[SI-LIST] Re: 90 degree turn in PCB tracks
- From: Scott McMorrow <scott@xxxxxxxxxxxxx>
- To: scott@xxxxxxxxxxxxx
- Date: Tue, 25 Jul 2006 15:05:51 -0400
Y'all,
I just went and pulled up the old app. note. you can find it here
http://www.onsemi.com/pub/Collateral/HB205-D.PDF
I think the problem people have had is in not putting the old Motorola
measurements in perspective. Clearly if you read the previous pages
you'll find that on pages 140 to 142 the design being tested is that of
a microwave hybrid divider on a substrate that was 62 mils thick and had
an Er of 5.3 (Er(eff)= 4.5 for microstrip). A hybrid divider takes a 50
ohm trace and divides it into two 100 ohm traces. If you do the
calculations, that 50 ohm trace was somewhere around 110 mils wide.
As we go on to page 144, we find the famous corner tests, where clearly
there are corner discontinuities shown by the TDR. If I were a betting
man, even though the material thickness and trace widths were not shown,
I'd bet that they are similar to the previous hybrid divider. So, we
have a case where there is a really fat trace and therefore a big
discontinuity. If we do the math, as I did previously, that
discontinuity should be:
t(discontinuity) = 85 x sqrt(2 x 4.5) x .11 = 28 ps.
In a TDR this time will be doubled, due to the round trip across the
corner, so we would expect a hump with a duration of approximately 50 to
60 ps. Even though the plots do not show a scale, I'd guess that around
a 50 to 60 ps pulse duration would be about right, when you add in the
additional TDR rise time, itself.
My conclusion is that there is nothing wrong with the old Motorola
application note. In fact, if you were to replicate this experiment
with 100 mil lines on a thick FR4 substrate today, you'd get pretty much
the same result. The problem is that people have generalized the data
to cases with smaller line widths without scaling. If you scale the
Motorola measurements by 1/22 (5 mils/110 mils), those corner
discontinuity impedance bumps will just blend into the noise. They
still exist ... but they are really, really tiny ... and not important
for modern digital board design ... just as Lee advocates.
regards,
Scott
Scott McMorrow
Teraspeed Consulting Group LLC
121 North River Drive
Narragansett, RI 02882
(401) 284-1827 Business
(401) 284-1840 Fax
http://www.teraspeed.com
Teraspeed® is the registered service mark of
Teraspeed Consulting Group LLC
Scott McMorrow wrote:
> Lee
> The experimental design used in Ultracad corner test is quite
> sub-optimal, given the incredible launch discontinuity. I'd be
> surprised if better than a 35 to 50 ps rise time edge actually makes it
> into the board. Then, given the length of the traces involved, I
> suspect that the actual bandwidth seen at the corners is somewhere
> around 3 GHz, at best. Unfortunately, there do not appear to be any
> structures available to determine the actual measurement bandwidth.
> Just because Todd Hubing was responsible for the measurements, does not
> mean that the experimental design is as good as an 18ps launch and 20
> GHz makes it sound. If you can get a fast launch into a board, and can
> use low loss materials, it is possible to see corners. Gus Panella and
> I did so about 8 years ago on a test vehicle that we designed using
> Rogers 4350 material.
>
> Having said that, 3GHz bandwidth is typical of the fastest standard
> logic signals on most conventional PCBs. Your assertion that corners do
> not matter for digital designs is true.
>
> The original source of the corner information comes from the Microwave
> literature. Gupta, et. al., Microstrip Lines and Slotlines, is a good
> reference. It has been well known that, in Microwave design, right
> angle bends are not a good thing. If you're trying to decrease return
> loss in a microwave design, you will most assuredly use radiused corners
> or the optimal chamfer. Why? Because microwave boards use low loss
> materials, like PTFE, and very wide trace width on thick substrates to
> reduce total power loss. The corner discontinuity on a 5 mil wide logic
> signal trace on a conventional high density PCB may not be an issue, but
> on a low loss microwave board that uses 50 to 200 mil wide microstrip
> traces, that silly little corner is a killer.
>
> The magnitude of the corner discontinuity is proportional to it's
> duration. A corner discontinuity lasts for approximately 85 x sqrt(2 x
> Er(eff)) x w ps:
>
> Where
> Er(eff) = is the effective Er of the material
> w = the width of the trace in inches.
>
> If we run the numbers,
>
> For a 5 mil stripline trace on FR4 with and Er(eff) of 4, the
> discontinuity lasts for 1.2 ps.
> For a 100 mil microstrip trace on Duroid with an Er(eff) of 2, the
> discontinuity lasts for 17ps.
>
> If we then use the rule of thumb that a discontinuity is important only
> when it approaches 1/10th of the risetime, then our little 5 mil corner
> has an effective operating bandwidth of .35/(1.2 e-12 x 10) = 29 GHz.
> Clearly out of the region where we are interested for digital logic.
>
> But for the 100 mil trace, we have an operating bandwidth of .35(17e-12
> x 10) = 2 GHz. This is quite frankly not a very good microwave design.
>
>
> Best regards,
>
> Scott
>
> Scott McMorrow
> Teraspeed Consulting Group LLC
> 121 North River Drive
> Narragansett, RI 02882
> (401) 284-1827 Business
> (401) 284-1840 Fax
>
> http://www.teraspeed.com
>
> Teraspeed® is the registered service mark of
> Teraspeed Consulting Group LLC
>
>
>
> Lee Ritchey wrote:
>
>> The paper published by Ultracad tests the effects of right angle bends in a
>> logic signal trace with an 18 pSec edge which is equivalent to about 20
>> GHz. There is no detectable effect. These tests were conducted by Todd
>> Hubing of UMR in a well equipped lab and are to be trusted. There have
>> been amy other similar tests done with the same result.
>>
>> I believe the notion that right angle bends are a source of problems stems
>> from an error in the Motorola ECL handbook published in 1974 and still in
>> print with the error. Now O Semiconductor publishes it.
>>
>> This may be one of those cases where simulation shows a change in the field
>> distribution around the right angle bend, which we expect. The question is
>> whether the change is significant. One of our jobs is to distinguish
>> between visible and significant.
>>
>> Once again, we have lies, damn lies and simulations. Simulations without
>> validation may well be more dangerous than no simulations at all in some
>> cases.
>>
>>
>>
>>
>>> [Original Message]
>>> From: Matthias Bergmann <MBergmann@xxxxxxxxxxxxxxxxxx>
>>> To: <si-list@xxxxxxxxxxxxx>
>>> Date: 7/25/2006 1:42:10 AM
>>> Subject: [SI-LIST] Antwort: Re: 90 degree turn in PCB tracks
>>>
>>> Hello,
>>>
>>> Following the advices I read the article about 90 degree bends on
>>> ultracad.com as well as the appropriate chapter in Eric Bogatins book.
>>> I am quite surprised about the conclusion that 90 degree corners don't
>>> matter. I remember that once in a simulation of a 50 ohm microstrip-lin=
>>> e
>>> with
>>> a chamfered 90=B0 corner in ADS Momentum, I never got a S11 better than=
>>> 15 dB
>>> at frequencies higher than 15 GHz, even 10 GHz made problems.
>>> The mentioned articles don't consider the frequency respectively just
>>> consider designs where the right-angle bend is electrically smaller tha=
>>> n a
>>> rising edge.
>>> Would be interesting to know how two 45 degree corners behave at higer
>>> frequencies.
>>>
>>> Regards, Matthias
>>>
>>> ____________________________________
>>>
>>>
>>> ___________________________________________________________________
>>> Diese E-Mail enth=E4lt vertrauliche und/oder rechtlich gesch=FCtzte
>>> Informationen. Wenn Sie nicht der richtige Adressat sind oder diese E-M=
>>> ail
>>> irrt=FCmlich erhalten haben, informieren Sie bitte sofort den Absender =
>>> und
>>> vernichten Sie diese Mail. Das unerlaubte Kopieren, jegliche anderweiti=
>>> ge
>>> Verwendung sowie die unbefugte Weitergabe dieser Mail ist nicht gestatt=
>>> et.
>>> -----------------------------------------------------------------------=
>>> -------------------------------
>>> This e-mail may contain confidential and/or privileged information. If =
>>> you
>>> are not the intended recipient (or have received this e-mail in error)
>>> please notify the sender immediately and destroy this e-mail. Any
>>> unauthorized copying, disclosure, distribution or otherwise use of the
>>> material or parts thereof is strictly forbidden.
>>> ___________________________________________________________________=
>>>
>>>
>>> ------------------------------------------------------------------
>>> 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 FAQ wiki page is located at:
>>> http://si-list.org/wiki/wiki.pl?Si-List_FAQ
>>>
>>> List technical documents are available at:
>>> http://www.si-list.org
>>>
>>> List archives are viewable at:
>>> //www.freelists.org/archives/si-list
>>> or at our remote archives:
>>> http://groups.yahoo.com/group/si-list/messages
>>> 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 FAQ wiki page is located at:
>> http://si-list.org/wiki/wiki.pl?Si-List_FAQ
>>
>> List technical documents are available at:
>> http://www.si-list.org
>>
>> List archives are viewable at:
>> //www.freelists.org/archives/si-list
>> or at our remote archives:
>> http://groups.yahoo.com/group/si-list/messages
>> 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 FAQ wiki page is located at:
> http://si-list.org/wiki/wiki.pl?Si-List_FAQ
>
> List technical documents are available at:
> http://www.si-list.org
>
> List archives are viewable at:
> //www.freelists.org/archives/si-list
> or at our remote archives:
> http://groups.yahoo.com/group/si-list/messages
> 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 FAQ wiki page is located at:
http://si-list.org/wiki/wiki.pl?Si-List_FAQ
List technical documents are available at:
http://www.si-list.org
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
