-
Notifications
You must be signed in to change notification settings - Fork 85
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
more balanced state-space realization when converting from tf #552
Comments
It is poorly conditioned, but it has exactly the same information as the transfer function. How does balancing influence future computations? Any operation performed on the system will lose accuracy, but perhaps it's useful if subsequent computations are improved by balancing. |
For a lot of computations, related to, e.g., eigenvalues/matrix exponentials there is a preliminary balancing step, so in many cases the poor conditioning isn't noticeable. I think the main thing is that one wants the states to have roughly the same nominal magnitude, for balanced truncation, etc. But this is also for convenient plotting and when one wants to inspect the matrices. The following is along the lines of what I think should be done. The implementation of
Compare for example
In the case below, the poorly balanced version gives a LAPACK error
|
How does the |
I wasn't aware of I have a feeling that there might be a way to exploit the companion form structure to get a very cheap implementation of the balancing, but |
Fixed, balancing is now on by default when converting tf -> ss
|
This gives very poorly conditioned state-space realization of Padé approximations of small delays. Which in turns gives poorly balanced Padé approximations of
DelayLtiSystems
with small delays.I think that we have discussed this poor balancing before.
The text was updated successfully, but these errors were encountered: