fixed point fft butterfly stage testing help

I am new to verilog/hardware arithmetic and seeking good advice on testing a fft butterfly stage that I have written.

module butterfly #(parameter size=16, Q=4)
(input signed[size-1:0] a_real,
input signed[size-1:0] a_imag,
input signed[size-1:0] b_real,
input signed[size-1:0] b_imag,
input signed[size-1:0] w_real,
input signed[size-1:0] w_imag,
output reg signed[size-1:0] x_real,
output reg signed[size-1:0] x_imag,
output reg signed[size-1:0] y_real,
output reg signed[size-1:0] y_imag
);



reg signed[size-1:0] w_b_real;
reg signed[size-1:0] w_b_imag;

always @(a_real or a_imag or b_real or b_imag or w_real or w_imag) begin

w_b_real = b_real*w_real - b_imag*w_imag;
w_b_imag = b_real*w_imag - b_imag*w_real;

w_b_real = w_b_real >> Q;
w_b_imag = w_b_imag >> Q;

x_real = a_real + w_b_real;
x_imag = a_imag + w_b_imag;
y_real = a_real - w_b_real;
y_imag = a_imag - w_b_imag;

end

endmodule

When writing the test bench, I came across several problems which I have no prior experience.
1. How to create fixed point signed numbers that can be fed as input which will not create overflows inside butterfly module
2. How to create the expected results to compare

Greatly appreciate any help on these regard.