More synchronization problems

[Simon]

Hi

I'm having some problems understanding synchronizing my design and reducing
the metastability. I was hoping someone could answer some of my questions.

In my design I have a main system clock called SYS_CLK, which comes from a
programmable timer. This clock runs at 32MHz. An input to my system is a
NRZI signal. The output is the bit values extracted from the NRZI signal.
Inside my system I have designed a phase locked loop (to extract the clock
from the NRZI signal) and I also have a couple more processes which, among
other things, look for certain bit sequences. My questions are:

1) Should everything be synchronized to SYS_CLK?
2) If I pass a signal to a comonent e.g. my phase locked loop, should that
signal be synchronized inside the component or outside the component?
3) Is my sychronizer correct (I have included the code at the bottom of this
message)?
4) If I synchronize a signal for one component, will I have to synchronize
the same signal again for another process?
5) From my 32MHz clock I generate an 8MHz clock for use with the phase
locked loop. Is the 8MHz clock already synchronized correctly? Can I then
just pass the 8MHz clock straight to the phase locked loop and use it
without any extra synchronization?

Thanks for any help with this as I'm really stuck. If anyone has any
further advice or pointers concerning this subject then please let me know.

Thanks again,



-- Synchronize CLK_IN with SYS_CLK
entity syncer is
Port ( RST : in std_logic;
SYS_CLK : in std_logic;
CLK_IN : in std_logic;
SYNC_OUT : out std_logic
);
end syncer;

architecture syncer_arch of syncer is
signal ff : std_logic;
begin

p1 : process(RST, SYS_CLK)
begin
if RST = '0' then
ff <= '0';
elsif rising_edge(SYS_CLK) then
ff <= CLK_IN;
end if;
end process main;

p2: process(SYS_CLK)
begin
if rising_edge(SYS_CLK) then
SYNC_OUT <= ff;
end if;
end process test;

end syncer_arch;

 

[tbx135]

As I understand what you wrote, it sounds like you are recovering clock and
data from an NRZI serial source. A recovered clock from this stream, using a
PLL, will have a certain amount of jittering, drift and wander. This may be
an issue - did you validate the circuit?

I don't have a clear picture of your data path, I am assuming you want to
resyncronize to SYS_CLK at some point, is this true? A FIFO may be a quick
way to accomplish this.

Next you have an asyncronous reset in your first process, a big no-no. The
second process has no reset statement, a bigger no-no. I'm suprised the
synthesis tool isn't whining about the second process.

Please clarify the above points.