ambrosa

Added VID/PID for Leadtek USB DVB-T stick 0x0413 0x6A03

(thanks to djnice)

This is the report made by nobody9 for his system equipped with a Nvidia 320M
Nvidia 320M is more powerful but probably can be problem watching 1080p @ 50Hz coded with H264 : max frame rate is 45 frames/sec.


qvdpautest 0.5.1
Intel® Core™2 Duo CPU P8600 @ 2.40GHz
NVIDIA GPU GeForce 320M (C89) at PCI:4:0:0 (GPU-0)

VDPAU API version : 1
VDPAU implementation : NVIDIA VDPAU Driver Shared Library 260.19.06 Mon Sep 13 04:58:44 PDT 2010

SURFACE GET BITS: 677.9 M/s
SURFACE PUT BITS: 854.746 M/s

MPEG DECODING (1920x1080): 77 frames/s
MPEG DECODING (1280x720): 167 frames/s
H264 DECODING (1920x1080): 45 frames/s
H264 DECODING (1280x720): 93 frames/s
VC1 DECODING (1440x1080): 62 frames/s
MPEG4 DECODING (1920x1080): 52 frames/s

MIXER WEAVE (1920x1080): 455 frames/s
MIXER BOB (1920x1080): 811 fields/s
MIXER TEMPORAL (1920x1080): 71 fields/s
MIXER TEMPORAL + IVTC (1920x1080): 64 fields/s
MIXER TEMPORAL + SKIP_CHROMA (1920x1080): 67 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 51 fields/s
MIXER TEMPORAL_SPATIAL + IVTC (1920x1080): 45 fields/s
MIXER TEMPORAL_SPATIAL + SKIP_CHROMA (1920x1080): 48 fields/s
MIXER TEMPORAL_SPATIAL (720x576 video to 1920x1080 display): 140 fields/s
MIXER TEMPORAL_SPATIAL + HQSCALING (720x576 video to 1920x1080 display): 139 fields/s

MULTITHREADED MPEG DECODING (1920x1080): 59 frames/s
MULTITHREADED MIXER TEMPORAL (1920x1080): 44 fields/s

*** VDPAU BENCHMARK AND PERFORMANCE OBSERVATIONS ABOUT HD CHANNELS ***

My system is a Intel Core 2 Duo E6600 6GB RAM with Nvidia 9600GT Green Power 512MB RAM.
Usually I play without any problem 1080p video with VLC or MPLAYER2 (-vc x11). These software use CPU for decoding video and play movie is very fine. CPU usage is about 70%

But I cannot play 1080 HD video stream or recoderd movie using openpliPC.
As you know openpliPC use xine libraries and these libs use VDPAU.
VDPAU use Nvidia GPU as coprocessor for decoding video stream. The idea is GPU is more specialized and powerful than CPU to do decoding process.
If I play an HD movie .ts with MPLAYER2 using VDPAU (-vc vdpau -vo ffmpegh264vdpau) CPU usage is about 4% (yes ! VDPAU is working) but HD movie has many freeze. It's impossible to watch them.

QUESTION: Nvidia 9600GT (used by VDPAU) doesn't meet minimum performance to decode HD channels ?

ANSWER: I've just pushed to openpliPC two userful sw: vdpauinfo and qvdpautest . With them you can have infos about VDPAU and answer to the question above.

From README:
4) If you are interested you can compile vdapauinfo and qvdpautest
(building qvdpautest require install qt4-dev-tools (100 MB))
./build_vdpauinfotest.sh

You can find binaries under libs/vdpauinfo-0.0.6/ and libs/qvdpautest-0.5.1/

Below is the result running qvdpautest in my system.
cd openpliPC/
cd libs/qvdpautest-0.5.1/
./qvdpautest

As you can see, my 9600GT has very poor VDPAU performance, less powerful than using simple CPU for decoding process
is not able to decode MPEG2 HD video with frame/sec. > 43
is not able to play H264 HD video with frames/sec. > 18
is not able at all to decode any MPEG4 video


So, if you have problems with HD channels, run qvdpautest and see benchmark.


My solution: I've just bought a Nvidia GTX550


qvdpautest 0.5.1
Intel® Core™2 CPU 6600 @ 2.40GHz
NVIDIA GPU GeForce 9600 GT (G94) at PCI:1:0:0 (GPU-0)

VDPAU API version : 1
VDPAU implementation : NVIDIA VDPAU Driver Shared Library 280.13 Wed Jul 27 17:15:20 PDT 2011

SURFACE GET BITS: 860.555 M/s
SURFACE PUT BITS: 851.924 M/s

MPEG DECODING (1920x1080): 43 frames/s
MPEG DECODING (1280x720): 97 frames/s
H264 DECODING (1920x1080): 18 frames/s
H264 DECODING (1280x720): 39 frames/s
VC1 DECODING (1440x1080): 55 frames/s
Profile unsupported.

MIXER WEAVE (1920x1080): 2126 frames/s
MIXER BOB (1920x1080): 3101 fields/s
MIXER TEMPORAL (1920x1080): 739 fields/s
MIXER TEMPORAL + IVTC (1920x1080): 443 fields/s
MIXER TEMPORAL + SKIP_CHROMA (1920x1080): 997 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 225 fields/s
MIXER TEMPORAL_SPATIAL + IVTC (1920x1080): 184 fields/s
MIXER TEMPORAL_SPATIAL + SKIP_CHROMA (1920x1080): 242 fields/s
MIXER TEMPORAL_SPATIAL (720x576 video to 1920x1080 display): 830 fields/s

MULTITHREADED MPEG DECODING (1920x1080): 43 frames/s
MULTITHREADED MIXER TEMPORAL (1920x1080): 672 fields/s

INFO
Yesterday nobody9 has applied my suggestion to change main install directory.

Now all Enigma2 files are installed under

/usr/local/e2/

Now it's simple to remove e2 :
sudo rm -Rf /usr/local/e2

See README for further infos.

@ambrosa one of the missing USBids for 2832 from v2.2.2:
0458:707f - Genius/Kye TvGo DVB T03
btw. did you manage to build 2.6.18 version that works?

Kernel 2.6.18 was released in August 2006. It's 5 years old (!!).
Really, is too old and probably RTL driver can be backported but this require a lot of efforts. And I'm not interested in this issue because I have not a DMM or Vu+

Say to DMM (and Vu+) to upgrade to newer kernel or stop to buy their products.

For AVerMedia (A835) is can be done?

I've a A835 stick also but it work bad. Zap very slow and very low sensibility (about 100 channels less tuned than A867)

Now I've finished my job around A867 and RTL2832 and I can consider to take a look to A835 driver.
But this will be a long work ....... stay tuned on January: I'm busy @ job

That would be great as this usb dvb-t is much cheaper !
But extremely slow....till now. We hope in a miracle !

Thanks for your contribution.

Dear attaros, a small comparison for you.
here in Italy:

AverMedia A835
price= 29 euro
zap time= PC about 2 sec , E2 about 8 sec
channels tuned= about 290

Avermedia A867 (the best)
price= 19 euro
zap time= PC about 1 sec. , E2 about 1 sec
channels tuned= about 410

Dikom USB DVB-T (RTL2832)
price= 19.50 euro
zap time= PC about 2 sec. , E2 about 2 sec
channels tuned= about 360


A835 is the worst

During these days I've made some little changes to drivers above.

Avermedia A867 (aka "Sky Digital Key" blue led version)

1) correct some minor bug
2) improved SNR precision
3) (thanks to Gennar1) final kernel 3.2.0 patch. Really this patch is made to adapt to newer V4L tree version that use a multi-frontend approach (MFE). With this patch driver compile fine also with 3.2.0 kernel (that include newer V4L with MFE) OR older kernel with newer V4L tree (with MFE) OR older kernel with older V4L (without MFE).

4) module name changed : now module is compiled as dvb-usb-a867.ko (former dvb-a867.ko)
Please adapt autoload in openembedded/conf/machine/etxx00.inc

module_autoload_dvb-usb-a867 = "dvb-usb-a867"

RTL2832

1) correct some minor bug
2) added SNR output in 16bit format: now is working fine
3) (as A867) (thanks to Gennar1) final kernel 3.2.0 patch. Really this patch is made to adapt to newer V4L tree version that use a multi-frontend approach (MFE). With this patch driver compile fine also with 3.2.0 kernel (that include newer V4L with MFE) OR older kernel with newer V4L tree (with MFE) OR older kernel with older V4L (without MFE).

4) module name changed (cosmethic): removed final 'u' to driver name. Now module is compiled as dvb-usb-rtl2842.ko (former dvb-usb-rtl2842u.ko).
Also patch filename now is dvb-usb-rtl2842.patch (former dvb-usb-rtl2842u.patch): please change openembedded/recipes/linux/linux-etxx00_3.1.0.bb

Add autoload in openembedded/conf/machine/etxx00.inc (before was missing):

module_autoload_dvb-usb-rtl2832 = "dvb-usb-rtl2832"

Note: RTL2832 looks to output Signal Strength ("AGC" in E2) as dBm value and not percentage. So users see E2 AGC moving from about 40 (strong signal) to 30 (poor signal)

Linux DVB-T drivers usually are NOT made directly by manufacturer but often they are made by reverse-engineering efforts from open source community.
And often they are a "mix" of code taken from different driver for similar stick.
STB world is NOT PERFECT, so there is need a good level of "flexibility"

BTW: I don't understand WHY is not possibile to have infos like screenshot attached below. Many skins have them.
Is a 6 lines patch, it's simple, no performance impact .... some user can found it useful ... why not ? Because "it's not following some standard ?"

As discuss here http://openpli.org/f...lem-with-plihd/, since some versions ago in PliHD skin (very nice !) is missing the SNR,AGC,BER infos and its coloured bars.

About 2-3 weeks ago all was fine. A friend of mine has an old Openpli (installed about 3 weeks ago) and infos are there.

Now infos are missing (for any type of signal, either DVBT, DVBS/S2). Feature or bug ?

Chipset support also DAB/DAB+ but it depends on stick brand (as Realtek says to me).

Linux driver support DAB and one stick (thanks Gennar1) looks to be http://www.terratec....Data.pdf</span>

Other people says "Please note that the DVB part of the Noxon DAB Stick (USB ID 0ccd:00b3) is also supported by the Linux driver code"

It's a pity that Enigma2 doesn't support DAB.

Dear nobody9,

I appreciate very much your effort.
Some weeks ago I asked to Pli team why they didn't release an E2 "PC version". Very useful for debugging.

You have answered to my question

Thanks again and I suggest you two improvements:

1) can run on Ubuntu 11.10 64bit machine ? If not I need to install in a VirtualBox machine. No problem about this anyway. Now 64bit are more common than 32bit.

2) I suggest you to open a GitHub repository and push your code into it using git. I've started to use GIT since 1 month ago and I'm an enthusiastic user (newbie level ahaha)
B rowsing the code other people can give you their suggestion and improvement.


I test it soon: It will be very useful for me to debug DVB-T USB stick driver code...

Good work.

About one week ago I've bought a DVB-T USB stick based on Realtek RTL2832U chipset.
I've found some driver for Ubuntu but they are old and someone has patched them introducing some wrong typecast in C code so they compile only for x86 64bit arch.

"Restart from scratch !" : I've emailed Realtek and they kindly give me their latest RTL2832 Linux driver rel 2.2.2. This driver is for kernel 2.6.x and V4L

I've started porting the code to kernel 3.x.x and I've made some code cleanup and minor bug correction.
My little project is here https://github.com/a...od_kernel-3.0.0 and you also can find in my GitHub repo the original RTL2832U 2.2.2 driver (two versions: full with 10 tuners and simplified with 3 tuner)

Since yesterday evening driver compile fine without any warning in Ubuntu 11.10 32bit and 64bit and device works fine.

And this morning compile fine also with OpenPLi for CT9x00 (based on 27-11-2011 repo)

These are some USB sticks RTL2832U based:
DIKOM USB-DVBT HD [ID 1b80:d394] (my own)
NILOX DVB-T Stick N15 [ID 1b80:d393]
TRUST 16738 [ID 0bda:2832]
TERRATEC Cinergy T Stick BLACK [ID 0ccd:00a9]
TX Hollywood DVTB [ID 1d19:1101]
DIGIVOX mini II v3.0 [ID 1d19:1101]

The driver "full" support:
Chipset 2832, 2836, 2840
DVB-T, DVB-C, DTMB
Tuners: E4000, FC0012 (most used), FC0013, FC2580, MAX3543, MX2063, MX2266, MXL5007T, TDA18272, TUA9001
All in one package: patch (see below) is about 2.5 MB

I've tested DIKOM USB-DVBT HD (USB2.0, RTL2832, FC0012, DVB-T) in my CT9100 with my driver and compared with Avermedia A867 stick:

Channels found (using Manual Scan, Tuner C, Complete, n, n, n; "Europe DVB-T" freq. set)
DIKOM: 312 (VHF ok), the same more or less than Avermedia A835
A867: 393 (VHF ok)

Zap time:
DIKOM: about 3-4 sec.
A867: about 1.5-2 sec.

Uptime after 10 minutes watching the same DVB-T HD channel:
DIKOM: 0.47
A867: 1.20

Signal quality feedback in E2:
DIKOM: none (no SNR, no AGC, no BER). I hope to fix in future.
A867: AGC, BER (no SNR)

Device temperature:
DIKOM: medium
A867: medium


The driver looks stable.


This is my local linux-etxx00_3.1.0.bb (made this morning) and I attach here my openembedded/recipes/linux/linux-etxx00/dvb-usb-rtl2832u.patch
I hope it can be included in OpenPli

DESCRIPTION = "Linux kernel for ${MACHINE}"
LICENSE = "GPL"
 
MACHINE_KERNEL_PR_append = ".7"
 
DEPENDS = "mtd-minimal-nand-utils"
RDEPENDS_kernel-image = "mtd-minimal-nand-utils"
 
SRC_URI += "http://www.et-view.com/download/linux-${PV}.tar.gz \
	file://${MACHINE}_defconfig \
	file://fix-proc-cputype.patch \
	file://dvb-usb-af9035.patch \
	file://dvb-usb-it9135.patch \
	file://tda18218-7mhz-lopass.patch \
	file://dvb-usb-a867.patch \
	file://PCTV-DVB-S2-stick-460e.patch \
	file://cxd2820r-enable-LNA-for-DVB-T.patch \
	file://cxd2820r-changed-condition-to-break-out-from-wait-lock-loop.patch \
	file://dvb-usb-smsdvb_fix_frontend.patch \
	file://dvb-usb-rtl2832u.patch \
	"
 
SRC_URI_append_et5x00 = " file://disable_early_fb.patch"
 
S = "${WORKDIR}/linux-${PV}"
 
inherit kernel
 
export OS = "Linux"
KERNEL_OBJECT_SUFFIX = "ko"
KERNEL_OUTPUT = "vmlinux"
KERNEL_IMAGETYPE = "vmlinux"
KERNEL_IMAGEDEST = "/tmp"
 
FILES_kernel-image = "/${KERNEL_IMAGEDEST}/${KERNEL_IMAGETYPE}.gz"
 
do_configure_prepend() {
	oe_machinstall -m 0644 ${WORKDIR}/${MACHINE}_defconfig ${S}/.config
	oe_runmake oldconfig
}
 
kernel_do_install_append() {
	install -d ${D}/${KERNEL_IMAGEDEST}
	install -m 0755 ${KERNEL_OUTPUT} ${D}/${KERNEL_IMAGEDEST}
	gzip ${D}/${KERNEL_IMAGEDEST}/${KERNEL_IMAGETYPE}
}
 
pkg_postinst_kernel-image () {
	if [ "x$D" == "x" ]; then
		if [ -f /${KERNEL_IMAGEDEST}/${KERNEL_IMAGETYPE}.gz ] ; then
			flash_eraseall -j /dev/mtd1
			nandwrite -p /dev/mtd1 /${KERNEL_IMAGEDEST}/${KERNEL_IMAGETYPE}.gz
			rm -f /${KERNEL_IMAGEDEST}/${KERNEL_IMAGETYPE}.gz
		fi
	fi
	true
}
 

I've added in main /etc/tuxbox/terrestrial.xml the ITalian freq. list for DVB-T

Can I suggest to include the new file in main distribution ?

It works fine with many USB DVB-T


Attached here

Hi folks.

I'm not a developer and I cannot help you but I wish to suggest you a feature in new OpenWebIf: access to local terminal (telnet/ssh) using web interface.

This can be useful if anyone needs to access to a remote STB but is not available a "pure" telnet/ssh client or the connection goes through a HTTP proxy server (like in my company: when I'm @ office I can use only HTTP connection) or anything else.

Hi,

maybe umount procedure is just a tiny bit of a more complex Devicemanager Plugin as seen in here:

http://www.newnigma2...65606#post65606

Just my two cents on this

Yes. In many images there is something like you suggest.
But I've not found something similar as above in OpenPLI

So I've start to code my plugin. I believe it's simple and fast.
But if anyone what use a more complete DeviceManager ... yes he can !