AnsweredAssumed Answered

ADS Momentum: Fundamental Questions

Question asked by germatron on Jul 9, 2008
Latest reply on Jul 10, 2008 by ALEPEREZ

Like • Show 0 Likes0
Comment • 5

Hello,

Please to meet you guys. I am a newbie RF Engineer assigned to work on 802.15.4 wireless devices. I just have a few fill-in-the-blank questions regarding momentum.

Below is a link to my first set of questions:

link

I also would like to know exactly how do I place my soldermask onto my board? In the entry layer box, there lists a soldermask option. When I click on it, and draw a polygon, where does the solder mask go?

My question is:

1: Suppose I have cond 1 and cond 2 layers. Does the soldermask cover cond 1 or cond 2 or both?

2: How do I draw the soldermask? Do I just extend a rectangle over my cond layer?

3: I tried listing my soldermask in my substrate definition. When I go to the metallization tab, i can map one solder mask layer as a strip/slot/via but not the other.

Is it better to list my soldermask as a substrate layer, map it as a strip or just draw it in layout?

I am designing a stacked patch antenna. the soldermask placed on the radiators during fab/production time is critical in my desing since the dielectric coating detunes my device.

Any help will be greatly appreciated.

Thanks.

No one else has this question

Outcomes

germatron Jul 9, 2008 4:50 AM

Oh and another thing. How does one know, what is the proper mesh size to apply to the simulation model? I leave it at the default of 20 cells per wavelength.

However, a co-worker suggested that a mesh size that is to fine will introduce false valleys into the s-parameter plots. I don't want my mesh too course, but I also want accurate infromation for my stacked patch design.

Actions

ALEPEREZ

Jul 10, 2008 12:44 AM

It would be very unusual that the solder mask lacquer would detune an antenna since it's thickness would normally be much less than the dielectric underneath. However, if this is really the case, then Momentum allows you to model it as with any other dielectric. However, it must be homogenous in the xy-plane and will be infinite in extent. In between the solder mask dielectric and the PCB, you will be able to map the metal strip layers (the so-called interface plane which is marked with a dashed line "----------").

You can make metal 3D thick but be careful as this really drives up simulation time and memory requirement and is usually not needed to get good results (at least for PCB type sizes). If the line widths and patch dimensions are large compared to plating thickness (>10), ignore thickness and use Sheet Metal model.

The meshing for an antenna should always be manually checked until resonance frequency is stable. You can start with Edge Mesh and 30 cells/wavelength, but for very high Q resonators, don't be surprised if you still see shifts going to 50, 100 or even 200 cells/wavelength. Usually, statistical variations in dielectric constants will dominate at some point.

For detailed help, I would recommend working with the EEsof technical support department in your region of the world.

Actions

germatron Jul 10, 2008 3:43 AM

Hello. Thank you for the response. My patch, a single radiator antenna, has a very narrow impedance bandwidth, so a LPI soldermask, 1 mil thick, does in fact detune the device about 10-40 MHz.

I have read in some papers, that since radomes and soldermasks cover the conductor, the field lines are confined closer to the board space. This has an effect of changing antenna parameters, depending on the soldermask/radome setup. So...

I modeled the soldermask as a substrate layer, on top of the conductor. It does affect the antenna performance, but I can probably compensate for it by adding a matching T network.

So I'll stick with that convention until something better comes up.

As for the mesh size, I see reduction in performance for coarser cell parameters. In momentum, the solver accounts for the field lines and current distribution on the patch unlike in schematic. A finer mesh size gives me more accurate results, since the cells subdivide the green's functions.

But I do not know to what extent, the mesh size is applicable. So i'll stick with the default of 20.

Actions

germatron Jul 10, 2008 4:14 AM

So I'm using ADS 2008. The ADS tutorial that Agilent provides uses the older 2003 engine.

In that setup, there exists the optimization tool. My package does not have that. Is it because my company did not purchase that particular license or does ADS 2008 lack it?

I'm trying to create a suitable model to import over to schematic. But to no avail, schematic doesn't recognize polygons or at least I don't think it does. So I would have to create a microstrip model from the component libraries. Has anyone had this experience?

Actions

ALEPEREZ

Jul 10, 2008 8:14 AM

The way optimization is done is using Layout Components (the Momentum > Component > * commands) which are parameterized using one of the two methods. The details require some training and/or support. Maybe the manuals are sufficient for you to pick up the steps. The LC is then inserted into a schematic and optimizations are setup the same way as for any other circuit optimization. A Momentum Optimization license is required, however.

Regarding the 20 cells/wavelength, that is the minimum for resonant structures and going to 30 or 50 should show a noticeable shift (improving the results by definition as current distributions are more accurately computed with a denser mesh).

Actions

5 Replies

germatron Jul 9, 2008 4:50 AM
Mark Correct
Correct Answer
Oh and another thing. How does one know, what is the proper mesh size to apply to the simulation model? I leave it at the default of 20 cells per wavelength.

However, a co-worker suggested that a mesh size that is to fine will introduce false valleys into the s-parameter plots. I don't want my mesh too course, but I also want accurate infromation for my stacked patch design.
Like • Show 0 Likes0
Actions
ALEPEREZ Jul 10, 2008 12:44 AM
Mark Correct
Correct Answer
It would be very unusual that the solder mask lacquer would detune an antenna since it's thickness would normally be much less than the dielectric underneath. However, if this is really the case, then Momentum allows you to model it as with any other dielectric. However, it must be homogenous in the xy-plane and will be infinite in extent. In between the solder mask dielectric and the PCB, you will be able to map the metal strip layers (the so-called interface plane which is marked with a dashed line "----------").

You can make metal 3D thick but be careful as this really drives up simulation time and memory requirement and is usually not needed to get good results (at least for PCB type sizes). If the line widths and patch dimensions are large compared to plating thickness (>10), ignore thickness and use Sheet Metal model.

The meshing for an antenna should always be manually checked until resonance frequency is stable. You can start with Edge Mesh and 30 cells/wavelength, but for very high Q resonators, don't be surprised if you still see shifts going to 50, 100 or even 200 cells/wavelength. Usually, statistical variations in dielectric constants will dominate at some point.

For detailed help, I would recommend working with the EEsof technical support department in your region of the world.
Like • Show 0 Likes0
Actions
germatron Jul 10, 2008 3:43 AM
Mark Correct
Correct Answer
Hello. Thank you for the response. My patch, a single radiator antenna, has a very narrow impedance bandwidth, so a LPI soldermask, 1 mil thick, does in fact detune the device about 10-40 MHz.

I have read in some papers, that since radomes and soldermasks cover the conductor, the field lines are confined closer to the board space. This has an effect of changing antenna parameters, depending on the soldermask/radome setup. So...

I modeled the soldermask as a substrate layer, on top of the conductor. It does affect the antenna performance, but I can probably compensate for it by adding a matching T network.

So I'll stick with that convention until something better comes up.

As for the mesh size, I see reduction in performance for coarser cell parameters. In momentum, the solver accounts for the field lines and current distribution on the patch unlike in schematic. A finer mesh size gives me more accurate results, since the cells subdivide the green's functions.

But I do not know to what extent, the mesh size is applicable. So i'll stick with the default of 20.
Like • Show 0 Likes0
Actions
germatron Jul 10, 2008 4:14 AM
Mark Correct
Correct Answer
So I'm using ADS 2008. The ADS tutorial that Agilent provides uses the older 2003 engine.

In that setup, there exists the optimization tool. My package does not have that. Is it because my company did not purchase that particular license or does ADS 2008 lack it?

I'm trying to create a suitable model to import over to schematic. But to no avail, schematic doesn't recognize polygons or at least I don't think it does. So I would have to create a microstrip model from the component libraries. Has anyone had this experience?
Like • Show 0 Likes0
Actions
ALEPEREZ Jul 10, 2008 8:14 AM
Mark Correct
Correct Answer
The way optimization is done is using Layout Components (the Momentum > Component > * commands) which are parameterized using one of the two methods. The details require some training and/or support. Maybe the manuals are sufficient for you to pick up the steps. The LC is then inserted into a schematic and optimizations are setup the same way as for any other circuit optimization. A Momentum Optimization license is required, however.

Regarding the 20 cells/wavelength, that is the minimum for resonant structures and going to 30 or 50 should show a noticeable shift (improving the results by definition as current distributions are more accurately computed with a denser mesh).
Like • Show 0 Likes0
Actions

ADS Momentum: Fundamental Questions

Outcomes

Related Content

Recommended Content