Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
16 years ago

Is the speed determined by fmax or delays?

Hi everyone,

This may seem stupid but I can't get a direct answer from my professors on this question.

One of them says its the fmax while one of them says its the delays.

Anybody mind shedding some light on this topic?

And if it's the delays, which one?

Appreciate any replies

Kind regards,

Chris

11 Replies

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor

    To further add to my post about fmax anatomy:

    Fmax is dependant on register Tsetup and reg to reg delay.

    --------x--------------x------- launching clk edge

    ----------x--------------x ------------ launched data transitions

    dd--------x---------------x------- latching clk edge

    dd----------x---------------x ------------ arrived data transitions

    .................AAAAAAAAASSHH...................S is setup window

    the arriving data transition has the interval window AAAA.. else it will violate setup at S.

    this window = clk period - SS -reg Tco(which is intrinsic)

    Remeber f = 1/clk period

    The AAAA.. interval is normaly not through direct wire but a comb. section. Thus you can break it up shorter to approach maximum posiible speed for the given device which then is limited by intrinsic reg tco and reg quality...

    arrival of data beyond at HHH violates hold time. this is due in practice to clk arriving late and is tackled by internal design to ensure fast clk tree.

    I said in practice because it might well occur due to excessive data delay but at such point fmax is already violated and so hold violation drops from discussion.