text to mif conversion

[Altera_Forum]

Hello, 

I am looking for something that would convert a text file (with fractional entries), with values to be stored in a LUT, into a mif file with hexadecimal entries.  For example, let a fractional value be 0.123456 For a 16-bit result, taking dec2hex( 0.123456* 2^14 ) would give out the hex conversion of decimal.  Overall, positive fractional values get converted into range 0 to 2^14, and 2^16 is added to the negative fractional values to get them in range 2^14 to 2^15.  If I now have both decimal AND fractional in a value, for example -15.123456 and I need to convert it to 16-bit-binary/4-bit-hexadecimal, how would I convert these to hex and maintain the range in which positive and negative values land?  Thank you in advance for your help.

[Altera_Forum]

I assume you use Matlab, here is short cut: 

 convert your data to signed integers: data = round(data *(2^15-1)/max(data)); %for 16 bits mif  then display as column: data'  copy and paste the column to a mif in quartus as follows: open a new mif with 16 bits width. change data view to signed then copy the data directly, save and enjoy...

[Altera_Forum]

Wow, this is great! Thanks a bunch Kaz.

[Altera_Forum]

I want to find a solution for converting images to .hex / .mif format. Can i change it by using the converter (please suggest to me) or do I need to make their own code by matlab or another software.? 

thank you in advance for any help from members..

[Altera_Forum]

Hi, 

 You better use a program to write the mif files as copy/paste is suitable for small size data.  Here is an example  Using Matlab to write 256 signed data to 16 bit mif:   fid = fopen('filename.mif','w'); fprintf(fid,'--MIF data generated by MATLAB\n'); fprintf(fid,'--Date: %s \n\n', date); fprintf(fid,'WIDTH=16;\n'); fprintf(fid,'DEPTH=256;\n'); fprintf(fid,'ADDRESS_RADIX=UNS;\n'); fprintf(fid,'DATA_RADIX=DEC;\n'); fprintf(fid,'CONTENT BEGIN\n'); for k = 1:256 fprintf(fid,'%i : %i;\n',k-1,data(k)); end fprintf(fid,'END;'); fclose(fid);     *****************

[Altera_Forum]

thanks kaz...this is really help me...

[Altera_Forum]

i have this full program,its work.but then when i want to change the spesific size,it have the error. 

----------------- name = 'image'; img = imread(strcat(name,'.gif')); g = im2double(img); g = g/max(max(g)); rgb = cat(3,g,g,g); imshow(rgb); out = ones(480,640); out(10+(1:460),90+(1:460)) = img; imshow(out);  x = out'; x = uint8(x(:)); n = length(x); y = reshape(x,8,n/8)'; z = y(:,8); for i=1:7  z = bitor(z,bitshift(y(:,i),8-i)); end  m = 0; n = 1; v = z(1); for i=2:length(z)  if (z(i)~=v)  m = m + 1;  value(m) = v;  runlen(m) = n;  n = 1;  v = z(i);  else  n = n + 1;  end end m = m + 1; value(m) = v; runlen(m) = n;  fprintf('total runs %d\n',sum(runlen)); fprintf('number of runs %d\n',length(runlen)); nz = length(find(value==0)); nf = length(find(value==255)); fprintf('zeros %d all ones %d\n',nz,nf);  dfv = 255; fid = fopen(strcat(name,'.mif'),'w'); str = 'WIDTH=8;\nDEPTH=38400;\n\nADDRESS_RADIX=HEX;\nDATA_RADIX=HEX;\n\n'; fprintf(fid,str); str = 'CONTENT BEGIN\n [0000..%04X] : %X;\n'; fprintf(fid,str,sum(runlen)-1,dfv); n = 0; for k=1:length(runlen)  if (runlen(k)==1)  str = sprintf(' %04X : %X;\n', n, value(k));  else  str = sprintf(' [%04X..%04X] : %X;\n', n, n+runlen(k)-1, value(k));  end  if (value(k) ~= dfv)  fprintf(fid,str);  end  n = n + runlen(k); end fprintf(fid,'END;\n'); fclose(fid); -------------- can you show me how to make the image 16bits width and 4096bits depth. which part should i alter? thank you.

[Altera_Forum]

Hi Zainab 

 If I understood you you want 16bits width and 4096 depth  fid = fopen('filename.mif','w'); fprintf(fid,'--MIF data generated by MATLAB\n'); fprintf(fid,'--Date: %s \n\n', date); fprintf(fid,'WIDTH=16;\n'); fprintf(fid,'DEPTH=4096;\n'); fprintf(fid,'ADDRESS_RADIX=UNS;\n'); fprintf(fid,'DATA_RADIX=DEC;\n'); fprintf(fid,'CONTENT BEGIN\n'); for k = 1:4096 fprintf(fid,'%i : %i;\n',k-1,data(k)); end fprintf(fid,'END;'); fclose(fid);

[Altera_Forum]

i'm sorry,kaz. 

 again, i try your suggestion.but then, i have this error.  Error in ==> pingu2 at 10 fprintf(fid,'%i : %i;\n',k-1,data(k));  what is 'data' means here?is it the name of file? thank you.

[Altera_Forum]

Hi Zainab, 

 data is the vector(array) in Matlab that contains your data to be written into named mif file. It is meant to be signed integer (no fractions) in the code (since data radix is made decimal).  if your data is not scaled then the following statement creates signed integers scaled to 10 bits signed: data = randn(1,2048); %unscaled floating data = round((2^9-1) * data/max(abs(data)));  in this case your array length is 2048 to replace "k"

[Altera_Forum]

Hi, 

 I use the Fixed-Point toolbox that is embedded in Matlab to convert data to hex format. With this you can select rounding modes(ceil, floor, fix, convergent, nearest, round), define overflow behavior (wrap, saturate), word length, fraction length, signedness, and data types (fixed, the built-in single and double). Moreover, you can display your object in the equivalent hex, bin, oct and int of the stored bits. I understood that you can build objects with up to 65536-bit length.  function generate_mif(filename, data, depth, width, frac, sign) % data is a vector fid = fopen(filename, 'w+'); fprintf(fid, 'WIDTH=%d;\n', width); fprintf(fid, 'DEPTH=%d;\n\n', depth); fprintf(fid, 'ADDRESS_RADIX=UNS;\n'); % UNS, BIN, HEX, ... fprintf(fid, 'DATA_RADIX=HEX;\n\n'); fprintf(fid, 'CONTENT BEGIN\n'); fi_data = fi(data, sign, width, frac, 'RoundMode', 'nearest'); for i = 1:depth fi_temp = fi_data(i); fprintf(fid, '\t%-4d: %s;\n', i-1, fi_temp.hex); end fprintf(fid, 'END;\n'); fclose(fid);  Test the function with this code:  filename = 'sin96.mif'; depth = 96; width = 36; frac = width-1; % range: [-1,+1) signed = true; data = sin(2*pi*0.05*(1:depth)); generate_mif(filename, data, depth, width, frac, signed); plot(data); grid on;  Explore this toolbox. It's very useful. :)