<lekernel>
what's the difference between "line out" and "line level out"?
<lekernel>
the wm codec has both
<lekernel>
atm we're using the "line out" and it works, so i guess we'll stick with that
<lekernel>
hi aw
<lekernel>
could you get samples of WM9707? we'll probably use it in the end
<aw>
lekernel, hi evening
<aw>
yes, i can order some of them.
<aw>
still reading your email. :-)
<lekernel>
yeah... you should read the IRC log too :)
<aw>
okay...i may change some capacitor with low ESR to have extremely lower ripples from power supplies to check.
<lekernel>
you think the noise might come from power supply caps?
<lekernel>
FYI: I'm using the WM codec without noise now
<lekernel>
I have fixed some problems in the FPGA design that prevented it from working correctly
<lekernel>
and now we have noise free audio
<aw>
yes, WM may have different performance noise reduction tech...yup? also fixed on fpga design?>>> sounds great.
<aw>
you used WM codec in RC1 not RC2, right?
<roh>
aw: ack. it was a fight!
<roh>
not for the wm.. that one came out easy... the video adc was hard
<roh>
i either need nozzles for that tqfp stuff or much more experience ;)
<roh>
aw: can you tell us about your hot-air experiences, give some tips?
<aw>
roh, yup, such skilled tech. you have is great. :-)
<lekernel>
yeah, I took a RC1 board
<aw>
well...last week i just used hot-air to rescue one damaged RC2 board with all ICs,
<lekernel>
roh: well, with the bigger nozzle it wasn't so hard
<lekernel>
and leaded solder doesn't have so much tendency of forming balls behind the pins
<aw>
lekernel, okay..the layout/schematic between rc1 and rc2 is quite different, so i may yes order WM to replace it on my rc2 for confirmation.
<lekernel>
aw: the WM chip requires a few modifications
<aw>
lekernel, yup, can imagine. :-)
<lekernel>
remove 1M resistor on crystal, add capacitor on pin 33 (for 3D effect, optional), add capacitors on pin 32
<lekernel>
I have uploaded the datasheet to the git repos. please have a look at it
<aw>
okay..
<lekernel_>
<lekernel> remove 1M resistor on crystal, add capacitor on pin 33 (for 3D effect, optional), add capacitors on pin 32
<lekernel_>
I have uploaded the datasheet to the git repos. please have a look at it
<lekernel_>
also, we might also want to route SPDIF to the internal connector and DNP the parts for the nonexistent headphones amp
<lekernel_>
you should also recompile the test program
<aw>
lekernel_, um..seems lot of works to be learn on my side. ;-)
<lekernel_>
btw i'm running the demo console in 800x600 resolution atm
<aw>
has this changed by SoC or s/w porting enough?
<lekernel_>
some dirty soc hacks
<aw>
umm.:-)
<lekernel_>
1024*768 wouldn't work though. well, maybe with more buffering
<aw>
lekernel_, do you think that we still need U23 if change into WM in the end? or not necessary.
<lekernel_>
the 4.3V regulator?
<lekernel_>
well... could be a good idea to protect against incoming power supply noise
<lekernel_>
at least we have more safety if we happen to pick up a noisy transformer
<aw>
I probably also order some low ESR capacitors to replace C215 10uF. As a recommended part there in datasheet. ;-) yeap, seems use it as a seperated circuitry part.
<aw>
i saw some doc the 'white' noise can come from bad ripples from power supply, but we didn't test though. :-)
<lekernel_>
power supply?
<lekernel_>
I checked it on scope... looks clean
<lekernel_>
but you can still double check :)
<aw>
yeah, double check is good after replace a low ESR capacitor. :-)
<xiangfu>
lekernel_: Hi. I am working on try to create the /dev/flash5 image.  then ender user can easy flash that by using urJtag. got more patches and wallpaper.png
<xiangfu>
lekernel_: I think I mis-understand your task about default wallpaper.
<xiangfu>
so I talk with wolfgang about my task on milkymist one.
<xiangfu>
something wrong with the "www.milkymist.org"?  always give me connect timeout. and
<lekernel_>
a clean solution can be copy-free (since the framebuffer is already mapped in memory and uses triple buffering) and you'd simply "lock" the current displayed framebuffer from the driver
<lekernel_>
i.e. you add a 4th buffer
<lekernel_>
in normal operation, that 4th buffer isn't used, it's the current triple buffering
<lekernel_>
but when fbgrab requests a still framebuffer you return the currently displayed one and you don't touch it again until fbgrab is done with it
<lekernel_>
but you can keep rendering using the other 3 remaining buffers
<lekernel_>
it's a bit more complex to implement, but it would make fbgrab not interfere at all with the rendering
<lekernel_>
also, keep in mind that the CPU is heavily used by the rendering already
<lekernel_>
so if you use a background task, your PNG encoding can take forever
<lekernel_>
about the ramdisk: you can write loads of stuff there, it's allocating on the ~100MB heap
<wolfspraul>
xiangfu - my m1 still has flickernoise 0.1 - what is the easiest way to upgrade?
<xiangfu>
lekernel_: thanks for the info.
<xiangfu>
wolfspraul: urjtag. I think.
<wolfspraul>
ok. need to disassemble my case first then...
<lekernel_>
got 1024x768 ...with still some bugs though
<wolfspraul>
hey my projector can only do 800x600
<wolfspraul>
and I bought it because I thought it was 'future-proof' with m1. but you seem to be too fast!
<wolfspraul>
my flickernoise 0.1 does not work with an Apple USB keyboad I have here
<lekernel_>
is that a keyboard with integrated hub?
<wolfspraul>
is that something that was improved in later flickernoises?
<wolfspraul>
wait, checking...
<wolfspraul>
yes, indeed
<wolfspraul>
when I plug it into LInux, I get 2 new devices, hub and keyboard
<lekernel_>
keyboards with integrated hubs are fucking extreme bloody pains in the ass that would make you hate USB forever should you even think about trying to develop support for them
<wolfspraul>
good catch!
<wolfspraul>
well, xiangfu can add it to his todo list :-)
<lekernel_>
and they're not supported atm
<wolfspraul>
(at the bottom though)
<lekernel_>
well it won't be an easy task
<wolfspraul>
no problem, I will find another keyboard
<wolfspraul>
yes I can imagine
<wolfspraul>
not the right priority now
<wolfspraul>
I will find another keyboard
<wolfspraul>
then I much rather have xiangfu implement usb keyboard support in the Ben NanoNote :-)
<wolfspraul>
without a hub, of course
<wolfspraul>
thanks that was an easy catch. I will find another keyboard.
<lekernel_>
god, USB is so stupid... how could one even dream about making so complicated and messy the simple task of transmitting key presses and mouse moves down a bit of electric wire?
<lekernel_>
I don't get it
<xiangfu>
btw. I have a wireless mouse and keyboard.  using one usb receiver. the mouse works fine. keyboard not working
<lekernel_>
xiangfu: also you're not freeing buf_p in fbgrab
<lekernel_>
and zeroing it should be unnecessary
<lekernel_>
you could also use an ioctl() interface that gives you the address of the framebuffer... and you wouldn't have to allocate memory anymore
<xiangfu>
lekernel_: thanks.
<Fallenou>
accessing the fb through the driver each time is an overhead
<Fallenou>
accessing directly the memory array is faster
<xiangfu>
is that ok, just read the framebuffer when mm1 do performance ?
<Fallenou>
look at how it is done in fb.c
<Fallenou>
(for the ioctl)
<lekernel_>
wolfspraul: for the resolution, 800x600 will be supported as well. ideally, we could even detect the supported resolutions with DDC and pick the highest.
<lekernel_>
and in fact it wouldn't make much difference with the rendering, only with the GUI
<lekernel_>
I don't believe the current SoC architecture would be able to run the complete rendering system at high resolutions and high frame rate
<lekernel_>
so the rendering would happen with a limited resolution internal framebuffer that is scaled up on the fly by the VGA controller
<xiangfu>
lekernel_: I will send another patch with fixed 'free' and 'close', I will look into the ioctl, direct read framebuffer.
<lekernel_>
ok
<xiangfu>
lekernel_: I add the close(fd) to the end of main_fbgrab. the problem.
<xiangfu>
everytime 'fbgrab a.png' finished the screen got no signal.
<lekernel_>
ah :)
<lekernel_>
uhm, yes
<xiangfu>
just like the close(fd), close the vga output.
<lekernel_>
actually the framebuffer driver isn't supposed to be acceded from two different tasks
<lekernel_>
it enables video out at open() and disables it at close()
<lekernel_>
you should be able to fix that by adding a reference counter to it
<xiangfu>
yes. when I run the fbgrab again. video out work again. before fbgrab finished.
<xiangfu>
lekernel_: ok. I will send out the patch to mailing list.
<Fallenou>
the resolution is held in a struct fb_var_screeninfo
<Fallenou>
dunno if you can get it from ioctl
<Fallenou>
i guess not :x
<Fallenou>
xiangfu:Â Â FBIOGET_VSCREENINFO
<Fallenou>
this ioctl will give you the good struct :)
<Fallenou>
to get the resolution
<xiangfu>
Fallenou: ok. try to implement in next fbgrab. 1. direct read fb; 2. get more info by using ioctl
<xiangfu>
Fallenou: thanks.
<lekernel_>
ok, now the funny part: use partial reconfiguration to let the software change the pixel clock
<lekernel_>
since it seems xilinx fucked up the normal way to reconfigure clock generators, I guess I'll have to spend the afternoon doing icky fpga hacking
<Fallenou>
lekernel_: will the rtems framebuffer driver have to change its resolution?
<lekernel_>
hmm... I guess I'll generate a differential bitstream with only the DCM configuration changed and then throw that into the ICAP to change the pixel clock
<Fallenou>
I mean dinamically
<lekernel_>
yes, sure. we want GUI menus to select the resolution, don't we?
<lekernel_>
and before on the fly scaling works, we might even have to switch back to 640x480 during rendering
<Fallenou>
ok so we need to implement the FBIOPUT_VSCREENINFO
<Fallenou>
ioctl
<xiangfu>
lekernel_: I have a very small patch to connect mm1 by serial console. but not sure if you would like put it in milkymist.git or flickernoise.git. blow is the small patch
<xiangfu>
+connect: /dev/ttyUSB1
<xiangfu>
+Â Â Â Â Â Â stty -F $< raw 115200
<xiangfu>
+Â Â Â Â Â Â while : ; do cat $< ; done & cat > $<
<xiangfu>
+
<xiangfu>
i add it to "milkymist.git/boards/milkymist-one/flash/Makefile"
<xiangfu>
I would like this small feature. since some one just bought on mm1. then he connect the usb cable. run 'make connect' he can got the boot log :)
<CIA-43>
milkymist: Sebastien Bourdeauducq master * rdcef87c / boards/milkymist-one/flash/Makefile : flash: connect target to get boot log (Xiangfu) - http://bit.ly/geWVza
<xiangfu>
lekernel_: thanks. :)
<xiangfu>
lekernel_: sorry. the my terminal make the 'tab' as 'space'. :(
<lekernel_>
and what's the problem?
<xiangfu>
lekernel_: after you apply my patch from "http://pastebin.com/pgiznE7u" in makefile the front of 'stty ...' and 'while ..' is space not 'tab'
<lekernel_>
argh
<lekernel_>
ah, and it should be a phony target too
<lekernel_>
we should put some with the video input
<lekernel_>
preferably with good looking dancers etc. :)
<kristianpaul>
i can find good looking dancers next week at labsurlab ;-)
<xiangfu>
after I update the rtems-milkymist.git. I got some error on network.
<xiangfu>
"No rx buffers left in the pool! We are in big troubles!"
<xiangfu>
ok. not always. after reboot. the ftp works fine again
<lekernel_>
yeah, there are still several ethernet bugs
<CIA-43>
rtems-milkymist: Sebastien Bourdeauducq master * r68f1136 / c/src/lib/libbsp/lm32/shared/milkymist_framebuffer/framebuffer.c : add open ref_counter to framebuffer driver (Xiangfu) - http://bit.ly/gAoF3v
<lekernel_>
xiangfu: could you re-send your fbgrab patch with tabs for indentation?
<Fallenou>
hum I've read somewhere that usually spaces are preferred
<Fallenou>
because tabs can be interpreted by your editor
<xiangfu>
lekernel: sended out. will more carefully next time.
<xiangfu>
Fallenou: linux kernel style is using 'tab'
<Fallenou>
tabs could show up as a different number of spaces depending on the editor
<Fallenou>
whereas spaces are always the same
<Fallenou>
xiangfu: so I guess it's the best :p
<xiangfu>
lekernel: btw. I saw you always manually apply the patch. I always using save the file to .eml, the 'git am PATCH_EMAIL.eml', if the patch is as expect :)
<xiangfu>
Fallenou: the only issue for me is the terminal always replace 'tab' with 'spaces'.
<lekernel>
I found that crap when looking for info about how to send rs232 breaks
<lekernel>
ELM electronics has a "break detector" circuit
<lekernel>
I guess they're programmed attinys/pic or something like that
<mwalle>
lekernel: lol
<mwalle>
lekernel: the gdb stub will need 6kb/2kb rom/ram
<lekernel>
8kb in total?
<mwalle>
yes
<lekernel>
bytes?
<mwalle>
bytes
<lekernel>
should be ok i think...
<lekernel>
we have that much spare BRAM afaik
<mwalle>
unfortunately it is hard to mux gdb packets and console.. atm i did a tx redirection with the help of the uart core
<mwalle>
there are some scripts that intercepts the serial port traffic and looking for gdb packets and split them from normal data, but that wont work out of the box
<mwalle>
gdb supports 'systemcalls' eg read and write to stdin/stdout/stderr, but that requires an application that is 'gdb aware'
<mwalle>
gn8 :)
<mwalle>
12.142ns sys_clk period with one breakpoint