# buried trace impedance

• posted

Suppose we have a 6-layer pcb, FR4, 10 mil dielectrics for the sake of argument. Ground plane is Layer 3.

L1 traces are simple microstrip, 20 mils above the plane, so something like Appcad gives the trace impedances, about 87 ohms for a 10 mil trace.

If the L2 trace were pure microstrip, 10 mils wide above a 10 mil diel with no overlay, the impedance would be 66.5 ohms.

But the L2 traces are buried microstrip. They are 10 mils above the ground plane but have 10 mils of epoxy glass on top. So the upper dielectric lowers the impedance.

I coded an equation from a Motorola appnote, but like most simple equations of this sort, it's not very good. For the situation above, a

10 mil buried trace, it predicts an impedance *above* 66.5 ohms!

Wadell's book suggests that the buried impedance will be about 0.85 of the non-overlay impedance, so my L2 trace will be maybe 56 ohms. Is this about right?

Anybody have or know know of a calculator app that does buried traces?

John

• posted

This one does:

If thing get really fancy and you have a buried differential line situation some day:

```--
Regards, Joerg

http://www.analogconsultants.com```
• posted

Thanks, but that's classic stripline, which Appcad or Txline can calculate. Buried microstrip has a ground below but none above...

---------------------------------- L1 (surface) ^ | diel trace v

-----------======---------------- L2 ^ | diel v ================================= L3 ground plane

so the impedance is lower than microstrip but higher than stripline.

John

• posted

Check out the field solver by Polar Instruments. You can download an evaluation from their web site that's fully functional.

John Lark> Suppose we have a 6-layer pcb, FR4, 10 mil dielectrics for the sake of

• posted

Some years ago, knowing how web sites don't always stay up forever, I implemented the equations from the site Jorge mentioned in an Excel spreadsheet, so I can do them locally. In fact, it's a sheet I added to another Excel PCB calculator I picked up off the web. It was put together by Chris Robertson, who has also written a book on printed circuit design. You may be able to find the spreadsheet somewhere on the web, but a Google search just now didn't turn it up for me. Like I was saying, web sites doen't always stay up forever...

By the way, my implementation of the equations, assuming a relative dielectric constant of 4.8 and a trace thickness of 1.4 mils, says the impedance of your buried line should be about 48.4 ohms; and the approximation limits all checked OK. 1.4 mils is a big enough percentage of the 10 mil width that lowering it noticably raises the impedance; half a mil thickness yields 51 ohms from the calculation in my sheet.

Cheers, Tom

• posted

Check out the free student version of Sonnet's EM Software, should be able to do it...

• posted

Oops. How 'bout this one then:

```--
Regards, Joerg

http://www.analogconsultants.com```
• posted

My usenet is on the fritz again, hopefully this gets through. Here is the buried version:

```--
Regards, Joerg

http://www.analogconsultants.com```
• posted

Cool. It says 60 ohms for my case, versus my estimate of 56. That's "precision" in this business.

It also appears that I can save the page from Firefox, as html, and run it forever offline.

Thanks!

John

• posted

That is close. My guess is that controlling the epsilon of the material could throw you a curve if it has to be more precise than that.

Don't know about forever. Might not be so easy to do 40 years from now...

Most welcome.

```--
Regards, Joerg

http://www.analogconsultants.com```
• posted
[...]

Here is the link to all of them:

Just keep in mind that the underlying formulae are approximations.

They also have useful inductance calculators:

```--
Regards, Joerg

http://www.analogconsultants.com```
• posted

On Tue, 19 Dec 2006 17:49:25 -0800, John Larkin wrote in Msg.

Don't know about Firefox, but in Opera you can even save entire webpages including images as a /single file/ for off-line use. Way cool!

robert

• posted

Supossing L1 is top layer. You can't control precise the epsilon between the L1 and L3. L1-L2 stack could be a factory made with reproducible epsilon but L2 to L3 will be prepreg with variable epsilon on the same width. Talking about impedance of 66.5 is pure fiction while the epsilon could be any between 3.9 and 4.8 You have to think exactly how is defined the layer stack. On the other hand L1must have ground too.

greetings, Vasile

• posted

John Larkin a écrit :

UltraCad does it. It gives me 52.4 for an Er of 4.7

```--
Thanks,
Fred.```
• posted

On 19 Dec 2006 16:56:18 -0800, "Tom Bruhns" Gave us:

Make sure that the PCB has been baked out for best results, as they are hygroscopic.

• posted

"Fred Bartoli" wrote in message news:458902e6\$0\$2164\$ snipped-for-privacy@news.free.fr...

Hi Fred:

I suspect you used 1 oz copper. As this is an inner layer, it is likely to be 0.5 oz, which will raise the impedance to 54.6 ohms according to UltraCalc.

The corresponding impedances using the emclab calculator are

59.6/61.6 ohms.

YMMV

Regards Ian

Who was it complaining about the use of Imperial units? ;-)

• posted

Firefox saves it as "embedded.html" with a folder on the side "embedded_files", which holds the graphics and such. It all works just like the web page.

For the simple microstrip, various impedance calculators seem to agree pretty well for narrow traces, but begin to differ seriously for wider ones. For a 200 mil wide trace over a 20 mil dielectric, some report the impedance as negative!

John

• posted

That's in the ballpark of the various calculations. But I'm a little wary of Ultracad... I've seen some nonsensical results from some of their programs. Their web site indicates they may have improved things, but it would cost \$150 to find out. Appcad or Txline are fine for all common cases except inner-layer traces.

John

• posted

That would make Maxwell turn around in his grave. Watch out for range and ratio limits. For any kind of approximation the authors should be honest and state those right next to the formula or calculator (Missouri-Rolla does).

```--
Regards, Joerg

http://www.analogconsultants.com```
• posted

Ian a écrit :

Yup. I just took the same figures as Tom for comparison purpose.

:-)

```--
Thanks,
Fred.```

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.