>> We have a Dalsa 64T camera which is 12 bit 1600 fps camera with 64x64
>> pixels, ... We are having a lot of problems getting this camera to
>> work faster than 400 fps for 64x64 pixels.
>
>When many manufacturers sell a camera as "1600 fps" it means in fact
>that one frame can be taken at a speed of 1/1600 second. However, this
>is certainly NOT the same as saying that 1600 frames can be sent
>through the system (camera, grabber card, computer) in one second. In
>other words, the 400 fps might well be a de facto maximum.
>
>Camera manufacturer do not like to point out this difference: The same
>camera, sold in Germany will be rated X while it will often be rated
>about 4x in the US.
>


I would echo the point that it is probably the capture that is at fault.

We have a Matrox Meteor RGB frame-grabber which is NOMINALLY capable
of grabbing realtime video in RGB (768 * 540 * 3 colours approx at 25 fps -
approx 31 MB per sec.  Pal's camera is equivelent to only ~ 9 MB/sec, or
13 if it actually uses a 16bit word for each 12 bit pixel.)
However to get it to give anything like this spec was extremely hard -
fast computer, fast RAM, etc are not enough - the exact motherboard
chipset and configuration was crucial.  Getting the details of what was
required to achieve this was EXTREMELY DIFFICULT!  Neither the board
nor the software manufacturers seemed to be able to tell us exactly
what was required.  We have eventually got it to go at ~10fps in RGB
or real-time (25 fps) in composite, and that is clearly all we are
going to succceed in getting (it is good enough for us).  But it
needed very little to be sub-optimal to drop it down to less than
half this value.  In other words, with everything perfect we can
get just about the transfer rate Pal needs, and it's very easy indeed
for performance to fall down to about what Pal was actually getting.
It has taken us nearly two years, two brand-new computers, and two
versions of the frame-grabber, to reach this point.

So my message would be to get the manufacturers to tell you exactly
what hardware they recommend, and if there are any quirks in setting
up that are needed (one Intel chipset needs an undocumented 'magic'
bit to be set by software, for example).  And even if you do set it
up just as they say you probably still won't get _quite_ the performance
they promise.  (But then, you try to get the rated number of pages
per second from your laser printer ....)

                                                Guy Cox

Dr. Guy Cox,   |                    ooOOOOOOoo
E.M. Unit, F09 |        #       oOOOO  |  |  OOOOo       #
Univ of Sydney |       ###    OOO|  |  |  |  |  |OOO    ###
NSW 2006,      |       ###  OOO  |  |  |  |  |  |  OOO  ###
Australia      |       ### OO |  |  |  |  |  |  |  | OO ###
Phone:         |      #####   |  |  |  |  |  |  |  |   #####
+61 2 9351 3176| =====#####============================#####=====
Fax:           |      #####                            #####
+61 2 9351 7682|    ~~#####~~~~~~~~~~~~~~~~~~~~~~~~~~~~#####~~