Reseting on an edge or one-shot

[Steven Menk]

I have a state machine where I need to reset with asyncronise signal
but don't want to wait I have a state machine where I need to reset
with asynchronous signal but don't want to wait for that signal to go
back low before the machine goes out of its initial state. In other
words I'd like to reset either on the edge or during the first clock
cycle in which the reset is high and continue to run until it goes low
then back high again...

How would I go about doing this?

Thanks.

 

[Steven Menk]

Sorry. I screwed up when cutting and pasting from Word:

I have a state machine where I need to reset
with asynchronous signal but don't want to wait for that signal to go
back low before the machine goes out of its initial state. In other
words I'd like to reset either on the edge or during the first clock
cycle in which the reset is high and continue to run until it goes low
then back high again...

How would I go about doing this?


Thanks.

 

[Mike Treseler]

I have a state machine where I need to reset
with asynchronous signal but don't want to wait for that signal to go
back low before the machine goes out of its initial state. In other
words I'd like to reset either on the edge or during the first clock
cycle in which the reset is high and continue to run until it goes low
then back high again...

How would I go about doing this?

I would start by drawing typical input
and expected output waveforms on a clock grid.
Then I would name and fill in the internal
state values at each clock tick.

-- Mike Treseler

 

[Steven Menk]

That's the problem. The signals are completely asyncronous. The only
thing that is given is that the reset signal will be at least 3 clk
cycles long. Other then that the timing and edge relationships are
completely arbitrary.

What I ended up doing is taking two D flip-flops, hooking the reset
signal to the input of the first, and the input of the second to the
output of the first:

ARST -> D0 Q0 -> D1 Q1 ->
clk -> CLK
-> CLK

For the reset signal I said:

RST <= Q0 and not Q1.

Then:

process(clk, RST)
begin
if RST = '1' then
CS <= S0;
elsif clk = '1' and clk'event then
CS <= NS;
end if;
end process;

Seems to work, but I have a lot more testing to do. Always fun to see
something work 2 millions times in a row, but break on the
2,000,001st...

 

[Zara]

ARST -> D0 Q0 -> D1 Q1 ->
clk -> CLK
-> CLK

For the reset signal I said:

RST <= Q0 and not Q1.

But now this reset is synchronous!

Then:

process(clk, RST)
begin
if RST = '1' then
CS <= S0;
elsif clk = '1' and clk'event then
CS <= NS;
end if;
end process;

So the following is more correc:

process(clk)
begin
if rising_edge(clk) then
if rst='1' then
CS <= S0;
else
CS <= NS;
end if;
end if;
end process;


Best regards,

Zara

 

[Steven Menk]

Wouldn't that give me a race condition? My RST will go high on one
rising edge of clk then low on the next.

 

[Zara]

Wouldn't that give me a race condition? My RST will go high on one
rising edge of clk then low on the next.

From your previous message:

ARST -> D0 Q0 -> D1 Q1 ->
clk -> CLK
-> CLK

For the reset signal I said:

RST <= Q0 and not Q1.

So I deduce you are synchronizing reeset signal by sampling it (Q0),
and then delaying it by one cycle (Q1). The new RST signal appears
only on the rising edge of ARST, so there should be no race condition,
unless ARST is toggling. But in such case, the reset pulses would be
OK, as they meet the specificationos of OP

Best regards,

Zara