Modal Analysis of Spherical Shell

Hello, I have a 9ft dia spherical steel tank with a 0.75 in thick wall. I want to know how its going to ring when i start slamming lots of loud things around it and against it, so i'm attempting modal analysis with CAElinux.

I have the 32 bit CAElinux2008 because my computer is 32 bit but i put the SalomeMecha 4.1.4 standalone pack on it.

I used the geometry tool to make my geometry.
54 dia sphere subtracted from a 54.75 dia sphere at origin=thinshell
thinshell scaled from origin by 25.4 =mmthinshell (thinshell in milimeters)
I then partitioned the mmthinshell with the XY,YZ,and XZ planes to help with meshing.

for meshing
i take my partitioned mmthinshell and did Netgen
1st order
moderate fitness
shell thickness is about 20 units, so i have tried max sidelength from 25 to 150 in making meshes.
yes optimize
which generally makes a mech onthe order of 200k-300k items
in a few minutes.

I switch to code aster module.
Wizards->modal_Analysis
3D->mesh->mmthinshellmesh
i'm in mm scale so for steel properties i'm using
E=2.100e5,NU=0.28,RHO=0.007850
I ask for 33 to 72 modes
I then edit the aster case setting memory to 512k and time to 17000s so they wont be an issue

Heres my problem. I get negative mode frequencies,
and my dominant modes are on the order of 1e-6 so i'm guessing any results i have gotten so far are not physical

NUMERO FREQUENCE (HZ) NORME D'ERREUR
4 2.84054E-07 1.72738E-21
3 -1.05203E-06 1.61798E-20
5 1.38651E-06 4.10544E-20
6 1.79279E-06 5.71857E-20
2 -1.82604E-06 5.71267E-20
1 -2.01754E-06 4.91058E-20
7 5.09970E+00 8.79957E-08
8 5.14657E+00 6.14791E-08
9 5.71300E+00 2.23035E-07
10 5.72267E+00 2.49166E-07
: : :
: : :

---

i also get this message with negative frequency

LA VALEUR MAXIMALE EST INFERIEURE A LA VALEUR DE CORPS RIGIDE ON LA MODIFIE, ELLE DEVIENT: -3.94784E-03
LA VALEUR DE DECALAGE EN FREQUENCE EST : -1.00000E-02

---

and i see these

<E><ALGELINE5_15>pour le concept MODES le mode num�ro 13
de frequence 6.554518 a une norme d'erreur de 0.000006 sup�rieure au seuil admis 0.000001 Cette erreur sera suivie d'une erreur fatale.

I've run about 30-50 different case over the past week, all crap. any ideas as to what i'm doing wrong?
What additional information do you need? the .comm the .mess?
I see people doing failure mode analysis of support column under load with code aster. Is Salome only good for meshing and post processing results, and the rest is done by manually editing the code aster .comm file?

your 68 Hz seems more reasonable then my 10e-3 hz.
did you have to lock down a face of the model to get those modes or did you leave the sphere in free space? did you load it with gravity?
I made the negative modes disappear by adding a solid 10in dia cylinder protruding a few inches to make a base to give me a flat face that could be anchored. I used 60mm as the max size so elements would be 3x1 with the wall thickness
the comm file is. which is just what the wizard makes with a face locked down. the mesh should be in mm (i'm guessing a 1x1x1 units cube in the geometry module, makes a 1x1x1 same units cube in the mesh module)
[code:1]
DEBUT();

MAIL=LIRE_MAILLAGE(UNITE=20,FORMAT='MED',);

MODELE=AFFE_MODELE(MAILLAGE=MAIL,
AFFE=_F(TOUT='OUI',
PHENOMENE='MECANIQUE',
MODELISATION='3D',),);

ACIER=DEFI_MATERIAU(ELAS=_F(E=2.1e5,
NU=0.28,
RHO=.007850,),);

CHMAT=AFFE_MATERIAU(MAILLAGE=MAIL,
AFFE=_F(TOUT='OUI',
MATER=ACIER,),);


BLOCAGE=AFFE_CHAR_MECA(MODELE=MODELE,
DDL_IMPO=(
_F(GROUP_MA='Group_1',
DX=0,
DY=0,
DZ=0,),),);

MACRO_MATR_ASSE(MODELE=MODELE,
CHAM_MATER=CHMAT,
CHARGE=BLOCAGE,
NUME_DDL=CO('NUMEDDL'),
MATR_ASSE=(_F(MATRICE=CO('RIGIDITE'),
OPTION='RIGI_MECA',),
_F(MATRICE=CO('MASSE'),
OPTION='MASS_MECA',),),);

MODES=MODE_ITER_SIMULT( MATR_A=RIGIDITE,
MATR_B=MASSE,
CALC_FREQ=_F(
OPTION='PLUS_PETITE',
NMAX_FREQ=85,),);

IMPR_RESU(MODELE=MODELE,
FORMAT='MED',
RESU=_F(MAILLAGE=MAIL,
RESULTAT=MODES,
NOM_CHAM='DEPL',),);

FIN();
[/code:1]

but my output looks more like
[code:1]

-- CODE_ASTER -- VERSION D'EXPLOITATION 9.04 --

COPYRIGHT EDF-R&D 1991 - 2008

EXECUTION DU : JE-04-MARS-2010 19:33:45

PLATE-FORME : LOCALHOST

NB MAX PROC OpenMP: 1

SYSTEME : LINUX

CPU : I686


****************snip*******************************

---

LE NOMBRE DE DDL

TOTAL EST: 108552

DE LAGRANGE EST: 102

LE NOMBRE DE DDL ACTIFS EST: 108399

---

L'OPTION CHOISIE EST: PLUS_PETITE

LA VALEUR DE DECALAGE EN FREQUENCE EST : 0.00000E+00

---

INFORMATIONS SUR LE CALCUL DEMANDE:
NOMBRE DE MODES RECHERCHES : 85

LA DIMENSION DE L'ESPACE REDUIT EST : 0
ELLE EST INFERIEURE AU NOMBRE DE MODES, ON LA PREND EGALE A 170



=============================================
= METHODE DE SORENSEN (CODE ARPACK) =
= VERSION : 2.4 =
= DATE : 07/31/96 =
=============================================
NOMBRE DE REDEMARRAGES = 3
NOMBRE DE PRODUITS OP*X = 257
NOMBRE DE PRODUITS B*X = 770
NOMBRE DE REORTHOGONALISATIONS (ETAPE 1) = 255
NOMBRE DE REORTHOGONALISATIONS (ETAPE 2) = 0
NOMBRE DE REDEMARRAGES DU A UN V0 NUL = 0

---

LES FREQUENCES CALCULEES INF. ET SUP. SONT:
FREQ_INF : 4.12075E-03
FREQ_SUP : 7.96893E-01

---

CALCUL MODAL: METHODE D'ITERATION SIMULTANEE
METHODE DE SORENSEN

NUMERO FREQUENCE (HZ) NORME D'ERREUR
1 4.12075E-03 6.34139E-10
2 4.13680E-03 5.70143E-10
3 2.02226E-02 2.34509E-08
4 5.78469E-02 6.19203E-10
5 2.38082E-01 9.60725E-10
6 2.38331E-01 4.25963E-10
7 4.41370E-01 1.08905E-09
.
.
.
85 7.96893E-01 1.33681E-09
NORME D'ERREUR MOYENNE: 0.18027E-08

---

VERIFICATION A POSTERIORI DES MODES

DANS L'INTERVALLE (-1.00250E-02, 7.98883E-01)
IL Y A BIEN 85 FREQUENCE(S)

---

[/code:1]

mili-hetz seems to low. perhaps i need to run some test cases on models with analytical solutions before i can jump to my desired case.
eficas looks powerful. I take it theres no translations available?

okay so anchoring the mesh at at least 1 location is important.
the other thing i noticed is for a beam problem.
Given E and rho and for a small square cross-section beam of length L,
v=sqrt(E/rho) is the soundspeed
and the major mode should be f=v/2L
well if
units of E is (MPa or or 1e6 kg/(ms^2))
units of rho is g/mm^3
and units L is mm
then
v is in units of (1e3 mm/s),
and f is in units of 1e3 1/s or kHz?

so using MPa and mm mesh gets you frequency in kHz?
and PA m mesh would get you hertz?