CCD-world: S16D interface...

[Sidik Isani]

The following was posted to CCD-world:

Hello List, Klaus -

  I got S16D+DMA ring buffer mode working for our CFH12K camera.
  Our system has worked reliably on a Sparc 20 with Solaris 2.5
  and 2.6.  Indeed, EDT's documentation has a few errors, and
  fails to point out some pitfalls.  I hope something here can
  help you out...

  It's been a while, but here are some quick notes off of the
  top of my head:

  - I would use "libedt" from www.edt.com to program DMA mode.
    All the calls look like edt_something().  Also look on their
    Web site for the latest dmalib.pdf (or whatever they call the
    document that shows how to use the library.)

  - If you use that, you *should* be able to switch to an EDT
    PCI16D later on and just re-compile your code.  You alpha
    machine has a PCI bus?  There's probably no driver for the
    PCI16D for that machine though, unfortunately.

  - EDT needs to upgrade old rev. boards or DMA mode will not work
    on sun4m architectures!  This was not covered under warranty.
    They will be able to tell you if your rev. is ok or not.  Try
    calling EDT.  They are a small company, and their telephone
    support is reasonably quick if the right person is in.  If
    *they* don't remember, I can look for e-mails exchanged with
    them on this subject, but they should be able to tell you.

  - Other known features of the EDT software:
    o edt_byte_count() is off by 2 (subtract 2 to get the right value.)
    o edt_byte_count() may return strange values during a transfer,
      because the update of the counter is not atomic.  (We detect the
      end of the readout by checking for the same byte count twice
      in a row, and then we know _that_ value is reliable.)
    o Your readout size should be a multiple of 16 _words_, in DMA mode.
      This is the size of the FIFO in the card, I believe.  If the last
      few pixels don't fill it up, you will loose them.  This only happens
      with odd subraster sizes, and we work around it by padding zeroes.
    o Make sure S16S_RTIMEOUT is 0 before configuring ring buffers.
      This read timeout does not work/apply(?) in DMA mode and should
      be set to zero.
    o Be sure you always exit DMA mode cleanly in all cases before going
      back to any regular I/O.  (edt_disable_ring_buffers()).
    o I had a few problems with handling of the ALARM signal with
      programs linked with libedt.  If you do anything that uses SIGALRM,
      maybe save/restore the signal mask and interrupt vectors each time.

  I'd check with EDT (the phone number is on their Web site) and find
  out if your board needs to be upgraded first...

  Good luck!

- Sidik <isani at cfht.hawaii.edu>

Klaus Reif typed:
> Date: Thu, 24 Feb 2000 11:54:41 +0100
> To: CCD-world at astro.ku.dk
> From: Klaus Reif <reif at astro.uni-bonn.de>
> Subject: CCD-world: S16D interface, circular buffer mode
> 
> Hi all,
> 
> we have to implement the control and data acqusition software
> for the Bonn University Simultaneous Camera (BUSCA,
> http://www.astro.uni-bonn.de/~ccd/busca/busca.html) on a Sparc10 
> (Solaris 2.5). (The software was developed and is running currently 
> on a DEC Alpha.) It seems that everything works fine except for the 
> circular buffer data acquisition mode of Sbus S16D DMA interface. 
> Tranferring data into a single buffer (tested up to 32MB) does work.
> But BUSCA does have 4 CCD systems with 4Kx4K pixels (each).
> To transfer the >128MB we need to make use of the circular buffer mode:
> On the DECAlpha we have realized a real time display of the CCD data
> and some calculations (data statistics) and the disk transfer are 
> running in parallel to the CDD readout. According to the S16D 
> documentation this mode is available. But we didn't get it working yet.
> The documantation is bad.
> 
> Who uses the S16D and got this particular mode working, so that I can ask
> him for further help? 
> 
> Thank you, Klaus Reif
> *******************************************************************
> Klaus Reif
> Sternwarte of the Univ. of Bonn                 Phone:    +49 228 737834
> Auf dem Huegel 71                                   FAX:     +49 228 733672
> D 53121 Bonn                                          Email:
> reif at astro.uni-bonn.de
> Germany                                                   WWW:
> http://www.astro.uni-bonn.de/~reif
> *******************************************************************
> 
> 
> 
> 
> 
> - -- CCD-world -- --
> CCD-world is fully moderated. Send posts to CCD-world at astro.ku.dk
> Standard replies will go to the list; address personal replies manually.
> For more information, please go to:  http://www.not.iac.es/CCD-world/
> 
- -- CCD-world -- --
CCD-world is fully moderated. Send posts to CCD-world at astro.ku.dk
Standard replies will go to the list; address personal replies manually.
For more information, please go to:  http://www.not.iac.es/CCD-world/