Main Content


Identify possible feedback data

Since R2023a


[fbck,fbck0,nudir] = checkFeedback(Data)
[___ = checkFeedback(Data,'InputName',inputname,'OutputName',outputname)


Data contains uniformly sampled input/output data in one of the following forms:

  • timetable in which the variables contain the input and output data.

  • Comma-separated pair of input/output numeric matrices u,y with dimensions Ns-by-Nu for u and Ns-by-Ny for y. Here, Ns is the number of samples, Nu is the number of inputs, and Ny is the number of outputs.

  • iddata set with Ny outputs and Nu inputs.

InputName and OutputName — Input and output channel names. Use these arguments especially when Data is a timetable that has more variables than you want to use, or when there. Also use OutputName to identify the output channels when the timetable data contains more than one output variable. Otherwise, the software interprets only the last variable as an output channel.

For example, specify the input and output signal variable names using fbck = checkFeedback(Data,'InputName',["u1","u3"],'OutputName',"y1").

fbck is an Ny-by-Nu matrix indicating the feedback. The ky,ku entry is a measure of feedback from output ky to input ku. The value is a probability P in percent. Its interpretation is that if the hypothesis that there is no feedback from output ky to input ku were tested at the level P, it would have been rejected. An intuitive but technically incorrect way of thinking about this is to see P as “the probability of feedback.” Often only values above 90% are taken as indications of feedback. When fbck is calculated, direct dependence at lag zero between u(t) and y(t) is not regarded as a feedback effect.

fbck0: Same as fbck, but direct dependence at lag 0 between u(t) and y(t) is viewed as feedback effect.

nudir: A vector containing those input numbers that appear to have a direct effect on some outputs, that is, no delay from input to output.

Version History

Introduced in R2023a

expand all

See Also