Send a PULSE on input change, asynchronous

[Jamie]

Trying to formulate and asynchronous design again. I need to generate
a low pulse to signal an IRQ and wake-up a microcontroller.

My conditions.

port(A, B, RIGHT, ENTER: in std_logic; PULSE: out std_logic

1. on A and B any change should send a pulse.
2. ENTER and RIGHT only a high-to-low (falling_edge) should send a
pulse

Any suggestions on this?

 

[Thomas Stanka]

Trying to formulate and asynchronous design again. I need to generate
a low pulse to signal an IRQ and wake-up a microcontroller.

My conditions.

port(A, B, RIGHT, ENTER: in std_logic; PULSE: out std_logic

1. on A and B any change should send a pulse.
2. ENTER and RIGHT only a high-to-low (falling_edge) should send a
pulse

Any suggestions on this?

There are a few problems concerning the width of the pulses, racing
conditions and so on. But in general I would use a simple xor to
detect changes.
First you have to delay each input to get a chance to see changing
values.

A_Delayed<=NOT NOT A; -- Force synthesis to keep this delay

Pulse_A<=A_delayed XOR A;
....
Pulse_Enter<=(Enter_delayed XOR Enter) AND NOT Enter;-- falling Edge

Pulse<=Pulse_A OR .. OR Pulse_Enter;

bye Thomas

 

[Jamie]

There are a few problems concerning the width of the pulses, racing
conditions and so on. But in general I would use a simple xor to
detect changes.
First you have to delay each input to get a chance to see changing
values.

A_Delayed<=NOT NOT A; -- Force synthesis to keep this delay

Pulse_A<=A_delayed XOR A;
...
Pulse_Enter<=(Enter_delayed XOR Enter) AND NOT Enter;-- falling Edge

Pulse<=Pulse_A OR .. OR Pulse_Enter;

bye Thomas

What option forces synthesis to keep the 'NOT NOT A' delay?

And, are all these statements under the same process()?

Thanks

 

[Mike Treseler]

Trying to formulate and asynchronous design again. I need to generate
a low pulse to signal an IRQ and wake-up a microcontroller.

My conditions.

port(A, B, RIGHT, ENTER: in std_logic; PULSE: out std_logic

1. on A and B any change should send a pulse.
2. ENTER and RIGHT only a high-to-low (falling_edge) should send a
pulse

Any suggestions on this?

Consider a synchronous process.

If this is out of the question, then consider
an HC14 and some diodes, caps, and resistors.
At least you can tune up the delays that way.


-- Mike Treseler

 

[Thomas Stanka]

(e-mail address removed) (Thomas Stanka) wrote:

[shorted to a line]

What option forces synthesis to keep the 'NOT NOT A' delay?

And, are all these statements under the same process()?

First I don't mean to insult you in any way.
If you can't answer this questions by your self, your IMHO not
experienced enough for doing asynchronous design in its real meaning.
Please start doing synchronos designs, you will likely fail getting
your synthesis tools doing the right thing in an asynchronous design,
if you even can't figure out how to instantiate two inverter. I never
done a real asynchronous design for myself and I'm not sure if I ever
want to because of the mass of problems I expect to encounter.

What option forces synthesis to keep the 'NOT NOT A' delay?

There are several ways forcing your synthesis tool keeping two
inverter, depending on which tool you are using.
Independent of your tool you could instantiate two inverter directly
from the vendor library and forbid your tool to systhesise these
libraryelments.
Or you could instantiate one inverter and manipulate the netlist after
synthesis, inserting a second.
But your tool will likely offer you one or two easier ways *g*.

And, are all these statements under the same process()?

They are all out of a process (concurrent statement), but there is no
problem
putting the lines in one process or split them over more processes.

bye Thomas