VerilogA issues -- transition function

I found some bugs in the code above. This is the revised code.

I think sometimes the derivates do not evaluate correctly, specifically:

(polarization_next-polarization_prev)/(time_next-time_prev)

I was hoping that there would be a bit of a lag between the polarization_next and polarization_prev when the polarization variable is changing but sometimes they are equal. Is there a better way to take a derivative? I was trying to use the ddt function but I think it only works for voltages and currents, not variables. I think part of the problem may be that the variables are updated too often and so sometimes can evaluate to the same value when the incremental change in transient time is small enough even though the polarization may be changing but not quickly enough to register a change.

For transient simulations for example the polarization_next and polarization prev variables may evaluate to be the same for times 1 s and 1.001 s but may be different at 1 s and 1.1 s. Is there a way to force verilogA to limit how often it evaluate the variables (i.e. limit the minimum resolution of the time step of the transient time)?

#################### revised code ######################################

/*

This file implements the extraction of csv data into a verilog variable and then implements a Preisach implemenatation of a FE capacitor SPICE
model based on the Preisach data.

*/

`include "constants.vams"
`include "disciplines.vams"

module test_func(p,n);

inout p,n;
electrical p,n;

//files for reading/writing
integer mcd, fp;

real data, v_present, v_step;

real preisach_density[0:1024];// all elements should add up to 1
real domain_states[0:1024];// all elements should be -1 or +1 (depending on polarity)
real polarization;//polarization (between -Pr and +Pr)

real polarization_prev, polarization_next;
real time_prev, time_next;
real v_prev, v_next;

integer row_offset, column_offset;

real current_calc;

integer i, j, k;//used in loops

integer rows, columns;//describes area scanned from applied voltage
integer rows_total, columns_total;//gives number of rows and columns in the grid ("rows" and "columns" should not exceed these values)

analog begin

@(initial_step) begin

//read data from csv file and transfer it to verilog-A array variable
//data stored in csv file should be preisach distribution
//csv file should have the same number of elements as arrays "preisach_density" and "domain_states"
mcd=$fopen("whole_preisach_dist_5_1_2017.csv","r");

while (!$feof(mcd)) begin

$fscanf(mcd,"%f", preisach_density[i]);
domain_states[i] = 0;

i = i+1;
end

v_step = 0.1; // resolution

rows_total = 3.2/v_step;//the maximum anticipated applied input voltage swing divided by the resolution of the grid cells
columns_total = 3.2/v_step;//the maximum anticipated applied input voltage swing divided by the resolution of the grid cells

v_next = V(p,n);

time_prev = 0;
time_next = 0;

// centers the indiciess to the beginning of the fourth quadrant of the Preisach distribution (i.e. half of row_total or columns total plus 1)
row_offset = 17;
column_offset = 17;

end



if (v_next < v_prev) begin // negative voltage derivative

rows = row_offset + V(p,n)/(-v_step);

begin
if(rows > rows_total)
rows = rows_total;
else if(rows < 0)
rows = 0;
end

//for all columns
for(j = 0; j < columns_total; j=j+1) begin

//for all rows with a smaller value than "rows"
for(k=0; k<rows; k=k+1) begin

domain_states[k*columns_total + j] = -1;
end
end
end

if (v_next > v_prev) begin // positive voltage derivative

columns = column_offset + V(p,n)/(v_step);

begin
if(columns > columns_total)
columns = columns_total;
else if(columns < 0)
columns = 0;
end

for (j=0; j<columns; j=j+1) begin

for(k=0; k < rows_total; k=k+1) begin

domain_states[k*columns_total + j] = 1;

end

end
end

//calculate the present polarization across the device
polarization = 0;
for (i=0; i<columns_total*rows_total; i = i+1) begin

polarization = polarization + preisach_density[i]*domain_states[i];
end

//used for calculating derivatives
polarization_prev = polarization_next;
polarization_next = polarization;
v_prev = v_next;
v_next = V(p,n);
time_prev = time_next;
time_next = $abstime;

begin
if (time_next == 0)

current_calc = 0;

else
current_calc = (polarization_next - polarization_prev)/(time_next - time_prev)*0.000000001;//0.000000001 is arbitrary

end

I(p,n) <+ current_calc;

end

endmodule

I found some bugs in the code above. This is the revised code.

I think sometimes the derivates do not evaluate correctly, specifically:

(polarization_next-polarization_prev)/(time_next-time_prev)

I was hoping that there would be a bit of a lag between the polarization_next and polarization_prev when the polarization variable is changing but sometimes they are equal. Is there a better way to take a derivative? I was trying to use the ddt function but I think it only works for voltages and currents, not variables. I think part of the problem may be that the variables are updated too often and so sometimes can evaluate to the same value when the incremental change in transient time is small enough even though the polarization may be changing but not quickly enough to register a change.

For transient simulations for example the polarization_next and polarization prev variables may evaluate to be the same for times 1 s and 1.001 s but may be different at 1 s and 1.1 s. Is there a way to force verilogA to limit how often it evaluate the variables (i.e. limit the minimum resolution of the time step of the transient time)?

#################### revised code ######################################

/*

This file implements the extraction of csv data into a verilog variable and then implements a Preisach implemenatation of a FE capacitor SPICE
model based on the Preisach data.

*/

`include "constants.vams"
`include "disciplines.vams"

module test_func(p,n);

inout p,n;
electrical p,n;

//files for reading/writing
integer mcd, fp;

real data, v_present, v_step;

real preisach_density[0:1024];// all elements should add up to 1
real domain_states[0:1024];// all elements should be -1 or +1 (depending on polarity)
real polarization;//polarization (between -Pr and +Pr)

real polarization_prev, polarization_next;
real time_prev, time_next;
real v_prev, v_next;

integer row_offset, column_offset;

real current_calc;

integer i, j, k;//used in loops

integer rows, columns;//describes area scanned from applied voltage
integer rows_total, columns_total;//gives number of rows and columns in the grid ("rows" and "columns" should not exceed these values)

analog begin

@(initial_step) begin

//read data from csv file and transfer it to verilog-A array variable
//data stored in csv file should be preisach distribution
//csv file should have the same number of elements as arrays "preisach_density" and "domain_states"
mcd=$fopen("whole_preisach_dist_5_1_2017.csv","r");

while (!$feof(mcd)) begin

$fscanf(mcd,"%f", preisach_density[i]);
domain_states[i] = 0;

i = i+1;
end

v_step = 0.1; // resolution

rows_total = 3.2/v_step;//the maximum anticipated applied input voltage swing divided by the resolution of the grid cells
columns_total = 3.2/v_step;//the maximum anticipated applied input voltage swing divided by the resolution of the grid cells

v_next = V(p,n);

time_prev = 0;
time_next = 0;

// centers the indiciess to the beginning of the fourth quadrant of the Preisach distribution (i.e. half of row_total or columns total plus 1)
row_offset = 17;
column_offset = 17;

end



if (v_next < v_prev) begin // negative voltage derivative

rows = row_offset + V(p,n)/(-v_step);

begin
if(rows > rows_total)
rows = rows_total;
else if(rows < 0)
rows = 0;
end

//for all columns
for(j = 0; j < columns_total; j=j+1) begin

//for all rows with a smaller value than "rows"
for(k=0; k<rows; k=k+1) begin

domain_states[k*columns_total + j] = -1;
end
end
end

if (v_next > v_prev) begin // positive voltage derivative

columns = column_offset + V(p,n)/(v_step);

begin
if(columns > columns_total)
columns = columns_total;
else if(columns < 0)
columns = 0;
end

for (j=0; j<columns; j=j+1) begin

for(k=0; k < rows_total; k=k+1) begin

domain_states[k*columns_total + j] = 1;

end

end
end

//calculate the present polarization across the device
polarization = 0;
for (i=0; i<columns_total*rows_total; i = i+1) begin

polarization = polarization + preisach_density[i]*domain_states[i];
end

//used for calculating derivatives
polarization_prev = polarization_next;
polarization_next = polarization;
v_prev = v_next;
v_next = V(p,n);
time_prev = time_next;
time_next = $abstime;

begin
if (time_next == 0)

current_calc = 0;

else
current_calc = (polarization_next - polarization_prev)/(time_next - time_prev)*0.000000001;//0.000000001 is arbitrary

end

I(p,n) <+ current_calc;

end

endmodule