Version 1.4 (R2006a) Fixed-Point Toolbox Software :: Fixed-Point Toolbox Release Notes (Fixed-Point Toolbox™)

Version 1.4 (R2006a) Fixed-Point Toolbox Software

This table summarizes what's new in Version 1.4 (R2006a):

New Features and Changes	Version Compatibility Considerations	Fixed Bugs and Known Problems	Related Documentation at Web Site
Yes Details below	Yes—Details labeled as Compatibility Considerations, below. See also Summary.	Bug Reports Includes fixes	No

New features and changes introduced in this version are

[Slope Bias] Math Support Added

Arithmetic using the +,-, .*, and * operators is now supported for objects with [Slope Bias] scaling. Refer to [Slope Bias] Arithmetic in the product documentation for more information.

In support of this feature, the following properties have been added to the fimath object:

ProductBias — Bias of the product data type
ProductFixedExponent — Fixed exponent of the product data type
ProductSlope — Slope of the product data type
ProductSlopeAdjustmentFactor — Slope adjustment factor of the product data type
SumBias — Bias of the sum data type
SumFixedExponent — Fixed exponent of the sum data type
SumSlope — Slope of the sum data type
SumSlopeAdjustmentFactor — Slope adjustment factor of the sum data type

Refer to Property Reference in the product reference documentation for more information.

Scaled Double Data Type Support Added to the fi Object

The fi object now supports the scaled double data type. The value ScaledDouble has been added to the DataType property of the numerictype object. The following values have also been added to the DataTypeMode property of the numerictype object:

Scaled double: binary point scaling
Scaled double: slope and bias scaling
Scaled double: unspecified scaling

Math operations are supported for fi objects with data type ScaledDouble.

Global DataTypeOverride Property Added to the fipref Object

The fipref object now has the property DataTypeOverride, which allows you to override fi objects with scaled doubles, singles, or doubles. Refer to Using fipref Objects to Set Data Type Override Preferences in the product documentation for more information.

Embedded MATLAB Subset Supports More Fixed-Point Toolbox Functions

The following Fixed-Point Toolbox functions are now supported by Embedded MATLAB subset:

Refer to Supported Functions and Limitations of the Fixed-Point Embedded MATLAB Subset in the product documentation for more information.

Embedded MATLAB Subset Does Not Support a CastBeforeSum Value of `'false'`

You can no longer set the fimath object property CastBeforeSum to false or 0 in Embedded MATLAB Function blocks. The reason for this restriction is that Embedded MATLAB subset does not produce the same numerical results as MATLAB when CastBeforeSum is false.

Compatibility Considerations

In the previous release, CastBeforeSum was set to false for default fimath objects in Embedded MATLAB subset If you have existing models that contain Embedded MATLAB Function blocks in which CastBeforeSum is false, you will now get an error when you compile or update your model. To correct this issue, you must set CastBeforeSum to true. To automate this process, you can run the utility slupdate either from the Model Advisor or by typing the following command at the MATLAB command line:

slupdate ('modelname')

where 'modelname' is the name of the model containing the Embedded MATLAB Function block that generates the error. slupdate prompts you to update this property by selecting one of these options:

Option	Action
Yes	Updates the first occurrence of `CastBeforeSum=false` in Embedded MATLAB Function blocks in the model and then prompts you for each subsequent instance found in the model.
No	Does not update any occurrences of `CastBeforeSum=false` in the model.
All	Updates all occurrences of `CastBeforeSum=false` in the model.

Note slupdate detects CastBeforeSum=false only in default fimath objects defined for Simulink signals in Embedded MATLAB Function blocks. If you modified the fimath object in an Embedded MATLAB Function block, update CastBeforeSum manually in your model and fix the errors as they are reported.

`'`round`'` Value Added to the fimath Object RoundMode Property

The RoundMode property value round has been added to the fimath object. The behavior of this rounding mode is identical to the MATLAB round function. For more information refer to RoundMode in the product documentation.

numerictype Object Syntax Change

Previously, if you created a numerictype object without specifying a value for the FractionLength property, the fraction length would be automatically set to 15. Now however, if you do not set the FractionLength property when creating a numerictype object, the scaling will remain unspecified. For example:

T = numerictype(1, 16)
 
T =
 

          DataTypeMode: Fixed-point: unspecified scaling
                Signed: true
            WordLength: 16

T.scaling

ans =

Unspecified

T.FractionLength

ans =

     0

Compatibility Considerations

Any instances of this syntax in your existing code will now return a different result.

Minimums and Maximums Now Logged After Quantization

Previously, the fi and quantizer objects logged minimums and maximums before quantization. They now log after quantization.

Compatibility Considerations

If your fixed-point data overflows and you want to log minimums and maximums for the full floating-point range, use the 'ScaledDoubles' or 'TrueDoubles' values of the fipref object DataTypeOverride property. For example, the following fixed-point variable overflows. The saturated minimum and maximum values are logged:

p = fipref;
p.LoggingMode = 'On';
p.DataTypeOverride = 'ForceOff';

a = fi(-2:2, true, 16, 15)
Warning: 3 overflows occurred in the fi assignment operation.
 
a =
 
          -1           -1            0      0.99997      0.99997

          DataTypeMode: Fixed-point: binary point scaling
                Signed: true
            WordLength: 16
        FractionLength: 15

             RoundMode: nearest
          OverflowMode: saturate
           ProductMode: FullPrecision
  MaxProductWordLength: 128
               SumMode: FullPrecision
      MaxSumWordLength: 128
         CastBeforeSum: true

logreport(a)

      minlog      maxlog   lowerbound   upperbound   noverflows   nunderflows
  a       -1   0.9999695           -1    0.9999695            3             0

Now set DataTypeOverride to 'ScaledDoubles'. Note that overflows are reported, but the data is not quantized. The minimum and maximum logs show the full possible range of the data without quantization:

p = fipref;
p.LoggingMode = 'On';
p.DataTypeOverride = 'ScaledDoubles';

b = fi(-2:2, true, 16, 15)
Warning: 3 overflows occurred in the fi assignment operation.
 
b =
 
    -2    -1     0     1     2


          DataTypeMode: Scaled double: binary point scaling
                Signed: true
            WordLength: 16
        FractionLength: 15

             RoundMode: nearest
          OverflowMode: saturate
           ProductMode: FullPrecision
  MaxProductWordLength: 128
               SumMode: FullPrecision
      MaxSumWordLength: 128
         CastBeforeSum: true

logreport(b)

      minlog   maxlog   lowerbound   upperbound   noverflows   nunderflows
  b       -2        2           -1    0.9999695            3             0

For an in-depth example of using logging and data type override to help set appropriate scalings for fixed-point quantities, see the Fixed-Point Toolbox demo "Fixed-Point Data Type Override, Min/Max Logging, and Scaling".