Agilent ADS Momentum crashes with strange errors

[buenos]

hi

i was trying to do an EM simulation in Agilent ADS Momemntum. After like 50 minutes it gave a warning and stopped. I have 2GB RAM, Core2Duo-2.2GHz.

ERROR : memory allocation failure, no more virtual memory available!


------------------- Momentum ---------------------

Version: Momentum MomEngine 8.40.075 (*) 350.500 Feb 19 2009
Simulation mode: MW
Design: all3traces

Simulation started on : Thu Dec 10 07:31:53 2009

Layout processing started
Layout processing finished
S-parameter simulation started
...initializing
...analyzing ports
...extracting layout
...expanding thick conductors
...reducing mesh
Frequency sweep started
Simulation frequency [1] = 10 MHz
...loading Green functions
WARNING: Could not create the dataset
 since no citi files could be found.
...saving project files

---------------- Momentum stopped ----------

whats are these citi files and why are they missing? or is it a fake warning, caused by the memory allocation error?

how to make this simulation work?

 

[albbg]

I have had the same problem in the past. I solved it by simplification of my network. I don't remember exactly all the actions I did, but I remeber that I merged the layer cond (my design was a 2 layers: tracks on cond + ground plane on cond2) and I removed all the ports leading to passive components (RLC) using in place of each couple of ports the component that will be connected to them.
The "citi" file is a file format used to store s-parametrs (like touchston that is another file format). I think the memory error stopped the file creation, then when the Momentum crashed the file was missing.
At the moment I cannot open your design.
Regards

 

[buenos]

hi

thanks for the answer.

this design is 4 layers, basically a signal goes on top, to a via, then continues on the bottom. there are 2 stitching vias next to the signal via, and 2 big holes on the 2 ground planes.
this is a simple model, it does not contain any RLC elements.
the zap file is an arcived ads2009 project. file>unarchive_project.
i want to evaluate the accuracy og the software by simulating 2 kinds of current return path discontinuities, a layer transition with stitching vias and a ground plane void.

if i cant even simulate this, then how would i simulate a trace with 30 antipads on a real board design? do i need a server-farm with 32GB for each proc?

 

[RealAEL]

The transitions from the traces to the vias are not correct. The layers mapped into the VIA structures only represent the outer vertical edge of the via. Therefore simply overlaying the ETCH_TOP/ETCH_BOTTOM traces only connects to these structure on the circumference of the via where the trace touches that part of the vertical edge.

You need to include circular metal on ETCH_TOP/ETCH_BOTTOM to represent these pieces so that connection is all around the edge.

You did this correctly at the outside contacts but not the vias in the middle.

Use View > 3D View > All... to see these views.

 

[rfa86]

I am having the same error
WARNING: Could not create a dataset since no citi files could be found.

I am trying to simulate the CPW. I am using slot for the conductor configuration . I have the ports as coplanar ports.

Could anyone help me what could be the porblem??


Many thanks
Rfa

 

[buenos]

hi
thanx

i have tried, i added the via cirles to the top/bottom layers, but the simulation still fails in the same way.

is it because i have only 2GB memory?

 

[joshua_1400]

hi when u try to simulate the structure that is some complexity on it your RAM capacity reach 2.2Gbyte approx. and this error occur.
one suggestion is reduce mesh size or eliminate edge mesh .
when i try, this error is disappear
suggestion two. split your design in some part and simulate them separately.
suggestion three try to simulate in other computer like 64bit sys. with 4gbit ram.

 

[buenos]

when simulating big return path discontinuities, we can not make small enough portions for the program.

what is the memory need for a 14 layer structure, 150mm trace going on a bus to 8 chips?
just approximately?
i am asking this because if it needs 10Tbytes, then obviously its not feassible. but if we can simulate everything with a PC with 4GBytes, then we just buy some DIMMs and all sorted.

what is this 2.2GBytes anyway?

why would it work on a 64bit OS?
to try it, we should purchase a 64bit windooz, or a linux.

 

[venu.gongal]

Hi

The error u r getting is because of insufficient RAM , try increasing RAM it should work.
or by comprimising on accuracy, reduce the mesh size

 

[venturer84]

I think you should try AWR's Axiem EM Simulator. I have found it to be significantly easy to use and faster than Momentum.

 

[buenos]

the problem with crashing at unsufficient ram, is that if for example 8GB is just enough to simulate a point to point 50mm net, then i would need 100Gbytes to simulate a 200mm net with 8 components on it, which is not feasible.

 

[venturer84]

How many number of Unknowns does your structure has?

 

[buenos]

it crashes at 2 projcts: 10000 and 47000 nodes in the mesh.

 

[kspalla]

try this.
1) put Half of the stucture and see if you still face the problem.
if it is memory then this case (halved design) simulation should go.
2) 10 Mhz is too low freq, limit the simulation to max freq and single pont, this vill check the freq range problem.
3) i will let you know the third if you share the design

 

[tqtstuff]

No comments for RAM problems, my machine runs on WinXp Pro x64, 16GB RAM, Core i7 2600 but I got the same error "...no citi files could be found"

Just untick "Edge mesh" in Mesh setup. It works perfectly.

Thanks to Joshua_1400's msg.

 

[vasikara]

No comments for RAM problems, my machine runs on Win 7 64 Bit, 8GB RAM, Core i5 2600

but I got the same error "...no citi files could be found"

Just untick "Edge mesh" in Mesh setup. It works perfectly.

 

[tqtstuff]

unticking "Edge mesh" cannot work for big project.
I've solved this problem by setting affinity for the hpeesofemx process. No matter what OS is.
No more complaint of "Could not create the dataset
since no citi files could be found.... 100% worked again

 

[buenos]

"setting affinity for the hpeesofemx process"
-what does this mean?

 

[tqtstuff]

"setting affinity for the hpeesofemx process"
-what does this mean?

Start ADS2009
Run task manager (Ctrl+Alt+Del)
In processes tab, find "hpeesofemx.exe" and right click it, choose "Set Affinity"
Untick to reduce CPUs if you have multicore. Simulation should work perfectly with CPU 0 & CPU 1 selected. You can raise CPU number later to check if it still work.(My case works with 6 in 8 CPUs).
Now I am working on XP 64 bit OS with ADS2009 without this stubborn error.
Good luck

 

[buenos]

thanks, i will try it.