SST convergence with ELDO RF

[e.Horus]

hello all,

I am trying to simulate a switching mode PA with ELDO RF, and I am having trouble with convergence in certain process corners. Particularly in FAST process and FAST process/low temperature corners. Any ideas? I am using th e following options

.option pstran
.OPTION NOASCII
.OPTION ENGNOT
.OPTION AEX
.OPTION NOWAVECOMPLEX
.OPTION PCS=1
.OPTION ALIGNEXT
.OPTION MINRACC = 1.0e-2
.option nomemstp
.sst fund1=f1 nharm1=21
.option SST_CONVERGENCE_HELP=TRANSIENT
.option SST_TRAN_NPER=10
.option SST_TSTART=20n
.option sst_max_liniter=500
.option sst_ovrsmp=2
.option sst_use_ntone_procedure=0

 

[Mazz]

Did you try with:
.option SST_CONVERGENCE_HELP=ADVANCED_NEWTON and sweep the input power with a small step (or use the SST_RAMPING_FACTOR=0.25, that is the same) ?
Maybe, being a switch PA, you need more time in process corner to stabilize the steady state, so use larger SST_TRAN_NPER
I hope it can help.
Mazz

 

[e.Horus]

Thank you for your advice.

but

1. I cannot get it coneverging with SST_CONVERGENCE_HELP=ADVANCED_NEWTON at all.
It will only converge (for some but not all corners) if I use SST_CONVERGENCE_HELP=TRANSIENT,
and I have to set the following option .option nomemstp (even to get convergence on the first point).

2. I tried increasin SST_TRAN_NPER to 20 with no success

 

[Mazz]

Usually next step at this point is to contact Mentor support and send them your bench.
If the simulator does not reach the convergence, it can be because the switching mode PA have a long transient due to the switching. Did you have a check with transient? Do you know that you can use the transient result for sst (I don't remember the option...).
I hope it can help.
Mazz

 

[e.Horus]

I am doing everything now in transient.
In sst you can use transient assisted sst, that is why I use the option

sst_convergence_help=transient.

How do I send the file to mentor?

thanks a lot for your help by the way

 

[Mazz]

Mentor has on line support for customers:
**broken link removed**
you should give info on your license and register there.
Mazz

 

[bozzo]

hello,

can you give more informations concerning the divergence ?
is it Newton or Krylov that fails ?
Maybe more harmonics can help ...

Note: .option sst_convergence_help=transient can be used
in conjunction with .option sst_convergence_help=advanced_newton

Bozzo.