Quartus® Prime Pro Edition User Guide: Design Recommendations

ID 683082
Date 7/08/2024
Public
Document Table of Contents

2.3.1.2. Using Asynchronous Resets

Asynchronous resets are the most common form of reset in circuit designs, as well as the easiest to implement. Typically, you can insert the asynchronous reset into the device, turn on the global buffer, and connect to the asynchronous reset pin of every register in the device.

This method is only advantageous under certain circumstances—you do not need to always reset the register. Unlike the synchronous reset, the asynchronous reset is not inserted in the datapath, and does not negatively impact the data arrival times between registers. Reset takes effect immediately, and as soon as the registers receive the reset pulse, the registers are reset. The asynchronous reset is not dependent on the clock.

However, when the reset is deasserted and does not pass the recovery (µtSU) or removal (µtH) time check (the Timing Analyzer recovery and removal analysis checks both times), the edge is said to have fallen into the metastability zone. Additional time is required to determine the correct state, and the delay can cause the setup time to fail to register downstream, leading to system failure. To avoid this, add a few follower registers after the register with the asynchronous reset and use the output of these registers in the design. Use the follower registers to synchronize the data to the clock to remove the metastability issues. You should place these registers close to each other in the device to keep the routing delays to a minimum, which decreases data arrival times and increases MTBF. Ensure that these follower registers themselves are not reset, but are initialized over a period of several clock cycles by “flushing out” their current or initial state.

Figure 21. Asynchronous Reset with Follower Registers

The following example shows the equivalent Verilog HDL code. The active edge of the reset is now in the sensitivity list for the procedural block, which infers a clock enable on the follower registers with the inverse of the reset signal tied to the clock enable. The follower registers should be in a separate procedural block as shown using non-blocking assignments.

Verilog HDL Code of Asynchronous Reset with Follower Registers

module async_reset (
        input   clock,
        input   reset_n,
        input   data_a,
        output   out_a,
        );
reg     reg1, reg2, reg3;
assign  out_a  = reg3;
always @ (posedge clock, negedge reset_n)
begin
    if (!reset_n)
        reg1    <= 1’b0;
    else
        reg1    <= data_a;
end
always @ (posedge clock)
begin
    reg2    <= reg1;
    reg3    <= reg2;
end
endmodule  //  async_reset

You can easily constrain an asynchronous reset. By definition, asynchronous resets have a non-deterministic relationship to the clock domains of the registers they are resetting. Therefore, static timing analysis of these resets is not possible and you can use the set_false_path command to exclude the path from timing analysis. Because the relationship of the reset to the clock at the register is not known, you cannot run recovery and removal analysis in the Timing Analyzer for this path. Attempting to do so even without the false path statement results in no paths reported for recovery and removal.

SDC Constraints for Asynchronous Reset

# Input clock - 100 MHz
create_clock [get_ports {clock}] \
        -name {clock} \
        -period 10.0 \
        -waveform {0.0 5.0}
# Input constraints on data
set_input_delay 7.0 \
        -max \
        -clock [get_clocks {clock}]\
        [get_ports {data_a}]
set_input_delay 1.0 \
        -min \
        -clock [get_clocks {clock}] \
        [get_ports {data_a}]
# Cut the asynchronous reset input
set_false_path \
        -from [get_ports {reset_n}] \
        -to [all_registers]

The asynchronous reset is susceptible to noise, and a noisy asynchronous reset can cause a spurious reset. You must ensure that the asynchronous reset is debounced and filtered. You can easily enter into a reset asynchronously, but releasing a reset asynchronously can lead to potential problems (also referred to as “reset removal”) with metastability, including the hazards of unwanted situations with synchronous circuits involving feedback.