[SI-LIST] Re: RF Layout - Via spacing
- From: wolfgang.maichen@xxxxxxxxxxxx
- To: olaney@xxxxxxxx
- Date: Thu, 22 Oct 2009 13:25:19 -0700
Hi Orin,
you right that different edge shapes will have slightly different
conversion constants between risetime and bandwidth. 0.35 for one-pole,
0.4 or larger for steeper frequency roll-off (e.g. Chebycheff filter), and
around 0.33... (I think actually closer to 0.34) for Gaussian edge shape.
Now I would make the argument that typical signals we encounter do not
look like a single pole response but rather more like a Gaussian response.
In any case, all the rules of thumb (like distance < lambda/10) we have
been talking about are being pretty coarse, so IMHO anything that makes
less than maybe 10 or 20% difference is not worth worrying too much about
as long as you don't do a much more detailed analysis.
And as a previous poster pointed out, the application at hand seems to be
RF and not a digital signal (I have to admit that I am so used to working
with digital circuits that I glossed over that bit of information as
well), so for that the point is moot anyway :-)
Wolfgang
olaney@xxxxxxxx
10/22/2009 11:28 AM
To
wolfgang.maichen@xxxxxxxxxxxx
cc
arvind.yad1983@xxxxxxxxx, si-list@xxxxxxxxxxxxx
Subject
Re: [SI-LIST] Re: RF Layout - Via spacing
Arvind:
Wofgang gave an excellent explanation. However, note that the physical
wavelength spoken of is not that of free space but is determined by the
velocity of propagation for your particular implementation, probably in
the range of 2/3 c. This makes lamda/20 (or any other) spacing on your
board tighter by that factor.
Also, to pick a small nit, the constant for conversion between risetime
and bandwidth is .35 not .33 (it's actually .34969915...), though the 6%
error is small enough. The number is based on the assumption that a
single pole dominates frequency rolloff and that Tr is measured between
the 10% and 90% points. Other assumptions yield other numbers.
Calculations based on the 20% and 80% points often found on logic data
sheets should use a constant of .22 (.2206356...) to keep an apples to
apples comparison with 10% and 90%.
Orin Laney
On Thu, 22 Oct 2009 10:29:17 -0700 wolfgang.maichen@xxxxxxxxxxxx writes:
> Hello Arvind,
> ok, that's different. What you describe is a transmission line
> structure
> called "coplanar waveguide with ground". This is a signal line with
> ground
> planes on either side (that would constitute the coplanar waveguide)
> plus
> a ground plane underneath. It's basically a mixture between a
> coplanar
> waveguide and a microstrip line.
>
> In order for the two ground strips on either side to be truly
> ground, it
> is important that they be closely tied to the actual ground plane
> (better:
> "reference plane") underneath. If you forgot to do that the ground
> strips
> would just act as two additional transmission lines (with a line
> impedance
> significantly larger than zero) coupled to the center line, but not
> a
> ground references (with impedance ideally = zero). So in a simple
> picture
> your line impedance would be higher than you would expect. Second,
> current
> entering into them would have no way of redistributing itself into
> the
> ground plane underneath, resulting in additional distortion. If your
> total
> line length approaches lambda/4 or more you would see resonance
> effects in
> those "ground" strips - you'd basically have a coupler with
> unterminated
> outputs. With vias spaced to far apart the same applies but not as
> severe.
>
> Any transmission line calculator (or 2D field solver) assumes that
> you tie
> every structure denotes as grounf very closely to ground. Only then
> your
> actual transmission line impedance will match the one you calculate.
>
> In order for a structure to act as ground the ground return path
> must be
> short against the shorest wavelength of interest (approximately
> given by
> the frequency 0.33/rise_time, i.e. by your rise time and NOT by e.g.
> you
> clock frequency or data rate). lambda/20 is quite conservative in
> that
> respect and in practice you will be able to get away with wider
> spacing
> (somewhere between lambda/4 and lambda/10). Also note that the vias
> add
> additional metal close to the signal trace which will lower your
> inductance compared to what your 2D solver will predict. To be safe,
> keep
> the vias at sufficient distance from the trace - about 3 trace
> widths
> would be the minimum - or even better, use a 2.5D or 3D field solver
> to
> simulate the actual structure including the vias.
>
> Wolfgang
>
>
>
>
>
>
> arvind yadav <arvind.yad1983@xxxxxxxxx>
> 10/22/2009 10:12 AM
>
> To
> wolfgang.maichen@xxxxxxxxxxxx, si-list@xxxxxxxxxxxxx
> cc
>
> Subject
> Re: [SI-LIST] RF Layout - Via spacing
>
>
>
>
>
>
> Hello Wolfgang ,
>
> Thanks for the reply .
>
> In my case i dont have a signal via . What i have is a RF signal in
> top
> layer and around that two strips of ground traces .
>
> In that GND track we have placed vias and some guidelines says to
> maintain lambda/20 rule between the two same gnd vias .
>
> Can you please explain this .
>
> Thanks
> Arvind.H
>
> On Thu, Oct 22, 2009 at 10:34 PM, <wolfgang.maichen@xxxxxxxxxxxx>
> wrote:
>
> Hello Arvind,
>
> the goal in designing a clean (reflection-free) signal path is to
> have
> homogeneous characteristic impedance all along the path (typically
> ZoP
> Ohm unless you are working with TV signals that use 75 Ohms).
>
> The characteristic impedance is determined by the ratio of
> inductance Lu
> per unit length and capacitance Cu per unit length:
>
> Zo=sqrt(Lu/Cu)
>
> A signal via and its closest return via (or vias) are just part of
> that
> path. Changing the distance d between signal and return via changes
> both
> capacitance C and inductance of that via structure (C decreases with
> d,
> and L increases with D), so you can use that to tune the impedance
> of the
> via structure. Ideally you'll achieve 50 Ohms although this is hard
> to do
> with just a single return via. In that ideal case (ZoP Ohms) the via
>
> structure becomes completely transparent to the signal, i.e. it only
>
> causes delay (delta_t = sqrt(C x L)) but no reflections.
>
> Designing a well-matched via structure is a challenge and typically
> need
> either a good 3D simulation tool or a few test boards to get it
> right at
> high data rates. Rules of thumb ar hardly sufficient although they
> can
> provide at least a goot starting point as well as show the "knobs"
> you can
> use to adjust the impedance (for via structure, there a are many
> knobs -
> via diameter and distance, stub or stub drilling, pad/antipad
> diamaters,
> and so on).
>
> The lambda/20 rule you mention comes from the fact that typically
> structures that are very short against the shorted wavelength
> (highest
> frequency) of interest only have negligible influence on the
> waveform,
> i.e. produce only minimal reflections even when they are mismatched
> (Zo <>
> 50 Ohms). This is of course just a crude rule of thumb.
>
> Whatthe lambda/20 rule achieves very nicely is that it forces you to
> place
> a return via close to every signal via. This is important - current
> is
> always flowing in a loop so if there is no return via close by, the
> return
> current has to "go looking" for the nearest return path which may be
> quite
> a detour - this will cause a large parasitic inductance in the path
> (because the current now encloses a large loop are) and resulting
> large
> reflection and reduced bandwidth.
>
> Wolfgang
>
>
>
>
>
>
>
> arvind yadav <arvind.yad1983@xxxxxxxxx>
> Sent by: si-list-bounce@xxxxxxxxxxxxx
> 10/22/2009 09:45 AM
>
>
> To
> si-list@xxxxxxxxxxxxx
> cc
>
> Subject
> [SI-LIST] RF Layout - Via spacing
>
>
>
>
>
>
>
>
> Hello All,
> I am working on a RF Layout. I looked into some design guidelines
> and had
> some doubt on gnd via spacing requirement .
>
> Guideline said that ë/20 distance has to be maintained between gnd
> vias
> that
> are stitched on either side of the RF signal
>
> Can any one please let me know the reason for this requirement ?
>
> I also would like to know what would be the gnd backoff distance
> from the
> RF signal and the reason .
>
> Thanks
>
> Arvind.H
>
>
> ------------------------------------------------------------------
> 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 technical documents are available at:
> http://www.si-list.net
>
> 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 technical documents are available at:
> http://www.si-list.net
>
> 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
>
>
>
____________________________________________________________
Best Weight Loss Program - Click Here!
http://thirdpartyoffers.juno.com/TGL2141/fc/BLSrjpTFoYcrd6wERphDjoRWK77us546r3N8GvaYgNKpsMhIAYdLjSDKwN6/
------------------------------------------------------------------
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 technical documents are available at:
http://www.si-list.net
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:
