I have pushed some commits to the dev-branch of the control package repository at https://github.com/gnu-octave/pkg-control which allows to enter imaginary coefficients when defining a transfer function.

However, especially when entering a transfer function as a combination of smaller subsystems, there is no possibility to distinguish between intermediate and final forms of the transfer systems. Therefore, also the final result is now allowed to have imaginary coefficients. Since many methods do not work correctly for imaginary coefficients, a warning is displayed whenever an imaginary coefficient is detected.

I think we could try
isnumeric
instead of
isreal

It is the sane in octave 8.4, 9.2.1 and 10.0.0 (control 4.0.1)

octave:3> pkg load control
octave:4> s = tf('s');
octave:5> FT = (s+4)/((s+4j)*(s-4j))
error: tf: numerator 'num' must be a real-valued, non-empty vector or a cell of such vectors
error: called from
    __adjust_tf_data__ at line 75 column 5
    tf at line 223 column 27
    __numeric_to_lti__ at line 36 column 16
    __sys_group__ at line 30 column 16
    plus at line 45 column 7
octave:6> ver
----------------------------------------------------------------------
GNU Octave Version: 8.4.0 (hg id: 78c13a2594f3)
GNU Octave License: GNU General Public License
Operating System: Linux 5.14.0-522.el9.x86_64 #1 SMP PREEMPT_DYNAMIC Sun Oct 20 13:04:34 UTC 2024 x86_64
----------------------------------------------------------------------
Package Name  | Version | Installation directory
--------------+---------+-----------------------
     control *|   4.0.1 | /home/dima/.local/share/octave/api-v58/packages/control-4.0.1

Dmitri.
--

The error also shows up in control 3.6.0.

Do you know with which version of the control package this kind of command (including complex coefficients) were running for you?

In  _adjust_tf_data at line 33 we have
  elseif (is_real_matrix (num))

but the value is a complex number so it returns false
and we need a true.

I don't see a is_double_matrix or an is_complex_matrix
so how should we fix this.

Hi. The required outputs are:

>> which tf

'tf' is a function from the file /home/luiz/.local/share/octave/api-v59/packages/control-4.0.1/@tf/tf.m

>> whos

Variables visible from the current scope:

variables in scope: top scope

  Attr   Name        Size                     Bytes  Class
  ====   ====        ====                     =====  =====
         ans         1x1                          8  double
         s           1x1                         34  tf

Total is 2 elements using 42 bytes

Hi,

I'm used to declare transfer functions following the below statement:

s = tf('s');

FT = (s+4)/((s+4j)*(s-4j))

However in the octave 9.2 I am getting the error messsage:

error: tf: numerator 'num' must be a real-valued, non-empty vector or a cell of such vectors
error: called from
    _adjust_tf_data_ at line 75 column 5
    tf at line 223 column 27
    _numeric_to_lti_ at line 36 column 16
    _sys_group_ at line 30 column 16
    plus at line 45 column 7
    teste at line 3 column 4

On previous versions these type of statements were working well.

I am using Octave 9.2 and the problem is appearing in Linux Mint 22 (Ubuntu 24.04) as well in windows 11.

The control toolbox version is 4.0.1


Just to compare, running the same code in Matlab I got the expected output

FT =

   s + 4
  --------
  s^2 + 16

Continuous-time transfer function.
Model Properties

Best Regards.