OK, found the problem.
The "Pattern" repeats every 1024 bytes !!!!!!!!
The original blinker code I used in the 232 version was sending a block marker then the block address then the data and finally a crc.
In the example 03 78 03 78 0A 55 AA 50 00 24 82 FF,
03 78 03 78 seems to be the crc of previous block then
0A 55 AA 50 is the block marker then
00 24 82 FF gives a block address of 0xFF822400.
I have edited the 232 blinker I used to remove all this, and will rerun and check it tonight.
I found out that my company use IDA and have managed to get a laptop (with it loaded) from them so I can now check the dump in IDA as well.
Hopefully I should have a good clean dump later tonight.
If we can get this version of the firmware dump cleaned, does anyone think it would be possible to get udumper "fixed" so it could deal with these cameras in the future ?.