`predplot` of identified state-space system emits a warning about transfer functions

Hi,

It may be not a bug but using predplot on an identified state-space system with a large enough prediction horizon h results in the following warning (h=30 with a state space system of nx=6 states):

┌ Warning: High-order transfer functions are highly sensitive to numerical errors. The result may be inaccurate. Consider making use of statespace systems instead
└ @ ControlSystemsBase ~/.julia/packages/ControlSystemsBase/hoyMe/src/types/SisoTfTypes/SisoRational.jl:12

I'm trying to understand why the state-space system (identified using subspaceid) is first converted into a transfer function. I'm also not sure why high h values leads to high order transfer functions even if the order of the original system was quite low.

Thanks for your help.

[baggepinnen]

I'm trying to understand why the state-space system (identified using subspaceid) is first converted into a transfer function.

It is not. The function observer_predictor forms a predictor system using transfer functions, which is then converted to statespace form and multiplied by the identified system. This happens to be okay since no operations are actually performed on the transfer function, but the warning does not know that

Alright thanks for the info!

Maybe it could be a good idea to supress this warning here? Or maybe emit and @info in such case to ignore the warning?

I will probably not be the only one to validate with a prediction horizon larger than 20...