16CH China/Ebay Conextant Fusion 878A WORKING!

[andrewa]

Thanks a lot for the tip! I will test it after the week long holiday along the beach, it smells to me of saulty water already!
I should note that my ZM is set to the default V4L version2 and I do not enable alarm but record images all the time. It is a company site and the linux box routes internet&VPN at the same time. That is why I need DVR to run under linux and cannot use the old HICAP50 board. Hope CPU capacity will not be exceeded with 8 cameras at high resolulion.
The 15 fps which You set is actually a kind of upper limit, what is the view speed while watching a monitor or montage? I set 2-3 fps in Monitor-General, but under each monitor in Montage view in streaming mode on Mozilla I read "0.96fps". I think that Captures_per_frame states the missed frames between two captures, which allow correct multiplexing of several camera streams, connected to one chip. With my value of 2 it means that the total number of processed frames is 8 cameras* (1frame captured + 2 frames missed) = 24 fps. I presume this is the result of using just one of the 4 chips on the DVR card. 3-4 fps is quite enough for my job, hope I can reach that speed by actually using all 4 chips. The box has 1GB RAM and I have set 512MB to shared memory. After finalizing it I will post a detailed report.
HICAP50 DVR board is discarded to a friend's farm

[alltech]

Hi everyone,

I have a simular card as this thread mentions. Only difference is its once conaxent chip no pci chip.

http://www.zoneminder.com/forums/viewto ... 10&t=17955

Anyone care to tell me what my settings should be for this card in /etc/modprobe.conf? BTW I put the settings in modeprobe.conf and got a viewable imagie..with horizontal lines and rolling images. I latter found that /etc/modeprob.d.conf/bttv.conf had the same settings.

What am I doing wrong? please read my link above!!

Thanks much!

[punch-card]

Hi everyone,

I have a simular card as this thread mentions. Only difference is its once conaxent chip no pci chip.

http://www.zoneminder.com/forums/viewto ... 10&t=17955

Anyone care to tell me what my settings should be for this card in /etc/modprobe.conf? BTW I put the settings in modeprobe.conf and got a viewable imagie..with horizontal lines and rolling images. I latter found that /etc/modeprob.d.conf/bttv.conf had the same settings.

What am I doing wrong? please read my link above!!

Thanks much!

Try this below for the modprobe.conf, and check your shmmax & shmall....
BUT the key for me was check V4L_MULTI_BUFFER and set CAPTURE_PER_FRAME to 4.... (see below)
I spend many nights trying to figure this one out....

alias char-major-81 videodev
alias char-major-81 bttv
options bttv gbuffers=16 card=131,131,131,131

===== /etc/sysctl.conf
# Controls the maximum shared segment size, in bytes
kernel.shmmax = 68719476736
# Controls the maximum number of shared memory segments, in pages
kernel.shmall = 4294967296

Have a look at Options->Config->V4L_MULTI_BUFFER and CAPTURES_PER_FRAME and see if that helps.

[alltech]

Sorry, where do I add/edit V4L_MULTI_BUFFER and set CAPTURE_PER_FRAME to??

[punch-card]

Sorry, where do I add/edit V4L_MULTI_BUFFER and set CAPTURE_PER_FRAME to??

Options->Config->V4L_MULTI_BUFFER and CAPTURES_PER_FRAME

From main page upper right....

[andrewa]

I palyed for a day with the card. Here are the results under ZM 1.24.2-6.fc14 (rpm from Fedora updates) with OS Fedora14 on DualCore 2.4GHz 1GB RAM, kernel shmall=shmmax=512MB:
- Inserted modeprobe lines and configured 8 B/W cameras at 640/480, PAL, grey, VFL2 (VFL1 is not available) on video0(ch0-1), video1(ch4-5), video2(ch8-9), video3(ch12-13), 10fps, MOCORD (monitor+record)
- Capture-per-frame=5 - at value 4 cross image occurs between chX and ch(X+1), at value 5 just a slight flicker is observed on some monitors from time to time, at value I have 6 perfectly separated images.
- with option V4L-Multi-Buffer turned on, cross-imaging occurs even at value CPF=6, I turned it off
- FPS on each monitor is reported 0.86 - 0.93. Setting the parameter fps=1 or =2 of each monitor has no effect - the reported FPS rate remains <1FPS though it seems that all 4 chips are used with 2 cameras connected to each chip.
- switching monitor mode to MOCORD (montor + record) has no impact.
After each change ZM was restarted with "service zoneminder restart".
ZM reports load 1.1 - 2.4 though all the time System monitor shows 50 - 60% load on the 2 cores, never exceeding 75% except while doing ZM restart.
I wonder what is the reason for the difference between Your FPS rates and mine. Could be ZM version 1.24 versus 1.23, could be the number of cameras attached (judging by the screenshots You have 4 cameras and they practically do not need multiplexing when attached to 4 separate btt chips). Are You sure cameras are observed with 15 FPS as set in Monitor-General? You report that the CPU cannot work for a long time under this load: how long does it still operate, what happens after that?
I am close to satisfaction. Only the FPS should be raised to 2-3 FPS for full happiness I wonder how to achieve it.

[jameswilson]

you have set your frames per capture to 5 so im amazed your even getting nearly 1 fps.
To get max fps out of it this needs to be set to 1. But this may have other issues that you are solving by pushing this to 5.

[andrewa]

I am also surprised that the max I get is only < 1 FPS from each one of the 2 cameras, connected to a bttv . I guess that 5 extra unused captures (to separate the streams coming from 2 cameras per 4 channel multiplexor) between the 2 used captures makes total 12 captures/sec. While the bttv claims 30 FPS. Pondering over the result I think that the multiplexor leaves 5 frames between each 2 adjacent channels for all 4 channels, not just between the set and used channels with cameras commected to them. So the calculation should be 4*(5+1)= 24 FPS/chip which is close.

Is there any difference between "captures" and "frames"?

I will make 2 more experiments:
1. Set and connect just 1 camera to one multiplexor and check if 20-30 FPS can be achieved decreasing the present 5 parasite captures between the cameras to 1 or even 0.
2. I will connect the cameras to channels 0-2-4-6-8-10-12-14 instead of 0,1--4,5--8,9--12,13--. Maybe in this case the parasite captures between adjacent channels could be reduced to 2-3 retaining a healthy pause of 4-6 parasite captures between each 2 cameras.
Another peculiarity is the constantly heavy load >1.2 accounted by ZM while the CPUs have 50-70% load. What does the figure reported by ZM mean?

[jameswilson]

a single chip can do 25 fps (PAL) single channel. But if you split that between 4 channels you would expect 6.25 fps per channel.
This will only happen if all cameras are totally sycned. They wont be. So id say 1.5-2 fps channel on a single 25fps chip is the max. If you want full framerate you need a chip per channel.

Try it on one channel and see.

[andrewa]

Connected just one camera to a chip with CPF=1 and got >20 FPS!
Connected 2 cameras to channels 0 and 2 (not adjacent channels like 0-1 or 1-2), decreased capture_per_frame to 2, activatet V4L and got nearly 4 FPS with stable image and half the CPU load.
This evening made the same experiment with a PICO2000 card at home with 2 cameras connected previously to ch0 and ch1: connected them to ch0 and ch2, made more aggressive settings (CPF=1) and the FPS reached 3.6! Which is nearly a stream, not the stills serie which I have previously observed with 0.9 FPS.
The conclusion is clear: whenever possible on cards with multiplexed chips, do not connect cameras to adjacent channels but leave at least one free channel in between to be able to gain high FPS and efficiency.
It is worth adding these considerations in the manual as the logic for the cycling of multiplexed chips in ZM is different from Windows based DVR soft, where it seems only the active channels are queried and capture_per_frame=1. There one calculates:
MAX FPS (camera) = MAX FPS (chip) / number of cameras
supposing all cameras have equal FPS values set. Windows DVR soft probably controls better chips with multiplexed input.