POLAR SI9000e and Sarturn PCB

[Carrington]

Hello!

I tried a demo of the last version of POLAR SI9000e, it have a utility to estimate the impedance of a via (via pad/anti-pad coaxial calculation). When I compare the result with the same utility in the sarturn PCB software I get a very different value, to the same via characteristics.

At least similar results should be obtained, true? Why does it happen?

SI9000e:--------------------------------------------------------------------
VP=0,2540mm (via pad diameter)
AP=1,0160mm (anti-pad diameter)
Er1=4,6000 (substrate dielectric)

Zo:38.74 Ohm
--------------------------------------------------------------------------------


Polar Application Note AP8178: http://www.polarinstruments.com/support/si/AP8178.html

Saturn PCB:-----------------------------------------------------------------
Via Hole Diameter=0,254mm
Internal Pad Diameter=0,255mm
Ref. Plane Opening Diam.=1,016mm
Remaining parameters by default.

Zo:99.21 Ohm
--------------------------------------------------------------------------------


Kind regards.

[Carrington]

Long time ago I contacted with Polar and Saturn, about this issue but I never received a response.
Surely both are correct, and the result does not match because they are estimating different parameters. But I don't get.   

[thm_w]

Via length is 0mm in the first one and 1.575mm in the second screenshot, right?

[penfold]

The relative dielectric constant is also different between the two

[jahonen]

Biggest fundamental difference between them is that Saturn PCB software is not a field solver, and uses an approximated formula for impedances so results are more or less approximate. Whereas Polar Instruments uses a field solver to calculate things by numerical solving of Maxwell's equations so it should produce more accurate results.

Regards,
Janne

[Carrington]

Via length is 0mm in the first one and 1.575mm in the second screenshot, right?

POLAR SI9000e for the via pad/anti-pad coaxial calculation not seem to take into consideration the length of the VIA

[Carrington]

The relative dielectric constant is also different between the two

Er is 4.6 in both (Polar and Saturn). Via check and via pad/anti-pad coaxial calculation appear to be unrelated. Via check/via Stub check provides a simple color coded go/no go check on the potential for signal distortion of a via or stub.

[Carrington]

Biggest fundamental difference between them is that Saturn PCB software is not a field solver, and uses an approximated formula for impedances so results are more or less approximate. Whereas Polar Instruments uses a field solver to calculate things by numerical solving of Maxwell's equations so it should produce more accurate results.

Regards,
Janne

Although that, there's something wrong, there is a big difference in the estimated values.

[Carrington]

Via length is 0mm in the first one and 1.575mm in the second screenshot, right?

POLAR SI9000e for the via pad/anti-pad coaxial calculation not seem to take into consideration the length of the VIA.
But you're right, I think they are two different views of the same problem.

[Carrington]

All this is starting to fit. They are two different views of the same problem (well, not exactly the same).
 - The first (Polar) only estimate the impedance of the via pad/anti-pad (coaxial) and separately estimate if the length of the via is OK for for a given (rise time, frequency or Bit rate).
 - The second (Saturn) seems to estimate the impedance of the total via. With all the changes over the structure of the same via (as a connector pin)? Or maybe it considered that the structure does not change and therefore only estimate that parameter for a multi layer board.
It must be something like that (but I have the feeling that still does not fit completely).


If someone has another point of view, go ahead, I'm all ears.
This can help:
http://www.polarinstruments.com/support/si/AP8178.html
http://www.polarinstruments.com/support/si/AP8166.html

Excuse my English, is not quite correct, sorry. And thanks to all.

[Carrington]

Is this the point of view of Polar respect to the vias? We must assume that for a multilayer board Antipad 1 is equal to 2?


Note: A, B, and C. Length of the section with impedance mismatch.


What is the geometry that Saturn considers to estimate the impedance?

[Carrington]

I am frequently asked why the Saturn PCB Toolkit has a different Z number here or there with respect to Polar and the answer is...I have no idea. I do not know what Polar uses to calculate impedance's on any of their tools. I use stated formulas given by the IPC or other industry standards is really all I can say. A board house called me once claiming that the Saturn PCB Toolkit's numbers were closer than what they got from Polar in their actual measurements.

Hi!
I gather that SPCD should be something like Satur PCB Designer. Woo hoo... Anyway, you are welcome to the forum. 
Thank you very much for your reply, but could you be a little more specific. For example: Where I can see these formulas?

[SPCBD]

Not sure where the rest of my reply went:

I am sometimes asked why the Saturn PCB Toolkit has a different Z number here or there with respect to Polar and the answer is...I have no idea. I do not know what Polar uses to calculate impedance's on any of their tools. I use stated formulas given by the IPC or other industry standards is really all I can say. A board house called me once claiming that the Saturn PCB Toolkit's numbers were closer than what they got from Polar in their actual measurements so they are pretty accurate.
As far as the via impedance, I do not see how Polar is correct but maybe I'm just not familiar enough with their tool. Via capacitance is dependent on the via pad with respect to a reference plane opening, this can be thought of as a parallel plate in a sense. Via inductance however is dependent on the hole diameter and I do not see where Polar is using this information in its calculations. I use an approximated via inductance formula but it shouldn't be that far off as I have modeled the more involved formulas and didn't see much of a change. Via Z is simply Sqrt(ViaL/ViaC).
Ken

[Carrington]

As far as the via impedance, I do not see how Polar is correct but maybe I'm just not familiar enough with their tool. Via capacitance is dependent on the via pad with respect to a reference plane opening, this can be thought of as a parallel plate in a sense. Via inductance however is dependent on the hole diameter and I do not see where Polar is using this information in its calculations. I use an approximated via inductance formula but it shouldn't be that far off as I have modeled the more involved formulas and didn't see much of a change. Via Z is simply Sqrt(ViaL/ViaC).
Ken

Ok, I now understand a little more. Thank you very much. 
"Sqrt(ViaL/ViaC)" So then Saturn estimates Z value in a similar way to a conecctor's pin.

[SPCBD]

Correct, there is nothing really special about a standard via. It acts just like a connector pin electrically.

[Carrington]

Correct, there is nothing really special about a standard via. It acts just like a connector pin electrically.

I'll try that Polar also get involved in this. I hope to have luck.

@ SPCBD, Thank you very much again.


More info:
http://www.saturnpcb.com/toolkit_help.htm ->

-> Via Capacitance:
This is the approximated capacitance of the via with the given inputs. This value is only available for multi layer vias as it uses the internal pad and reference plane opening for the calculation.

-> Via Impedance:
This is the approximated AC impedance of the via with the given inputs. This value is only available for multi layer vias as it uses via capacitance for the calculation. via_zo = sqrt(via_l/ (via_c * 0.001))

In the link above is all that I needed to know about Saturn. Why I did not look this before? 

[KenPolar]

The Polar Si9000e is a 2-D frequency dependent field solver used to accurately compute frequency dependent parameters such as Z, R, L, G, C, dielectric and copper losses for trace transmission lines on printed circuit boards. For vias, however, accurate calculations require a 3-D solver. No exceptions. This is because via environments are not uniform along their length. 
So if you want impedance for a via, you have three choices open to you;
 * Guess. Not a good choice.
 * Use a 3-D solver. If you have one (expensive) and if you can drive the solver to define the structure (complex) and if you have the time available, then this is your choice for accurate 3-D characteristics.
 * Or, as Eric Bogatin says, "Sometimes a good approximation today is worth more than an accurate answer tomorrow." And that's what the Si9000e gives you, a good approximation of impedance.
So you make the choice between these three.
As for the basis of the calculation, Si9000e users will see an "i" information button to the right of the calculation leading to a website description of that.
As for why other tools provide different answers... You'll have to ask them that question, what is the basis of their calculation etc. We don't comment on other suppliers' tools, only on our own.
Incidentally, Si9000e users. Look at the Via Check tool above the via impedance calculator (it also has an "i".) This will give you a very good clue as to whether you really need to concern yourself with via impedance - before you spend a gazillion dollars and a day of your life on a 3-D solver, unnecessarily.
Lastly, to address the comment about contacting Polar with questions; simply visit our website www.polarinstruments.com and look to our world-wide locations listed near the bottom of the index column on the left of your screen. Use the phone and a helpful human will answer you immediately during office hours. No menus and no wait times, we actually DO welcome your calls and questions.  :-)
I hope this helps and thanks for reading. Ken Taylor.

[Carrington]

Ken Taylor is the president of Polar Instruments Inc. Awesome! 
Thank you very much for your reply and welcome to the eevblog forum.


Ken Taylor: http://video007.com?725

[EEVblog]

Ken Taylor is the president of Polar Instruments Inc. Awesome! 
Thank you very much for your reply and welcome to the eevblog forum.

Yes, it's awesome when the companies respond and interact with the community on the forum level, and to do that at the President level gets Polar a big 

[Carrington]

About vias there are more articles, but these are interesting IMHO:

http://m.eet.com/media/1135917/324403.pdf
http://www.lamsimenterprises.com/SPI2008_43_Rel_Permittivity_Var_Via_Structure_Poster.pdf
http://www.ieice.org/proceedings/EMC09/pdf/22Q1-3.pdf

[Carrington]

Field solver, how closely are the results:



Amazing, Polar is on the average.

Source: http://www.alspcb.com/pdfs/controlled_impedance.pdf

[xzswq21]

Hi all
I have a question!
in a 4 layers PCB, we have two Prepreg and one Core. for example the dielectric coefficient for the Prepreg is 4.2 and for the core is 4, now how should I calculate the impedance of a via?!

[rx8pilot]

I just started a trial of Polar Si8000 - really nice.

Ken - if you are still listening....

Are you willing to consider a subscription approach to these pieces of software? When I need it, I really need it - but I would guess that may only be for 5-6 designs per year. Those designs are likely to be crammed into a relatively short period of time. The cost is non-trivial for a tool that we will use so infrequently.

It is really easy to use and offers the designer a ton of structure options. As a periodic user, it is hard to justify the cost.