Description

ovega14

opened on Jul 7, 2022

Issue body actions

As it stands, a function that takes zeroes, poles, gain into numerator, denominator transfer function coefficient arrays only lives in scipy.signal or can be imported from control.matlab, and used to create a transfer function my_tf as follows:
my_tf= tf(*zpk2tf(...))
Why not instead can python-control have a method that directly takes zeroes, poles, gain into a transfer function, i.e.
my_tf=zpk2tf(..)?

Activity

1. bnavigator

   added

   enhancement

   on Jul 7, 2022
2. sawyerbfuller commented on Jul 16, 2022

   

   sawyerbfuller

   on Jul 16, 2022

   Contributor

   More actions

   Hi @ovega14 I agree it would be very nice to have this functionality outside of the matlab module.

   Its not clear to me how we could implement that functionality inside TransferFunction because the three-argument case is already taken by TransferFunction(num, den, dt). Might be possible to include a keyword to make it explicit like TransferFunction(zeros, poles, gain, [dt,] usezpk=True), but implementing that logic seems like it would be awkward.

   Probably most expedient is to just move the import of scipy's zpk2tf (and might as well bring in zpk2ss while you're at it) into the LTI module rather than the matlab module, and then they will be available everywhere, including the matlab module.

   This might be an opportunity to also invent a more pythonic additional long name for them as well, e.g. zeros_poles_gain_to_tf, but they are so apt and convenient I am not sure that is necessary.

   Feel free to contribute a pull request.

   Reacted by Ben Greiner
3. murrayrm commented on Jul 16, 2022

   

   murrayrm

   on Jul 16, 2022

   Member

   More actions

   Whoever takes on implementing this should make sure that zpk2tf and zpk2ss eventually create a NamedIOSystem, versus a lower level TransferFunction or StateSpace object. That will allow the output system to be compatible with the I/O system functionality (including named input/output signals for connecting things together). A diagram of the dependencies can be found here.

   Reacted by Ben Greiner
4. ovega14 commented on Jul 21, 2022

   

   ovega14

   on Jul 21, 2022

   Author

   More actions

   Should we import zpk2tf and zpk2ss from scipy into the LTI module? If so, then these functions will not directly return a NamedIOSystem, nor a TransferFunction or StateSpace object. As a reminder, these functions return numerator and denominator coefficient arrays of a transfer function given zeroes, poles, and gain. Should we then redefine the two functions (zpk2tf and zpk2ss) to conform to this expected python-control behavior?
5. bnavigator commented on Jul 21, 2022

   

   bnavigator

   on Jul 21, 2022

   Contributor

   More actions

   Yes functions in the control namespace should return a python-control conforming system. Hence @murrayrm's request to return NamedIOSystem. This requires a wrapper rather than a simple import.
6. murrayrm

   linked a pull request that will close this issueadd zpk() function #816on Dec 16, 2022
7. murrayrm

   closed this as completedin #816on Dec 17, 2022

Issue actions