Content-Type: text/plain;
charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
C.C.
=20
This data looks like it was derived from simulations. The transitions =
from the start of the data and at the junctions of the different =
impedances have no aberrations at all. And I see differences in the 85 =
Ohm section. When it is imbedded between the two 100 Ohm segments its =
impedance is essentially flat over its length. When it is pulled out it =
has a change of 1 Ohm over its length. Whereas both 100 Ohm sections =
have a 1 Ohm change. Why the difference? If it has enough losses to =
show a 1 Ohm change of impedance in one plot I would expect to see it in =
the other.
=20
Do you have any data that shows results where the ringing of the launch =
continues to the DUT section? As you said, tune out bad launches?
=20
Tom Dagostino
=20
Teraspeed Labs
9999 SW Wilshire Street
Suite 102
Portland, OR 97225
=20
tom@xxxxxxxxxxxxxxxxx=20
www.teraspeedlabs.com <http://www.teraspeedlabs.com/>=20
=20
971-279-5325 office
503-430-1065 cell
=20
From: C.C. Hwang [mailto:cchwang2013@xxxxxxxxx]=20
Sent: Tuesday, May 05, 2015 4:48 PM
To: Alfred P. Neves
Cc: Tom Dagostino; jun zhang; si-list; istvan.novak@xxxxxxxxxxx
Subject: Re: [SI-LIST] Re: some questions about a PCB test coupon to =
measure the impedance curve and crosstalk curve
=20
Hi Al,
=20
I put one test case under =
http://ataitec.com/test/impedance_test_case.zip
=20
The DUT (85 ohm differential) sees 2 ohm difference with or without 1.5" =
100 ohm differential trace before it.
=20
Ching-Chao Huang
=20
On Tue, May 5, 2015 at 1:50 PM, Alfred P. Neves <al@xxxxxxxxxxxxxxxxx> =
wrote:
With .07dB/inch/GHZ typical loss you have approximately 2dB of loss at =
20GHz for a 1.5inch and this loss impacts a peeled impedance profile by =
2ohms, really? I doubt it. =20
=20
As for =E2=80=9Ctuning=E2=80=9D out the launch do you have any organized =
data using carefully based standards or channels to prove that claim? =
Again, most launch designs will resonate past the .6inch electrical =
distance, so it will be very difficult to de-embed.
=20
We check de-embedding on our Channel Modeling platforms using THRU =
structures with intentional excess capacitive and inductive pathology, =
then compare it to pristine launch design. Most of the de-embedding =
schemes except measured-modeled (see Keysight/WRT 32G Test fixture =
Tutorial on our website) and the new Simbeor THz approach simply has =
major issues with crappy or even mediocre launch designs (Yuriy has data =
and reports to support that and has done an incredible amount of work =
validating his EDA tool). A good launch is roughly defined to have =
better than -10dB of return loss at Fstop.
=20
All de-embedding schemes are compromises and have an effective bandwidth =
efficiency, it would be helpful to outline the limitations and =
assertions with hard data and reports.
=20
- Al
=20
=20
=20
Products for the Signal Integrity Practitioner
Alfred P. Neves
Chief Technologist
=20
Office: 503-679-2429
www.wildrivertech.com
=20
On May 5, 2015, at 1:21 PM, C.C. Hwang <cchwang2013@xxxxxxxxx> wrote:
Hi Tom,
If the launch and lead-in trace have little discontinuity and loss =
(i.e., ideal 50 ohm lossless transmssion line), then indeed TDR will =
give true impedance reading at DUT. The ~2 ohm increase for ~1.5" =
microstrip attenuation is from many typical designs we saw. One of =
ISD's main advantages is that it can "tune out" bad launches.
Ching-Chao Huang=20
On Tue, May 5, 2015 at 11:15 AM, Tom Dagostino <tom@xxxxxxxxxxxxxxxxx> =
wrote:
Ching-Chao
As Al Neves pointed out if the DUT is swamped out by ringing from poorly =
designed launches or the transitions from the co-planar to DUT you will =
not be able to de-embed the fixture from the DUT.
I'm not sure why you predict the DUT impedance will be higher for TDR =
based measurements, especially after a short 1.5" lead-in. If there are =
low discontinuities and losses there should be no issue. If the =
discontinuities are bad then some well-known techniques can be employed =
such as the peeling algorithms in IConnect from Tektronix.
Tom Dagostino
Teraspeed Labs
9999 SW Wilshire Street
Suite 102
Portland, OR 97225
tom@xxxxxxxxxxxxxxxxx
www.teraspeedlabs.com
971-279-5325 office
503-430-1065 cell
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] =
On Behalf Of C.C. Hwang
Sent: Tuesday, May 05, 2015 10:40 AM
To: al@xxxxxxxxxxxxxxxxx
Cc: zhangjun5960@xxxxxxxxx; si-list; istvan.novak@xxxxxxxxxxx
Subject: [SI-LIST] Re: some questions abou a PCB test coupon to measure =
the impedance curve and crosstalk curve
Hi Jun,
Your DUT is 0.6" and the rest of fixture is 2", so ISD should give you =
very good de-embedded results. (If your DUT is 0.01", then you'll need =
to be
careful.) We have routinely used ISD to extract a 0.5" to 1" connector =
from a 12" board. On the other hand, your DUT is a short trace (0.6"), =
so it will be affected more by local fiber weave effect. If you are =
looking for good average impedance, you can consider increasing the =
length of your DUT.
Due to fixture's trace attenuation, TDR will give you higher impedance =
at DUT than it actually is. We saw ~2 ohm higher impedance at DUT due =
to ~1.5" lead-in trace attenuation. This can be easily verified through =
simulation. Good de-embedding is the only way to get correct impedance =
of DUT.
Ching-Chao Huang
On Tue, May 5, 2015 at 9:01 AM, Alfred P. Neves <al@xxxxxxxxxxxxxxxxx>
wrote:
A typical poorly designed SMA resonates for >100psec, and since =1=C3=A2=E2=82=AC
of microstrip prop delay is around 150psec your going to run into =trouble.
Since your fixture probably goes single ended then coplanar coupledregion
my guess is you have a significant impedance bump on the coupling =
also. Our platform SE to coupled DIFF is not trivial to design and =was a
challenge.
We do not de-embed when doing crosstalk analysis like ICR, ICN, PEXT,
etc., for compliance.
- Al
Products for the Signal Integrity Practitioner
Alfred P. Neves
Chief Technologist
Office: 503-679-2429
www.wildrivertech.com
On May 5, 2015, at 3:55 AM, Istvan Novak <istvan.novak@xxxxxxxxxxx>wrote:
co-planar
An 0.6-inch long trace is very challenging to measure...
Regards,
Istvan Novak
Oracle
On 5/5/2015 3:11 AM, jun zhang wrote:
Hi experts,
I have designed a PCB test coupon to measure the impedance curve
and crosstalk curve from 0-10GHz.
My test trace is about 0.6". At each end of the trace, there is a
thetrace about 1" connected to SMA connector. Therefore the total
length of the coupon is about 2.6".
I have some question below:
1. When testing TDR, do you think 0.6" is too short to be
displayed? If
tocharacteristic can be displayed, I think I need not to de-embed the
influences of the co-planar traces at both ends. Am I right?
2. When testing TDR, I have two ways. The first is using TDR
instrument
VNA tomeasure it directly; the second way is tranforming SDD11 obtained
by
moretime domain.The second way is more convinient for me. Which method
is
ISDsuitable for displaying the impedance for the 0.6" test trace? .
3.When testing crosstalk, I will de-embed the co-planar trace to
reveal only the characteristic of the test trace. I plan to do
de-embedding by
onlyoffered by AtaiTec Co. My concern is still that whether this
method can give me accrate de-embedding result in the case that my
test trace is
of0.6" long. I think if the trace is longer, maybe the influence of
multi-reflection from co-planar trace will be minor because of the
loss
the trace. Under this consideration, I think the longer the test
trace length is, the better accurate de-embedding results we can
obtain. Am I right?
Hope to your reply
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject
field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu