[Elphel-support] Internal CF Cards

[Andrey Filippov]

On Tue, May 10, 2011 at 6:43 AM, Andreas Bean <office at beanbox.com> wrote:

> Andrey,
>
> I have seen that two CF cards can be put inside the cam.
> With this setup, does the cam achive the max write speed of 10Mbyte/sek?
> Whats the maximum supported size?
>

Andreas,

Maximal sustained rate that the camera can record is 16MB/s, limited by
ETRAX on-chip IDE controller. With the CF cards there is another issue (
described in http://www.linuxdevices.com/articles/AT5102023409.html ) - most
CF cards report DMA operation, while they actually do not have it - it is
caused by a bug in the controller chip made by Silicon Motion. In most
applications they can safely hide that bug (falsely report DMA) because it
does support UDMA, and UDMA implementation has a bug in ETRAX. When the new
CF card (not yet tried and "blacklisted") is connected to the camera, camera
reads the device ID information and assume tries to use it in DMA mode, and
it fails. It is possible to put this make/model into blacklist (as we did
with many cards) so the software will access it only in PIO mode, but that
will be both slow and eat up CPU resources, influencing other applications.

So far we use just Sandisk Extreme III series of CF cards that are known to
support DMA mode.


>
> Btw, I have read the development blog. It's very interesting for me,
> since we have build a panoramic camera too, as you know probably.
> I tried the leveling and image stabilization a couple of month ago,
> unsuccesseful.
> We tried two different IMU units and several filter algorithms for the
> image stabilization. None of the worked satisfactory, actually.
>

We are working now with Analog Devices ADIS16375 IMU, the new FPGA code
(I'll release it after when finish testing) supports logging of the IMU data
at high rate that IMU can provide (>2 Ksamples/sec)  combined with the
serial GPS data (5-10Hz), supporting optional precise GPS timing pulse
(otherwise using the beginning of the NMEA sentence for time-stamping). The
same FPGA module also accepts accepts/logs external input (i.e.
odometer/pedometer pulses) and inter-camera synchronization (to precisely
match local time with that of a master camera). All the data for these four
channels is buffered and transferred to the system memory over the DMA
channel without additional load on the camera CPU - with the ADIS-16375
running at the full rate (2.4Ksamoples/sec) and the record size fixed to 64
bytes/sample (other logged sources have much lower bandwidth) it is up to
160KB/sec of position/orientation raw data to be recorded - still much less
than the images.

We do not plan to run any image stabilization algorithms in the camera - as
the data bandwidth is manageable, we will do that in the post-processing
where it is easier to experiment with different filtering algorithms. But we
do not have that software yet, so we'll definitely appreciate help in this
area - as soon as we'll have the real data recorded, we'll post it.

Andrey



>
> Andreas
>
> _______________________________________________
> Support-list mailing list
> Support-list at support.elphel.com
> http://support.elphel.com/mailman/listinfo/support-list_support.elphel.com
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://support.elphel.com/pipermail/support-list_support.elphel.com/attachments/20110510/e1d3bf82/attachment-0002.html>