| Author |
 Topic  |
|
|
tomed
France
28 Posts |
 Posted - Jun 04 2015 : 18:48:56
|
Dear Enrico,
Please i have got an issue and i would need your help.
Are you agree with this meshing parameters:
E63 N125 N126 w=0.7 h=35.0e-3 nhinc=3 nwinc=5 rh=10 rw=15
Also, when i changed the length of the trace by changing the position of nodes (125 or 126), i got different results by using the same meshing parameters. Is that correct?
In fact, i've learned that for transmission lines the per-unit-length parameters can be calculated on a cross-section of the line (2D).
By the way why Fast Henry 2 gave different results for the same geometrical and meshing characteristics except the length of this line?
Thank you
Appreciate your time and continuous support
Best Regards,
Tomed |
|
|
Enrico
545 Posts |
Posted - Jun 09 2015 : 19:25:00
|
quote:
Are you agree with this meshing parameters:
E63 N125 N126 w=0.7 h=35.0e-3 nhinc=3 nwinc=5 rh=10 rw=15
The meshing parameters are meaningless if you don't refer them to a specific frequency. At low frequency w.r.t. skin effect, you don't need any meshing, for instance. As a general rule, anyway, I can say that rh and rw values you used are too high ratios. Also, read on..
quote:
Also, when i changed the length of the trace by changing the position of nodes (125 or 126), i got different results by using the same meshing parameters. Is that correct?
Yes this is absolutely correct, as FastHenry2 is a 3D solver, not a 2D solver. So you see the different effect of the line end points, i.e. the field is not uniform in the plane orthogonal to the line, and more so near the ends of the line.
quote:
In fact, i've learned that for transmission lines the per-unit-length parameters can be calculated on a cross-section of the line (2D).
This is true only provided that your cross-section is small enough w.r.t. the maximum wavelength of interest (that, note, is NOT the maximum frequency of a digital signal, but the maximum frequency of the harmonic you are still interested in. A square signal in principle has infinite harmonics, in the real world it will depend on how 'square' it is).
If this condition is respected, you can 'simulate' the line with a very long line (w.r.t. the cross section) and calculate the per-unit-length parameters from this line, dividing by the length of this test line. E.g. if your line is 100mm long, you divide by 0.1 to have the H/m of the transmission line.
quote:
By the way why Fast Henry 2 gave different results for the same geometrical and meshing characteristics except the length of this line?
See above
Best Regards,
Enrico
|
 |
|
|
tomed
France
28 Posts |
Posted - Jun 09 2015 : 22:08:33
|
Dear Enrico,
Thank you for your feedback.
Sorry for the missmatch, the maximum frequency of interest is 500MHz so delta min (skin effect) = 2.96µm.
I'll continue to review and understand your explanation and will recome back to you.
Appreciate your time and continuous support
Best regards,
Tomed |
|
|
|
tomed
France
28 Posts |
Posted - Jun 10 2015 : 17:10:47
|
Dear Enrico,
Thank you for your help.
I've reviewed your feedback and have some questions.
By .equiv and .external commands, we define a loop for one geometry of trace(width, thickness, length) and define one network's port respectively.
quote:
By the way, changing the length of this trace, will impact the surface of the loop , is that correct?
quote:
What is the impact of changing the surface of this loop on R&L calculation?
quote:
Is there any link with R&L changing by modifying the length of this trace?
Thank you
Appreciate your time and continuous support
Best Regards,
Tomed |
|
|
|
Enrico
545 Posts |
Posted - Jun 10 2015 : 19:27:00
|
If you need the per-unit-length parameters, you must close the strip in a loop over the gnd plane - sorry I gave it for granted after the explanations I gave you also on other threads. All my previous remarks apply in this case.
You can check a similar example case in my old thesis' work, under the 'literature' page, "Caratterizzazione Elettromagnetica di Packages Ultraminiaturizzati Applicati a Memorie Flash-EEPROM", pages 29-30.
It is in Italian, but you may get the gist of it. In those pages I evaluated FastHenry against some inductance values from a reference paper (you find it cited) at different frequencies (i.e. where skin effect plays a role)
Best Regards,
Enrico |
|
|
|
tomed
France
28 Posts |
Posted - Jun 16 2015 : 23:02:06
|
Dear Enrico,
Thank you for your feedback.
Seems your PH D topic has dealt with higher frequencies than my study.
quote:
Is there any way to get the meshing parameters of the example of your PH D?
Appreciate your time & support,
Thank you
Regards,
Tomed |
|
|
|
Enrico
545 Posts |
Posted - Jun 18 2015 : 18:10:56
|
Could not find the original file any more. I quickly assembled a simulation input file for you, based on that geometry:
* A FastHenry analysis of a microstrip, as shown at page 48
* of E. Di Lorenzo thesis's work "Caratterizzazione Elettromagnetica
* di packages ultraminiaturizzati applicati a memorie Flash-EEPROM",
* Politecnico di Milano, 1998
*
* Parameters are:
* microstrip width: 200um, thickness 10um
* gnd plane widht: 2mm, thickness 10um
* gap from top of gnd plane to bottom of microstrip: 100um
* conductivity is 5.6e7 1/(m*ohm)
*
* Note that this is a 3D simulation, to find the parameters per unit length
* of a microstrip. So we defining a microstrip long enough for the end of line
* effects to be negligible, in this case 10mm is enough
* To convert from H to H/m and from R to R/m, you must then scale
* the results multiplying by 100 (from 10mm to 1m)
.Units um
* Make z=0 the default z coordinate and define the default conductivity.
* Note that the conductivity is in units 1/(um*Ohms), not 1/(m*Ohms)
* since the default units are microns, so sigma is scaled accordingly.
.Default z=0 sigma=5.8e1
* The nodes of the gnd plane (z=0 is the default)
N1 x=0 y=0
N2 x=0 y=0 z=10000
* The nodes of the ustrip
N3 x=0 y=110
N4 x=0 y=110 z=10000
* The segments connecting the nodes
* GND plane (being in fact a 2D simulation, no need of any
* filament along the z direction, so this actually is a segment,
* and not a full-blown gnd plane definition)
E1 N1 N2 w=2000 h=10 nhinc=7 nwinc=100 rw=1
* ustrip
E2 N3 N4 w=200 h=10 nhinc=7 nwinc=15
* connections at far end
.equiv N2 N4
* define the ports of the network
.external N1 N3
* Frequency range of interest.
.freq fmin=10e3 fmax=10e9 ndec=1
* All input files must end with:
.end
I also strongly recommend you to read also the paper from which I took the example:
A. R. Djordjevic, T. K. Sarkar, "Closed-Form Formulas for Frequency-Dependent Resistance and Inductance per Unit Length of Microstrip and Strip Transmission Lines" IEEE Trans. on Microwave Theory and Techniques, vol. 42, no. 2, February 1994.
as this paper quite clear in explaining the behavior of the microstrip when frequency increases, and the role of skin effect.
Best Regards,
Enrico
|
|
|
|
tomed
France
28 Posts |
Posted - Jun 19 2015 : 17:04:11
|
Dear Enrico,
First of all, thank you so much for your great help.
I've took this paper also, i will study it as well.
I've implement your code into Fast Henry, and i've seen that the track is centered over the ground plan.
quote:
Is the segment centered around one value without using the synthax G for the ground plan definition?
Appreciate your time and continuous support
Best Regards,
Tomed |
|
|
|
Enrico
545 Posts |
Posted - Jun 19 2015 : 18:05:22
|
I'm not sure I fully understand your question. Yes, I'm not using G definitions but only a segment, as commented in the file, as the current flow is essentialy along the Z axis; and yes, segments are centered at the nodes defining the end points, so you must consider the thickness when calculating the gap between the gnd plane and the line.
Am I answering you?
Best Regards,
Enrico |
|
|
|
tomed
France
28 Posts |
Posted - Jun 20 2015 : 18:59:43
|
Dear Enrico,
Thank you for your quick feedback.
I guess we're on the right direction to make me understand as well.
I'll try to be more clear regarding my concerns:
In your input file:
quote:
1)
x direction is for the width
y direction is for the thickness
z direction for the length
Am i right?
2)
the maximum frequency of interest is 10GHz, so the minimum wavelength is 30mm, you've chosen a length of 10mm, which is minimum wavelength/3,
quote:
why have you chosen 10mm for the microstrip length?
quote:
why the length of the ground plan is also 10mm?
3)
for the segment position (ground plane and also microstrip),
ground plane, x range [from -1000µm to +1000µm] & y range [from 0 to 10µm]
microstrip, x range [from -100µm to +100µm] & y range [from 110µm to 120µm]
quote:
Do you agree with the range definition?
quote:
why the center of the Cartesian coordinate system is fixed to the center of the ground plane width?
4)
regarding nwinc & rw definition for the ground plane , the default value is 2,
quote:
what happened when we fix rw=1?
quote:
Do all the segments define get the same width?
quote:
why have you choose a even number nwinc=100?
5)
As you know i'm studying microstrip line, and i'm interesting to get the per unit length (p.u.l) parameters.
With your input file that i will give my best to understand all, i've learnt that no need to define real ground plan(with the letter G), just need to define a segment as ground plan and also define the geometry of the microstrip, to stay in 2D study since the p.u.l parameters are calculated in 2D with the quasi-static assumption.
By using Fast Henry, to return back to p.u.l parameters, i have to divide the results obtained by the length define.
quote:
Am i on the right way to understand?
quote:
how to correctly define the microstrip length(when we do not know it) for Fast Henry input file, since we need it to return back to p.u.l parameters?
Also, in the paper "A. R. Djordjevic, T. K. Sarkar, "Closed-Form Formulas for Frequency-Dependent Resistance and Inductance per Unit Length of Microstrip and Strip Transmission Lines" IEEE Trans. on Microwave Theory and Techniques, vol. 42, no. 2, February 1994" you've cited, i haven't found the length of the ground plane and also the length of the microstrip.
quote:
Is there any assumption regarding this length for Fast Henry input file?
quote:
Do we get more accuracy on p.u.l parameters by choosing a greater length?
I hope my concerns is right now as clear as possible, if not, please would you like to let me know?
Thank you
Appreciate your time and continuous support
Best Regards,
Tomed
|
|
|
|
Enrico
545 Posts |
Posted - Jun 23 2015 : 08:43:50
|
quote:
1)
x direction is for the width
y direction is for the thickness
z direction for the length
Am i right?
Yes you're right. Actually in my previous post there was a typo (now corrected), I wrote 'Y' instead of 'Z'.
quote:
2)
the maximum frequency of interest is 10GHz, so the minimum wavelength is 30mm, you've chosen a length of 10mm, which is minimum wavelength/3, why have you chosen 10mm for the microstrip length? why the length of the ground plan is also 10mm?
These are actually two questions.. the answer to the first is that you can choose any length you want, provided the strip is long enough in the Z direction to actually appear as a 2D structure, that would have infinite length along Z. However it cannot be too long, otherwise simulation times in 3D would be not practical. This is a good compromise. Note that the ground plane must have the same length exactly because you are simulating the 2D structure in 3D; and the gnd plane is not an optional element of the microstrip, it is part of the microstrip itself.
quote:
3)
for the segment position (ground plane and also microstrip),
ground plane, x range [from -1000µm to +1000µm] & y range [from 0 to 10µm]
microstrip, x range [from -100µm to +100µm] & y range [from 110µm to 120µm]
:Do you agree with the range definition?
why the center of the Cartesian coordinate system is fixed to the center of the ground plane width?
If you mean, where the line and the ground plane are centered and extend, I would say:
ground plane, x range [from -1000µm to +1000µm] & y range [from -5 to +5µm]
microstrip, x range [from -100µm to +100µm] & y range [from 105µm to 115µm]
The actual position in space is irrelevant, only the distances are important (i.e. relative positions of the elements). I choose those coordinates only out of my personal convenience. You could center the origin between the strip and the gnd plane, or everywhere else.
quote:
regarding nwinc & rw definition for the ground plane , the default value is 2,
what happened when we fix rw=1?
Do all the segments define get the same width?
why have you choose a even number nwinc=100?
rw=1 would mean that all segments would have the same width. This is why I didn't care to use an odd number for nwinc in that case.
quote:
5)
As you know i'm studying microstrip line, and i'm interesting to get the per unit length (p.u.l) parameters.
With your input file that i will give my best to understand all, i've learnt that no need to define real ground plan(with the letter G), just need to define a segment as ground plan and also define the geometry of the microstrip, to stay in 2D study since the p.u.l parameters are calculated in 2D with the quasi-static assumption.
By using Fast Henry, to return back to p.u.l parameters, i have to divide the results obtained by the length define.
Am i on the right way to understand?
how to correctly define the microstrip length(when we do not know it) for Fast Henry input file, since we need it to return back to p.u.l parameters?
Yes. See the answer to point 2, also regarding your last point.
Best Regards,
Enrico
|
|
|
| |
Topic |
|
|
|
impact of length on per-unit-length calculation
