Extracting a subset of a binary number to form a new binary number

Hi, I am new using fixed point designer. I'm trying to pick out a subset of a fixed point number

This is what I need to do verbally:

"a" is '010' and I would like to form a new fixed number, "b", by taking the upper two bits of "a", ie. '01',

a=fi(2,1,3,0); b=fi(0,1,2,0);

b= uppder two bits of a

Just wondering what the best way is to do this. Do I use the quantize command, or do I need to do bitshift? Is there a quick way to pick out the first two bits of "a"?

Thank you very much.

Answer 1

[Andy Bartlett]

Bin fine for interpretted mode, but not code gen or Simulink

If you only need to perform the operation in MATLAB interpretted mode, then using bin as Walter showed is fine.

If you need to have support for code generation, or to use this in a Simulink context such as with MATLAB Function block or MATLAB System block, then bin will not work. You'd see an error like this.

??? Property 'bin' is not supported in code generation.

Cast via fi supports code generation

If you wish to extract bits in code generation or in a Simulink context, then casting via fi is a good approach. A discussion of this approach can be found here.

If you are operating directly on a Simulink signal, then the Extract Bits block is available.

Example function extractMSBits

I have attached an example function, extractMSBits, for extracting the most significant n bits from an input using fi casting.

function y = extractMSBits(u,nBitsKeep) %#codegen
    %extractMSBits extract n most significant bits of the input
    %
    %Usage
    %  y = extractMSBits(u,nBitsKeep)
    %
    %  u is a fixed-point or integer variable
    %  nBitsKeep is the number of MSBits to return
    %
    %  The least significant bits are simply discarded. Discarding
    %  is mathematically equivalent to rounding toward floor.
    
    % Copyright 1994-2020 The MathWorks, Inc.
    
    % Change built-in integer to fi equivalent
    %
    u2 = castIntToFi(u);
    
    assert(isfi(u2))
    assert(isfixed(u2))
    
    ntu = numerictype(u2);
    
    isSigned = ntu.SignednessBool;
    wordLength = ntu.WordLength;
        
    assert( nBitsKeep <= wordLength );
    assert( isSigned < nBitsKeep );
    assert( nBitsKeep == floor(nBitsKeep) );
    
    nBitsDrop = wordLength - nBitsKeep;
    fixedExponent = ntu.FixedExponent + nBitsDrop;
    
    if ntu.SlopeAdjustmentFactor ~= 1 || ntu.Bias ~= 0
        nty = numerictype( ...
            isSigned, ...
            nBitsKeep, ...
            ntu.SlopeAdjustmentFactor, ...
            fixedExponent, ...
            ntu.Bias);
    else
        nty = numerictype( ...
            ntu.SignednessBool, ...
            nBitsKeep, ...
            -fixedExponent);
    end
    
    fmWrapFloor = fimath('RoundingMethod', 'Floor', ...
        'OverflowAction', 'Wrap');
    
    y2 = fi( u2, nty, fmWrapFloor );
    
    y = removefimath( y2 );
end

The attached script, example_extractMSBits, will exercise the function extractMSBits using random choices of input data type and input value. The following is the random output for one call.

>> example_extractMSBits

Term                     bin_SI          type                   RealWorldValue
a                        1111100100000   numerictype(0,13,11)   3.890625
extractMSBits( a, 13)    1111100100000   numerictype(0,13,11)   3.890625
extractMSBits( a, 12)    111110010000    numerictype(0,12,10)   3.890625
extractMSBits( a, 11)    11111001000     numerictype(0,11,9)    3.890625
extractMSBits( a, 10)    1111100100      numerictype(0,10,8)    3.890625
extractMSBits( a,  9)    111110010       numerictype(0,9,7)     3.890625
extractMSBits( a,  8)    11111001        numerictype(0,8,6)     3.890625
extractMSBits( a,  7)    1111100         numerictype(0,7,5)     3.875
extractMSBits( a,  6)    111110          numerictype(0,6,4)     3.875
extractMSBits( a,  5)    11111           numerictype(0,5,3)     3.875
extractMSBits( a,  4)    1111            numerictype(0,4,2)     3.75
extractMSBits( a,  3)    111             numerictype(0,3,1)     3.5
extractMSBits( a,  2)    11              numerictype(0,2,0)     3
extractMSBits( a,  1)    1               numerictype(0,1,-1)    2

[Walter Roberson]

Good point about code generation -- I did not take that into account at all.

Answer 2

[Walter Roberson]

a=fi(2,1,3,0)

a = 
     2

          DataTypeMode: Fixed-point: binary point scaling
            Signedness: Signed
            WordLength: 3
        FractionLength: 0

b=fi(0,1,2,0)

b = 
     0

          DataTypeMode: Fixed-point: binary point scaling
            Signedness: Signed
            WordLength: 2
        FractionLength: 0

b.bin = a.bin(1:2)

b = 
     1

          DataTypeMode: Fixed-point: binary point scaling
            Signedness: Signed
            WordLength: 2
        FractionLength: 0

[scc28x]

Ah, I see. Use bin() to access the bits. Thx.