<philippe_fouquet>
Whi if i don't powered the RTC_VCC the fel mode don't start?
<libv>
Turl: oh, ok, i was under the impression that such emails were sent with the date they were actually sent
<libv>
perhaps this changed with a newer git version
<libv>
or perhaps i never did notice that
FDCX has joined #linux-sunxi
FergusL has joined #linux-sunxi
<philippe_fouquet>
no body with RTC_VCC
avsm has joined #linux-sunxi
shineworld has joined #linux-sunxi
shineworld has quit [Changing host]
shineworld has joined #linux-sunxi
shineworld has quit [Read error: Connection reset by peer]
<libv>
hramrach: please check the previews of wiki changes you make
Akagi201_ has joined #linux-sunxi
shineworld has joined #linux-sunxi
shineworld has joined #linux-sunxi
<hramrach>
what do you have problem with?
<libv>
missing ]
<libv>
but i'm about to fix it
Akagi201 has quit [Ping timeout: 240 seconds]
<hramrach>
I do not see any
<libv>
around plastic tool.
Andy-D has joined #linux-sunxi
<libv>
i usually catch those while looking at a preview, or when i make a minor change, shortly after i committed it :)
ssvb has quit [Remote host closed the connection]
<libv>
great. thanks heaps. now i get to copy paste.
hipboi_ has joined #linux-sunxi
hipboi has quit [Read error: Connection reset by peer]
popolon has joined #linux-sunxi
Akagi201_ has quit []
shineworld has quit [Quit: Leaving]
montjoie[home] has quit [Ping timeout: 260 seconds]
montjoie[home] has joined #linux-sunxi
kaspter1 has quit [Remote host closed the connection]
bengal has joined #linux-sunxi
avsm has quit [Quit: Leaving.]
arete74 has quit [Quit: leaving]
arete74 has joined #linux-sunxi
Zboonet has joined #linux-sunxi
Renard has joined #linux-sunxi
bengal has quit [Ping timeout: 245 seconds]
Skaag has quit [Ping timeout: 268 seconds]
Akagi201 has joined #linux-sunxi
Gerwin_J has joined #linux-sunxi
akaizen has quit [Ping timeout: 272 seconds]
kill_-9_1 has joined #linux-sunxi
Andy-D has quit [Ping timeout: 252 seconds]
Zboonet has quit [Remote host closed the connection]
Black_Horseman has quit [Remote host closed the connection]
avsm has joined #linux-sunxi
codekipper has quit [Quit: Page closed]
deasy has joined #linux-sunxi
rafaelMOD has joined #linux-sunxi
akaizen has joined #linux-sunxi
JohnDoe_71Rus has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org]
hipboi_ has quit [Ping timeout: 245 seconds]
hipboi_ has joined #linux-sunxi
ricardocrudo has joined #linux-sunxi
ricardocrudo has quit [Remote host closed the connection]
ricardocrudo has joined #linux-sunxi
avsm has quit [Quit: Leaving.]
cubear has quit [Quit: Leaving]
mmarker has joined #linux-sunxi
Gerwin_J has quit [Quit: Gerwin_J]
<kill_-9_1>
What should I do to the Kurio 7S?
bsdfox has quit [Ping timeout: 246 seconds]
Gerwin_J has joined #linux-sunxi
bsdfox has joined #linux-sunxi
naobsd has quit [Quit: Page closed]
Kevinxp has joined #linux-sunxi
Kevinxp1 has joined #linux-sunxi
Kevinxp has quit [Ping timeout: 246 seconds]
Kevinxp has joined #linux-sunxi
Kevinxp1 has quit [Ping timeout: 246 seconds]
<Kevinxp>
hi guys, i need help with my tablet with a13 processor, i have succesfull runing debian with micro sd, but only boot if have uart1 conected to pc ,, if remove uart conector from board , it cant boot , i dont understan .why ..
<kill_-9_1>
kevinxp: Do you need to type something to UART before it boots?
<Kevinxp>
nop i dont need type. .
<Kevinxp>
if have uart conected it boot normaly ..
<Kevinxp>
but if not, cant boot ..
<kill_-9_1>
kevinxp:Does Android works without UART?
<Kevinxp>
yes.
<kill_-9_1>
kevinxp: Did you try to add a resistor to ground for Tx.
<kill_-9_1>
?
<Kevinxp>
no i dont try .
<kill_-9_1>
kevinxp: Did you try the right script.fex?
<Kevinxp>
i have converted script.bin to .fex
<kill_-9_1>
kevinxp: The one for your device?
<Kevinxp>
uhm i tried with A13 olinuxino and other that found in nanda
<Kevinxp>
ah, i boot without uart, and when conect uart in terminal say Unknown command 'U-Boot' - try 'help' sun5i#
<Kevinxp>
i tried re-build spl with a13 olinuxino , a13 olinuxino_fex , sun5i , but nothing..
<kill_-9_1>
kevinxp: Normal. Add resistors ( your Rx and Tx pins are connected together)
<kill_-9_1>
so the U-boot message is redirected as a command
<Kevinxp>
ok a resistor ( tx to gnd, rx to gnd ?.)
<Kevinxp>
resistor what value ?..
<libv>
kevinxp: which tablet is this?
<libv>
kill_-9_1: did you get a uart already?
<Kevinxp>
hahaha is a qbex slimpad 7... HEY !! the resistors works !!
<philippe_fouquet>
for me pullup to 3.3V on UART0_RX and mothing on UART0_TX
<philippe_fouquet>
with 10k
<kill_-9_1>
libv: I have just a array of pins.
afaerber has joined #linux-sunxi
<Kevinxp>
already used ( tx to gnd with 1K ) (rx to gnd with 1k ) its works !!
<libv>
kevinxp: follow the new device howto on our wiki
<Kevinxp>
ok.. sorry for my english .. i speak spanish .. its hard understand.
<Kevinxp>
question, always need use a resistor in uart ? or have other method
<kill_-9_1>
kevinxp: I started learning Spanish this year, I do know only writing the date and a few basic things. The resistor should normally be in the PCB. Just Chinese saving every penny.
<rafaelMOD>
kevinxp: its good to put a 100R in series with TX and RX, for current limiting and a 10k pullup on RX, and allways check if RX and TX are not short-circuited!!
naobsd has joined #linux-sunxi
<Kevinxp>
ok thnks. .
enrico_ has joined #linux-sunxi
<kill_-9_1>
There are always weird things done by Chinese,...
<kill_-9_1>
(in modern tech)
<Kevinxp>
kill_-9_1: good !, i need understand more english i can read and write but so so .. sometimes need a google translate for some verbs
<Kevinxp>
ok this tablet no have uart port, i have solder the pin from processor
<kill_-9_1>
kevinxp: You are crazy.
<libv>
kevinxp: and then you are wondering _why_ you have problems?
<kill_-9_1>
( soldering a pin like that can destroy things)
<libv>
kevinxp: please stop wasting our time
<Kevinxp>
sorry libv but my time is dont your time. .
<Kevinxp>
i have succesfull run debian ..
<Kevinxp>
its no a loss time ..
<kill_-9_1>
kevinxp: Can you document what you did somewhere , with photos (bad quality is acceptable)?
<libv>
kevinxp: you are here asking questions, about some bad hacks you did, and you did not document your device in any way, and you did not describe how you so badly attached a uart
<Kevinxp>
men , you staring with document ??
<libv>
yes.
<Kevinxp>
or were born knowing that _ .
<libv>
we've seen more than a hundred devices go past
<Kevinxp>
ok its your ..
<libv>
and we have everything nicely documented
<Kevinxp>
i dont !.
<kill_-9_1>
kevinxp: If you doesn't know english well, write in your preffered language.
<Kevinxp>
sorry ..
<libv>
kevinxp: because you did not have the patience to read our wiki
<libv>
it couldn't be more splattered across our front page
<kill_-9_1>
(better than nothing)
<libv>
but you keep using a language barrier as an excuse
<libv>
language barriers can be overcome, with patience and google translate
<libv>
so instead you went ahead, and you did some really stupid shit, and now you are here asking questions as to why your device doesn't boot
<libv>
you're lucky you didn't fry anything (essential)
naobsd has quit [Quit: Page closed]
<Kevinxp>
kill_-9_1: thnks ! allready missing the resistors .
<libv>
Turl: i am quite amazed how he got to that stage
<libv>
i am even more amazed that he thought he could just ask "why does my device not boot" without mentioning any of that
Skaag has joined #linux-sunxi
<libv>
i am actually glad that he refuses to follow the ndh
<libv>
i hate to undo all the damage he clearly is capable of doing
<kill_-9_1>
I wonder how he soldered such small pins,...
<libv>
badly
Renard has quit [Remote host closed the connection]
<mripard_>
still, the only issue he had was a missing pullup. It's pretty impressive
<libv>
but then, with the right head, you can solder anything that attracts solder
<libv>
mripard_: yeah, it is pretty impressive that it still boots
<libv>
but i wonder what else he killed by doing that
<Turl>
chinese chips built like tanks:p
<mripard_>
probably not much. The package is much larger that the chip itself
jinzo has joined #linux-sunxi
<libv>
oh, the cut in the package is not what i worry about
<libv>
it's the leads to the pcb that vanished that worry me
<libv>
and what he did before he reverted to cutting open the chip itself
<kill_-9_1>
He is the exact example of what NOT to do...
<libv>
in many ways, indeed
<hramrach>
kevinxp: I would think the Chinese use u-boot without uart support. It does not stop booting because of noise on the uart pins then
<kill_-9_1>
libv: I think I would mention him somewhere in the Wiki. (what not to do in the bringing UART section)
<libv>
kill_-9_1: no, it would give people ideas
<kill_-9_1>
He even took pictures.
<libv>
kill_-9_1: how is things progressing with your kurio 7s?
<libv>
did you open up the device already?
naobsd has joined #linux-sunxi
<libv>
did you manage to get decent pictures of it?
<kill_-9_1>
libv: Half-opened then reclosed. ( there is two types of screws, and I don't have the tools to open the second part)
<Kevinxp>
omg its true !! yeah ! the noise cant do to boot
<kill_-9_1>
As I'm a student, not much time now, may continue in holidays or week-end if I have time.
<hramrach>
libv: you see, there is a part next to the CPU on the PCB so by cutting the package you make more room for soldering ;-)
<hramrach>
and by pulling the pins from PCB you avoid soldering them together with the nearby pins ;-)
<kill_-9_1>
hramrach: but that is not recommended.... :P
<libv>
hramrach: micro-sd breakout
<hramrach>
I used that, yes
<hramrach>
but it sire is not that much fun :DDD
<Kevinxp>
hahah
<kill_-9_1>
libv: for the uSD breakout, I can't do it because I also use the uSD for booting.
<hramrach>
we have nice howto for booting from USB
<kill_-9_1>
hramrach: Tried it long ago. May retry.
<kill_-9_1>
(with CB2,BananaPi and CT settings)
<hramrach>
I could not boot from SD because of noise on uart ...
<libv>
hramrach: what noise?
<libv>
is this the same as the short reported yesterday?
<wens>
rafaelMOD: oh nice
<hramrach>
no. whatever pins used for uart are connected *somewhere* on the board
<hramrach>
and with uart enabled in u-boot you cannot boot the 86vs
<hramrach>
actually, the rx pins are maybe not connected when only the tx pin is used in camera or something
<libv>
they tend to be connected somewhere for a reason
<hramrach>
so pretty much the same as this picture
<libv>
cutting them is not the sane solution
<hramrach>
they may not be connected at all. just soldered to the board because it's easier to just solder everything than pick
<hramrach>
but yes, there are methods of using an UART which do not require making changes to the PCB
<rafaelMOD>
wens: was hard to get them!
<rafaelMOD>
wens i was doing the schematics with XP221
<rafaelMOD>
they are pin to pint compatible
<hramrach>
pint of what?
<rafaelMOD>
:)
<libv>
heh, something tells me that my semitime g2 kernel failed for the same reason my mele a210 failed: the drv gpio pins for usb
<rafaelMOD>
what do you think about putting some hardware schematic KiCAD project files on the wiki?
<libv>
duplicate in one case, empty in the other
<libv>
rafaelMOD: for your project?
<libv>
rafaelMOD: make a device page
Kevinxp has quit [Quit: Page closed]
<rafaelMOD>
libv: the device is not ready yet, only schematics
<libv>
rafaelMOD: then wait until the device is ready with putting the schematics online :)
<rafaelMOD>
libv: still wip on the layout, its an A23 coreboard
<libv>
rafaelMOD: what else does it do?
<libv>
because we sure would love more hackable a23 boards here
<libv>
all we have is some crappy tablets
<rafaelMOD>
libv: its a coreboard deidcated to audio, specially the i2s interface. probably ready until the end of this year.
<wens>
rafaelMOD: do you have standard/sample layout of a23 board?
<hramrach>
it could have, like all the USB ports wired somewhere. Is there anything else the A23 has but the tablets don't?
<rafaelMOD>
wens: no I dont!
<rafaelMOD>
when the board is ready i am going to send some samples for you guys
<rafaelMOD>
also with a baseboard
<philippe_fouquet>
I don't know if interest some body but I have android 4.4 ready for A20 but my last jobs it to work with fel mode
<hramrach>
There are probably people interested
<libv>
rafaelMOD: that should really help a23 support, thanks
<hramrach>
maybe put some links to repos and images to the wiki?
<hramrach>
I am not an android fan but having a working repo as a starting point is always useful
<wens>
rafaelMOD: ah, too bad
<wens>
anyway, will wait for the axp223 datasheet upload
<wens>
then i'll compare against axp221
<rafaelMOD>
wens: whant me to send it to you by email?
<rafaelMOD>
wens: its even translated to english
<libv>
wens: it is already on dl.linux-sunxi.org
<rafaelMOD>
libv: i am getting: "403 Forbidden"
<libv>
ok
<libv>
will fix
<kill_-9_1>
philippe_fouquet: Interesting. (got FINALLY KitKat running on a Raspberry Pi , on my side)
<libv>
fixed permissions now
<rafaelMOD>
libv: the core will have only 512MB of RAM, need to keep it cheap, 4GB of nand and thats all
<hramrach>
I doubt A23 supports more
<libv>
rafaelMOD: what will it have for display?
<kill_-9_1>
rafaelMOD: No problem for NAND ( SD boot,).
<hramrach>
did not see any a23 board with 1G so far so debugging would be insane
<rafaelMOD>
libv: none, sorry, need to keep it cheap, and is for an audio product, not a mini-pc
<libv>
ok :)
<libv>
still, if it has uart, we should be able to do some hacking on it
<hramrach>
hmm, routing the pins for a display connector would be probably needless pain if you do not intend to use it
<philippe_fouquet>
kill_-9_1: for my part i have kitkat aosp ether patch valid when I have lot of time for put it on repo but before I must to try to used fel mode that it fail
<kill_-9_1>
rafaelMOD: For an audio project, a PIC would be cheaper (with USBTiny and a SMSC LAN controller)
<hramrach>
sounds more menacing than software developer manual for Netburst processors
<kill_-9_1>
(apparently, the Pi is still the king for high performance hardware, 24 GFlops 3D and 21Gflops DSP)
<kill_-9_1>
*cheap
<kill_-9_1>
hramrach: Error HTTP/503
<hramrach>
Samsung processors are probably fast because they do not cut costs on memory bus.
<hramrach>
but then they cost more than the Chinese chips, too
<hramrach>
also they use PoP so you get consistent bus configuration. no per-board memory controller tuning
<kill_-9_1>
hramrach: And BCM did choose to cut space for the ARM and PoP.
<kill_-9_1>
(BCM2835=1GByte BCM21154=2Gbyte)
ssvb has joined #linux-sunxi
<hramrach>
I cursed this PoP thing because it makes it impossible to make a Samsumg SoC device with a decent amount of ram. And you could actually use that ram there
_massi has quit [Quit: Leaving]
<hramrach>
but in practice AW chips are limited to .5G with two!! external chips
<kill_-9_1>
hramrach: Why? 512MB is decent.
<hramrach>
depends on the use. For decent desktop I need like 8G
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
<hramrach>
I can split the desktop tasks somewhat if the device takes 5W but that only gets you so far
<kill_-9_1>
hramrach: The AMD Opteron A1100 manages 128G.
<kill_-9_1>
(ARMv8)
bonbons has joined #linux-sunxi
<hramrach>
I do not need 128G unless I was trying to set up a VPS farm
<hramrach>
but yes, my desktop fits within that limit
<kill_-9_1>
hramrach: The Opteron A1100 manages DDR3 and DDR4.
<hramrach>
at this CPU size an Intel i3 of the last generation would be probably a better bet, though
<kill_-9_1>
hramrach: In performance, the A1100 is still winning ( 4 or 8 cores)
<kill_-9_1>
( doesn't need a chipset, unlike Intel)
<ssvb>
hramrach: which chips are limited to 512MB? a13?
<hramrach>
and probably a23 as well
astr has quit [Quit: Ex-Chat]
Skaag has joined #linux-sunxi
<ssvb>
hramrach: the a23 datasheet says "Support 1GB address space"
<hramrach>
nice. A10 is supposed to support 2GB but no such device exists. I will believe it wheh I see it
<kill_-9_1>
ssvb: There is the IO registers space,...
<ssvb>
hramrach: A10 supposedly can support 2GB when using two ranks
<ssvb>
hramrach: but everyone seems to only use single rank configuration
<hramrach>
and olimex did not get the dual rank configuration working
<ssvb>
hramrach: it could be a software problem, because some of the DRAM controller configuration knobs might be needed to set up correctly
<hramrach>
it can be any number of problems
<hramrach>
and since there is no docs and no reference design that is a problem
<ssvb>
hramrach: or it could be some hardware errata, but we don't have any errata lists available to us :-)
<hramrach>
or it may be that the board should have some different passive parts for which there is no reference design
<hramrach>
were the AMD processors even released?
<ssvb>
hramrach: but as for A23, it looks like it has DA0-DA15 address lines (A10 and A13 only had DA0-DA14)
<ssvb>
hramrach: so the 1GB RAM size limit seems to be believable
<ssvb>
hramrach: getting a better device with a 32-bit memory interface resolved this
<hramrach>
sadly I have no idea what memory bus the tablets have
<hramrach>
but memory bandwidth tends to be the most limiting factor with these mobile SoCs and the A10/A20 HD display issues show
<kill_-9_1>
It is bizzare that the cheaper Raspberry Pi has a better memory bus...
<kill_-9_1>
(and the more expensive Exynos)
leviathanch has joined #linux-sunxi
<ssvb>
kill_-9_1: afaik, Raspberry Pi has a 32-bit memory interface, so its theoretical bandwidth is roughly comparable to A10/A20 (also with a 32-bit memory)
<hramrach>
I tried running a flash game on a rk3188 tablet and it was really slow
<hramrach>
but without looking furthere into it I cannot tel if it's the flash player, network, memory, graphics or what
<kill_-9_1>
ssvb: When using the Pi in 4K@60fps with turbo mode, there should be a low latency controller,... ( for a 2011 chip...)
<ssvb>
kill_-9_1: however, the ARM core in the Raspberry Pi has some sort of a slow path to the DRAM controller and its performance sucks
<kill_-9_1>
ssvb: I crunch 1GBps from the VPU.
<kill_-9_1>
(not ARM)
<ssvb>
kill_-9_1: hmm, 1GBps is not really much
<hramrach>
hmm, it says 32bit in the log
<ssvb>
kill_-9_1: but yes, the VPU/GPU parts should have faster access to dram
bertrik has joined #linux-sunxi
bertrik has quit [Changing host]
bertrik has joined #linux-sunxi
Gerwin_J has joined #linux-sunxi
konradoo77 has quit [Ping timeout: 252 seconds]
<kill_-9_1>
ssvb: 1GBps with 4K at the same time(30fps through)...
<kill_-9_1>
I'm too lazy to compute the used bandwith from video,...
<hramrach>
I have the very same problem with rk and a23 kernels. The makefiles for manufacturer drivers reference files in different directory and the kernel build system is not designed to handle that. So it does not build
<ssvb>
kill_-9_1: 4K is how many pixels by how many?
<hramrach>
probably around 4kx2k
<hramrach>
3840x2160
<ssvb>
then it is 3840 * 2160 * 4 bytes per pixel * 30 frames per second = ~1GB/s
konradoo77 has joined #linux-sunxi
<ssvb>
roughly the same as driving two 1080p monitors at 60Hz (and Allwinner chips can do this)
<hramrach>
which ones?
<hramrach>
people had problems driving even a single 1080p display from a10 or a20
<hramrach>
the ML is full of complaints about that
<kill_-9_1>
ssvb:Is zerocopy useful for performance?
amitk has quit [Quit: leaving]
<ssvb>
kill_-9_1: the Raspberry Pi does not have any memory bandwidth advantage over A10/A20 with 32-bit dram, the theoretical bandwidth is comparable and the practical memory bandwidth is in the same ballpark
avsm has quit [Quit: Leaving.]
<ssvb>
kill_-9_1: yes, zerocopy is naturally useful, because we don't want unnecessary copies
<hramrach>
hmm, ok. so we use some hardware that was there for some video acceleration to get any graphics output at all. Makes the hardware unavailable for the actual acceleration but gets you a usable desktop
<hramrach>
that's a cool hack
<ssvb>
hramrach: the A20 does not have this problem in the first place :-)
<hramrach>
how about sun5i?
<hramrach>
do you have some data for that?
<ssvb>
hramrach: luckily, sun5i has no HDMI and can't use really high resolutions to trigger this problem
<ssvb>
hramrach: oh, A10s has HDMI, so somebody probably needs to run some tests
<diego71>
ssvb: hi. I'm testing the zq parameter, but looks like it's always = 0x7b, even if I set to different value i u-boot
<ssvb>
diego71: hmm, do you set 'odt_en=3'?
<diego71>
ssvb: no
Skaag has quit [Quit: Leaving.]
<ssvb>
that's it
<diego71>
ssvb: when i tried, kernel paniced during boot
Skaag has joined #linux-sunxi
<ssvb>
diego71: it probably means that the selected 'zq' value was unreliable
<ssvb>
diego71: which one have you tried?
<diego71>
3c and 6c
<ssvb>
diego71: maybe try to temporarily reduce the dram clock speed for making it bootable
<diego71>
ssvb: what does it mean odt_en=3 ?
<diego71>
(and = 0)
pietrushnic has joined #linux-sunxi
<ssvb>
diego71: it has two bits, one is enabling ODT for DQ lines (data) and another is enabling ODT for DQS lines (strobe)
<ssvb>
diego71: you could test different combinations of these bits, but odt_en=3 just enables ODT for everything
<diego71>
uhm, so zq values with odt_en=0 have some meanings?
<ssvb>
diego71: odt_en=0 means that you don't want to enable ODT
<ssvb>
diego71: I suspect that the 330 ohm RZQ in your board makes the optimal 'zq' value a bit different from the good values for Cubieboard/Cubietruck (with 240 ohm RZQ), so starting with '3c' initially degrades reliability
<ssvb>
diego71: maybe '4f' would be better
<diego71>
ssvb: it's a wild guess or an educated guess?
<diego71>
(is it ...)
<ssvb>
diego71: 330/240 = 1.375
<ssvb>
diego71: to get the same impedance as with /3 and /12 divisors for 240, they need to be adjusted
<ssvb>
diego71: yes, it is a somewhat educated guess :-)
afaerber has joined #linux-sunxi
<ssvb>
diego71: actually /12 is a bit problematic, because it needs to be corrected to /16.5 under these assumptions, which is out of range
<diego71>
ssvb: meanwhile i'm testing your u-boot on a13. It looks it works fine with default dram settings ...
<ssvb>
diego71: maybe the two-digit 'zq' value is not going to work well for A20-Olinuxino at all
<diego71>
ssvb: 4f seems good. At least it boots :)
buZz has quit [Read error: Connection reset by peer]
<ssvb>
diego71: for A20-Olinuxino it might be necessary to directly program ZDATA and bypass ZQ calibration
buZz is now known as Guest49513
tomboy64 has quit [Remote host closed the connection]
tomboy64 has joined #linux-sunxi
Renard has quit [Read error: No route to host]
<ssvb>
diego71: the result of ZPROG->ZDATA conversion is shown in the brackets after 'zq' parameter in a10-meminfo log
<ssvb>
diego71: this value can be used for the 'zq' parameter too
<diego71>
ssvb: uhm. I'm not sure to have understand it. Usually you set the divisor, and then the dram controller find out the right value?
<ssvb>
diego71: the DRAM controller is using the 20-bit ZCTRL settings (4 coefficients, 5 bits each)
<ssvb>
diego71: these coefficients are used for the on-die resistors inside of A20
<ssvb>
diego71: but because these on-die resistors are supposed to be not very accurate, they are calibrated against the external high precision resistor
Andy-D has joined #linux-sunxi
<diego71>
that is the 240ohm resistor?
<diego71>
(330 on olinuxino)
<ssvb>
diego71: yes
Andy-D has quit [Remote host closed the connection]
<ssvb>
diego71: so the two 4-bit divisors from 'zq' are used to find the matching four 5-bit coefficients for the internal on-die resistors
<ssvb>
diego71: but if we load these four 5-bit coefficients directly, then the external resistor is not used and we don't care if it is 240 or 330 ohm
<ssvb>
diego71: now the main question is how big is the deviation of these on-die resistors for different chips
<diego71>
ssvb: so we need the coeffient because different revision of olinuxino have different resistor...
<ssvb>
diego71: not quite, these on-die resistors are inside of A20 chip (that's why they have the "on-die" name)
<ssvb>
diego71: but due to manufacturing differences, in theory different A20 chips may have slightly different on-die resistors
<diego71>
ssvb: still I don't full understand (and that's an understatement) how the divisor works during the calibration ...
<ssvb>
diego71: the divisor is applied to the external resistor
<diego71>
ssvb: that is used as a reference, right? So it should have just 1 correct value...
<ssvb>
diego71: and then the A20 chip tries to pick the right ZCTRL coefficients to get the closest match to this value
Renard has joined #linux-sunxi
<ssvb>
diego71: regardless of how it is done, the external resistor is used as a reference for finding the right coefficients, and then these coefficients are used with the on-die resistors
<ssvb>
diego71: we can bypass this step and load the ZCTRL coefficients directly
Skaag has quit [Quit: Leaving.]
<ssvb>
diego71: but in this case we are relying on the accuracy of the on-die resistors and on the fact that they must be similar enough in different A20 chips
<diego71>
I have another olinuxino. At the end of the calibration, I can run a test on the other, and looks if autocalibration give the same value
<diego71>
s/calibation/round of test/
<ssvb>
diego71: this makes sense, thanks
<ssvb>
diego71: actually the results of ZQ calibration are also temperature dependent, so that's another factor to take into account
<diego71>
ssvb: zq calibration is done at u-boot stage, right?
<ssvb>
diego71: yes
<ssvb>
diego71: it is also possible to enable periodic re-calibration (during refresh), and I see that the ZCTRL value changes at runtime (as seen by running a10-meminfo)
<hramrach>
can the coefficients be read back?
<ssvb>
diego71: however the reliability becomes bad
<ssvb>
hramrach: not "write back", but just report the value which can be written to the 'dram_para' struct
<ssvb>
diego71: anyway, based on the latest meminfo dumps and also on inspecting the rockchip dram sources, it looks like everyone is loading ZDATA, so it may be probably a preferred option in the end :-)
<ssvb>
diego71: in the case if ZQ calibration itself is implemented poorly, then calibrating against the external high precision resistor may result in inaccurate values
<hramrach>
the thing is that if you do calibration against 330 ohm resistor instead of 240 you know that the calibration is off so you can recalculate the coefficients based on that
<ssvb>
hramrach: yes, in principle we can detect the nominal of the RZQ resistor based on reading back the calibrated data
<ssvb>
hramrach: this can be used for A20-Olinuxino board revision identification :-)
<ssvb>
hramrach: and maybe added as an extra safety check to u-boot
<hramrach>
I think I can get an A10 olinuxino somewhere but presumably that is not interesting
<hramrach>
lime
<diego71>
hramrach: why not?
<hramrach>
does it also have odd resistors?
<hramrach>
I guess I could find out by running the tools on it
deasy has joined #linux-sunxi
Skaag has joined #linux-sunxi
konradoo77 has quit [Ping timeout: 245 seconds]
konradoo77 has joined #linux-sunxi
netlynx has quit [Quit: Leaving]
<ssvb>
hramrach: what is your revision of A10 lime?
<ssvb>
hramrach: you can download schematics and check the resistor nominals there
<ssvb>
hramrach: all A10 lime revisions are different
<hramrach>
heh
<hramrach>
I have to look at the board
<ssvb>
is it in a box?
akaizen has quit [Read error: Connection reset by peer]
akaizen has joined #linux-sunxi
npcomp has quit [Ping timeout: 246 seconds]
npcomp has joined #linux-sunxi
paulk-collins has quit [Quit: Ex-Chat]
Skaag has quit [Quit: Leaving.]
npcomp has quit [Ping timeout: 245 seconds]
pwhalen has quit [Ping timeout: 260 seconds]
npcomp has joined #linux-sunxi
<libv>
yes, i slam people with NDH just because i am evil
<libv>
i have no other reason to do so
Renard has quit [Remote host closed the connection]
pwhalen has joined #linux-sunxi
konradoo77 has quit [Ping timeout: 240 seconds]
konradoo77 has joined #linux-sunxi
Skaag has joined #linux-sunxi
npcomp has quit [Ping timeout: 240 seconds]
pietrushnic has quit [Ping timeout: 250 seconds]
npcomp has joined #linux-sunxi
Black_Horseman has joined #linux-sunxi
Black_Horseman has joined #linux-sunxi
Renard has joined #linux-sunxi
bonbons has quit [Quit: Leaving]
pietrushnic has joined #linux-sunxi
Skaag has quit [Quit: Leaving.]
tgaz has quit [Ping timeout: 260 seconds]
pwhalen has quit [Ping timeout: 255 seconds]
tgaz has joined #linux-sunxi
dack has quit [Remote host closed the connection]
konradoo77 has quit [Ping timeout: 245 seconds]
PulkoMandy has quit [Quit: Vision[0.9.7-H-20140108]: i've been blurred!]
leviathanch has quit [Ping timeout: 240 seconds]
pwhalen has joined #linux-sunxi
Guest49513 has quit [Read error: Connection reset by peer]
buZz has joined #linux-sunxi
buZz is now known as Guest83239
cubear has quit [Quit: Leaving]
Skaag has joined #linux-sunxi
<libv>
*sigh*
kill_-9_1 has quit [Quit: Connection closed for inactivity]