WOLFRAM SYSTEM MODELER

UpSample

Upsample the clocked Real input signal and provide it as clocked output signal

Wolfram Language

In[1]:=
SystemModel["Modelica.Clocked.RealSignals.Sampler.Utilities.UpSample"]
Out[1]:=

Information

This information is part of the Modelica Standard Library maintained by the Modelica Association.

This block upsamples the clocked Real input signal u and provides it as clocked output signal y.

To be more precise: The clock of y is factor-times faster than the clock of u. At every tick of the clock of u, the value of y is set to the value of u, at intermediate ticks of clock y, the value of y is set to zero. The first activation of the clock of y coincides with the first activation of the clock of u. By default, the upsampling factor is inferred,that is, it must be defined somewhere else. If parameter inferFactor = false, then the upsampling factor is defined by the Integer parameter factor.

For signal interpolation an upsample should be followed by an appropriate filter (anti-imaging).

Examples

The following example samples a sine signal with a periodic clock of 20 ms period, and then upsamples the resulting clocked signal with a factor of 3:

UpSample1_Model.png     UpSample1_Result.png
model simulation result

The upsampling factor is explicitly given at the upSample2 block, which also makes the factor visible in the icon. For the upSample1 block the factor is inferred. As can be seen, upsample1 introduces 3 additional clock ticks for the output y. Note, the up-arrow in the icon of the upSample blocks indicates that the clocks of their outputs is faster than the clocks of their inputs.

The following example samples a sine signal with a periodic clock of 20 ms period, upsamples the resulting clocked signal with a factor of 3 and applies varies filters on this signal:

UpSample2_Model.png
model
UpSample2_Result1.png
simulation result 1
UpSample2_Result2.png
simulation result 2

This model shows various possibilities to operate on an up-sampled signal: When filtering the upsampled signal with filter block FIR1 using FIR coefficients {1,1,1}, then the result is identical to a super-sampled signal (see signal FIR1.y in figure "simulation result 1"). On the other hand, when filtering the upsampled signal with filter block FIR2 using FIR coefficients {1/3, 2/3, 1, 2/3, 1/3}, then the result is a linearly interpolated super-sampled signal (see signal FIR2.y in figure "simulation result 2"). The same result can be achieved with block SuperSampleInterpolated (see signal superSampleIpo1.y in figure "simulation result 2"). The only difference are the first clock ticks, since the FIR2 signal is initialized a bit differently.

Parameters (2)

inferFactor

Value: true

Type: Boolean

Description: = true, if upsampling factor is inferred

factor

Value: 1

Type: Integer

Description: Upsampling factor >= 1 (if inferFactor=false)

Connectors (2)

u

Type: RealInput

Description: Connector of clocked, Real input signal

y

Type: RealOutput

Description: Connector of clocked, Real output signal (clock of y is faster as clock of u)

Used in Examples (2)

UpSample1

Modelica.Clocked.Examples.Elementary.RealSignals

Example of an UpSample block for Real signals

UpSample2

Modelica.Clocked.Examples.Elementary.RealSignals

Example of an UpSample block for Real signals combined with FIR filter blocks