[SciPy-User] Incoherent results with signal.impulse

[Warren Weckesser]

Cédrick FAURY wrote:
>>> / Hello,
>>>       
>> />/
>> />/ I have scipy 0.7.1, python 2.6, and when I do :
>> />/
>> />/ n = scipy.array([1])
>> />/ d = scipy.array([0.01, 0.2, 1.0])
>> />/ T, yout = scipy.signal.impulse((n,d))
>> />/
>> />/ it gives incoherent results for yout.
>> />/
>> />/ And that doesn't occurs with [1.0, 2.0, 1.0] denominator.
>> />/
>> />/ Is it a bug ?
>> />/ I'm doing something wrong ?
>> />/ Is anybody knows a solution ?
>> /
>>
>> scipy.signal.impulse assumes that the state matrix A is diagonalizable, 
>> so it does not give a correct result when A is defective.  I would call 
>> that a bug. :)
>>
>> The attached file contains the function impulse_response() that uses a 
>> different method to compute the impulse response.  If run as a script, 
>> the code at the bottom of the file plots impulse responses computed by 
>> impulse_response() and by scipy.signal.impulse() for your example, and 
>> for two other values of the leading coefficient of your denominator.
>>   
>>     
> Thank you very much, it works fine now !
> Actualy, if i'm right, the solution is to use lsim2 ?
>   

Yes, it uses lsim2, with the input U all zeros, and with the initial 
condition set to the B matrix (plus the optional X0, if given).

Warren
> Cédrick
> _______________________________________________
> SciPy-User mailing list
> SciPy-User@scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>