線上影音
Home > ANSYS Q3D 教學 > Q3D vs. Maxwell for TSP Application
本文始於2015,分別以R16的版本為例,說明Q3D與Maxwell做電容式觸控面板模擬的差異。This article is intended to discuss the difference between Q3D and Maxwell for TSP application, specially for C calculation. Use Q3D2015 and Maxwell2015 for example.
-- update section 7 on 2017.10
-
Introduction
-
Create TSP Model
-
Assign Material
-
Q&A
8.1 For Maxwell, how to assign "solution frequency" for C?
8.2 Why does Maxwell solve real thickness 1um faster than Q3D?
8.3 Can Maxwell solve R, C at AC (with skin effect) or transition region (DC~AC)?
The same as Q3D, or you can copy all objects (ctrl+A, ctrl+C) from Q3D and paste (ctrl+V) them to Maxwell directly.
the same as Q3D
-
Solution Type and Excitation
4.1 Set Solution Type and Design Unite
[Maxwell 3D] \ [Solution Type],選[Electric:Electrostatic]
選[Electric:Electrostatic],等一下步驟4.2設定Excitation會出現Voltage/Charge/Floating/Charge Density供選擇,步驟6的模擬結果才會出現[寄生電容]值。如果這裡一開始選的是[Magnetic:Magnetostatic],步驟4.2設定Excitation會出現Voltage/Current/Current Density供選擇,步驟6的模擬結果則是[寄生電感]值。
Modeler \ Units,選[mm]
4.2 [Maxwell 3D] \ [Excitations] \ [Assign] \ [Voltage]
top-layer ITO are named as Rx, Rx_1, Rx_2,...,bottom-layer ITO are named as Tx, Tx_1, Tx_2,...
Rx=R1=0V, Rx_1=R2=0V, ..., Tx=T1=1V, Tx_1=T2=1V, ...
對導体設定電壓或電流後,Maxwell才有辦法對object,進行靜態電場的分析。 一個design中,建議至少有兩種電壓值,比方0V, 1V;避免所有net都同一個電壓值(0V),會導致meshing很久切不到target [Percent Error]。這是因為Maxwell是透過object的電壓/電流設定,計算能量場於design的分佈情況,如果每個object電壓相同, 彼此間沒有電位差(電力線分佈),很難算收斂整個系統的場分佈情況
4.3 [Maxwell 3D] \ [Parameters] \ [Assign] \ [Matrix]
這裡有勾選的電壓,其所代表的net才會進行分析,模擬結果才會 在step6看到他們的自電容與互電容值。
[Maxwell 3D] \ [Analysis Setup] \ [Add Solution Setup]
Maxwell裡的[Percent Error],就像是HFSS裡的[Maximum Delta S],決定了模擬的精度,其數值越小,meshing refine的次數會越多,所切出的四面體(四面錐體)越多,模擬所耗的記憶體越大。
-
Result Matrix and Simulation Time
6.1 Maxwell 3D \ Validation Check
6.2 Maxwell 3D \ Analysis All
6.3 [Maxwell 3D] \ [Results] \ [Solutions Data]
或是在[Project Manage]內,選Analysis\Setup1按滑鼠右鍵選[Profile] or [Convergence]即可
即使把percent error調的夠小(for example, 0.1% twice),Maxwell解出來的Cm略大於Q3D解的值,原因是兩邊的厚度不同。
同樣解1um真實厚度的題目,Maxwell比Q3D with HPC的速度略快,但Maxwell解real 1um比Q3D解thin conductor with HPC的速度慢 。
雖然Maxwell解靜態電場電容只用單核,但速度不會比Q3D+HPC多核慢,這是讓筆者比較surprise的
Both of Maxwell and Q3D are
-- Quasi-static solver with FEM
-- support "floating net"
Maxwell與Q3D的電容C求解結果很接近,但
-- Q3D不用設air box, Maxwell需要
-- Q3D可以by solver type或CG, AC_RL做HPC distribution求解加速, 但Maxwell不行
-- Maxwell沒有thin conductor,所以一定要放大等效厚度求解(至少1um以上厚度)
-- 對於Maxwell,CG與RL的求解結果,無法整合成SPICE或S-parameter model
-- Maxwell只能解DC電容,而Q3D可以解DC與其他頻點的電容
Maxwell比起Q3D在觸控應用上,目前的優點是
-- 可以看空間場, plot 3D E field
-- 有對稱boundary可以利用
Maxwell比起Q3D在觸控應用上,目前限制是
-- 無法export SPICE或S-parameter model (只解C@DC)
-- 沒有thin conductor技術
-- HPC在Maxwell對觸控的應用上, 目前幫不上忙。 (freq. distribute觸控用不上)
8.1 For Maxwell, how to assign "solution frequency" for C?
Ans:Maxwell solves [Electrostatic] at DC, so it does not need to assign solution frequency.
8.2 Why does Maxwell solve real thickness 1um faster than Q3D?
Ans: I am not sure what is the root cause, but note that Maxwell solves static C with tetrahedra (FEM) mesh, but Q3D solves C with triangle (BEM) mesh.
8.3 Can Maxwell solve R, C at AC (with skin effect) or transition region (DC~AC)?
Ans:Maxwell has an [Eddy Current] solve type to calculate skin effect, but it is only for Rac, not for C and transition region. In the future, maybe R19 or R20, AC solver for CG will be added in Maxwell.