Is the comm.Raise​dCosineTra​nsmitFilte​r stateful?

When I reuse an instance of comm.RaisedCosineTransmitFilter multiple times with the same vector, I get several different results until it stabilizes.

If I use a new instance each time, the result is the same. So it looks like the filter object is stateful, which I suppose makes sense, but I guess I expected the impulse definition to be essentially just a declaration of coefficients, not necessarily the state variable.

The first few times you run the below lines with the 4-long impulse vector, you get a different plot, and on ones with a 3-long impulse vector, the first time you run it, the plot is 8 samples shorter than the previous one, as I would expect, but subsequently running it several more times, the plot changes shape.

K>> rctFilt=comm.RaisedCosineTransmitFilter()

rctFilt = 

comm.RaisedCosineTransmitFilter with properties:

                     Shape: 'Square root'
             RolloffFactor: 0.2000
       FilterSpanInSymbols: 10
    OutputSamplesPerSymbol: 8
                      Gain: 1

K>> clf;plot(real(step(rctFilt, [1 0 0 0]')));xlim([0 32])
K>> clf;plot(real(step(rctFilt, [1 0 0 0]')));xlim([0 32])
K>> clf;plot(real(step(rctFilt, [1 0 0 0]')));xlim([0 32])
K>> clf;plot(real(step(rctFilt, [1 0 0 0]')));xlim([0 32])
K>> clf;plot(real(step(rctFilt, [1 0 0 ]')));xlim([0 32])
K>> clf;plot(real(step(rctFilt, [1 0 0 ]')));xlim([0 32])
K>> clf;plot(real(step(rctFilt, [1 0 0 ]')));xlim([0 32])

I'm just trying to figure out how the transmit filter differs from the receive filter.

[Mason Huynh]

I see exactly the same issue, which seems unreasonable as Raised Cosine Transmit Filter is a filter, and the output of each instance should not depend on previous instances.