Main Content

Compare Numerical Response of Sum Block and Sum in MATLAB® Function Block

Open Script

This example shows how to generate simulation inputs and use them to exercise a model over its full operating range. In this example, generate test data to simulate a model and compare the numerical response of the Sum block, and sum implemented in a MATLAB® Function block in theex_testsummodel.

Open the model.

model ='ex_testsum'; open_system(model); set_param(model,'SimulationCommand','update');

Specify Data Attributes and Generate Data

Use thefixed.DataSpecificationobject to specify input data attributes. In this example, create twoDataSpecificationobjects, one with a double-precision data type, and the other using a single-precision data type. The interval of values generated by the first object is from 1 to 64, and the interval of values generated by the second is from 1 to 32.

dataspec1 = fixed.DataSpecification('double','Intervals', {1 64}); dataspec2 = fixed.DataSpecification('single','Intervals', {1 32});

TheDataGenerator对象生成numericall的组合y-rich values. To use the output data in a Simulink® model, set the format of the output to'timeseries'.

datagen = fixed.DataGenerator; datagen.DataSpecifications = {dataspec1, dataspec2}; [tsdata1, tsdata2] = outputAllData(datagen,'timeseries');

Set Up Model and Simulate

Apply the attributes of theDataSpecificationobjects to the Inport blocks in the model.

applyOnRootInport(datagen.DataSpecifications{1}, model, 1); applyOnRootInport(datagen.DataSpecifications{2}, model, 2);

Load the generated timeseries data into the model and simulate.

set_param(model,'LoadExternalInput','on',...'ExternalInput','tsdata1, tsdata2',...'StopTime', string(tsdata1.Time(end))); simout = sim(model);

Visualize Output

Visualize the output of the simulation, and compare the numerical behavior of the two implementations of the sum operation.

% Get the unique values in the generated data for each set of data.[x, y] = datagen.getUniqueValues; d = abs(simout.yout{1}.Values.Data - simout.yout{2}.Values.Data); X = reshape(tsdata1.Data, numel(x), []); Y = reshape(tsdata2.Data, numel(x), []); D = reshape(d, numel(x), []); figure;% Plot the difference between outputs as a function of the input values.surf(X, Y, D,'EdgeColor','none'); gridon; view(2); axistight; xlabel('In1'); ylabel('In2'); colorbar; title('abs(MATLAB Function block output - Sum block output)');

从情节,你可以看到that the difference between the two implementations increases as the values of the numeric inputs get larger. This difference is due to the difference in the data type of the accumulator in the two implementations.

Compare Numerical Response with Single-Precision Accumulator

TheAccumulator Data Typeparameter of the Sum block is set toInherit: Inherit via internal rule. In this case, the data type used for the accumulator is a double-precision floating-point type. Set theAccumulator data typetosingleand compare the output again.

set_param([model,'/Sum'],'AccumDataTypeStr','single') simout = sim(model);

Visualize the output. When the accumulator type of the Sum block is set tosingle, the implementations return the same result at all values.

[x, y] = datagen.getUniqueValues; d = abs(simout.yout{1}.Values.Data - simout.yout{2}.Values.Data); X = reshape(tsdata1.Data, numel(x), []); Y = reshape(tsdata2.Data, numel(x), []); D = reshape(d, numel(x), []); figure; surf(X, Y, D,'EdgeColor','none'); gridon; view(2); axistight; xlabel('In1'); ylabel('In2'); colorbar; title('abs(MATLAB Function block output - Sum block output)');