[Elphel-support] linescan mode

[Andrey Filippov]

On Tue, May 11, 2010 at 5:32 AM, Michael Aschauer <m at ash.to> wrote:

> Dear Andrey,
>
> I had a brief chance to testdrive it - also should do that more, - but and
> seems to work!
>
> I have two main question / issues:
>
> * there are two error lines at the end of each frame. Is there a technical
> reason for this? Would it easily be possible to get rid of them before
> compression?
>

Michael,

the code now is under active development, I started to work on this as I
needed to make camera work with large frames - it is for the multi-sensor
arrangements (like stereo or panoramic), and the linescan mode was the
simplest thing to use to generate large frames, I had jsut to fix a few
errors. And there definitely could still be some bugs -some I'm aware of and
working on them, others - are still hiding. So if you can send be examples
with comments, I would really appreciate that.


* variable framerate? I run the linescan with variable framerate (which
> worked well with elphel 333). But I have some troubles to figure out how to
> do that correctly. That's not just with photofinish mode, but also in normal
> mode.
>

The simplest way to slow down the linescan mode is to use two parameters
(they appear in parsedit.php links in the camera if you use current code),
here is the link to the source:
http://elphel.cvs.sourceforge.net/viewvc/elphel/elphel353-8.0/packages/web/353/php_top/parsedit.php?view=markup

There are two paramteres (with "tooltip" when you hold mouse pointer over)

VIRT_HEIGHT - number of scan lines per each "frame" (two lines in your
case). The scan line duration is constant, by default software sets it
to a minimal value allowed by the sensor, but you may increase it.

VIRT_KEEP - set to 1 to instruct software to keep VIRT_HEIGHT value.
When 0 the VIRT_HEIGHT is calculated automatically.
And of course the period of 2 line-stripes (including the invisible to
the total of VIRT_HEIGHT) ca not be less than exposure time. So if
exposure increases, the frame rate goes down.



> I currently do the linescan on the host computer, streaming a ROI of
> 2592x48 - depending on quality settings this is possible up to 424 frames
> per second (and strangely the height of 48 seem to deliver best performance.
> it significally drops if below that 48px)
>
> setting the framerate via FPSFLAG and FPSLIM1000 works as expected with
> auto-exposure turned on. but if exposure is set manual (which is what I
> use), it seems to me exposure time is actually determining the framerate
> (and not the FPSLIM1000 setting). I work around that with internal
> triggering the camera but that has it's on glitches (for instance
> performance strangely drops if exposure time is below 3ms)
>

Can you provide more details? What does "performance strangely drops" mean -
frame rate? Image quality? Something else? You can upload image samples with
explanations to our wiki (and make a page about this project).


> and none of both options seem to work well in linescan mode and manual
> exposure....
>

I'll it it.

Andrey

>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://support.elphel.com/pipermail/support-list_support.elphel.com/attachments/20100511/1acf0eed/attachment-0002.html>