How to get two different clock

[Amit]

Hello group,

Is this possible or does it make sense to create an element which its
output become a clock of other elements of a circut? I need to clock
two different entities. One gets a clock as milisecond and one as
second!



Thanks,
Amit

 

[Szymon Janc]

Amit napisa³:

Is this possible or does it make sense to create an element which its
output become a clock of other elements of a circut? I need to clock
two different entities. One gets a clock as milisecond and one as
second!

Use clock-enable signals to hold your elements. And read about 'gated
clock' and problems connected with it.

 

[David Spencer]

Is this possible or does it make sense to create an element which its

output become a clock of other elements of a circut? I need to clock
two different entities. One gets a clock as milisecond and one as
second!

There is no simple answer to this really. The purists will argue that such
things should never be done. However, in the real-world it is in general OK
providing a couple of facts are true:
1. The technology you are using allows the generated clock to be distributed
as if it was a normal external clock. In the case of FGPAs, this typically
involves instantiating a global clock buffer to force the generated clock to
be routed on dedicated clock routing nets.
2. Your design doesn't rely on any timing relationship between the clocks.
In other words, when interfacing between logic using the two clocks one
needs to follow normal multi-clock domain procedures.

Whether it's a good idea to use generated clocks or not is a different
matter, and depends on what you are trying to do. If, for example, you have
a design that features two unconnected blocks of logic - one running at
100MHz and the other at 50MHz, it would make sense to generate the 50MHz
clock by dividing the 100MHz one. However, in your case you are probably
better running everything off the system clock and implementing a prescaler
that generates a one-clock wide enable every millisecond (and another every
second) that are then used to gate your low-speed entities.

 

[Ralf Hildebrandt]

Amit schrieb:

Is this possible or does it make sense to create an element which its
output become a clock of other elements of a circut? I need to clock
two different entities. One gets a clock as milisecond and one as
second!

Clock gating is the key to low-power designs, but it is not simple as
clock-skew and data transfer between clock domains has to be considered.

Use clock buffers and clock constraints for your gated clocks, if
available. If not use at least a constraint during synthesis.

Even with only a clock constraint (and without a suitable clock buffer)
it _may_ be possible to get a running design at a FPGA. But it will be
very slow.

If you don't need clock gating, try to avoid it.

Ralf

 

[Amit]

Clock gating is the key to low-power designs, but it is not simple as
clock-skew and data transfer between clock domains has to be considered.

Use clock buffers and clock constraints for your gated clocks, if
available. If not use at least a constraint during synthesis.

Even with only a clock constraint (and without a suitable clock buffer)
it _may_ be possible to get a running design at a FPGA. But it will be
very slow.

If you don't need clock gating, try to avoid it.

Ralf

Thank you all for your response. To be honest with you, I'm still
confused about the clock so please let me explain what I have
understood so far then I will be more than happy if you corrent me:

The most confusing part for me is the master clock. I think I have to
create an entity as ent_CLOCK for instance? then defining some
constraints for this clock someting like a counter to check number of
clock pulses. Since it must divide the time in 1000 slices and send
them to entity_1 indvidually and when its COUNTER = 1000 then it
should clock the entity_2 too.

Is my assumption right?

Regards,
Amit

 

[Mike Treseler]

Amit wrote:

The most confusing part for me is the master clock. I think I have to
create an entity as ent_CLOCK for instance?

The master clock is a crystal oscillator soldered
on the circuit board. It is just an input
port to the synthesis entity.

For simulation, I might model this in my testbench
with a process using wait statements like this.

begin -- process tb_clk
clk_s <= '0';
if now < rst_time then
rst_s <= '1'; -- rst once with clk running
else
rst_s <= '0'; -- then low forever
end if;
if done_s then wait; end if; -- Stop clock
wait for clk_cy/2; -- clock low
clk_s <= '1';
fixed_stim; -- auxiliary outputs
wait for clk_cy/2; -- clock high
end process tb_clk;

for details see:
http://home.comcast.net/~mike_treseler/

then defining some
constraints for this clock someting like a counter to check number of
clock pulses. Since it must divide the time in 1000 slices and send
them to entity_1 indvidually and when its COUNTER = 1000 then it
should clock the entity_2 too.

The counter output is used as a clock *enable*,
Not a clock. Here is an example:

http://home.comcast.net/~mike_treseler/count_enable.vhd

-- Mike Treseler

 

[Shannon]

Amit might be trying to create a simple divided clock. Here is an
example of one of the ten thousand ways of making a divided clock.
This is a divide by 4 clock: (DIVISOR := 4)

PROCESS (MClk, Reset)
BEGIN
IF (Reset = '1') THEN
clk_per_cur <= (OTHERS => '0');
ELSIF (rising_edge (MClk)) THEN
clk_per_cur <= clk_per_next;
END IF;
END PROCESS;

clk_per_next <= (OTHERS => '0') WHEN clk_per_cur = (DIVISOR-1) ELSE
clk_per_cur + 1;


Shannon

 

[Amit]

Amit might be trying to create a simple divided clock. Here is an
example of one of the ten thousand ways of making a divided clock.
This is a divide by 4 clock: (DIVISOR := 4)

PROCESS (MClk, Reset)
BEGIN
IF (Reset = '1') THEN
clk_per_cur <= (OTHERS => '0');
ELSIF (rising_edge (MClk)) THEN
clk_per_cur <= clk_per_next;
END IF;
END PROCESS;

clk_per_next <= (OTHERS => '0') WHEN clk_per_cur = (DIVISOR-1) ELSE
clk_per_cur + 1;

Shannon

THANKS ALL FOR YOUR HELP. I TRULY APPRECIATE IT.

REGARDS,
AMIT