George,
I'd have to agree with Michael. There are so many variables that go into
the puzzle to give you accurate TDR "results", that you really have to get
specific to rationalize what you're seeing.
For example:
What frequency of operation were you theorizing the Dk to? And was
that the frequency you did the TDR to? What is the actual material you
have for dielectic? You only said high Tg expoxy, that still gives you a
few choices dependant on resin vs. glass contents. (You're right in the
suppliers "estimate", ever seen actual test data for the material?) What
kind of test pattern was the TDR test done on? (IPC specified?) Or, was
it done on traces on the actual board? (Bad idea, unless they were
designed to meet IPCs TDR requirements)
All these add up, including the dielectric thickness, etch width (BTW, I
would have thought you would've come up with 8.5 mils or 9.5 for the
trace width, not an exact 9 mils), etc, etc,etc to give you what you've
found.
What kind of certification did the fab house give you with the boards?
Have they any explaination for you? I'm sure they must have someone
who can answer so you variations.
Good luck,
Mitch Morey
Sr PCB Designer
NSI Communications
San Diego, CA
[log in to unmask]
>>> "Mcmaster, Michael" <[log in to unmask]> 08/24/98
07:47am >>>
The problem is What's Theoretical? Every published equation I've ever
seen is only an approximate solution to solving the theoretical
solution. Every equation out there should have an accompanying
statement that describes where it is valid. Mathematically, the best
results I've seen come from using a field solver. I use a program
called Apsim RLGC, but there are many others out there.
Even if you are using a field solver, you need to make sure your inputs
are valid. Even a 100% theoretical solution gives the wrong answer if
the input variables aren't correct. Most oscilloscopes for measuring
TDR have a sampling edge rate of approximately 10 GHz. The dielectric
constant for FR4 you used is the value at 1 MHz. The exact value of the
dielectric constant of FR4 at 10 GHz is difficult to come by. There a
number of test methods out there all of which seem to give conflicting
results. Empirically, I've found that at dielectric constant of 3.5 to
3.7 works well in matching measured values. If I model your structure
with a Dk of 4.5, I get 49.9 ohms. With a Dk of 3.7 I get 55.0 ohms.
> ----------
> From: George Franck[SMTP:[log in to unmask]]
> Reply To: TechNet E-Mail Forum.;George Franck
> Sent: Monday, August 24, 1998 5:50 AM
> To: [log in to unmask]
> Subject: [TN] [Fab] Emperical Impedances
>
> Technet,
>
> Is it typical to have actual impedance reading which vary (10% or so)
> from
> those predicted from theoritical data?
>
> For example, in a cross section of a Stripline impedance coupon, I
> measured
> the following dimensions.
> ========= Ground plane
> ========= 11.9 mil Dielectric (4.5 Dk)
> ========= Circuit Layer 1.2 mil thick, 9.0 mil wide
> ========= 11.8 mil Dielectric (4.5 Dk)
> ========= Ground plane
>
> Material is High Tg Epoxy. The Dk is supplier estimate. Theory would
> predict a value of 49 Ohms. Actual TDR measurements show an
impedance
> of
> 54.7.
>
> George Franck
> Raytheon Systems
> Falls Church Virginia
> "My Opinions are just that."
>
>
################################################################
> TechNet E-Mail Forum provided as a free service by IPC using
LISTSERV
> 1.8c
>
################################################################
> To subscribe/unsubscribe, send a message to [log in to unmask] with
> following text in the body:
> To subscribe: SUBSCRIBE TechNet <your full name>
> To unsubscribe: SIGNOFF TechNet
>
################################################################
> Please visit IPC's web site (http://www.ipc.org) "On-Line Services"
> section for additional information.
> For technical support contact Hugo Scaramuzza at [log in to unmask] or
> 847-509-9700 ext.312
>
################################################################
>
################################################################
TechNet E-Mail Forum provided as a free service by IPC using LISTSERV
1.8c
################################################################
To subscribe/unsubscribe, send a message to [log in to unmask] with
following text in the body:
To subscribe: SUBSCRIBE TechNet <your full name>
To unsubscribe: SIGNOFF TechNet
################################################################
Please visit IPC's web site (http://www.ipc.org) "On-Line Services"
section for additional information.
For technical support contact Hugo Scaramuzza at [log in to unmask] or
847-509-9700 ext.312
################################################################
################################################################
TechNet E-Mail Forum provided as a free service by IPC using LISTSERV 1.8c
################################################################
To subscribe/unsubscribe, send a message to [log in to unmask] with following text in the body:
To subscribe: SUBSCRIBE TechNet <your full name>
To unsubscribe: SIGNOFF TechNet
################################################################
Please visit IPC's web site (http://www.ipc.org) "On-Line Services" section for additional information.
For technical support contact Hugo Scaramuzza at [log in to unmask] or 847-509-9700 ext.312
################################################################
################################################################
TechNet E-Mail Forum provided as a free service by IPC using LISTSERV 1.8c
################################################################
To subscribe/unsubscribe, send a message to [log in to unmask] with following text in the body:
To subscribe: SUBSCRIBE TechNet <your full name>
To unsubscribe: SIGNOFF TechNet
################################################################
Please visit IPC's web site (http://www.ipc.org) "On-Line Services" section for additional information.
For technical support contact Hugo Scaramuzza at [log in to unmask] or 847-509-9700 ext.312
################################################################
|
