<GitHub164>
j1soc/master 7633b71 Cristian Paul Penaranda Rojas: Update to 14.6 ISE
GitHub164 has left #qi-hardware [#qi-hardware]
<kristianpaul>
eww
<kristianpaul>
that stills works :p
sb0 has joined #qi-hardware
fengling has joined #qi-hardware
atommann has joined #qi-hardware
xiangfu has joined #qi-hardware
<wpwrak>
kristianpaul: great, eh ? :)
<wpwrak>
xiangfu: heya ! how's life ? in the last days, i made a bit of progress with anelok. after the boost converter rework, it can run on battery power. at least the basic system (MCU and OLED). so i'd recommend applying the rework to the devices you have as well.
<xiangfu>
wpwrak: great. I am slow on anelok. I will try that.
<wpwrak>
xiangfu: we may see problems with the memory card, though. the worst-case power demand of these cards can go up to 300 mW, while the CR2032 can at best deliver a bit more than 100 mW.
<wpwrak>
what i don't know yet is how bad they are in real life. presumably a lot nicer than that, but only measurements will tell.
<wpwrak>
if they need more power than the battery can supply, then the current design won't work and we'll have to switch to AAA or such :-(
<wpwrak>
(current design) that is, mainly the mechanical design. the electrical side would almost be identical. to if one doesn't mind an ugly case, the same PCB could be used. at least in theory :)
<wpwrak>
for experimenting with the memory card, i'd like to integrate some things from your code. i think it may be easiest if i just cherry-pick, so you don't have to generate pull requests. does that sound okay for you ?
<xiangfu>
sure.
<xiangfu>
Yes. free feel to cherry-pick
<wpwrak>
great. i'll do this during the long easter weekend.
fengling has quit [Ping timeout: 272 seconds]
fengling has joined #qi-hardware
atommann has quit [Ping timeout: 250 seconds]
archang has quit [Ping timeout: 246 seconds]
archang has joined #qi-hardware
xiangfu has quit [Ping timeout: 272 seconds]
archang has quit [Ping timeout: 264 seconds]
xiangfu has joined #qi-hardware
atommann has joined #qi-hardware
xiangfu has quit [Ping timeout: 264 seconds]
xiangfu has joined #qi-hardware
wolfspraul has joined #qi-hardware
wolfspraul has quit [Client Quit]
wolfspraul has joined #qi-hardware
jekhor__ has joined #qi-hardware
xiangfu has quit [Ping timeout: 250 seconds]
xiangfu has joined #qi-hardware
xiangfu has quit [Ping timeout: 255 seconds]
xiangfu has joined #qi-hardware
xiangfu has quit [Remote host closed the connection]
pcercuei has joined #qi-hardware
<DocScrutinizer05>
wpwrak: ((presumably a lot nicer than that)) uSD? don't bet on that, constant writes on uSD heat them up to temperatures where you start to worry
<DocScrutinizer05>
then otoh *reads* are prolly way nicer, due to operation principles
<DocScrutinizer05>
how about CR2045?
<DocScrutinizer05>
wouldn't change THAT jmuch on your design, only 1 1.3mm higher case, right?
<DocScrutinizer05>
s/ 1 / a /
<qi-bot>
DocScrutinizer05 meant: "wouldn't change THAT jmuch on your design, only a 1.3mm higher case, right?"
jwhitmore_ has joined #qi-hardware
jwhitmore_ has quit [Ping timeout: 252 seconds]
jekhor__ has quit [Ping timeout: 264 seconds]
jekhor has joined #qi-hardware
wolfspraul has quit [Quit: leaving]
wolfspraul has joined #qi-hardware
wolfspraul has quit [Quit: leaving]
wolfspraul has joined #qi-hardware
jwhitmore_ has joined #qi-hardware
atommann has quit [Quit: Leaving]
fengling has quit [Quit: WeeChat 1.0]
<wpwrak>
(much nicer) well, the idea is to use the memory card only very infrequently, especially writes. and we can go slowly - do a sector, take a break for a few dozen milliseconds, etc.
<wpwrak>
cr2045 seems a bit on the exotic side, isn't it ? digi-key don't even have that type
<wpwrak>
2450 seems to be somewhat common. internal resistance is still pretty high, though
<wpwrak>
about 12-15 ohm for cr2450 vs. 20 ohm for cr2032
<wpwrak>
if we assume an ideal cell voltage of 2.8 V, then max power of 2032 is ~100 mW and of 2450 ~160 mW
<wpwrak>
but yes, it could be a possibility luckily, anelok's battery holder is on a separate board, so it's very easy to accommodate other cells :)
<wpwrak>
i hadn't even thought of other coins. thanks !
<DocScrutinizer05>
yw
<DocScrutinizer05>
((much nicer)) the suspected problem case is page erase
<DocScrutinizer05>
you hardly can mitigate that
<DocScrutinizer05>
only way: provide buffer energy ("Hasso! Schlafende!") sufficient for one page erase
<wpwrak>
for the first round of devices, i could just buffer writes in internal flash, then commit them to the memory card when there's plenty of power. not ideal, but one would hardly notice
<wpwrak>
and i did the calculations for buffering worst-case power. it should actually be possible, but you need a fat supercap for it.
<wpwrak>
not the usual coin type or other small ones (internal resistance way too high) but there are some larger blocks that have a low resistance. not cheap, though, some USD 5+ apiece @ 1000 at digi-key
<DocScrutinizer05>
or you siply go for two cr2015 in series
<DocScrutinizer05>
wouldn't even need major rework in battery holder, I guess
<DocScrutinizer05>
double voltage, half the current
<DocScrutinizer05>
needs buck converter then
<DocScrutinizer05>
well, they're 'cheap'
<wpwrak>
2016 ? would need to change the converter, yes. the current one is only step-up
<DocScrutinizer05>
go for buck/boost type, should be "compatible"
kyak has quit [Ping timeout: 252 seconds]
<wpwrak>
resistance is the same as 2032, so one could indeed draw more juice
<DocScrutinizer05>
evidently, since you only need half the current at twice the voltage for same energy
<DocScrutinizer05>
of course, if the cr2016 had tiwce the ESR than CR2032, it would be a zero-sum equation
<wpwrak>
yup
kyak has joined #qi-hardware
kyak has joined #qi-hardware
<wpwrak>
but if i make major changes, i'd probably go to AAA then. ESR drops by some two orders of magnitude
<wpwrak>
i actually tried to feed my current boost critter with a low voltage, but it didn't like that for some reason. may be a lab effect, though - they specifically warn about the inductance of long feed lines from the lab power supply
<wpwrak>
the chop should be good down to 0.85 V to start, 0.5 V when running. idle current is too high, though, 0.5 mA, but there's a close cousin from TI that's much better
<wpwrak>
(lab effect) also, they're weasling in their reference circuit: the caps they specify are 4.7 uF, so they look all nice and efficient, but then they mention that this is effective capacitance, so you really need a bigger cap because real-life caps usually don't have their nominal capacitance when you actually make them do something
<wpwrak>
that's another pending rework item. and the we'll see if the critter can handle an alcaline from the next supermarket. i mean, even logitech can do it in their mice, so how hard can it be ? ;-)
<wpwrak>
i want to stay mainstream. the sort of batteries you can find in a village you visit for the first time and where you don't speak the local language :)
<wpwrak>
i.e., AA >> AA > CR2032. but then AA would be rather big.
<DocScrutinizer05>
go for AAA then
<DocScrutinizer05>
even CR2032 are sort of tricky
<DocScrutinizer05>
and friggin expensive
<wpwrak>
around here they're about 50 us-cents apiece if you buy a bunch. else, about 80 us-cents. so they must be single-digit cents in the rest of the world :)
<paul_boddie>
Argentina provides the "distressed economy" test, I guess. ;-)
<wpwrak>
guess that's happens when beavis and butthead grow up and become ministers
jekhor has quit [Ping timeout: 252 seconds]
<paul_boddie>
Wonder how much they are here... 29 NOK!
<paul_boddie>
I was reading about Argentine political history not that long ago. Enlightening and rather sad.
<wpwrak>
elections are in october. there will be a reckoning.
<wpwrak>
NOK 29 = USD 3.6. that's for how many CR2032 ?
<paul_boddie>
Is that going to be merged upstream at some point?
<pcercuei>
that's the plan
<pcercuei>
but I need larsc to finish the pinctrl driver first :)
<pcercuei>
(since this updated driver requires that you have a pinctrl driver)
<paul_boddie>
Linux dependency hell! ;-)
valhalla has quit [Ping timeout: 256 seconds]
valhalla has joined #qi-hardware
dandon_ has joined #qi-hardware
dandon has quit [Ping timeout: 244 seconds]
dandon_ is now known as dandon
wej has quit [Ping timeout: 250 seconds]
wej has joined #qi-hardware
jwhitmore_ has joined #qi-hardware
wej has quit [Ping timeout: 245 seconds]
jekhor has joined #qi-hardware
wej has joined #qi-hardware
jekhor has quit [Ping timeout: 272 seconds]
<paul_boddie>
Is it worth following device tree work for jz47xx? If so, where should I be following?
<larsc>
yes
<larsc>
non-devicetree is not going to be supported in the future
<paul_boddie>
I see that there's some jz4770 DT stuff, and maybe the MIPS people did some jz4780 DT stuff as well.
<whitequark>
wpwrak: that's because coin cells use methanol for electrolyte
<whitequark>
they're not really "battery" cells
<whitequark>
they're more like fuel cells with a semipermeable membrane
<whitequark>
and metallic li
jwhitmore_ has quit [Remote host closed the connection]
<pcercuei>
there's some jz4740 stuff as well
sb0 has quit [Read error: Connection reset by peer]
<paul_boddie>
I found a 3.18 branch in gcwnow for DT, but is that the "most upstream" place?
<pcercuei>
qi-hardware is the "most upstream" for jz4740, gcwnow is the "most upstream" for jz4770, MIPS/CI20_Linux is the "most upstream" for jz4780 :)
<paul_boddie>
I meant for DT. I didn't think the qi-kernel repo was being updated any more.
<pcercuei>
ah, then it's the jz4780 tree from the MIPS guys
<pcercuei>
we started moving our jz4770 tree to DT some months ago, but some parts (e.g. clocks driver) aren't easy
<paul_boddie>
OK. I'll have a look at what the MIPS people have for DT.
<paul_boddie>
Short of that, I'm tempted to do some dirty hacks to see if the "old" jz4740 PWM driver can be tweaked to work with jz4730 pin assignments.
<paul_boddie>
I think the timer/counter registers are the same between the jz4730 and jz4740.
<pcercuei>
if you want to do it cleanly, bump larsc - IIRC he has a pinctrl driver ready for jz4740
<pcercuei>
the pin assignments are done in DT so it would probably work on jz4730 as well
<pcercuei>
(e.g. we use the exact same pinctrl driver for jz4770 and jz4780 now)
<pcercuei>
if the subsystem didn't change too much it may even be used for older versions
<paul_boddie>
I think the confusing thing when looking at the old (2.6) code is, apart from kernel coding changes, the way some things like PWM are done.
<paul_boddie>
It's hard to tell from vendor code, though, as they dump their internal knowledge to C.
<paul_boddie>
But since the timer/counter unit looks the same, we can probably ignore weird vendor stuff and go the tried and tested route.