Transcoder / Encoder - technically, how do they differ ?

[TeeeRex]

I haven't got any particular problem I just thought I would ask the question to increase my knowledge on the subject.

How does a transcoder (shrink or whatever) manage to do it's work so quickly against an reencoder (CCE/TMPGenc) - WHAT exactly is shrink removing from the bitstream to make it smaller ?

[BJ_M]

many transcoders dont have to recalculate the very complex motion vectors or the relationship beteeen diffeent GOB's bitrate or I frame placement on scene changes or a few other intensive computations .. they just take exsisting data and lower the bit rate (higher comression) , they prob. also throw out some data to do so..

at least thats my understanding how some of the high speed transcoders work .

[FOO]

MPEG2 is stored as the frequency transform of the pictures
High frequency for fine detail., Low for gross detail.
The values are also quantized according to perceived importance
for example 8 bit number for important ones, 5 bits for less important.
I think the shrink process involves lowering the quantization
and eliminating high frequency components.
Hence the loss of fine detail

Note that way they don't have to do the frequency transformation
or the motion search which are expensive in time. (DCT)

[TeeeRex]

Thanks for the replies FOO and BJ_M - it's starting to make sense - this relates to the Q values in one pass VBR etc ...

Cheers,

TeeeRex