Yes, we know and admit that, in MAD9, as in all other programs, there are bugs! They will be fixed or documented here on a best effort basis as time and manpower allow. The lists below do not claim to be exhaustive or to fully document the bugs. However we hope it will give users clues as to where they may expect some problems.
Meanwhile, if you think you've found a bug, either in the program or its documentation, please click here.
Although the HGAP and FINT attributes are defined for SBEND and RBEND elements, they have no effect (their effects have not been implemented yet).
There are a number of bugs in these commands. The program may crash or give inconsistent results at different orders of calculation. Under study ....
MAD9 allows you to define a bend magnet with, say, ANGLE=0 (i.e. no change in direction of the reference orbit) but K0 non-zero (so there is a magnetic field and particle orbits are bent). Exploiting this possibility (useful, e.g., for magnets that separate two beams into different rings) leads to problems with the calculation of dispersion.
MAD9 allows you to define an element sequence and then track particle backwards through it using switches on TWISS and other commands. At present these options have some bugs.
Take care to specify a beam explicitly before using STATIC.
Count of number of calls made can get confused.
No default emittance. Perhaps there should be one?
Asking for APERTURE on ATTLIST seems to crash MAD9.
qf: quadrupole,L=2,aperture={1.7,2};
l: line=(qf,qf);
select, line=l,full;
attlist,file=term,line=l,column={NAME,L}; // is OK
select,
line=l,full;attlist,file=term,line=l,column={NAME,L,aperture};
// core dump
even if APERTURE is given as a scalar.
This concerns the generation of a thin lens version in MAD8 format. All elements which are defined as multipole elements with a finite length in MAD9 and have a parameter dependence for their multipole strength lose the parameter dependence during the conversion.
This has been considerably improved in the development version. A Mathematica package is available to complete the restoration of the symbolic dependences of multipole strengths.
In large part, this appears to be due to the fact that they are much more general than in previous versions of MAD. With present day computers this seems to be problematic only for the largest and most complex accelerator structures.
Relatively minor errors (like omitting a semi-colon) can cause MAD9 to crash. We
need more robust error recovery.
These are bugs that we think we have fixed in the development Version 9.5/5 being tested in CERN. They are still present in the last public release, Version 9.5/1.
RBENDs whose K0 attribute was not specified by the user were given a magnetic field that produced a bending greater than the ANGLE attribute. This resulted in unwanted
K0 = 2 Sin[angle/2]/ L
and eliminates the closed orbit deviations. The RBEND
element is now closer to what is
advertised in the manuals for both MAD8 and MAD9.
However it is not the same as the "RBEND" in MAD8.
To change the sign of K0 for cases like Ring2 of the LHC, a formula like the above must be given on input.
These small differences are due to the way MAD8 models an RBEND using an SBEND.
Various headers for ALF* functions interchanged.
LIST command repeats the format specification twice on one line in the TWISS tables.
In some circumstances.
MAD home page
Last updated: 26/9/2001