sonnet em simulation tools rapidshare
I am first day user of em simulation so please dont be angree for stupid questions .)
Well i need to design HF rfid antenna and it was possible to simulate spiral rectangular structure. But netlist got from Spice does show different model compared to reference design from Sonet .
https://www.edaboard.com/viewtopic.p...ht=rfid+sonnet
Due to project requirements i have to design antenna sized as 10x25 mm . Output shows closed induction values for frequencies in range (was taken from 10 to 16 MHz). Of course some parameters could be unrealistic such as thicknes and strip width set to 0.1 mm . But that was first try to get errorl and warning less result.
The questions are - how the structure will be reproducable when PCB is done using conventional process ? What is the minimum recommended spacing between wires and wire width (not restricted by technology capailities but dependent on reproducability and coil Q parameter )? What other measures have to be taken to avoid mistakes and go for production at least in 2 attempts?
I have also some questions regarding generated netlist Are ports in structure directly related to connection numbers in netlist . And what is the meaning of connection point 0 in netlist . is it ground ? or something else .
netlist print is below :
* From Project: rfid1
* Data File Written: 12/09/2005 17:56:55
*< HDATE 12/09/2005 17:53:11
*< MDATE 12/09/2005 17:53:11
* Spice Data
* Limits: C>0.1pF L<1000000.0nH R<100000.0Ohms K>0.01
* Analysis frequencies: 10.0, 11.0 MHz
.subckt SonData 1 2
C1 1 0 2.9662883505037pf
C2 2 0 3.7181761877835pf
R1 1 2 53064.266550196
L1 1 3 3489.0103767686nh
RL1 3 2 2.6923714768782
.ends SonData
* Analysis frequencies: 11.0, 12.0 MHz
.subckt SonData 1 2
C1 1 0 2.9692064796869pf
C2 2 0 3.7219410391898pf
R1 1 2 63187.815009912
L1 1 3 3482.6126035446nh
RL1 3 2 2.8368098257633
.ends SonData
* Analysis frequencies: 12.0, 13.0 MHz
.subckt SonData 1 2
C1 1 0 2.9724127469545pf
C2 2 0 3.7260445713257pf
R1 1 2 74271.877736121
L1 1 3 3476.1946692413nh
RL1 3 2 2.9639719590486
.ends SonData
* Analysis frequencies: 13.0, 14.5 MHz
.subckt SonData 1 2
C1 1 0 2.9768707177611pf
C2 2 0 3.7317126565688pf
R1 1 2 89476.377167733
R2 2 0 96155.287991752
L1 1 3 3466.4476075203nh
RL1 3 2 3.0942799812765
.ends SonData
* Analysis frequencies: 14.5, 16.0 MHz
.subckt SonData 1 2
C1 1 0 2.9827748424865pf
C2 2 0 3.7391715019619pf
R1 2 0 86204.906852378
L1 1 3 3456.3063115934nh
RL1 3 2 3.2286871356919
.ends SonData
Thanks
I have done new analysis . :
http://rapidshare.de/files/11468680/rfid3.rar.html
any suggestions, opinions are wellcome .
size 30 x 30 , strip width 10 millstrip gap 10 mill. Done for Philips SLI HF ICODE RFID IC.
Hi Artem -- Just got back from some more travel and have gotten rested up a bit.
I can not answer your manufacturability questions. Those questions should be very easy for your PCB manufacturer to answer. Different PCB manufacturers will have different answers.
As for the SPICE netlist, the
subckt SonData 1 2
line indicates that port 1 will be attached to node 1 and port 2 will be attached to node 2. References to node 0 indicate a connection to ground, you are correct.
Notice how the RL1 increases with frequency. This is proabably due to skin effect. The R1 is very large and proabably has no influence on the circuit. You can remove it by modifying the RMAX parameter. The two capacitors and one inductor are essentially constant with frequency. That means you have a good lumped model up to the highest frequency you analyzed, as long as a kind of average value for skin effect loss is OK.
When you have a good lumped model, the values of the lumped model correspond too what you physically have. You have close to a 3475 nH inductor, etc. Most SPICE lumped models do curve fitting. The lumped circuit that comes out does not look anything like the circuit you analyzed.
In Sonnet, to make sure you have a good lumped model, as described above, must make sure the lumped element values are about the same over the entire band of interest.
Hi Rautio
Thank you for answer. The first netlist was simulated wrong . I mistakenly did set the layer thickness (ground - 0 was OK but 0 - 1 was thin 1mm) very low and enabled high impendance appearance in Spice netlist generation settings. Newly modelled netlist shows consistency you mentioned. Variation 3031 - 3037 nH. This one was simulated using 10 mill for wire width and 10 mill for interwire distance . Just i am curious - can thikness of layer affect induction value - for 1 mm i got 2000nH for this one - 3000 nH ?
.subckt SonData 1 2
C1 1 0 1.028731pf
C2 1 2 0.611176pf
C3 2 0 3.09894pf
L1 1 3 3037.315nh
RL1 3 2 1.175132
.ends SonData
* Analysis frequencies: 11.0, 12.1 MHz
.subckt SonData 1 2
C1 1 0 1.028858pf
C2 1 2 0.648262pf
C3 2 0 3.100026pf
L1 1 3 3035.684nh
RL1 3 2 1.239232
.ends SonData
* Analysis frequencies: 12.1, 13.3 MHz
.subckt SonData 1 2
C1 1 0 1.029964pf
C2 1 2 0.676796pf
C3 2 0 3.102368pf
L1 1 3 3034.126nh
RL1 3 2 1.302037
.ends SonData
* Analysis frequencies: 13.3, 14.62 MHz
.subckt SonData 1 2
C1 1 0 1.030913pf
C2 1 2 0.699761pf
C3 2 0 3.104751pf
L1 1 3 3032.609nh
RL1 3 2 1.359147
.ends SonData
* Analysis frequencies: 14.62, 16.0 MHz
.subckt SonData 1 2
C1 1 0 1.030506pf
C2 1 2 0.719227pf
C3 2 0 3.105877pf
L1 1 3 3031.13nh
RL1 3 2 1.40352
.ends SonData
I did not use em simulation software before. Could we trust the simulation result ?
I mean is the lumped element closed values in frequency range is only condition for correct simulation?
The new simulationn 3D picture is in attachment.
Hi Artem -- Going from 10 mil (about 0.254 mm) thick to 1 mm thick is a big change. If you keep the same 10 mil gap, you need more sheets in the thickness model.
For your original circuit, 10 mil gap and 10 mil thick, usually 2 sheets is plenty. This puts a sheet of current on top and a sheet of current on the bottom to model thickness. We recommend 2 sheets when gap and thickness are about the same.
For 1 mm thick, I think you might need at least four sheets because the gap is 1/4 the thickness. The way to find out is to include more sheets and see if the answer change is significant.
There are many different thickness models, and they all have advantages and disadvantages, as discussed in several threads elsewhere on this forum. Keep in mind that when anyone says that their model is nearly perfect, they are speaking as salesmen, not as engineers.
NEVER trust any software or what any person (including me) says! (In God we trust, all others must have proof!)
First, all EM analyses give the wrong answer. Every single one of them. This is absolutely 100% guarenteed. Anyone who says an em analysis gives the right answer is doing sales. When they say the answer is right to within X%, and they have some data to backup the X%, now they are doing engineering.
The way to figure out X for Sonnet is to repeat the analysis and cut the cell size in half. If you are using thickness, double the number of sheets. The percent difference between the two results is X.
This called convergence analysis. If thickness is important, it is critical to refine the thickness meshing too!
Just had another thought. The box sidewalls might be close enough to lower the resonant frequency. You should also try doubling the size of the box (keep the cell size the same) and see how things change.