[SI-LIST] Re: AW: Ground vias around signal via
- From: <Wolfgang.Maichen@xxxxxxxxxxxx>
- To: <ricbrook@xxxxxxxxx>, <si-list@xxxxxxxxxxxxx>
- Date: Fri, 17 Oct 2014 08:00:52 +0000
To add some more "meat" as requested: I ran some simple simulations in Sonnet
Lite and would like to share the results:
A 3D view is here:
http://pdamusician.com/electronics/return_via_two_vias_3d_view.gif
Simulation space is a 1 inch x 1 inch metal box; this is large compared to the
trace widths and dielectric thicknesses (see below) so even though the return
current can run through the box walls they are far enough away to not be very
effective at high frequencies (several GHz).
So layer setup is (limited by Sonnet Lite's restriction to 2 inner layers):
Top cover (metal)
Top dielectric (eps_r = 4, thickness 12 mils)
Signal Layer 0 (metal)
Center dielectric (eps_r = 4, thickness 100 mils)
Signal Layer 1 (metal)
Bottom dielectric (eps_r = 4, thickness 12 mils)
Bottom cover (metal)
Top and bottom cover act as reference (return) planes. The center dielectric
being much thicker than top and bottom dielectric assures that a trace on layer
0 mostly couples (= has return current flowing in) the top cover, and a trace
on layer 1 mostly couples to the bottom cover. So the return current will need
to transition from top cover to bottom cover at the box center.
I set up the signal path with a trace entering on the left on layer 0,
transitioning to a trace on layer 1 in the center of the box. Trace width is 20
mils which gives a characteristic impedance close to 50 Ohms, so with a
"proper" return path one would expect only very low insertion losses at all
frequencies.
I added two return vias between top and bottom cover, close to the signal via
(I did not spend time to optimize the via positions, just placed them at 20 mil
spacing from the signal via on either side). Here are S11 and S21 (return and
insertion loss) from 10 MHz to 20 GHz:
http://pdamusician.com/electronics/return_via_two_vias_sparams.gif
Performance is very acceptable (S21 better than -3dB) up to > 10 GHz, above
that some box resonances are visible.
Now I repeated the simulation, but with the vias placed further away (250 mils
on either side) and completely removed (so the box walls will be the only way
for the return current). A 3D view of the wide spacing is here:
http://pdamusician.com/electronics/return_via_two_vias_wide_spacing_3d_view.gif
The results are strikingly worse in both cases:
http://pdamusician.com/electronics/return_via_two_vias_wide_spacing_sparams.gif
http://pdamusician.com/electronics/return_via_no_vias_sparams.gif
Performance limit drops to < 4 GHz and there is almost complete reflection (S11
approx. 0dB) over wide ranges above that.
I would say this demonstrates the importance of having return vias close by
quite impressively. It also shows that you can get away with an "ill defined"
return path (box wall return rather far away, approx. 0.5 inches in this case,
which is still not excessively long!) for low frequencies, in the present case
meaning a few 100 MHz. This is what others have stated previously (think
"lambda/10" guideline).
So yes, "it depends", but I would certainly dispute a blanket statement that
return vias are never necessary or useful.
Anyone wanting to run their own simulations, here are my Sonnet model files
(they need only ~1 MB simulation memory so can easily run even on the
non-registered free version):
http://pdamusician.com/electronics/return_via_two_returns.son
http://pdamusician.com/electronics/return_via_two_returns_wide_spacing.son
http://pdamusician.com/electronics/return_via_none.son
Wolfgang
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On
Behalf Of Rick Brooks (ricbrook)
Sent: Thursday, October 16, 2014 9:13 PM
To: leeritchey@xxxxxxxxxxxxx; scott@xxxxxxxxxxxxx;
Yishan.Li@xxxxxxxxxxxxxxxxxxxx
Cc: Gert.Havermann@xxxxxxxxxxx; aaditya.kandibanda@xxxxxxxxx; si-list
Subject: [SI-LIST] Re: AW: Ground vias around signal via
The counter argument is that anyone who says that GND stitching vias do nothing
or are not needed under any conditions, is also suggesting a "rule" which
should be proved for all possible cases.
Honestly, I do not remember anyone saying on this thread that having close GND
stitching vias is required in every conceivable place or scenario.
I have certainly seen specific designs where the location and number of GND
stitching vias made a big difference with signals containing even mode.
Obviously, that does not mean they are a must have for every via on every board
that is built or shipped.
It also does not mean that I am willing to share the actual data, because, like
others, there may be confidential aspects to it.
I have also seen conditions where right angle bends on traces creates no
visible problems.
That does not mean you should use right angle bends, or that there aren't an
infinite number of cases where right angle bends would be a disaster.
As always, "it depends"
This forum is for people to put forth their ideas and experience.
I, for one, welcome their comments, with or without proof.
cheers
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On
Behalf Of Lee
Sent: Thursday, October 16, 2014 10:16 AM
To: scott@xxxxxxxxxxxxx; Yishan.Li@xxxxxxxxxxxxxxxxxxxx
Cc: Gert.Havermann@xxxxxxxxxxx; aaditya.kandibanda@xxxxxxxxx; si-list
Subject: [SI-LIST] Re: AW: Ground vias around signal via
This discussion seems to be terribly theoretical. We all know that there are
thousands, perhaps tens of thousands of assemblies shipped every day with
differential signals as high as 28 Gb/S on them that work just fine without the
need to put in "ground vias" nearby.
What comes to my mind when I read discussions such as this is that old Burger
King commercial where Clara Peller asks "where's the beef?"
For all of the postulated problems mentioned in these discussions, "where's the
proof?"
We do too much speculating on this forum and not enough proving!
My position on this whole thing as well as many others that appear on this
discussion group is, if you are going to put forth a rule, be prepared to offer
the proof that the rule is valid as well as where it is valid. If you are not
prepared to do this, it is a disservice to those who are asking for advice to
make such a posting.
-----Original Message-----
From: Scott McMorrow
Sent: Thursday, October 16, 2014 3:29 AM
To: Yishan.Li@xxxxxxxxxxxxxxxxxxxx
Cc: Gert.Havermann@xxxxxxxxxxx ; aaditya.kandibanda@xxxxxxxxx ; si-list
Subject: [SI-LIST] Re: AW: Ground vias around signal via
Gert has good advice about the distance of the ground via to the signal via,
but I would like to modify it a bit. A well-referenced set of ground planes
will have ground stitch vias that are separated by no more than 1/10 the
wavelength of the Nyquist frequency of the highest differential bit rate, or
the bandwidth of the signal edge rate of the fastest single ended driver.
So for some numbers in the English system.
Let Dk = 4
Tdelay = 170 ps/inch
For DDR3/4 with 100 ps driver edge rates BW = .35/100ps = 3.5 GHz lambda = 285
ps
1/10 lambda = 28.5 ps
In Dk = 4 ground vias should be separated by a maximum of 167 mil to maintain a
good return path for these singled-ended DDR signals.
For 10 Gbps
Nyquist is 5 GHz
lambda = 200 ps
1/10 lambda = 20 ps117
In Dk = 4 ground vias should be separated by a maximum of 117 mil for 10 Gbps
signalling to maintain a good return path.
For 28 Gbps
Nyquist is 14 GHz
lambda is 71 ps
1/10 lambda is 7.1 ps
In Dk = 4 ground vias should be separated by a maximum of 42 mil for 28 Gbps
signalling.
These recommendations apply to the region in proximity of the signal transition
vias. They serve to tie the ground cavities together, provide shielding for
the power/ground cavities, eliminate resonances in the signal passband and
first harmonic, and reduce via-to-via crosstalk. If the signal transition is
at the balls of a semiconductor device, or in the pin field of a connector,
there are "usually" enough ground vias in these regions to meet these
requirements. If that is the case, then no additional grounds are required.
But, there are often cases that we encounter at Teraspeed Consulting where
these rules are violated. Here are some common areas to look at.
Via transitions around dc blocking capacitors.
Boards with outer layer buildup microvias, where drilled vias do not carry the
package or connector grounds down through the board.
Areas with asymmetric stripline crossing power splits on the distant plane side
of the stripline (the stripline is close to ground.)
In all these regions it is necessary to close the return path loop with ground
vias as described above. Use the above as a guide to current and future
designs.
best regards,
Scott
Scott McMorrow
Teraspeed® Consulting - A Division of Samtec
16 Stormy Brook Rd
Falmouth, ME 04105
(401) 284-1827 Business
http://www.teraspeed.com
On Thu, Oct 16, 2014 at 3:39 AM, LI Yishan <Yishan.Li@xxxxxxxxxxxxxxxxxxxx>
wrote:
> Hi, Aaditya:
> According to 3D simulation, ground vias around signal via deeply
> effect high frequency impedance. If your signal is low speed signal,
> it seems the ground vias are not necessary.
>
> Best regards
> Li Yishan
> -----Original Message-----
> From: si-list-bounce@xxxxxxxxxxxxx
> [mailto:si-list-bounce@xxxxxxxxxxxxx]
> On Behalf Of Havermann, Gert
> Sent: 2014年10月16日 14:59
> To: aaditya.kandibanda@xxxxxxxxx; si-list
> Subject: [SI-LIST] AW: Ground vias around signal via
>
> Aaditya: My opinion is, if we have a proper return path, they are not
> necessary.
>
> -> As Wolfgang said, GND vias are needed for proper return path
> -> because
> without any GND via there will be no proper return path. You don't
> have to put multiple of those around the signal via. with proper
> placement one is enough even for very high speeds.
>
> Aaditya: Any situations they are useful other than shielding?
>
> -> Placing many stitching vias doesn't necessarily provide any shielding.
> That’s an old Myth. Take a look at waveguide filters. The placement
> of screws into the dielectric (air) looks and functions very similar
> to a PCB where the energy also travels in the dielectric. Stitching
> vias can act as a filter, meaning that some frequencies are not
> shielded but guided to the outside.
>
> Aaditya: How will they help? When do we need them?
>
> -> As already said, use GND vias to provide a proper GND return. as a
> -> rule
> of thumb there should be a return via within the range of 1/8
> wavelength
> (Nyquist) to the signal via.
>
> BR
> Gert
>
>
> ----------------------------------------
> Absender ist HARTING Electronics GmbH, Marienwerderstraße 3, D-32339
> Espelkamp; Registergericht: Amtsgericht Bad Oeynhausen; Register-Nr.:
> HRB 8808; Vertretungsberechtigte Geschäftsführer: Dipl.-Kfm.
> Edgar-Peter Düning, Dipl.-Ing. Torsten Ratzmann,
> Dipl.-Wirtschaftsing. Ralf Martin Klein
>
> -----Ursprüngliche Nachricht-----
> Von: si-list-bounce@xxxxxxxxxxxxx
> [mailto:si-list-bounce@xxxxxxxxxxxxx]
> Im Auftrag von Aaditya K
> Gesendet: Mittwoch, 15. Oktober 2014 19:42
> An: si-list
> Betreff: [SI-LIST] Ground vias around signal via
>
> Hello Experts,
> I have a question on ground vias placement around signal via.
>
> My opinion is, if we have a proper return path, they are not necessary.
> Am I correct?
>
> Any situations they are useful other than shielding?
>
> How will they help? When do we need them?
>
> Please help.
>
> Thanks
> Aaditya
>
>
> ------------------------------------------------------------------
> 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
>
>
> ------------------------------------------------------------------
> 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
-----
No virus found in this message.
Checked by AVG - www.avg.com
Version: 2014.0.4765 / Virus Database: 4040/8400 - Release Date: 10/16/14
------------------------------------------------------------------
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
------------------------------------------------------------------
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
Other related posts:
