What is the best way to clock data in on one clock edge and out on another?

[simon.stockton]

Dear All,

I have a requirement to convert the edge of a clock that the data is
valid on. I want to clock data into an entity on the falling edge of a
clock and clock data out of an entity on the rising edge of the same
clock.

Is the code below sufficient to achieve this functionality? Any
pitfalls that I might fall into?

Regards,

Simon

***********************************************
entity clock_edge_conversion is
port (
word_stream_out : out std_logic_vector ( 9 downto 0 );
reset : in std_logic;
clock : in std_logic;
word_stream_in : in std_logic_vector ( 9 downto 0 )
);
end clock_edge_conversion;
architecture RTL of clock_edge_conversion is
begin
process ( clock, reset )
variable temp_word : std_logic_vector( 9 downto 0 ) := B"00" & X"00";
begin
if ( clock'event and clock = '0' ) then
temp_word := word_stream_in;
end if;

if ( reset = '0' ) then
word_stream_out <= (others => '0');
elsif ( rx_rocket_io_clock'event and rx_rocket_io_clock = '1' ) then
word_stream_out <= temp_word;
end if;
end process;
end RTL;
***********************************************

 

[KJ]

You'll probably have a hard time making it through synthesis. You'll
need to...
1. Break it into two processes, one triggered by the rising edge of
clock, the other by the falling edge.
2. Change 'temp_word' from a variable to a signal so that it can be
transferred between the two above mentioned processes.
3. The 'initial value' that you assign to temp_word is usually not
synthesizable. Add an if condition to initialize it to what you want
based on the signal 'reset' much like you have for "word_stream_out".
4. Consider using 'rising_edge(x)' and falling_edge(x) functions that
are a part of ieee lib instead of x'event and x = '1'. It's more
readable.
5. I'm assuming that "rx_rocket_io_clock" is a typo and that you meant
to say "clock" since there is no signal called "rx_rocket_io_clock"

KJ

 

[Kolja Sulimma]

4. Consider using 'rising_edge(x)' and falling_edge(x) functions that
are a part of ieee lib instead of x'event and x = '1'. It's more
readable.

The second form is also a simulation mismatch.
A transition of x from 'U' to '1' (as it can occur at simulation start)
is not a rising edge, but it is an event on x with x = '1'.
The same is true for transitions from 'H' to '1' and 'X' to '1'.

Kolja Sulimma

 

[simon.stockton]

Thanks for the reply,

The code does actulaly synthesis (withouth the typo). However, is it
possible that the data on word_stream_out could not be the same as the
data on word_stream in, with this current implementation?

If I seperate it into two processes (one on the rising edge and another
on the falling edge) and with a temp_word signal between them, will
this give me the functionality that I require?

Simon

 

[Peter Alfke]

There is a very simple, albeit wasteful solution:
Just cascade two flip-flops. The first one clocked on the falling
edge,(usually indicated by a bubble on the clock input), its Q driving
a second flip-flop's D, and this second flip-flop is clocked by a
rising-edge clock
This contraption accepts data right before the falling edge, and
outputs the data right after the subsequent rising edge. And the output
stays stable until the next rising edge. Ideal behavior.
It uses a total of four latches, and runs only at half the conventional
clock rate, but it uses conventional circuit elements. Efficiency be
damned!
Peter Alfke