[Gate-users] Noise parameters

David GUEZ davguez at free.fr
Wed Feb 13 01:14:59 CET 2008


Hi ashley,

Ashley Devarim a écrit :
> Hi David,
>
> Thank you very much for your answer, would you mind please if you take a look at my questions below :
>
>
>   
>> Hello,
>> actually, I don't understand what you want to set up... The setLambda parameter (7.57 mus in your exemple) is equal to the mean time between to consecutives noise events in a Poisson distributed noise source. The inverse of this value is the noise frequency, thus.
>>     
>
>
>
> 1) but the difference time distribution is an exponential and the setLambda parameter is the power of the distribution, even in this case the setLambda parameter would be simply the inverse of the noise frequency ?
>
>   
Actually, you can deal with any kind of time interval distribution. The 
one given in the example correspond to an exponential distribution, 
because a Poisson's process imply such model of time interval 
distribution. So if your noise source follow the Poisson law, you can 
keep this distribution (and adapt the setLambda parameter to your noise 
frequency). If not, a generic way to define your own distribution is to use

/gate/distributions/insert File

and specify a file that you prepared and which will contain your 
distribution (table 8.1, page 97 in the v3.1.2 GATE user guide is 
regrouping parameters for file based distribution).

> 2) For example if I have a noise values varying between 0 and 90 keV  and mean value = 60 keV , what would be the setMean and setSigma parameters in this case ?  
>
>
>   
Here again, the example correspond to a Gaussian distributed noise 
(Gaussian in terms of energy spectrum, and Poissonian in terms of time 
distribution, so).
If this doesn't corresponds to the noise you want to modelize, you can 
always use the general purpose File distribution to fit your needs.
In the example you gave, you are specifying the min/max value of the 
distribution, as well as its mean value. This is not sufficient to 
characterize the energy spectrum. If this spectrum can be modelized by a 
Gaussian, an exponential or a flat distribution, you can use one of the 
specific forms of distributions given in the table 8.2 already 
mentioned. In any other case, you should use the generic File 
distribution method.

>> Concerning, the eventID number, it is due to the fact that noise events are not produced by physical interaction of particles with detectors (which can be attached to a particle emission event number, which is eventID). The noise events are described as independent source of detector signal, not correlated to any physical particle emission, and are then marked with the special event number (-2).
>>     
>
>
>
> 3) here the reason for my question was,  that the noise events participate to the coincidences events like a particle emission events and I was wondering if this is happening in real life ? I thought that noise should be mixed with the signal and not seen as separated event ( like energy blurring )?  what do you think please ?
>
>
>   

Actually, it depends on what you mean by "coincidence"
If you're using the coincidence feature of gate 
(/gate/digitizer/coincidence/...) the module doesn't care about the 
eventID and so, the noise events will be merged with "physical" events 
for coincidences (this, because the time flag of the noise events are 
distributed along the total time of your simulation).

But if you're using a test on eventID to determine if two events are in 
coincidence (e.g. eventID of data1 == eventID of data2), so, this test 
can't deal with noise events, because eventID is actually a marker of 
the particle source creation marker.


> Kind Regards
>
> Ashley
>
>
>
>   
>

Best regards,
D. Guez
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opengatecollaboration.org/mailman/private/gate-users/attachments/20080213/8c2aa43e/attachment.htm>


More information about the Gate-users mailing list