Microstrip calculators inaccurate?

[PetAd]

Hi there!

For a school project I need to calculate microstrip impedance and write a little essay. I checked the IPC-2141A guide for the calculations, and the errate too. I made an m file for Matlab to calculate the impedance, but I need to check the results. And here comes the problem. The internet is full of calculators, but with the same numbers the results are very different.
My numbers are:
Dielectric constant =  4.58
Dielectric height     =  10mil
Signal line width     =  12 mil
Copper thickness   =  1Oz = 1.4mil
I gathered the following results from different calculators:

Saturn PCB Design Toolkit
 57.99Ohm
http://www.microwaves101.com/encyclopedias/microstrip-calculator
 63.74Ohms
http://www.multek.se/engelska/engineering/pcb-structures-2/microstrip-impedance-calculator-2
 60.185Ohms
http://www.finetune.co.jp/~lyuka/technote/ustrip/
 60.93Ohms
http://chemandy.com/calculators/microstrip-transmission-line-calculator-ipc2141.htm
 60.26Ohms
http://www.eeweb.com/toolbox/microstrip-impedance/
 46.6Ohms
My result is 69.28Ohms, more likely wrong

[KJDS]

The high end microwave simulators use Hammerstad and Jensens equations.

Covered in some detail here.

http://brd4.braude.ac.il/~alomes/Introduction2EMC/Applicational_Notes/Single%20Microstrip%20Line%20by%20Stefan%20Jahn%202007.pdf

[nctnico]

For general purpose use (to calculate 50 Ohm traces) I often use Wcalc:
http://wcalc.sourceforge.net/cgi-bin/microstrip.cgi

The online calculators are only somewhat accurate for 50 an 60 Ohm traces. Beyond that you need the Hammerstad and Jensens equations KJDS pointed to.

[KJDS]

IIRC Appcad uses H & J

[TonyStewart]

1. Dielectric constant is a large 10 or 20% depending on matl, so TDR testing at factory is advised.
2. Dielectric constant not constant, declining slowly above where loss tangent starts to rise. e.g. 1GHz for FR4
3. Track thickness adds another tolerance to width called etchback factor.

[Carrington]

Marked in red, without tolerances. Marked in dark blue, with a tolerance of a little more than 10% for Er:

[PetAd]

The high end microwave simulators use Hammerstad and Jensens equations.

Covered in some detail here.

http://brd4.braude.ac.il/~alomes/Introduction2EMC/Applicational_Notes/Single%20Microstrip%20Line%20by%20Stefan%20Jahn%202007.pdf

I made calculations with Matlab and H&J methods, now i got
Z0 = 63.45Ohm
Effective permittivity  = 3.325
Eff. line width is 12.1259 mil

Wcalc result is 60.4811 Ohm, closer with this method, but not the same again.

1. Dielectric constant is a large 10 or 20% depending on matl, so TDR testing at factory is advised.

Where I work, there are some good DPO, maybe i can check some pcb-s.

[T3sl4co1l]

I would be more impressed with the assignment if it involved an analytical or computational approach to derive the formulas from theory, and/or actual experimental tests involving fabricated boards of various geometry (assuming of course you have access to the VNA fast enough and accurate enough to perform the requisite tests).

Neither of which sounds like "short essay" material.  Indeed, masters' theses have been written on less.  So...

On the plus side, for such a, well, weak paper -- on an otherwise rather advanced and involved topic -- whether your conclusion is ultimately correct or not, probably won't matter very much, just as long as you've presented your case clearly from thesis to conclusion, based upon the references used to inform that conclusion.  Or to put it slightly differently: even if you use shitty references, as long as your conclusion is right, it's only partially your fault (choice of reference, or failure to sufficiently research the subject, is still your fault, but it's not usually as big a part as doing the writing itself).

Tim

[westfw]

When I was looking at some TI C3000 designs, I couldn't get any of the online calculators to match up with the dimensions that TI had in their reference design, either :-(