Re: Do we need a faster readout?

[Tom Droege <tdroege@veriomail.com>]

Chris and all,

This all started when it took Michael 300 seconds to read out the images to 
disk with the computer he had available.  In this case, it does matter a 
lot.  Further, the time it takes in QBasic is just marginal for not wasting 
time.  So I think the exercise was worth the effort.  I also expect to get 
out of this effort the ability to write an arbitrary .fts header from the 
QBasic program using CFITISIO.  This I think is also worth the effort.

I expect to be using the QBasic program for a long time.  That way I can 
make changes and test them without negotiating with a programmer somewhere 
to get something compiled in.  I am sure that you do not want to do this 
for me at the frequency that I often do it.

The shutter speed it part one of this effort.  Timing the shutter is messed 
up by the present mystery 10 second feature of Windows/late DOS.  Doug is 
working on this.

My QBasic routine attempts to do the same sort of thing that you describe 
below.  I can eventually do it pretty well.  QBasic has an interrupt 
feature.  Well not exactly an interrupt, but it is possible to interrupt 
the program every second and go to a scheduler.  This works except when you 
are timing shutter operations which are the only thing that is time 
critical on the order of 1 second.  When I get around to it I could change 
my QBasic program to poll.  It all depends on other software appearing that 
makes me not want to take the effort.  ;^)

I note that you have programmed in some nice features, like the shutter 
protection.  These things are worth doing.

I do understand the read out time bottleneck.  The Mark IV was designed to 
be modular.  It will be possible to reduce the read out time to of order 10 
seconds in the future by changing one PC board.  The parts are available 
which will allow this.  I just did not want to push the first systems.  If 
it turns out that this will make a big difference in the future I will 
design a new ADC board.

Tom Droege

At 06:14 PM 2/6/01 -0800, you wrote:



>All this talk about reading images out of the RAM card and
>saving to disk really is looking to save time where it does
>not matter much.  What we should be concerned with is the
>shutter open/closed ratio.  That's all that matters.  Read
>out speed matters only in so far as it effects the above ratio.
>
>The way I look at it, the image moves through an assembly line.
>While it takes Ford 200 man hours to assemble a car they can
>produce more then one car every 200 hours.  We can do the same
>thing.  My real-time controller attempts to use Ford's method.
>The software moves an image through a pipeline:
>
>1) Shutter opens, Image in on the CCD array,
>2) Shutter closes, Image is moved from CCD to RAM Buffer Card,
>3) At any later time, Move image from RAM Buffer Card to internal
>    arrays, one for each color
>4) Move image arrays to the current list of "output locations"
>    these may be any or all of the following:
>    a) FITS file
>    b) A soft copy display device (i.e. some window on
>       some computer's screen)
>    c) A (possibly reduced size) tiff, gif, or jpg file.
>    d) Some program (like a data reduction pipeline).
>
>It tracks which location contains an image or is empty.  For
>example I can't open the shutter if the CCD still contains
>an not yet read out image.  I can't read out the CCD if the
>RAM Buffer Card has not yet been read into the internal
>array.  I can't read out the RAM card if the internal array
>has not been saved.  So in the worst case we have to delay
>the shutter opening a few seconds while the pipeline clears.
>
>If the exposure times are reasonable we have as I said
>"all the time in the world" as all this gets done when the
>shutter is open.  A 100 second exposure gives us 100 seconds
>to mess with the last image.  Even QBASIC is fast enough.
>If you are taking 60 second exposures I really don't see
>why the readout and save to disk operation needs to be fast.
>
>The only bottle neck in the system is the CCD readout time.
>For a survey you'd like to have the shutter open 100% of the time.
>Increasing the shutter open/closed ratio is as good as
>increasing the lens' f-ratio.
> >