqou: yes
<whitequark> it has a Python API
<rqou> i though only pcbnew had that?
<whitequark> right
<rqou> eeschema could use one too
<rqou> also, does it actually work with an X server?
<rqou> *without
<rqou> i thought all the data structures were annoyingly wx-contaminated?
<rqou> whitequark: troll idea: kernel-rs :P
<rqou> link this into your program and you don't even need Linux anymore! :P
<rqou> not-sure-if-more-or-less-troll-idea: pvguest-rs
<rqou> hey, if people spin up a VM just to run one program, you don't need a kernel for that :P
<whitequark> rqou: so you know how xen ships with an ocaml pvguest scaffold?
<rqou> what
<rqou> i did not know that
<whitequark> and that was like a decade before mirage
<whitequark> mirage people, on the other hand, are utterly serious about it
<florolf> rqou: you just reinvented unikernels?
<whitequark> yes lol
<whitequark> well
<whitequark> technically mirage people and xen people are mostly the same
<gruetzkopf> so i looked into line drivers some more
<rqou> i didn't realize anybody though this was a serious idea
<florolf> rqou: here's a funny chimera: https://en.wikipedia.org/wiki/OSv
<gruetzkopf> VDSL2 line drivers apparently do like 150MHz bandwitdh at 10x voltage gain
<whitequark> florolf: oh that's just the linux version of rump kernels
<rqou> goddammit i thought I was joking
<florolf> whitequark: wait, wasn't rump kernels built the other way around? as in "here, take our drivers to build your kernel"?
<florolf> also, this doesn't use any linux kernel code but cheerfully steals from freebsd and other codebases and then emulates the linux syscall interface
<florolf> not sure if they adapted the freebsd linux emulation or hacked something together themselves
<whitequark> lol
<openfpga-github> [Glasgow] whitequark pushed 2 new commits to master: https://github.com/whitequark/Glasgow/compare/e9ab696666f9...b122273eb699
<openfpga-github> Glasgow/master b122273 whitequark: Fix LED current comment on schematic.
<openfpga-github> Glasgow/master 67f3dd9 whitequark: Implement ADC support.
<awygle> whitequark: re: alarm response speed, we can respond much faster on Rev C if we wire alarm to one of our plethora of additional FPGA pins
<tnt> esden: Hi. Just found out about your icebreaker-bitsy board, are you planning to sell it ?
<esden> if we can get a bootloader working on it then yes... we have still some trouble with getting usb working properly in the bootloader, I am focusing on the full size icebreaker for the moment. But I am definitely looking forward to have the bitsy in the store too.
<rqou> o/ esden
<tnt> esden: Ah good to know. What kind of issues ? I'm looking into making USB working on a UP5k as well.
<rqou> didn't notice you were here too
<esden> I might actually make a few development version ones that can only be programmed using ftdi for now...
<whitequark> awygle: not quite
<rqou> tnt: afaik the problem is that the UP fabric is slow AF (cc daveshah)
<esden> rqou: I am everywhere just like you :P
<daveshah> Yeah, that's the basic problem
<whitequark> right now I actually read out which port had an alarm, and then only disable that LDO
<daveshah> But Luke also got it to just pass timing in icecube, but it still didn't work
<daveshah> We suspect PLL jitter could be an issue
<tnt> rqou: right. I would still have thought that 48M should be doable :p
<esden> tnt: we thought too...
<tnt> ATM I only have an UPDuino and ... well ... layout is suboptimal :p
<whitequark> lol upduino
<esden> lol ... yeah ... the good ol upduino...
<whitequark> I would be surprised if that *did* work
<esden> I am pointing people to it to show how to not do layout
<balrog> oh you meant the hashtag was popular :P/
<balrog> oops
<balrog> is the v2 any better?
<daveshah> Slightly
<tnt> daveshah: What's the original clock rate ? (if you're using the pll, I guess it's not 48M native)
<tnt> There is a v2 ?
<daveshah> But they went for ground and power planes on a two layer board instead of two ground planes
<esden> I might need to rethink the icebreaker bitsy ... and just drop in a small FTDI chip on there after all :/
<daveshah> And so the ground plane is almost useless
<daveshah> tnt: for the bootloader? 48MHz
<whitequark> wtf
<daveshah> That's my memory of the v2 design
<daveshah> Or maybe it was two ground planes but next to no stitching
<daveshah> I can't remember
<whitequark> thelatter IIRC
<gruetzkopf> my first cavity resonator?
<esden> the alternative for the icebreaker bitsy is to remove the usb connector and double down on programming it using the ftdi multitool I made...
<esden> we would gain a few more IO usable in the user design
<whitequark> ew, ftdi
<gruetzkopf> i strongly dislike using ftdi parts
<esden> and if we can squeeze out usb out of the ice40UP eventually then we can rethink it
<daveshah> whitequark: actually looking again I'm pretty sure it's ground on the bottom and 3.3v on the top
<whitequark> daveshah: depressing
<tnt> esden: oh, that'd be a shame. native usb is so cool :p
<gruetzkopf> y u no 4layer
<whitequark> not sure native usb is worth it for up5k
<daveshah> gruetzkopf: because its cheap crap
<whitequark> or is it just for the bootloader?
<esden> everyone hates ftdi... noone is willing to make an actual alternative to it... you would only need a few million dollars and we could have a risc-v and usb-hs competitor to FX2 and FTDI :P
<esden> </troll>
<whitequark> I mean, if you can barely squeeze it in with nothing else, squeezing it with an user design is next to impossible
<whitequark> esden: there's fx3 with a cortex
<whitequark> but $$$
<esden> not a cortex ... unfortunately
<tnt> whitequark: unfortunately in my case, in qfn I don't have much choices.
<esden> it is an old ARM8 or something
<whitequark> oh ARM9
<esden> or so yeah
<gruetzkopf> i've thrown tiny little STM32F103 at this exact problem
<gruetzkopf> i'm comfortable with arm7 and arm9, though
<esden> it sucks ... makes me angry to think about it... I really wish it was cortex and I could use BMP to debug it ... I would be much more likely to like that chip
<daveshah> STM32 is not a bad option, if someone can be bothered to write the programming code for ice40
<esden> that would be very simple to do actually...
<whitequark> daveshah: I have programming code in 8051 assembly :P
<esden> I could use the same stm32 as on the bmp...
<gruetzkopf> (one of my programmers was a 2.50 'stlink' in it's earlier life)
<gruetzkopf> so a stm32f103 which is unqualified for 16k codespace and unqualified for usb, but still works
digshadow has quit [Remote host closed the connection]
<daveshah> whitequark: oh, that's useful to know
<daveshah> Thanks
<whitequark> you're just blasting the bitstream as bytes from MSB to LSB using SPI
<daveshah> What programming time do you have for SRAM
<daveshah> Just curious how Fx2 compares to mpsse
<whitequark> I have a 6 MHz SCLK
<whitequark> I could actually use the GPIF for this and make it much faster but I decided not to bother because it takes around 800ms
<daveshah> Sounds similar to FTDI
<daveshah> Not exactly a problem
<whitequark> basically, yeah
m_t has quit [Quit: Leaving]
<rqou> 800ms is a _little bit_ slow
<rqou> but yeah, totally usable
<whitequark> rqou: like I said, I could make it almost an order of magnitude faster
<whitequark> but decided not to bother and do something actually useful instead
<whitequark> ok well, more like 4 times faster
wpwrak has joined ##openfpga
cpresser has quit [Ping timeout: 240 seconds]
<whitequark> awygle: hm https://imgur.com/a/1fbLl21
<rqou> dat overshoot
<awygle> ringy-ringy
<rqou> limit your slew rate
<awygle> are these level shifter outputs?
<whitequark> yes
<whitequark> I mirrored i2c on fpga pins
<whitequark> the good news is that my concerns about slew rate being insufficient are unfounded
<awygle> lol
<rqou> is the crosstalk a measurement artefact or real?
<awygle> you're probing this decently, right? you don't have a long inductive ground path on the scope or anything?
<rqou> lol
<whitequark> I kinda do actually
<whitequark> uhhhh let me rig it up in a marginally less shitty way
<rqou> i need to buy some of those spring ground things (my probes are second-hand and don't have them)
<whitequark> I have them but in a different country
<whitequark> also I hate them
<awygle> fortunately they are tuppence a bag
<whitequark> such a pain to not short anything unrelated
mumptai has joined ##openfpga
<rqou> at least most of my probes have the colored rings :P
<rqou> most important part :P :P
<awygle> also those spring clips are steel so you can't solder them to stuff
<whitequark> well, you can
<whitequark> but you need active flux
<whitequark> like hydrochloric acid
<rqou> um...
<awygle> i wish digital shit spec'd its source impedance
<rqou> that sounds terrible for your boards though
<whitequark> rqou: you only need to wet them once
<whitequark> after that they'll be more or less solderable
<rqou> that still sounds bad for your soldering iron tips
<rqou> or do you use a shit iron for this?
<whitequark> iunno never tried
<whitequark> I just know it's possible to solder steel
<whitequark> I think most people actually use a blowtorch
<rqou> ah ok
<whitequark> awygle: ugh these green LEDs are seriously hurting my eyes
<whitequark> who the fuck specs these things
<whitequark> wait that was me
<rqou> interestingly, I don't really keep around $5 firestarter soldering irons
<awygle> well, us
<awygle> #azonenbergwasright
<rqou> lol
<whitequark> #azonenbergwasbright
<awygle> lmao
<awygle> inaccurate but deeply amusing
<whitequark> wait that sounds grim
<awygle> oh no wonder i feel like garbage i haven't eaten
* awygle lunch
<rqou> lol
<rqou> currently getting lunch here
<rqou> hmm random realization: ice40 does not have distributed ram
<rqou> i guess xray really needs to get more working
<daveshah> xray is working fine...
<daveshah> It's the place and route that's not
<rqou> oh, i thought that was under the same umbrella
<rqou> but yeah, that's what I meant
<daveshah> I believe xray refers just to the documentation efforts
<daveshah> Well, all I'm able to say right now is wait and see
<rqou> i mean, you keep hijacking the things i want to work on but can't because ENOTIME
<cr1901_modern> rqou: Sure it does, distributed RAM == "using the LUTs for RAM" :)
<awygle> no dibs in oss
<azonenberg> awygle: No DIBs? Only BMPs?
<rqou> cr1901_modern: except there's no way to dynamically edit contents
* azonenberg hides
<whitequark> awygle: oh whoa you fixed all revB issues
<whitequark> super cool thanks
<cr1901_modern> dynamically edit the contents?
<rqou> how do you propose storing the data into the luts?
<awygle> yeah you can do LUT ROM or FF RAM but not LUT RAM
<rqou> ^
<cr1901_modern> Oh, terminology fail. I meant LUT + FFs
<daveshah> Or you can dynamically rewrite bitstream
<awygle> whitequark: yup!
<rqou> yeah you can always do that but it's inefficient AF
<rqou> daveshah: but that's slow and you can't easily edit yourself while running
<cr1901_modern> I'm pretty sure that's what "distributed RAM" is even on Xilinx FPGAs, unless there are extra config bits specifically to use CLBs as a RAM more efficiently
<rqou> there are
<azonenberg> cr1901_modern: yes, each LUT6 can be used as a 64-bit SRAM
<daveshah> cr1901_modern: you can actually use the LUTs as ram
<daveshah> You can add a write address and data port
<azonenberg> much more efficient than using FFs for storage + LUTs for address decode
<awygle> lmao poor cr1901_modern
<azonenberg> Not all LUTs can do it, only SLICEMs
* whitequark has actually read the xilinx docs and knows this :P
<daveshah> All tiles in the ECP5 can do RAM
<rqou> whitequark: why the obsession with no BGAs?
<daveshah> 3 out of the 4 slices in a tile can be used together to make a 64x16 RAM
<whitequark> rqou: it's no use making it OSHW if few can actually build it
<awygle> Can you use it as a LUT at the same time?
<rqou> lern2bga?
<whitequark> and, well, there's the perceived complexity of BGA
<daveshah> *16x4
<daveshah> My errot
<whitequark> I mean it's not like *actually* soldering a QFN board is any easier
<azonenberg> whitequark: perceived, rather than actual, difficulty is not a good reason to not use a part
<daveshah> I'd say qfn is slightly easier
<rqou> awygle: um, pretty trivially i think? just put the inputs into the address port
<azonenberg> if you have data pointing to actual poor assembly yield in a hobbyist environment thats different
<daveshah> Easier to fix mistakes
<whitequark> azonenberg: it's not MY perceived difficulty though
<azonenberg> daveshah: i'd take a 1mm bga over a 0.5mm qfn in terms of yield
<whitequark> convincing others that BGA is actually easy is not a hill I want to die on
<daveshah> On a production process, 100%
<daveshah> Even at home with practice
<azonenberg> whitequark: see, this is the advantage of building boards to solve my problems and then open sourcing them because there's no harm in doing so
<azonenberg> Rather than because i want to build a community etc
<rqou> oh yeah apparently even 0.5mm QFP can have yield issues in real production
<rqou> fortunately those are pretty easy to rework
<azonenberg> rqou: in a production process i would take 1mm bga over 0.5mm qfp without even thinking
<daveshah> Definitely, large fine pitch qfp are a nightmare
<whitequark> azonenberg: I do not really want to build a community but I do not want to unnecessarily create issues for that either
<azonenberg> i literally only use qfp if i absolutely need THAT part and there are no viable qfn/bga alternatives
<whitequark> QFP is shit
<whitequark> fuck QFP
<whitequark> also there are "thermally enhanced QFP"
<rqou> why shit?
<whitequark> that's all disadvantages of QFN and none of the advantages
<whitequark> who even came up with those
<gruetzkopf> BGA ftw
<whitequark> azonenberg: but I think you're reading my opposition to BGA way stronger than it actually is
<whitequark> most likely, I will actually use the BGA part in the end and just say "anyone who wants QFN can build this slightly inferior revB version"
<cr1901_modern> I almost bought a skillet today to buy BGA, but then I realized I forgot an important bit of research
<cr1901_modern> to try&
<cr1901_modern> Is 420F enough for the balls to melt on the BGA package?
<azonenberg> cr1901_modern: skillet reflow is a *terrible* idea
<rqou> not a convection oven set to "cookies"?
<azonenberg> ^
<cr1901_modern> Yea, but it was cheap and right there
<azonenberg> cr1901_modern: No it's not
<azonenberg> that's about 215C
<azonenberg> SAC305 melts at 219
<cr1901_modern> i.e. I was out at a local store
<whitequark> LOL
<azonenberg> So no chance of it working
<cr1901_modern> azonenberg: Well, good I dodged a bullet
<rqou> sfe apparently used to use a skillet way back in the day
<cr1901_modern> Glad I went w/ my gut
<whitequark> rqou: wtf
<azonenberg> You can get hotter skillets
<whitequark> rqou: what was their budget
<rqou> $0?
<azonenberg> they kinda work, but its a terrible choice since it annoying and cant do 2 side placements etc
<rqou> :P
<azonenberg> rqou: 450F, peak temp of most cheap toaster ovens, is about 230C which is barely enough to melt sac350
<azonenberg> sac305*
<rqou> sfe started in a dorm room too afaik
<azonenberg> its below the ideal process window, but good enough for prototyping
<cr1901_modern> But it _can_ melt sac305?
<azonenberg> 450F / 230C does melt sac305, it just doesnt get to the optimal temperature for a really good solder joint
<whitequark> lol what the fuck cypress
<whitequark> your usb 3 chip *still* doesn't have Hs-mode I2C
<azonenberg> and it melts kinda slowly since the air temp is only a few degrees above the solder melting point
<azonenberg> and heat transfer is dependent on delta T between the endpoints
<cr1901_modern> azonenberg: Like I said, I _almost_ bought the skillet. But realized last minute that "I should've done more research".
<rqou> for a single bga you can always try blasting it with a rework station
<rqou> whitequark: because nobody uses Hs-mode?
<azonenberg> i really dont understand all of the obsession with strange smt assembly techniques
<whitequark> rqou: most times when I did that I delaminated the board before anything reflowed
<cr1901_modern> Trying to save money here rqou
<azonenberg> when putting it in an oven and turning it on until solder melts is so simple and practically idiotproof
<whitequark> I literally had more success with a blowtorch
<rqou> azonenberg: isn't a rework station a standard technique too? (well, for rework)
<azonenberg> about the only way you can screw up is setting the temperature wrong, or not placing the part on the right spot on the board, or grossly mis-printing the paste
<whitequark> azonenberg: Galden FTW
<azonenberg> rqou: yes but its a ton more work and gives less good results
<whitequark> can't even get the temp wrong :P
<azonenberg> whitequark: if i had vapor phase maybe :p
<whitequark> azonenberg: do you want some?
<cr1901_modern> azonenberg: Could you get a pic of your oven from the garage at some point tonight?
<whitequark> I have like 2L
<azonenberg> whitequark: maybe at the new lab
<rqou> anyways, i do intend to move to using a reflow oven eventually
<cr1901_modern> azonenberg: Or failing a pic, the make/model?
<rqou> probably for the hopefully-not-vaporware "Guren" project
<cyrozap> I don't know why people think BGA is so hard. Literally all you have to do is 1) Go to macrofab.com (or similar PCBA company website). 2) Upload your PCB artwork and BOM. 3) Gib moni. 4) Receive assembled PCBs, complete with those tricky-dicky BGA parts.
<rqou> lolol
<whitequark> cyrozap: $$$
<whitequark> also try that if you're in like Brazil or Russia
<balrog> cr1901_modern: what's the quality of macrofab BGA been like?
<cr1901_modern> balrog: I didn't use it
<balrog> I meant cyrozap. oops
<awygle> cr1901_modern: BLACK+DECKER TO1675B 6-Slice Convection Countertop Toaster Oven, Includes Bake Pan, Broil Rack & Toasting Rack, Stainless Steel/Black Convection Toaster Oven https://www.amazon.com/dp/B0060VQFQ6/ref=cm_sw_r_cp_apa_i_HLl8Ab2R9QDYY is what I bought, it is working well so far
<awygle> Wow all right amazon
<cr1901_modern> awygle: Well if you do Controleo w/ that oven, I can just ask you q's if I get stuck :P
<azonenberg> Yeah that is not the same oven as mine
<azonenberg> But looks equivalent
<azonenberg> same key specs
<awygle> I haven't done the install yet although I do have the parts
<cyrozap> balrog: I actually haven't sent a BGA design to them yet, though I will once I finish the one I'm working on. They have PnP machines, so I assume they'll do a better job making those boards than I ever would.
<cr1901_modern> azonenberg: There's no cookies setting...
<balrog> cyrozap: they use a paste printer
<balrog> which aiui works down to 0.4 but not below
<azonenberg> cr1901_modern: Thats just a preset, mine is a fancy mcu-controlled one
<awygle> I got publicly called out by FOSSi and had to assembly smolfpga in a hurry :-P
<balrog> 0.4mm that is
<azonenberg> cookies just means "all heating elements on, convection on"
<awygle> In this oven the equivalent is "toast" +"fan"
<azonenberg> they have various presets that map to enables for top heat, bottom heat, fan, and rotisserie motor
<rqou> wait azonenberg your toaster isn't asic-powered?
<rqou> no "BAGEL" pin?
<cr1901_modern> Where's the fan button :)?
<azonenberg> rqou: i havent opened it up so i have no idea what the internal controller looks like
<awygle> It's on the second dial
<azonenberg> it is quite possibly a mask rom asic
<balrog> azonenberg: more likely some weird mcu
<awygle> "bake"
<awygle> With a picture of the fan
<awygle> So "toast" "bake" "stay on"
<awygle> Is what I use
<rqou> lol
<cr1901_modern> Okay I think I can follow those insns
<awygle> Beyond that, just make sure to center the board front/back and it should be good
<cr1901_modern> I take it you bought the convection version of controleo?
xdeller__ has joined ##openfpga
<awygle> Yup
<whitequark> awygle: should we add slew rate limiting resistors
<awygle> whitequark: probably
<awygle> at least space for them
<whitequark> right
cyrozap-ZNC has joined ##openfpga
<openfpga-github> [Glasgow] whitequark opened issue #35: Slew rate limiting resistors https://github.com/whitequark/Glasgow/issues/35
bitd has quit [*.net *.split]
xdeller_ has quit [*.net *.split]
sielicki has quit [*.net *.split]
cyrozap has quit [*.net *.split]
<awygle> if your spring grounds are in HK i can run some tests when mine gets here, just to be sure
bitd has joined ##openfpga
<whitequark> yeah please do
<awygle> do you have a stance on RPACKs vs discrete resistors?
<openfpga-github> [Glasgow] whitequark commented on issue #35: This will also add some very limited pin overcurrent protection. Say at 200 ohm, the buffers should tolerate ±10V on up to two signal pins (not Vio or Vsense though). https://github.com/whitequark/Glasgow/issues/35#issuecomment-387203448
<whitequark> awygle: definitely arrays
<awygle> mk, cool
<whitequark> there isn't space for 8 discrete resistors in that area of the board
cyrozap-ZNC has quit [Client Quit]
<awygle> that is true
cyrozap has joined ##openfpga
<whitequark> you'll still have to move the DAC/LDO I think
<whitequark> a little bit
<awygle> sure, nbd
<rqou> do arrays still have crosstalk issues?
<whitequark> arrays have crosstalk issues?
sielicki has joined ##openfpga
<rqou> i though they could?
<cyrozap> whitequark: Yeah, if you're outside of the US MacroFab isn't as handy, but Seeed Studio does PCBA as well. As for the cost, I put a very high dollar value on my limited free time, so the amount I pay MacroFab is much less than the opportunity cost of taking the time to build PCBs myself. Plus by having MF build it, I don't have to worry about me borking the assembly and I don't need to own and operate
<cyrozap> a reflow oven in my apartment.
<openfpga-github> [Glasgow] whitequark opened issue #36: Measure FXMA108 slew rate https://github.com/whitequark/Glasgow/issues/36
<whitequark> cyrozap: I am not really willing to tolerate the turnaround time of PCBA houses
<whitequark> they have high lead time, and the shipping is slow as hell too
<awygle> you can get crosstalks with RPACK but ime it's never been a huge issue.
<cyrozap> whitequark: Maybe you could pipeline your projects ;)
<awygle> for me, paying for PCBA dramatically increases the cost of prototyping
<rqou> awygle: ah ok, that's nice
<daveshah> I am fortunate enough to have a desk in a PCBA house right now - https://factoryhub.at
<rqou> I've personally avoided resistor arrays after borking their footprints too many times
<daveshah> It's what we might test your board on whitequark
<awygle> and while yes, assembly does take time, it's a different headspace than programming or design work, so it doesn't come out of the same time "bucket" if that makes sense
<whitequark> daveshah: oooh interesting
<rqou> also I tend to have resistors on hand but not arrays
<awygle> daveshah: that looks cool
<whitequark> daveshah: let me know before you do it, I will update revB design files
<awygle> rqou: yeah i had at one point ~200 22R resistors
<daveshah> whitequark: it's not really our kind of place, the main line, they currently are 0603+
<awygle> for this purpose
<daveshah> But they want to move to 0402
<awygle> glasgow is 0603
<rqou> i think i have ~$9.9k 22R resistors
<rqou> :P
<rqou> oh woops no $
<daveshah> awygle: perfect, then we can get passives from stock
<awygle> i only have a reel of 10k currently
<rqou> just 9.9k resistors
<whitequark> I know someone who wanted 80 0R resistors
<rqou> yeah, one reel
<whitequark> but he accidentally ordered 80 reels instead
<awygle> oh i have a reel of 0R also
<whitequark> but probably not eighty
<rqou> i mean, reels aren't that expensive
<whitequark> or maybe it was twenty? some stupid large amount
<rqou> and I'm never going to use it up
<whitequark> he sold it to a local retailer at something like 1000% markup though
<rqou> ~$40 for 10k parts
<whitequark> rqou: why did you buy a reel of 22R
<whitequark> I understand a reel of 0R or 100R
<rqou> usb i think?
<whitequark> a *reel*?
<whitequark> why?
<daveshah> 22r is generally quite useful
<rqou> or maybe I don't actually have that
<daveshah> As a termination and slew rate limiting value
<rqou> I'd have to check
<whitequark> hm sure
<awygle> yeah 22R is my default series termination slash slew rate resistor (those are really the same thing)
<daveshah> But IMO 22r arrays are more useful than a reel of single resistors
<awygle> the problem is you can't stock arrays usefully
<awygle> because if you stock a 4-position array you'll inevitably hit a 5-bit bus
pie_ has joined ##openfpga
<whitequark> just use two? :P
<awygle> well, yes :p
<awygle> i use discretes for prototyping and then arrays for production. but i had a boss that hated arrays on principle so i always ask people.
<daveshah> I would use arrays for prototyping too for a big bus
<whitequark> hm there are 8-position arrays too
<awygle> yeah, if it was like 32-bit i would definitely use arrays at that point
<awygle> i have a card on my desk here that puts _every_ pin of an FMC connector through an 0402 resistor footprint...
<whitequark> awygle: so I'm thinking a 0402 8-position array
<whitequark> they're cheap and just the right size
<awygle> sgtm
<openfpga-github> [Glasgow] whitequark commented on issue #35: 0402 8-position array fits better (it's about the same size as FXMA108). https://github.com/whitequark/Glasgow/issues/35#issuecomment-387208118
<whitequark> hmmm what should I work on next
<whitequark> a CLI tool for setting the IO voltage.
<whitequark> awygle: oh your glasgow has already departed HK
<awygle> woo!
<whitequark> you know what would be weird?
<whitequark> send glasgow to glasgow
<awygle> lol
<whitequark> rqou: call your board Moscow or something
<awygle> Edinburgh
<whitequark> (well you have to choose a city you've never been within a hundred kilometers of)
<whitequark> awygle: Slough
<rqou> nah, Guren defeats the Glasgow :P
<whitequark> people will think i'm a weeb or something
<whitequark> wait
<rqou> while having unusable software that only one person can use :P
<cr1901_modern> whitequark: Uhhh
<awygle> haha
<cr1901_modern> we are long past that point :P?
<whitequark> thatsthejoke.jpeg
<rqou> unfortunately it won't come with a fanservicey waifu :P
<whitequark> rqou: ok yes that is a good name
<awygle> i don't actually remember much fanservice or waifuery in that show
<rqou> Kallen ends up in fanservicey scenarios a few times
<rqou> hmm, but now I'll have to figure out how to get red PCBs without being interpreted as a sfe product
<azonenberg> whitequark: why distant cities?
<whitequark> azonenberg: well, I've never been anywhere near Glasgow
<azonenberg> back when i was in grad school i picked nearby cities in upstate NY as names
<azonenberg> saratoga, cairo, etc
<whitequark> so to keep that sequence, rqou would have to pick a city he hasn't been near
<azonenberg> for CPU cores
<whitequark> >cairo
<azonenberg> Lol
<azonenberg> Cairo, NY
<azonenberg> Not the one in Egypt
<azonenberg> Although i'm sure the local city was named after the more famous one
<whitequark> yeah it sucks how americans name their cities like that
<whitequark> there's Moscow, ID
<azonenberg> There is a Vancouver in WA as well
<azonenberg> any time people talk about it i think about the one in BC
<whitequark> and Washington
<whitequark> I mean
<whitequark> who the fuck names the city the same as a state
<whitequark> or the other way around
<whitequark> I almost went to the city instead of the state once
<awygle> Cairo is in American Gods
<whitequark> oh yeah that book was good
<awygle> agreed
<awygle> also i blame you for the fact that i bought the first five amber books on saturday
<whitequark> hahaha
<awygle> (they were at my local used bookstore)
<awygle> i also bought all four volumes of azumanga daioh
<awygle> and have already completed them
<awygle> (they are short)
<zkms> whitequark: i was in Moscow, ID once
* awygle should really get back to work...
<whitequark> zkms: I want to go there just for amusement
<whitequark> it doesn't actually sound like a very interesting place
<cr1901_modern> Moscow ID == The Hollywood, FL of Moscow, Russia?
<whitequark> what's Hollywood, FL
<whitequark> oh the actual hollywood is in LA
<whitequark> god how does anyone figure anything out
<cr1901_modern> There's a city called Hollywood in Florida. No one cares about it b/c- yea, you got it
<whitequark> can you just rename every city to its zip code
<awygle> context
<awygle> like everything else
<awygle> Vancouver WA _is_ too close to Vancouver BC though
<whitequark> awygle: so apparently you can address any individual house in the US using a 5-4 format zip code
<cr1901_modern> the 4 is lot number
<whitequark> ok, lot
<whitequark> almost the same thing
<whitequark> why don't most people use their 5-4 zip code?
<awygle> whitequark: if you use the 5-4 zip code format your package gets automatically sorted to your particular mail carrier and arrives a day sooner on average
<awygle> (for USPS)
<whitequark> yeah. and you gave me a 5 format zip code :P
<awygle> yes :p
<awygle> i don't remember the last four
<cr1901_modern> Most ppl don't
<whitequark> lol ok
<awygle> because e.g. when using my credit card i only need the 5
<awygle> i look it up when i order things that i'm in a rush for though
<cr1901_modern> I know mine b/c... well, I've looked it up enough times that it stuck
<whitequark> i hate it when credit card processors want your zip code
<awygle> or like, it's stored on my amazon
<whitequark> i'm in hong kong assfucks
<whitequark> it doesn't have zip codes
<azonenberg> whitequark: oh, and my favorite
<azonenberg> I went to school right near Albany, NY (the one everybody knows about)
<azonenberg> was single at the time and browsing dating sites, saw this cute girl who listed her location as "Albany"
<azonenberg> we messaged back and forth a bit, decided to meet up
<azonenberg> i suggested some popular locations near a bus route and she acted all confused like she didnt know where they were
<azonenberg> like, a major mall downtown
<azonenberg> eventually we realized she lived in Albany, OR :p
<rqou> lolol
<awygle> heh
<rqou> wait a sec, azonenberg is a human? not an ace robot/machine? :P :P :P
<whitequark> azonenberg is human i confirm
<awygle> indeed
<rqou> (i make this joke despite the fact that i've met azonenberg's $WIFE)
<awygle> well, appears human at least
<whitequark> lol awygle
<awygle> you never really know for sure
* awygle REALLY should get back to work...
<whitequark> who's going to run voight kampff on azonenberg
<qu1j0t3> as long as they run a metal detector first
digshadow has joined ##openfpga
Bike has quit [Ping timeout: 260 seconds]
<gruetzkopf> Doesn't save us from Human-Form cylons or Westworld-stlye hosts
<genii> Or bodysnatchers.
* genii makes more coffee
<rqou> hmm some weird power quality issues going on here
<rqou> hmm and everything is fine again
<rqou> i wonder if this is related to some "utility work" going on a block or so away
<rqou> apparently a UPS is good for things other than passive-aggressive non-handling of our apartment's power problems :P
<whitequark> lol
<rqou> although this UPS is a little bit undersized still for what i want, and it probably has the typical shitty batteries
<rqou> <troll>where's ESR's "disruptive" UPS lolol?</troll>
<zkms> do they still use lead acid in UPSes
<rqou> yes
<awygle> Was that a thing?
<rqou> lead acid is still cheap
<rqou> awygle: was what a thing?
<zkms> depressing
<awygle> Disruptive ups
<rqou> but it's not going to go anywhere
<rqou> *) it's ESR
<rqou> *) ESR totally doesn't "get" how OSHW works
<awygle> Of course not, esr doesn't create things
<rqou> *) it's ESR
cpresser has joined ##openfpga
<balrog> "so little deep-cycle endurance" lol
<whitequark> "Batteries – should be more sophisticated than lead-acid; LiFePO (lithium iron phosphate) is suggested instead;
Panasonic has joined ##openfpga
<whitequark> "
<whitequark> uhhhhh
<balrog> in FIRST Robotics we deep cycle them
<whitequark> does he actually understand why people use lead-acid
<rqou> idk if he does
<balrog> literally the same batteries that are used in APC 1500VA units
<rqou> idk if i do either tbh
<rqou> my understanding is *) cheap *) easy to float charge
<balrog> running them down from 100% to around 50-60% in under 3 minutes
<balrog> rqou: yep
<whitequark> rqou: also: no memory effect if used correctly, and handles very high currents intermittently with no current at all
<whitequark> i.e. car starter
<balrog> also they do deep cycle reasonably well, as long as "deep" means you don't run them below 50-60% of charge
<rqou> unfortunately the batteries are still a little bit smaller than i would like
<rqou> ~15 minute runtime
<rqou> so i actually poked around esr's project and i don't see any detailed design of the power electronics section
<rqou> you know, the hard part
<rqou> maybe it's in the cloud eda that they're using, but i don't care enough to look much further
<whitequark> looool cloud eda
<rqou> yeah they're using upverter
<rqou> > *) ESR totally doesn't "get" how OSHW works
<rqou> even though i'm slightly interested in the idea of a better UPS, i definitely wouldn't build one alongside ESR
<whitequark> there are four cats on my bed rn and they all try to walk on me
<rqou> lol
<rqou> cute :P
<rqou> or not, if you're actually trying to work
<balrog> rqou: they're using cloud eda?
<rqou> balrog: upverter
<awygle> upverter isn't terrible
<rqou> sure, but it's not how most of the "OSHW ecosystem" works
<cr1901_modern> Why are the cats multiplying?
<rqou> "well little timmy, when a boy cat and a girl cat love each other very very much..." :P :P :P
<whitequark> cr1901_modern: my roommate forgot to give the female cat contraceptives once
<cr1901_modern> Ahhh
<rqou> whitequark: the cats aren't spayed?
<cr1901_modern> rqou: https://www.youtube.com/watch?v=-iDtR1yKJM0 Obligatory
<whitequark> rqou: the male cat is castrated because he was being an annoying little shit
<whitequark> now he's a fat slightly less annoying very big shit
<cr1901_modern> fat cats are good
<whitequark> IME when you sterilize cats they get fat and slightly depressed
<whitequark> so, since contraceptives work pretty well on female cats, I decided we're going with that
<cr1901_modern> Are you keeping the new kittehs?
<whitequark> only one
<rqou> whitequark: where can you get cat contraceptives?
<whitequark> rqou: literally any pet shop?
<rqou> huh
<whitequark> they work on male cats too btw
<rqou> oh wat
<rqou> how?
<whitequark> just kind of worse
<whitequark> um
<whitequark> progesterone disrupts the HPG axis
<rqou> why don't we have good human male contraceptives?
<whitequark> specifically it inhibits LH/FSH production
<whitequark> we do
<rqou> we do?
<whitequark> they just lead to erectile dysfunction
<cr1901_modern> >slightly depressed. Well, I hope that's not the case w/ current cat. She was a stray who just "showed up". No interest in outside since she got here. I take her chubbiness as a sign she's content.
<rqou> aah
<whitequark> well that and having both low T and E2 levels does get people depressed directly
<rqou> controversial question: do animal contraceptives work on humans?
<whitequark> and osteoporosis on large timescales
<whitequark> um.
<rqou> can this be used to bypass idiotic republicans?
<whitequark> they're literally the same drug but in like 100x less concentration
<whitequark> also you can just buy them at an internet pharmacy
bitd has quit [Quit: Leaving]
<rqou> why isn't there a "master document of how to get around republicans trying to screw you over"?
<whitequark> there is you just aren't aware of it :P
<rqou> ok, maybe
<whitequark> speaking of male contraceptives, we have other, better options
<whitequark> but they tend to lead to permanent infertility
<whitequark> those silly pesky humans tend to dislike it
<whitequark> well, I opt for the radical option. no balls, no problems :P
<awygle> lol
<rqou> i mean, i know about vasectomies
<whitequark> no, I mean drugs
<Zorix> snip snip
<awygle> there's a "superglue in your vas deferens" drug iirc
<rqou> i thought that's still in clinical trials or something?
<whitequark> I don't think it was superglue, it's a gel
<awygle> depends where you live
<whitequark> and yes, it's not approved yet
<whitequark> not by FDA anyway
<awygle> right. iirc it's common in india?
<awygle> such a weird thing for people to get hung up on anyway. condoms. they exist.
<whitequark> not sure what the status is
<whitequark> some people have latex allergies
<awygle> that's true
<whitequark> if we're talking about india, most people are pretty poor
<awygle> but that's not usually who you hear talking about male contraceptives
<whitequark> also, they're just plain out inconvenient and sort of gross
<awygle> it's usually dudes terrified they're going to be "trapped" into child support
<rqou> hey, i heard that unlubed condoms are really useful for all sorts of miscellaneous purposes :P
<whitequark> imean, for once, it's a reasonable concern, isn't it
<whitequark> if those people *want* to not have children, you ought to encourage it
<rqou> iirc i remember reading a story on the internet (so it must be true! /s) about a sound studio trying to make a dripping water sound
<awygle> can't argue with that :p
<rqou> and they eventually made a convincing-enough one by putting a condom on a microphone and dipping it into some motor oil
<rqou> but then they didn't clean up and left it in the corner of a room
<rqou> :P
<whitequark> rqou: so, how often do you do that
<rqou> the executives the next day were not amused :P
<whitequark> ##opencondoms
<rqou> lolol
<cr1901_modern> I'll take "things that didn't happen for $100 Alex"
<rqou> i've never needed anything like that
<rqou> i mean, it's a plausible story but idk how true it is
<qu1j0t3> motor oil vs condom who will win
<whitequark> motor oil against latex
<whitequark> ... what else do they make condoms of? I immediately thought about nitrile but that's gloves
<whitequark> come to think of it, nitrile gloves provide way better sensitivity than latex ones
<cr1901_modern> sheepskin
<zkms> polyurethane?
<whitequark> zkms++
<whitequark> oh wtf they DO make nitrile ones
<rqou> oh what
<rqou> i did not know that
<whitequark> female condoms only apparently
<rqou> are they purple? :P
<whitequark> because they are oil resistant specifically
<rqou> huh
<whitequark> i.e. so that you don't need to rely on your partner being sane and asking
<whitequark> it's kind of fucked up
<rqou> yeah
m_w has quit [Remote host closed the connection]
<rqou> actually whitequark: why are lab nitrile gloves usually purple?
Bike has joined ##openfpga
<whitequark> rqou: not a faintest clue
<whitequark> raw nitrile is yellowish
<whitequark> i think it's just tradition
<awygle> i have black nitrile gloves *shrug*
<sorear> Unfortunately, explicit communication violates American mores
<rqou> i used to have gray but now i have purple
<whitequark> oh right, america
<rqou> the internet claims that medical nitrile gloves are traditionally blue to indicate "not latex"
<sorear> rqou: same pair?
<rqou> which were traditionally off-white
<rqou> sorear: no, totally different
<rqou> idk where purple comes from though
m_w has joined ##openfpga
<whitequark> rqou: there's also transparent vinyl
<rqou> there is somebody on the internet claiming that purple in some cases indicated "not sterile"
<rqou> i've never seen anybody using transparent gloves
<whitequark> hm but i have plenty nonsterile blue ones
<zkms> ugh
<whitequark> transparent vinyl is definitely used
<rqou> oh wait i have
<zkms> transparent vinyl gloves are shit tier IME
<rqou> food service is usually transparent vinyl i think
<sorear> Solve the solvent compat problem with asbestos
<zkms> feel so weird and icky
<whitequark> they're not really good but they're cheap
<rqou> hence use in food service?
<whitequark> and they're more puncture-resistant than thin nitrile i think
<zkms> i like nitrile
<whitequark> i love nitrile
<rqou> yeah my default is nitrile
<whitequark> if you don't work with >68% nitric they're perfect
<awygle> hot take: file permissions are bad, actually
<rqou> i also have some butyl rubber for working with RFNA now
<zkms> thicc nitrile for heavy duty stuff like working on airplanes, thin nitrile for anything else
<whitequark> butyl rubber is good
<whitequark> thicc nitrile
<rqou> after an unintented xanthoproteic reaction made my parents "not amused" :P
<whitequark> why do they care
<whitequark> it's your life
<rqou> lol
<rqou> i still have a good relationship with them? :P
* whitequark has literally just used glasgow as a flashlight
<whitequark> fucking green leds
<zkms> flashlight?
<pie_> market glasgow as a flashlight
<zkms> that green LED is fit for a navigation light
<whitequark> zkms: yeah
<whitequark> hahaha
<pie_> ironically, youd probably make more money off it
<zkms> just need a model with red LED for the other side
<rqou> market glasgow as a fleshlight :P :P :P
* rqou is banned
<whitequark> doubt many people want a $70 green flashlight
<whitequark> that only runs off usb
<pie_> whitequark, add "smart" and a shitty default firmware
<whitequark> zkms: it does actually have a red LED already
<whitequark> butit's much less bright
<whitequark> which is weird because I matched them ...
<whitequark> ... ohhhh
<awygle> i wish i had a good red light flashlight
<zkms> matched them wit h what measurement
<whitequark> awygle: I matched them by luminous intensity
<rqou> human eye sensitivity curve?
<whitequark> yexactly'
<zkms> nice
<zkms> also red LEDs are better imo
<zkms> more traditional and better on the eyes
<whitequark> the red LED is for error
<rqou> pink :P
<whitequark> pink is expensive
<rqou> yeah i know
<rqou> is it made with two dies?
<awygle> candela are supposed to include human eye sensitivity
<awygle> isn't it?
<awygle> aren't they?
<rqou> i have no idea
<pie_> i dont imagine they would unless theres two definitions
<whitequark> oh no idea
<pie_> i think candela is just a nor....actually
<pie_> now i am also uncertain
<pie_> isnt it the typical unit for light intensity?
<Bike> wikipedia says they do
<pie_> photometry is weird
<pie_> theres like 20 different quantities
<Bike> i'd say that's complicated for an SI base unit but i remembered amperes exist
<rqou> lol
<whitequark> hahaha
<rqou> you can be a weirdo and use cgs electromagnetic units
<rqou> like measuring capacitance in centimeters
<Bike> nice
<pie_> and theyre all different and useful probably
<pie_> but they all sound like synonyms
<rqou> ah, just like electromagnetism
* rqou has a final soon
<Bike> i was confused enough when i was reading ancient schematics referring to µµF
<rqou> (still cramming)
<whitequark> Bike: is that pF
<Bike> 10^-12 yeah
<Bike> i don't think that prefix existed at the time?
<awygle> is cgs the "all constants are 1" system
<rqou> i think so?
<whitequark> isn't that natural units?
<awygle> hm, maybe
<whitequark> and cgs is centimeter gram second
<Bike> maybe depends on which constants
<Bike> planck units or bust, etc
<whitequark> but it also has different EM constants
<rqou> probably ^
<awygle> i just remember dealing with Real Units for years and then taking an honors physics class and them being like "na everything is 1 don't worry about it"
<rqou> lolol
<rqou> i've heard astrophysics works like that
<awygle> "construct a system of units such that every non-dimensionless constant is 1"
<whitequark> astrophysics: everything that isn't hydrogen or helium is a metal
<rqou> what
<cr1901_modern> azonenberg: So, I'm clearly 6 months behind, but now that Silego was acquired, how is the new management feeling toward openfpga?
<whitequark> they call all elements except H or He a metal
<Bike> yeah astro people do that
<rqou> how does that work?
<Bike> to be fair, that means metals are like, a tenth of a percent of all matter, or something
<Bike> so fuck em anyway
<whitequark> yeah they dont really care about metals a lot
<pie_> fuckin metals poisoning muh nuclear reactions
<whitequark> tats why
<Bike> i mean have you seen what bullshit you have to do to make carbon out of hydrogen
<Bike> it's basically impossible
<Bike> carbon is a myth
<pie_> make what out of hydrogen?
<Bike> metal
<pie_> what is that
<rqou> i mean, in that case oganesson is even more of a myth
<awygle> metallic hydrogen is a thing
<pie_> ( Bike you were supposed to say "carbon", to which i would reply "what" :PP)
<Bike> fuck jokes
<Bike> i got reminded how silly constants are
<Bike> look at me i'm the magnetic flux quantum, fundamental quantity of all reality, as it was specifically in 1990
<awygle> carbon is the triple-alpha process, right?
<Bike> yea
<pie_> you "just" have to put a lot of alpha in one place!
<awygle> yeah that's pretty crazy
<awygle> He3 is my favorite fusion reaction
<pie_> Bike, :O :D
<awygle> it's actually aneutronic and doesn't have any nasty side chains
<Bike> i could have swore i played one with actual triple alpha once
<pie_> Bike, ok this seems really cool xD
<Bike> http://dimit.me/Fe26/ maybe this one?
<awygle> p-b11 is a lie
<rqou> huh, why does ##openfpga not have a fusor yet?
<pie_> rqou, i was thinking about it
<rqou> azonenberg/whitequark please get on it
<rqou> or i guess pie_ too
<pie_> but then pie_ happened.
<whitequark> fusors are boring
<whitequark> well
<awygle> polywell! (to the tune of "monorail")
<whitequark> pyroelectric fusion is mildly interesting
<rqou> when you have four turbopumps you should be able to "just" whip up a fusor in a weekend, right? :P :P :P
<whitequark> sure
<whitequark> I probably have most parts
<pie_> psh fusors are like "hello world" right
<pie_> that should be like the #homecmos starter project
* pie_ crashes
* whitequark restarts pie
<pie_> pie_ is installing updates, please return in 6 hours
<pie_> Bike, do unstable nuclei emit stuff on decomposition in this?
<pie_> id be sad if it was just powers of two
<Bike> in the first one? don't think so
<pie_> aw
<openfpga-github> [Glasgow] whitequark opened issue #37: Allow flashing CYP_MEM with firmware https://github.com/whitequark/Glasgow/issues/37
<pie_> Bike, so doing random moves on the Fe26 one is kind of neat/fun/interesting
<pie_> kind of makes me want to make a screensaver of it on a larger board
<openfpga-github> [Glasgow] whitequark pushed 1 new commit to master: https://github.com/whitequark/Glasgow/commit/ad62d88fff3d42b114027a591becca1e44efca01
<openfpga-github> Glasgow/master ad62d88 whitequark: Add a `glasgow voltage` command-line verb....
<whitequark> awygle: hm what next
<awygle> whitequark: have you run the MPSSE core yet?
<awygle> (just curious)
<whitequark> awygle: nope
<whitequark> don't have the FIFO intrface yeet
<whitequark> or an adapter board for that matter. or a JTAG target
<awygle> hardware-wise... USB enumerates, FPGA bitstream over SPI works, does I2C to the EEPROMs work?
<whitequark> yes
<whitequark> to all
<whitequark> AFE works
<awygle> ADC, DAC, power all good. seems like the only hardware interface left to validate is the FIFO
<whitequark> yep
<whitequark> so i guess that's it
<whitequark> the timings on it are. nast.
<rqou> wait whitequark all of your commands are interpreted on the PC right? not a command-line parser on the device itself?
<rqou> hmm yes
<rqou> anyways, it'd be really nice if we had a freestanding not-gnuified readline
<azonenberg> rqou: new project?
<rqou> ENOTIME
<azonenberg> apparently there are several projects that are gpl'd specifically because they wanted to use gnu readline
<azonenberg> So a non-gnu readline library would be very helpful
<rqou> every time i dig into this it's a massive horror of legacy crap
<rqou> of stuff pretending to be an electromechanical typewriter connected to a modem
<azonenberg> lol
<rqou> see for example the _multiple_ CVEs that have resulted from terminal emulators making mistakes while implementing tab stops
* azonenberg continues to dream of a PoE keyboard that sends UDP packets full of UTF-8 keystrokes ready to feed right into an application
<rqou> would the IME run in the keyboard too?
<cr1901_modern> azonenberg: libedit is the BSD version of GNU readline
<rqou> would the keyboard have a screen for that?
<azonenberg> I'm not familiar enough with weird input methods for languages that have more characters than keys (say, chinese)
<azonenberg> to be able to come up with a good idea
<azonenberg> But it would definitely work fine for anything using latin, cyrillic, hebrew, or greek charsets
<azonenberg> or probably korean
<rqou> at least it might improve the situation of "herp derp i'm a vidya! what's an IME?"
<rqou> afaik korean has some basic input editing
<rqou> afaik korean unicode is also a shitshow
<zkms> in-keyboard IME would be....interesting
<cr1901_modern> hrm, upon further reading it looks like libedit is _NetBSD_'s response to readline, which means it's tied to their source tree
<cr1901_modern> azonenberg: here's your BSD readline http://thrysoee.dk/editline/
<rqou> either way it's not a drop-in replacement
<whitequark> libedit is used in llvm
<whitequark> it's not tied to netbsd
<whitequark> awygle: oh wtf
<whitequark> we could have tied SLOE and SLRD
<azonenberg> zkms: in general i'm a fan of having hardware be relatively self contained and speaking a sane API
<rqou> you mean HID reports are not a sane API? :P
<awygle> fpga-accelerated IME?
<awygle> whitequark: i'm SLOE today, what would that have done?
<whitequark> lol
<whitequark> saved a pin
<awygle> ah
<whitequark> on fpga
<awygle> a noble goal indeed
<azonenberg> So for example, a pointing device should report buttons and axis events, with a flag for relative vs absolute position
<rqou> you can already do that
<azonenberg> (which i think is basically how existing mouse stuff is handled)
<rqou> it's called HID reports
<whitequark> otoh on revB we dont really have any use for it
<azonenberg> But a keyboard should produce unicode characters
<azonenberg> not arbitrary scan codes that have no meaning outside the original ibm pc :p
<rqou> but afaik if your HID reports don't do the "expected" thing the OS will get confused
<azonenberg> And just define a framing format around it for modifier keys
<azonenberg> Yes, you need to not use hid for this
<azonenberg> this has to be a new os keyboard handling api :p
<whitequark> azonenberg: that... is too simplified
<rqou> azonenberg: um, we've fixed that? now it's arbitrary scan codes that have no meaning outside the USB HID protocol :P
<azonenberg> I didnt say it was easy to retrofit
<azonenberg> rqou: lol
<azonenberg> what i mean is, the application or os should not care what physical button you pressed
<whitequark> e.g. compose key
<azonenberg> us dvorak and us qwerty for example
<whitequark> i very much want to know what physical button i pressed so that compose works
<azonenberg> it doesnt matter which physical button i pressed
<azonenberg> G is G
<rqou> again, what should IMEs do?
<rqou> e.g. compose keys, "han/eng" keys, etc.
<rqou> also "half/full" keys
<azonenberg> Presumably be handled in the keyboard
<whitequark> you will need a display then
<azonenberg> IME popups might be a great use case for that little display the new apple stuff has
<zkms> what unicode character corresponds to shift/fn/ctrl/alt/whatever else
<azonenberg> Fn is a keyboard-level handler that changes what scancode is sent afaik
<azonenberg> i dont think the OS can even see that
<zkms> uh
<azonenberg> or did you mean numeric function keys?
<whitequark> yes but then it sends scancodes for like "volume up"
<zkms> it definitely can
<rqou> obviously U+2325 ⌥ and U+2318 ⌘ duh :P
<zkms> i just tried with xev
<zkms> Fn absolutely generates stuff there
<whitequark> not on my laptop
<whitequark> interesting
<azonenberg> interesting, it does on my thinkpad
<zkms> oh is this some acpi hell thing
<azonenberg> i guess some are handled at higher levels than others
<zkms> it might be which is actually really sad in so many ways
<azonenberg> anyway, you'd need to define some kind of framing around it
<rqou> also, what scancodes should you use for the "internet explorer" key or the "microsoft excel" key? :P :P
<azonenberg> a bit for control vs text
<whitequark> rqou: just make it send Win+R iexplore Return
<azonenberg> then either a utf8 text sequence of arbitrary length generated by the keyboard IME
<rqou> lolol whitequark
<whitequark> azonenberg: that won't work
<azonenberg> Or a control code for a non-printable key (backspace, arrow, etc)
<whitequark> it has to be self-synchronizing
<azonenberg> self synchronizing how?
<azonenberg> manchester coded text? :P
<whitequark> well, utf-8 itself is self-synchronizing
<rqou> huh so on my keyboard the "microsoft word" key doesn't seem to work
<azonenberg> i envisioned a udp packet per keystroke / ime event
<azonenberg> rqou: those are "special appN" keys
<rqou> the "internet explorer" key eventually ends up becoming XF86HomePage
<whitequark> ah ok
<azonenberg> which you define a function for in the os
<zkms> whats wrong with just sending scancodes on press/release
<rqou> also why does XF86HomePage exist?
<zkms> i dont understand the problems with that
<azonenberg> whitequark: again the dream is a keyboard that's connected to a PoE host in the PC
<azonenberg> zkms: because scancodes (should) have no significance outside keyboard firmware
<azonenberg> thus the os shouldnt deal with it
<azonenberg> keyboard layouts, etc are all hardware problems
<azonenberg> The OS should just see text
<rqou> huh oh wat: XF86Calculator
<rqou> also the "windows media center" key turns into XF86Tools
<cr1901_modern> >also why does XF86HomePage exist
<cr1901_modern> rqou: B/c some XFree86 dev wanted to copy microsoft :P?
<rqou> but why doesn't my "microsoft word" key work? :P
<cr1901_modern> B/c nobody wrote "XF86 Word"
<rqou> "staroffice" didn't count?
<cr1901_modern> It might :P (Idk what was going through their heads. Xf86 was dead before I ever touched a Linux system.)
<rqou> ok, t-2hrs till final, time to start making my cheat sheet
* azonenberg read that as "chest seal"
<azonenberg> :P
<awygle> i have eliminated like 400 lines of code today
<awygle> i feel good
<azonenberg> :)
* azonenberg has been battling ARM JTAG all day
fouric has quit [Ping timeout: 256 seconds]
<azonenberg> i learned a ton about low level ADIv5 and Cortex-A9 debug architecture
<azonenberg> But i'm no closer to a working AHB transaction than i was in the morning :p
<awygle> that's because knowing is only half the battle