是否有办法通过变量参数化地改变两层之间的高度并扫描变量

您好，Momentum不能使用参数作为基材定义。
以下是安捷伦在Momentum类中教授的一种解决方法：对于emModels，您可以添加string类型的单元格参数“Substrate”，并使用该字符串参数指向不同的（静态）基板名称。
有一个Momentum示例“Sweep_Substrate_Parameters_wrk”来显示该方法。
此方法可实现自动化，但设置时需要一些时间。
在您的情况下，手动切换基板可能更容易。
最好的问候VolkerEdited：volker_muehlhaus于2013年5月7日上午8:16



     以下为原文

  Hello,

Momentum can't use parameters for substrate definitions. 

Here is a workaround that Agilent teach in the Momentum class: for emModels, you can add a cell parameter "Substrate" of type string, and use that string parameter to point to different (static) substrate names. There is a Momentum example "Sweep_Substrate_Parameters_wrk" to show that method.

This method enables automation, but it takes some time to set up things. In your case, it might be easier to switch between the substrates manually.

Best regards
Volker

Edited by: volker_muehlhaus on May 7, 2013 8:16 AM

谢谢您的回复。
我想这就是我想要的。
运行一个模拟两个电感器（一个在另一个之上）花了将近7小时运行。
因此，我认为如果这将以多种厚度运行，则需要几天才能运行。
我不知道网格是否设置不好或者其他东西，但是在端口之间计算S21非常慢。
附件是我的项目。
我删除了* .dds数据集，否则它将是68Mb。
你能告诉我如何优化这个，这样我就可以快速提取S21 - > Z21-？
那么我可以计算互感= 1e9 * imag（Z21）/（2 * pi * freq）磁耦合k = M / sqrt（Lp * Ls）其中Lp可以是一个电感的自感，而Ls可以是另一个电感的自感
电感器。
这个SIM卡用了8个小时。
有没有办法减少时间？



     以下为原文

  Thank you for your reply.  I think that is what I want.  Running one simulation two inductors (one on top of each other) took almost 7 hrs to run.  So I figure if this is going to run with several thicknesses it will take several days to run.

I don't know if the mesh are not setup well or something but it is very slow to compute S21 between ports.  Attached is my project. I deleted the *.dds data sets otherwise it would have been 68Mb.  Can you tell me how to optimize this so I can extract quickly S21 -> Z21-?

So then I can compute

Mutual Inductance = 1e9 * imag(Z21)/(2*pi*freq)
Magnetic coupling k = M/sqrt(Lp*Ls) where Lp can be the self inductance of one inductor, and Ls for the other inductor.

this sim took 8hrs.  Is there a way to reduce its time?

附件

• TEST_ANTENNAS_wrk.7zap177.9 KB
  

漫长的模拟时间是因为您已手动将网格更改为相当极端的设置：粗金属+边网格+极短的分段长度。
你有没有检查过这个型号真的需要吗？
虽然我习惯在Sonnet中使用精细网格，但我认为在这种情况下，你可以在Momentum中使用不那么极端的网格。
例如，厚金属模拟的侧壁电流已经允许“边缘”电流，因此边缘网格不如Sonnet重要（Sonnet使用没有侧壁电流的厚金属模型）。
真正推动网格复杂性的是您设置的30μm最大子部分长度。
你尝试使用不太极端的价值吗？
另一种加速模拟的设置：您可以使用Momentum RF而不是Momentum Microwave模式，因为这是一个电气小型模型。
我主要研究片上电感器，其中默认设置的Momentum RF模式（厚金属模型，边缘网格关闭，20个子/λ）提供非常好的准确性。
我也为你的PCB电感器尝试了这些设置：使用Momentum RF和默认设置，13MHz电感为877nH，而超细网格仿真为945nH，模拟在不到30秒内完成。
快，但电感差异为7％ - 相当多。
我同意你可能想要更精细的网格，但不要太极端。
现在的问题是：你需要电感和耦合的精度是多少？



     以下为原文

  The long simulation time is because you have manually changed the mesh to rather extreme settings: thick metal + edge mesh + extremely short subsection length. Did you check if this really needed for this model? 

Although I'm used to fine mesh in Sonnet, I think you can get away with less extreme mesh in Momentum, for this case. For example, the side wall currents of the thick metal simulation will already allow "edge" currents, so that edge mesh is less important than in Sonnet (Sonnet uses thick metal model without side wall current). 

What really drives up the mesh complexity is the 30µm max subsection length that you have set. Did you try using a less extreme value? 
Another setting to speed up your simulation: you can use Momentum RF instead of Momentum Microwave mode because this is an electrically small model.

I'm mostly working on on-chip inductor, where Momentum RF mode with default settings (thick metal model, edge mesh off, 20 subs/lambda) gives very decent accuray. I tried those settings for your PCB inductors also: With Momentum RF and default settings, the 13MHz inductance is 877nH, compared to 945nH from your super-fine-mesh simulation, and simulation is finished in less than 30 seconds. Fast, but that is 7% difference in inductance - quite a lot. I agree that you might want finer mesh, but don't go too extreme.

Now the question is: To what precision do you need inductance and coupling?

亲爱的Volker，谢谢你的回复。
我会检查我的SIM卡，但我想我说30密耳（30/1000英寸）存根。
这在厚导体和边缘电流方面产生了精确的结果。
我会放松网格说100密耳等...并按照你的建议。在互感方面，我需要它尽可能准确地模拟真实的交易。
我正在进行从一个感应天线到另一个感应天线的非接触式功率传输，并且确定设计是否符合规格的一部分是验证从一个天线到另一个天线的电压传输。
我有一个物理设备我可以校准模型，但要开始我不想非常粗糙，我想找到一个地方，我可以通过快速模拟得到准确的结果，如果这样的设置存在。我将首先尝试重现
你的设置和准确性。
当我在模拟中设置30密耳的短截线时，我这样做是为了确保我将测量值​​与在点上真正匹配的真实VNA相匹配。
我想我可以把它减少到100或150密耳的短截线，看看我偏离测量值有多糟糕，但代价是更快解决。我也会尝试20subs / lambda。
看来这个存根将是13.56MHz（感兴趣的操作频率）。
不过，我会尝试一下。再次，我需要良好的互感精度才能代表真实的生活模型。
非常感谢你的有用建议。罗伯特



     以下为原文

  Dear Volker,

Thank you for your reply.  I will check my sims but I think I said 30 mils (30/1000 inch) stubs.  This yielded accurate results in terms of thick conductor and edge current.  I will relax the mesh to say 100 mils etc.. and follow your suggestions.

In terms of the mutual inductance, I need it to be as accurate as possible simulating the real deal.  I am doing contactless power transfer from one inductive antenna to another, and part of figuring out if a design will meet the specs is verifying the voltage transfer from one antenna to the other.  I have a physical device I can calibrate the model, but to start I don't want to go very coarse, I want to find a ground where I can get accurate results with fast simulations if such setting exists.

I will first try to reproduce your settings and accuracy.  When I set the 30mils stubs in the simulation I did that to ensure I matched the measured values with a real VNA which really matched on the dot.  I supposed I can reduce it to 100 or 150 mils stubs and see how bad I deviate from the measured value at the expense of solving faster.

I will also try the 20subs/lambda.  It seems this stubs will be huge at 13.56MHz (operational frequency of interest).  I will try that nevertheless.

Again, I need good accuracy in the mutual inductance to be able to represent a real life model. 

Thank you very much for your helpful suggestions.

Regards,

Robert

嗨，罗伯特已经模拟了数百个RFIC电感器，我习惯于微米但你的电感器尺寸当然是密耳的。
对不起该错字。在我看来，限制网格尺寸为30mils是完全矫枉过正的。
在极少数情况下，必须手动限制分段长度。
20 subs / lambda似乎很大，但这足以捕获电流密度随波长变化的变化。
对于电小电感器，这将导致大的子部分，因为电流/相位在几何结构上没有太大变化，并且网格由几何细节而不是波长确定。
所以是的，20 subs / lambda应该没问题。为了确认，我现在将你的超细网格模拟结果与默认网格设置（20 subs / lambda，edge mesh off）进行比较，两者都使用* Microwave *模式。
这些结果非常非常接近：在两种情况下电感为877nH，电阻为450mOhm（默认网格）与460mOhm（超细网格）。
所以我上一篇文章的不同之处在于RF模式与微波模式，而不是网格密度。
网格密度对结果的影响非常小。对模拟方法的另一种思考：上面，你提到你想要看到基板（距离）变化的影响。
我会通过合理快速的网格划分运行该分析 - 即使绝对电感或耦合值为5％，您也可以轻松看到该参数扫描的趋势。最好关注沃尔克



     以下为原文

  Hi Robert,

having simulated hundreds of RFIC inductors, I'm used to microns but your inductor dimensions are mils, of course. Sorry for that typo.

Limiting mesh size to 30mils is total overkill in my opinion. There are few cases where subsection length must be limited manually. The 20 subs/lambda seem large, but this is enough to capture the wavelength-dependent variation of current density. For an electrically small inductor, this will result in large subsections because the current/phase does not change much across the geometry, and mesh is determined by geometry details rather than wavelength. So yes, 20 subs/lambda should be fine.

To confirm, I now compared your super-fine-mesh simulation results with default mesh settings (20 subs/lambda, edge mesh off), both using *Microwave* mode. These results are very, very close: inductance 877nH in both cases, resistance 450mOhm (default mesh) vs. 460mOhm (super-fine mesh). So the difference in my previous post was from RF Mode vs. Microwave mode, not from the mesh density. The effect of mesh density on results is really small.

Another thought on simulation methodology: Above, you mentioned that you want to see the effect of substrate (distance) change. I would run that analysis with reasonably fast meshing - even if the absolute inductance or coupling values are 5% off, you will easily see the trend for that parameter sweep.

Best regards
Volker

亲爱的沃尔克先生，谢谢您的见解！
它非常有教育意义。
我在RF模式下运行模拟，并且确实比使用20 / lambda的微波模拟运行得更快。
然后我尝试了30 / lambda并且跑得相当快。
我现在将设置工具扫描不同的基板，并将功率传输视为位置和间隙的函数。
这是我的最终目标。
感谢您的帮助。
我真的很感激.Robert



     以下为原文

  Dear Mr. Volker,

Thank you for your insight!  It has been very educational.  I ran the simulation in RF mode and indeed ran faster than microwave simulation with 20/lambda.  I then tried 30/lambda and also ran rather fast. 

I will now set up the tool to sweep through different substrates and see the power transfer as a function of position and gap.  This is my ultimate goal.  Thank you for your help.  I really appreciate it.

Robert