AnsweredAssumed Answered

Can Momentum do it the job?

Question asked by CYBERPHOX on Oct 26, 2006
Latest reply on Nov 2, 2006 by ALEPEREZ

Like • Show 0 Likes0
Comment • 3

Hi All,

Over the past few days I have getting pretty frustrated :-[ with Momentum (ADS2006) and am beginning to wonder if we have made the right decision in getting this tool all together.

What I would like to know is if anyone has been able to simulate resonably complex substrate layouts with at least 4 layers? Not simple inductors or hairpin filters etc.

My substrate ~8x6mm has and 2 die an RF (about 30 bond wires) and baseband (about 90 bond wires). There are a few passives too. Basically a typical SiP.

My workstation is a brand new 64Bit Dell 8), 4 GB RAM, 3.2GHz processor, WinXP 64, I have setup 100GB virtual ram. There are very few other apps installed on this machine too, basically a clean installation.

It would be good to know what other engineers have been able to do.

Thanks,

No one else has this question

Outcomes

BONGJADORMIO Nov 2, 2006 7:40 AM

Hi!

You want to model the whole module?
For sure you can use momentum in your application. However your substrate is quite large 8x6mm with 4 layers plus complex line-ups in the layout. Its difficult to simulate that into a single simulation setup.

What I normally do in that kind of situations was, I broke the substrate into 4 equal parts such that I can reduce the physical size as well as the number of ports. Such that the momentum can able to hundle the memory size of my simulation, "make you circuit less complicated".

Model the substrate first then use TUDelf model to model your bondwires. That should be a lot of work for sure.

In may cased I had been able to modeled the whole PA module 8x8 mm with 4 layers using momentum simulator of ADS2005A.

Using momentum you need more patience for setting up your simulation bech but it will give you acurate results if you made you settings correct.

Actions

CYBERPHOX Nov 2, 2006 8:11 PM

Yes, that would be the ideal scenario. But I think in reality I will need to use some engineering judgement and break things down to simulate the bits that I need. But the dangerous thing is, if it is over simplified it then is not the same as the real module.

A big issue with my circumstance is that the layout does not originate from ADS layout, instead it was done by a 3rd party in Cadence. So when you import this, the physical shapes are complex. Particularly the ground plane clearances, these are circular. The mesher does not like these!

Over the past few days I had some progress, thanks to the help from the Agilent application engineer and the help desk.

Hopefully in our next module we will do it all in ADS!

You mentioned TUDelf model for the bond wires, where is this found? I was thinking of using either the Philips models provided or try using an 'air bridge' structure.

Thanks for your feedback. It is always nice to know what other people are able to do.

regards,

Actions

ALEPEREZ

Nov 2, 2006 10:08 PM

I have seen customers simulate IC structures with dozens of very thin layers (nm range). The issue is with geometric complexity and overall electrical size. You can partially clean up overly complex polygons by doing Booleans of those polygons with themselves and using the Preference>Entry/Edit setting "Final Minimum Vertex Distance".

The TU Delft bond wires are on the bottom of the Passive-RF Circuit palette.

Actions

3 Replies

BONGJADORMIO Nov 2, 2006 7:40 AM
Mark Correct
Correct Answer
Hi!

You want to model the whole module?
For sure you can use momentum in your application. However your substrate is quite large 8x6mm with 4 layers plus complex line-ups in the layout. Its difficult to simulate that into a single simulation setup.

What I normally do in that kind of situations was, I broke the substrate into 4 equal parts such that I can reduce the physical size as well as the number of ports. Such that the momentum can able to hundle the memory size of my simulation, "make you circuit less complicated".

Model the substrate first then use TUDelf model to model your bondwires. That should be a lot of work for sure.

In may cased I had been able to modeled the whole PA module 8x8 mm with 4 layers using momentum simulator of ADS2005A.

Using momentum you need more patience for setting up your simulation bech but it will give you acurate results if you made you settings correct.
Like • Show 0 Likes0
Actions
CYBERPHOX Nov 2, 2006 8:11 PM
Mark Correct
Correct Answer
Yes, that would be the ideal scenario. But I think in reality I will need to use some engineering judgement and break things down to simulate the bits that I need. But the dangerous thing is, if it is over simplified it then is not the same as the real module.

A big issue with my circumstance is that the layout does not originate from ADS layout, instead it was done by a 3rd party in Cadence. So when you import this, the physical shapes are complex. Particularly the ground plane clearances, these are circular. The mesher does not like these!

Over the past few days I had some progress, thanks to the help from the Agilent application engineer and the help desk.

Hopefully in our next module we will do it all in ADS!

You mentioned TUDelf model for the bond wires, where is this found? I was thinking of using either the Philips models provided or try using an 'air bridge' structure.

Thanks for your feedback. It is always nice to know what other people are able to do.

regards,
Like • Show 0 Likes0
Actions
ALEPEREZ Nov 2, 2006 10:08 PM
Mark Correct
Correct Answer
I have seen customers simulate IC structures with dozens of very thin layers (nm range). The issue is with geometric complexity and overall electrical size. You can partially clean up overly complex polygons by doing Booleans of those polygons with themselves and using the Preference>Entry/Edit setting "Final Minimum Vertex Distance".

The TU Delft bond wires are on the bottom of the Passive-RF Circuit palette.
Like • Show 0 Likes0
Actions

Can Momentum do it the job?

Outcomes

Related Content

Recommended Content