rising edge of the clock and data

[john]

Hello,

I am trying to get the following timing waveform

http://img337.imageshack.us/my.php?image=timingwaveformla7.png


The timing waveform suggests that data should get ready atleast a half
cycle before the rising edge of the DCLK. I am unable to do it. My
program gets the data ready exactly at the rising edge of the clock. I
know that I can not delay the data. Please advice!

Regards,
John

 

[Pieter Hulshoff]

I am trying to get the following timing waveform
http://img337.imageshack.us/my.php?image=timingwaveformla7.png
The timing waveform suggests that data should get ready atleast a half
cycle before the rising edge of the DCLK. I am unable to do it. My
program gets the data ready exactly at the rising edge of the clock. I
know that I can not delay the data. Please advice!

Is that your input or your output timing? If it's your input timing, then I
would assume you are to generate that signal from your testbench, in which case
you can use all the delay elements you want. What kind of clock is that DCLK
anyway or is that just an incorrectly drawn interruption in the clock cycle?

Kind regards,

Pieter Hulshoff

 

[john]

Hello,
Its the output waveform. DCLK is a 50 percent duty cycle 1MHz clock.

Regards,
John

 

[David Spencer]

Hello,
Its the output waveform. DCLK is a 50 percent duty cycle 1MHz clock.

Regards,
John

Are you generating both DCLK and Data as outputs? If so, they hopefully your
internal clock is running faster than 1MHz. Assuming it is, you just need to
design your logic that generates the signals to ensure that Data changes
before DCLK. I would be surprised if you really need half a cycle (500 ns)
setup time. All you need to ensure is that you meet the documented setup and
hold times for the device being driven.

 

[Pieter Hulshoff]

Its the output waveform. DCLK is a 50 percent duty cycle 1MHz clock.

If your output is both clock and data, consider inverting the clock output.
Alternatively, use a negative edge clocked FF at the output with a postive edge
clocked FF right before it (so you have a FF to FF transfer without logic in
between; should work fine at half clock cycle).

Kind regards,

Pieter Hulshoff

 

[john]

Hi,
Can you provide me with RTL view of the logic that you adviced? It
will be easy for me to follow.
Thanks
John

 

[john]

If your output is both clock and data, consider inverting the clock output..
Alternatively, use a negative edge clocked FF at the output with a postive edge
clocked FF right before it (so you have a FF to FF transfer without logic in
between; should work fine at half clock cycle).

Kind regards,

Pieter Hulshoff

Hi,

Should I add the positive edge FF before the negative edge clock for
CLOCK output or the data output?

John

 

[Pieter Hulshoff]

> Hi,
> Can you provide me with RTL view of the logic that you adviced? It
> will be easy for me to follow.
> Thanks
> John

Sure:


clock inversion solution:

clk_out <= NOT( dclk );


FF solution:

pre_ff_reg: PROCESS IS
BEGIN
WAIT UNTIL clk = '1';
dout_pre <= data_internal;
END PROCESS pre_ff_reg;

output_ff_reg: PROCESS IS
BEGIN
WAIT UNTIL clk = '0';
dout <= dout_pre;
END PROCESS output_ff_reg;

Add reset code depending on your wants and needs.


Kind regards,

Pieter Hulshoff

 

[john]

Hi,

I tried your advice. Please go to the following link to see the output
waveform,

http://img142.imageshack.us/my.php?image=waveformza5.png


As you can see from the waveform, that everything is happening
according to plan except one "DCLK" cycle is apperaing just after the
Tag. The Data is getting stable before the rising edge of the DCLK.
The Tag looks OK too. But I need to get rid of this one clock cycle
because it might serial out the zero first and the whole number will
be wrong. Any advice!

Regards,
John

 

[john]

Hello,

Your advice did solve the data problem but the "Tag" signal is not
synchronized at all. I need the "DCLK" to be OFF (zero) when "Tag" is
high for one clock cycle then "Data" (SER_OUT) has to be at the output
for half of the "DCLK" clock cycle before the "DCLK" gets to the
output pin of the FPGA. I got this waveform as you can see in the
following figure 1

http://img100.imageshack.us/my.php?image=waveformfq8.png

I got figure 1 using the state mode of the logic analyzer. But when I
use the timing mode I get everything same except the Tag repeats
itself several time during serial outing the one set of data which is
wrong. Please advice! My code is following

serialcount : process ( clk_b, Reset_counter)
begin
if ( Reset_counter = '1') Then
p2s_counter <= ( others =>'0');

elsif rising_edge (clk_b) then
p2s_counter <= p2s_counter + 1;
ser_out_sig_1<= ser_buff ( to_integer ( p2s_counter ) );
else
end if;
end process;

-- Data Clock Delay
serial_count_delay : process (clk_b, Reset_counter)
Begin
if ( Reset_counter = '1') then
elsif rising_edge(clk_b) then
ser_out_sig_2 <= ser_out_sig_1;
else
end if;
End process;

-- Data Clock Second Delay
process (clk_b, Reset_counter)
Begin
if ( Reset_counter = '1') then
elsif falling_edge(clk_b) then
ser_out <= ser_out_sig_2;
else
end if;
End process;


div : process ( clk1x)
begin
If ( Reset_counter = '1') Then
clk_divider <= "00000000000";
elsIf rising_edge (clk1x) then
clk_divider <= clk_divider + 1;
end if;
end process;

process(Reset_counter)
begin
If ( Reset_counter = '1') Then
Tag_out_0 <= '1';
Tag_out_1 <= '1';
elsif falling_edge (clk_b) then
Tag_out_1 <= Tag_out_0;
Tag_out_0 <= '0';
end if;
Tag <= Tag_out_1;
end process;

Regards,
John

 

[jens]

Some random pieces of advice:

1. Simulate the design.

2. Use only one clock domain. Synchronize all signals with that
clock.

3. Simulate the design.

4. Where is Reset_counter coming from? If it's coming from the clock
domain, don't use it as an asynchronous signal. If it isn't coming
from the clock domain, change it so it is.

5. Simulate the design.

6. Look at the logic analyzer's documentation about state vs. timing
mode (hint: use timing mode for timing analysis). Then only use the
logic analyzer once you're happy with the simulation results.

7. Check your sensitivity lists, some are not complete.

8. Simulate the design.

9. Format the code using indentation so that it's legible.

10. Simulate the design.

 

[john]

The Reset_counter is generated by the state machine running at much
higher frequency than the clk_b. How can I make them synchronous?
John

 

[jens]

Divide down the high frequency clock to generate a low frequency clock
enable. Use the high frequency clock as the clock for the entire
design (if possible). Use the high frequency clock enabled by the low
frequency clock enable to do everything you need to do at the low
frequency, something like this:

if rising_edge(high_freq_clk) then
if low_freq_clk_en = '1' then
-- do stuff here
-- Reset_counter can be used here and will be synchronous
end if;
end if;

 

[Daniel Reidal]

Divide down the high frequency clock to generate a low frequency clock
enable. Use the high frequency clock as the clock for the entire
design (if possible). Use the high frequency clock enabled by the low
frequency clock enable to do everything you need to do at the low
frequency, something like this:

if rising_edge(high_freq_clk) then
if low_freq_clk_en = '1' then
-- do stuff here
-- Reset_counter can be used here and will be synchronous
end if;
end if;