[net.unix] A second DMA

bruce@ctcgrafx.UUCP ( Bruce Ross x275 ATVT ) (09/12/85)

*** RELINE YOUR PLACE WITH THIS MESSAGE ***

    Unix Users Great and Small:

      I'm the owner of a Cadmus 9000 unix box which is connected to
    a Raster Technologies model One/80 graphics unit via DMA link.
    It works very well (ever since I got the bugs out the driver).
    I am in need of a high speed parallel port to communicate with
    a high resolution frame grabber.  A single frame of video rep-
    resents 3.9 Mbytes of RGB data. My frame grabber is being
    modified currently to interface with a DEC type 16 bit DMA link
    using a 68000 CPU board for control and data formatting. I would 
    like an opinion (many, in fact) on the best way to proceed.  My
    options are these :

    1) Forget the parallel link and use one or more serial ports.
    At first this seemed like a great idea until I calculated that
    under the best conditions at 9600 baud a single frame could 
    take up to 80 minutes to transfer.   Not a great idea !!!!

    2) Switch the existing DMA controller (using a kludge data
    switcher) between the Raster 0ne/80 and the grabber interface.
    Too inelegant!!!!

    3) Call Cadmus and ask them to do the job and charge us big $$$
    for it.  Their reply was for them to modify their current DMA
    driver to handle multiple devices in slave mode.  We would like
    to have interprocessor link mode so that either the cadmus or the
    frame grabber host would be able to issue interrupts for the other
    to respond.   This idea did not appeal to Cadmus. 

    Based on what Cadmus wants to do with this situation and my 
    relative ignorance of such things, is it possible (or desirable)
    to take the device driver in the form that it is offered and
    modify it to do the interprocessor link ?   What it seems like
    to me is modifying the device driver so that it will handle 
    the 0ne/80 normally (slave mode) as well as the interprocessor
    link mode using an IF or CASE statement within the driver itself.

    I welcome any suggestions regarding an outright solution to the
    problem or a modification to any of the above ideas which I may 
    have overlooked or anything else which may be helpful.  Please
    address any flames to /dev/null or other dark places.
    Paid consultants are also a possibility (you must have excellent
    credentials).  Please send all replies to !ctcgrafx!bruce or !root.

    Your help will be greatly appreciated.

berger@datacube.UUCP (09/20/85)

Since you did not specify much of a path, I will place a response
here...

Another alternative would be to use one of our frame-grabbers that
plugs directly into the Q-bus.  We have a 768 x 512 x 8 bit frame-grabber
board set that can plug directly into a cadmus 9000.  I even have some
driver code for it!

One of our oems uses 3 of these QVG-123 board sets for true 24 bit color
digitization and display.

We also make various other frame-grabber / image-processing boards
for VME, Multibus and the IBM PC.

(Hope this is not too much of a commercial!:)

			Bob Berger 

Datacube Inc. 4 Dearborn Rd. Peabody, Ma 01960 	617-535-6644
	
ihnp4!datacube!berger
ima!inmet!mirror!datacube!berger
decvax!cca!mirror!datacube!berger
decvax!genrad!wjh12!mirror!datacube!berger
{mit-eddie,cyb0vax}!mirror!datacube!berger

bruce@ctcgrafx.UUCP ( Bruce Ross x275 ATVT ) (09/23/85)

> *** RELINE YOUR PLACE WITH THIS MESSAGE ***
> 
>     Unix Users Great and Small:
> 
>       I'm the owner of a Cadmus 9000 unix box which is connected to

Regarding this message, I would like to clarify the return address:

!bruce!ctcgrafx!ittatc!decvax!philabs

will work.

Bruce Ross