[Halld-mc] EM background warning

[Richard Jones]

Blake,

The source generator has to sample a multi-dimensional bremsstrahlung 
cross section formula that has to be computed by iteration.  To throw 
random events with this exact weight would require me to be able to find 
the indefinite integral of the cross section and then find its inverse 
function -- sounds hard.  The easier thing, which I did, is to find a 
simple analytical formula that is reasonably close in it shape to the 
exact one, which I know how to integrate and invert.  So I throw 
according to the simplified one, and then reweight the final 
distribution by the ratio of the exact and approximate cross sections by 
computing their ratio for each event and doing a random discard based on 
that ratio.  If the approx. function is a good choice, this ratio comes 
out close to 1 most of the time, but there are corners of phase space 
where the approximation is poor: there the ratio can be orders of 
magnitude bigger than its average value.  The random discard method 
relies on chosing an effective maximum value for the ratio and then 
discarding based on a random number cast Unif[0,ratio_max].  In those 
rare cases where the ratio comes out bigger than ratio_max, there is a 
problem: what to do?  What I do is, print the warning message, move 
ratio_max to the new maximum, and carry on.  Obviously if this happens 
too much during the run you will end up with a background energy 
distribution that is gradually converging (growing more accurate, but 
less numerically efficient) the longer it runs.

My experience is that for reasonable values of the beam threshold and 
endpoint energies, the value I chose for ratio_max is good enough that 
the changes as ratio_max is adjusted upward will not be statistically 
significant.  But I left the warning in there anyway.  If you see the 
warning once per event, or every few events, please let me know and I 
will try to figure out what is driving the ratio to diverge for your 
simulation conditions.

tmi?  you asked...
-Richard J.


Blake Leverington wrote:
> Hi,
>
> I'm looking at the log files from a simulation where I've turned on the 
> em background. What does the warning: "Warning from beamgen: freq = 
> 0.00022101 greater than freqMaximum=0.0001999999995" indicate?  Any ideas?
>
> -Blake
> _______________________________________________
> Halld-mc mailing list
> Halld-mc at jlab.org
> https://mailman.jlab.org/mailman/listinfo/halld-mc
>   

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 3639 bytes
Desc: S/MIME Cryptographic Signature
Url : https://mailman.jlab.org/pipermail/halld-mc/attachments/20090924/953df13f/attachment.bin