a multichannel capture problem

a multichannel capture problem

[Ming Liu]

Hello,

As a newbie in V4L2, I am trying to capture two video channels 
simultaneously (we can live with about 0.05s difference between two 
channels). Currently, I am using PV-143 (a cheap card with one bt878 
chip). After some research, I found:

1. bt878 can only handle one channel at one time, so I will not be able 
to reach a speed higher than 15 frames/second per channel.
2. Since I can not synchronize the two cameras, I will have much lower 
frame rate if I try to switch between two channels.

My problems are:

1. Is there any example codes that I can follow to estimate the frame 
rates using switch channel approach? Can I do this by using standard 
V4L2 APIs, or I will need to deal with the driver?
2. If I choose a multi-chip PCI based video capture card, is there any 
limit from the bandwidth? Are there any sample codes available?
3. In the program point of view, is there difference between using 
multi-chip card and several single chip cards?

Thank you in advance.
Sincerely yours
Ming

--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list

Re: a multichannel capture problem

[Ming Liu]

Any comments will be helpful.
I will really appreciate them.

Sincerely yours
Ming

Ming Liu wrote:
> <div class="moz-text-flowed" style="font-family: -moz-fixed">Hello,
>
> As a newbie in V4L2, I am trying to capture two video channels 
> simultaneously (we can live with about 0.05s difference between two 
> channels). Currently, I am using PV-143 (a cheap card with one bt878 
> chip). After some research, I found:
>
> 1. bt878 can only handle one channel at one time, so I will not be 
> able to reach a speed higher than 15 frames/second per channel.
> 2. Since I can not synchronize the two cameras, I will have much lower 
> frame rate if I try to switch between two channels.
>
> My problems are:
>
> 1. Is there any example codes that I can follow to estimate the frame 
> rates using switch channel approach? Can I do this by using standard 
> V4L2 APIs, or I will need to deal with the driver?
> 2. If I choose a multi-chip PCI based video capture card, is there any 
> limit from the bandwidth? Are there any sample codes available?
> 3. In the program point of view, is there difference between using 
> multi-chip card and several single chip cards?
>
> Thank you in advance.
> Sincerely yours
> Ming
>
>
> </div>
>

--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list

Re: a multichannel capture problem

[Daniel Glöckner]

On Mon, Sep 08, 2008 at 05:09:33PM -0400, Ming Liu wrote:
> 1. Is there any example codes that I can follow to estimate the frame 
> rates using switch channel approach?

Motion (http://www.lavrsen.dk/twiki/bin/view/Motion/WebHome ) is said
to support multiple inputs of one card "at the same time".
I never tried it, though.

> Can I do this by using standard V4L2 APIs, or I will need to deal with
> the driver?

You can do this with V4L2.

Some time ago I made some experiments changing the input at random times
using direct hardware access while capturing. IIRC the chip will skip at
least one complete frame before it continues to capture. 

> 2. If I choose a multi-chip PCI based video capture card, is there any 
> limit from the bandwidth?

Of course. A few days ago we had a discussion about bandwidth issues when
using 5+ bt8xx cards to capture 640x480 in YUV 4:2:0. When there are
other chips using the PCI bus (f.ex. the harddisk controller), the limit
will be lower.

> Are there any sample codes available?

Motion can handle multiple devices at the same time as well.

> 3. In the program point of view, is there difference between using 
> multi-chip card and several single chip cards?

No.

  Daniel

--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list

Re: a multichannel capture problem

[Ming Liu]

Thank you so much, Daniel.
Now I know where to dig.

Sincerely yours
Ming

--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list

Re: a multichannel capture problem

[Vinicius Kamakura]

>> Can I do this by using standard V4L2 APIs, or I will need to deal with
>> the driver?
>
> You can do this with V4L2.
>
> Some time ago I made some experiments changing the input at random times
> using direct hardware access while capturing. IIRC the chip will skip at
> least one complete frame before it continues to capture.
>

what do you mean by direct hardware access?

- vk

--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list

Re: a multichannel capture problem

[Daniel Glöckner]

On Tue, Sep 30, 2008 at 05:25:27AM -0300, Vinicius Kamakura wrote:
> On Tue, Sep 09, 2008 at 09:07:27PM +0200, Daniel Glöckner wrote:
> > Some time ago I made some experiments changing the input at random times
> > using direct hardware access while capturing. IIRC the chip will skip at
> > least one complete frame before it continues to capture.
> >
> 
> what do you mean by direct hardware access?

I mmap'ed /sys/class/video4linux/video0/device/resource0 and toggled
bit 5 of IFORM.

  Daniel

--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list

Re: a multichannel capture problem

[Vinicius Kamakura]

2008/9/30 Daniel Glöckner <daniel-gl@gmx.net>:
> On Tue, Sep 30, 2008 at 05:25:27AM -0300, Vinicius Kamakura wrote:
>> On Tue, Sep 09, 2008 at 09:07:27PM +0200, Daniel Glöckner wrote:
>> > Some time ago I made some experiments changing the input at random times
>> > using direct hardware access while capturing. IIRC the chip will skip at
>> > least one complete frame before it continues to capture.
>> >
>>
>> what do you mean by direct hardware access?
>
> I mmap'ed /sys/class/video4linux/video0/device/resource0 and toggled
> bit 5 of IFORM.
>
>  Daniel
>

Isn't that the same as using the VIDIOC_S_INPUT ioctl?
Or is there a performance gain (less field/frame skipping) on doing that?

- vk

--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list

Re: a multichannel capture problem

[Daniel Glöckner]

On Wed, Oct 01, 2008 at 03:27:39PM -0300, Vinicius Kamakura wrote:
> Isn't that the same as using the VIDIOC_S_INPUT ioctl?
> Or is there a performance gain (less field/frame skipping) on doing that?

I wanted the lowest latency and the least impact on the hardware.

Doing

  m[1]^=0x20;

should be faster than calling the kernel to do it.

It was a quick hack..

  Daniel

--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list