[SI-LIST] Re: SATA board to board connector question

  • From: "Loyer, Jeff" <jeff.loyer@xxxxxxxxx>
  • To: <olaney@xxxxxxxx>
  • Date: Mon, 27 Aug 2007 14:07:31 -0700

Thanks for your kind words regarding the article.  I'd point out that my
results were based on the effects not only of Nelco N4000-SI (with a
corresponding Er = 3.4, Df = 0.007), but also of more common (cheap) FR4
materials with Er = 4.2 and Df as high as 0.017.
 

The only assertion I was questioning was that microstrip is more lossy
than stripline - do you have any data supporting this?

Jeff Loyer 



________________________________

From: olaney@xxxxxxxx [mailto:olaney@xxxxxxxx] 
Sent: Monday, August 27, 2007 1:33 PM
To: Loyer, Jeff
Cc: si-list@xxxxxxxxxxxxx
Subject: Re: [SI-LIST] Re: SATA board to board connector question

 

No question, there are many decent materials on the market.  They fab
like FR4, and often are called FR4 for marketing purposes.  FR4 is a
concept, not a product spec per se, so the wise designer will do the
homework and specify a vendor part # for consistent high performance
results.  The fact remains that there is nasty FR4 and there is nice
FR4.  If nothing but FR4 is called out, and your board house buys at
lowest cost, you are rolling the dice regarding what you'll get back.
The idea that you can run umpty Gb/s over something called FR4 is
gratifying, but incomplete.  The drop in dielectric constant with rising
frequency for the nasty stuff is well known, as is the rise in the loss
tangent, for classic, nasty FR4.  Digital signals are forgiving, and
when combined with equalization, extensive simulation and other tricks,
can be made to do amazing things on "FR4", or polyimide, or oak leaves
laminated with racoon poop if need be.  But if it was easy, the demand
for SI consultants would be rather lower than it is.  The very need for
care and expertise underscores that the effect of the laminate is
nontrivial.  

 

Having rescued clients by scrapping the tricks and simply using a decent
microwave capable material for critical circuit portions, I stand behind
my observations.  I also totally believe the results in your article.
Of course, the material upon which the article is based is "Nelco
N4000-13 SI, a readily available (though relatively expensive, compared
to FR-4) material."  Good stuff.  I like the article.  Your advice
concerning the transmission path implementation is right on: "Proper PCB
stackup and trace geometry design are key elements in the fight to lower
losses".  So is a short path.  The best advice for a SATA connection is
simple: use the right connector and put the phy chip next to it.  That
way, even those using the nasty type of FR4 (whose name is Legion, for
they are many) have a fighting chance of success, without need to do
double back flips to get there.  

 

Regards,

            Orin

 

On Mon, 27 Aug 2007 09:57:16 -0700 "Loyer, Jeff" <jeff.loyer@xxxxxxxxx>
writes:
> Hi Orin,
> Do you have some specific data on the difference in performance 
> between
> microstrip and stripline?  While I agree with your first premise 
> (that
> stripline introduces the parasitics of the via to get down to the
> layer), my studies (see url below) have indicated equal loss between 
> the
> 2 mediums (stripline vs. microstrip), with the caveat that 
> microstrip
> often allows wider traces with resultant lower copper loss.  But,
> impedance control on the inner layers (with no plating) is usually
> better, and crosstalk is reduced.
> 
> I've often heard similar arguments to your "snorkeling" metaphor, 
> but
> haven't seen data supporting it.
> 
> This leads me to favor lower layers of stripline for long lengths 
> of
> high-speed signals, when I have a choice (a rare luxury).
> 
> Perhaps you've had some other experience or have data to share?
> 
> url to my study on microstrip vs. stripline loss:
> http://pcdandm.com/cms/content/view/2572/95/
> 
> Note to all: there have been many postings on "Microstrip vs. 
> Stripline"
> on this forum, worthy of perusing if you're interested.
> 
> Disclaimer:
> The content of this message is my personal opinion only and although 
> I
> am an employee of Intel, the statements I make here in no way 
> represent
> Intel's position on the issue, nor am I authorized to speak on 
> behalf of
> Intel on this matter.
> 
> Jeff Loyer
> 
> -----Original Message-----
> From: si-list-bounce@xxxxxxxxxxxxx 
> [mailto:si-list-bounce@xxxxxxxxxxxxx]
> On Behalf Of olaney@xxxxxxxx
> Sent: Monday, August 27, 2007 8:56 AM
> To: nareshd@xxxxxxxxxxxxx
> Cc: goister@xxxxxxxxx; si-list@xxxxxxxxxxxxx
> Subject: [SI-LIST] Re: SATA board to board connector question
> 
> Usually I prefer microstrip over stripline for signals as fast as 
> SATA. 
> The reasoning boils down to two considerations.  One is that 
> transitions
> between buried stripline and other layers brings its own set of
> problems,
> and the other is that in a world where (censored) insist on using 
> FR4 in
> the microwave region, the less contact a high speed signal has with 
> the
> stuff, the better.  Air is a reasonable microwave dielectric.  The 
> FR in
> FR4 means "fire retardent", which is not a microwave concept.
> Microstrip
> steps in the poop on only one side of the trace, where stripline is 
> like
> snorkling in it.  This makes a difference in how much signal 
> quality
> gets
> eaten per unit length.
> 
> Don't get me wrong.  Stripline is excellent microwave practice when
> proper dielectrics are used and connectorization is competently
> implemented.  But, that level of attention seems reserved for a
> different
> world filled with directional couplers, 90 degree hybrids, and 
> other
> exotic animals never seen on this forum.
> 
> Orin
> 
> On Mon, 27 Aug 2007 15:43:26 +0530 "Dhamija Naresh-B07930"
> <nareshd@xxxxxxxxxxxxx> writes:
> > I also have a Question related to SATA signal routing in PCB.
> > 
> > Which routing is the best for SATA signals  -  On Outer Layers OR 
> 
> > in
> > inner layers.
> > I am finding the difference in performances of two.
> > 
> > - Naresh
> > 
> > -----Original Message-----
> > From: si-list-bounce@xxxxxxxxxxxxx 
> > [mailto:si-list-bounce@xxxxxxxxxxxxx]
> > On Behalf Of Sihan Goi
> > Sent: Monday, August 27, 2007 3:34 PM
> > To: olaney@xxxxxxxx
> > Cc: si-list@xxxxxxxxxxxxx
> > Subject: [SI-LIST] Re: SATA board to board connector question
> > 
> > Thanks. One more question. You mention to carry the ground pins 
> > together
> > with the signals on the same connector. Why is this so? I do have
> > another connector that carries the power signals. Does this mean 
> I
> > should not route any other ground signal(s) with this power 
> > connector,
> > and rely solely on the signal connector to provide the ground 
> plane 
> > for
> > the daughterboard?
> > On 8/24/07, olaney@xxxxxxxx <olaney@xxxxxxxx> wrote:
> > >
> > >  Separate power is the usual approach, and not a problem.  Some 
> of 
> > the
> > 
> > > smaller Samtec connectors sound like a good choice for constant 
> Z 
> > 
> > > board to board, among many others.  Most formal backplane 
> > connectors 
> > > probably have way too many pins for your need.  It looks like 
> you 
> > have
> > 
> > > everything on track for a successful design.  Keep the list 
> posted 
> > on 
> > > the more significant problems and progress, so that everybody 
> can
> > learn from your experience.
> > > And make those app engineers take you out to lunch.
> > >
> > > Regards,
> > >          Orin
> > >
> > > On Fri, 24 Aug 2007 12:07:16 +0800 "Sihan Goi" 
> > <goister@xxxxxxxxx>
> > writes:
> > >
> > > Thanks.
> > >
> > > Basically the daughterboard acts simply as kind of an extender 
> of 
> > the 
> > > SATA signal. So all the logic circuitry is on the main board, 
> with 
> > the
> > 
> > > SATA signals coming out of the SATA controller chip, and to the 
> 
> > > daughter board via the high speed connector direct(no cables). 
> > These 
> > > signals are then routed to an actual SATA HDD 7pin connector on 
> 
> > the 
> > > daughterboard, to be connected to a SATA HDD. SATA power to the 
> 
> > HDD 
> > > will most likely be routed with another connector, probably a 
> > regular 
> > > pin header/socket. That shouldn't be a problem right?
> > >
> > > Regarding the fat pads, I'll probably be using 0402 components 
> as 
> > much
> > 
> > > as possible to reduce this problem, particularly for the 10nF 
> > > capacitors required for the SATA signals. I believe this is part 
> 
> > of 
> > > the SATA recommendations too.
> > >
> > > On 8/24/07, olaney@xxxxxxxx <olaney@xxxxxxxx> wrote:
> > > >
> > > >  Yes, the connector manufacturer recommendations.  Any 
> connector 
> > 
> > > > with the right impedance and adequate data rate specs will do. 
>  
> > The 
> > > > same connector family can be used for many purposes, so it is 
> 
> > not a 
> > > > matter of the signaling system used as much as the electrical 
> 
> > > > performance and the mechanical need.  Extra pins can be 
> ignored 
> > or 
> > > > used for other purposes.  I don't know if you are trying to 
> > connect 
> > > > your boards connector to connector direct, or through a short 
> 
> > cable.
> > 
> > > > In any instance, keep the I/O chip close to the connector, and 
> 
> > avoid
> > 
> > > > changes in trace widths other than what might be recommended 
> for 
> > 
> > > > dealing with proximity effects (sometimes the signal traces 
> are 
> > 
> > > > tapered where they run under connector dielectric).  One 
> common 
> > 
> > > > signal integrity killer is where the traces are carefully 
> > designed, 
> > > > yet run through coupling caps or other passives using the fat 
> 
> > SMT 
> > > > pads that production loves so much.  Sometimes production 
> balks 
> > at 
> > > > the practice of using wider traces or narrower parts to 
> minimize 
> > the
> > 
> > > > discontinuity in microstrip width.  Given the choice between a 
> 
> > > > design that works and one that can be built without hand 
> > operations 
> > > > or other accommodations, they have a tough time making up 
> their
> > minds.  As an engineer with proper test equipment and adequate 
> time
> > (That's all of us, right?  Right?), you can often meet both goals.
> > > >
> > > > Anyway, you might as well check out formal SATA connectors to 
> 
> > > > understand them electrically before widening your search.  The 
> 
> > > > GSSGSSG layout is not absolutely necessary for short 
> distances, 
> > as 
> > > > crosstalk can be controlled with adequate pair separation.  I 
> 
> > would 
> > > > not assume that a 7 pin connector is the goal.  Think of it as 
> 
> > two 
> > > > impedance controlled pairs plus whatever ground paths are 
> > designed 
> > > > in to ensure impedance and crosstalk control.  The ground need 
> 
> > not 
> > > > be discrete pins on par with the signals, in the same way that 
> a 
> > 
> > > > coax shield is not carried by a pin like that of the center 
> > > > conductor.  However, make sure that the ground is carried with 
> 
> > the 
> > > > signals through the same connector!!!  Any ground connections 
> 
> > > > provided elsewhere in the system might make an ohmmeter happy, 
> 
> > but
> > relying on them = death at high frequencies.
> > > >
> > > > Molex might be another vendor to include in your list.  Don't 
> be 
> > shy
> > 
> > > > about using vendor field app engineers -- that's what they're 
> 
> > paid 
> > > > for, and it's job security for them.
> > > >
> > > > Orin
> > > >
> > > > On Thu, 23 Aug 2007 23:07:46 +0800 "Sihan Goi" 
> > <goister@xxxxxxxxx>
> > > > writes:
> > > >
> > > > Thanks. It's kinda confusing when I go to samtec, FCI or 
> > amphenol 
> > > > website and they don't show what kind of applications the high 
> 
> > speed
> > 
> > > > connectors are meant for, or their impedance rating or 
> > whatever...
> > > >
> > > > Anyway, I'm also wondering about how the signals should be 
> > routed to
> > 
> > > > the connector. SATA signals on regular SATA connector are as 
> > follows
> > 
> > > > GND A+ A- GND B+ B- GND
> > > >
> > > > Should I be following this topology or does it even matter? 
> I'm 
> > 
> > > > guessing most of these high speed connectors won't have 
> exactly 
> > 7 
> > > > pins. They're usually spec'ed for 2/3/4 pairs or even more. 
> I'm 
> > 
> > > > guessing I only need a 2 pair connector since I only have 2 
> > > > differential SATA signal pairs. I'm wondering if I even need 
> to 
> > 
> > > > route the GND with the high speed connector or can it be from
> > another regular connector somewhere else...?
> > > >
> > > > Lastly, when you say "pay attention to the manufacturer
> > recommendations"
> > > > which manufacturer do you mean? The connector manufacturer?
> > > >
> > > > Again, thanks for your reply!
> > > >
> > > > On 8/23/07, olaney@xxxxxxxx < olaney@xxxxxxxx> wrote:
> > > > >
> > > > > Basically, yes, but give your design some margin.  For 3 
> Gb/s 
> > go 
> > > > > for 5 GHz  or more of frequency response.  Biggest headache 
> 
> > tends 
> > > > > to be keeping the impedance constant where the traces enter 
> 
> > the 
> > > > > connector launch area.
> > > > > Pay attention to the manufacturer recommendations.
> > > > >
> > > > > Orin Laney
> > > > >
> > > > > On Thu, 23 Aug 2007 16:08:09 +0800 "Sihan Goi" 
> > <goister@xxxxxxxxx>
> > > > > writes:
> > > > > > Hi,
> > > > > > I have a design where I have to route SATA signals from a 
> 
> > main 
> > > > > > board to a daughter board. The daughter board will likely 
> 
> > have 
> > > > > > nothing except for the 7 SATA signals(4 data and 3 GND) 
> > > > > > connected to a regular SATA HDD connector (unless some 
> > passives 
> > > > > > are needed?).
> > > > > >
> > > > > > In my previous PATA design, I used a normal 44pin IDE type 
> 
> > > > > > connector pair(pin header + socket), and this worked well 
> 
> > for 
> > > > > > PATA. I'm guessing this will not work so well with SATA 
> > though. 
> > > > > > What kind of connectors would work for SATA1/2? I know the 
> 
> > > > > > differential impedance is 100ohms, and that the trace 
> length 
> > 
> > > > > > difference must be within 5 mils. I'm guessing I have to 
> get 
> > a 
> > > > > > high speed connector that has the same impedance and is 
> able 
> > to 
> > > > > > support 3GHz speeds? Is that all I need to be looking 
> for?
> > > > > >
> > > > > > Thanks.
> > > > > >
> > > > > > --
> > > > > > - Goi Sihan
> > > > > > goister@xxxxxxxxx
> > > > > >
> > > > > >
> > > > > > 
> > ----------------------------------------------------------------
> > > > > > --
> > > > > > 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
> > > > > >
> > > > > >
> > > > > >
> > > > > >
> > > > >
> > > >
> > > >
> > > >
> > > > --
> > > > - Goi Sihan
> > > > goister@xxxxxxxxx
> > > >
> > > >
> > >
> > >
> > > --
> > > - Goi Sihan
> > > goister@xxxxxxxxx
> > >
> > >
> > 
> > 
> > --
> > - Goi Sihan
> > goister@xxxxxxxxx
> > 
> > 
> > 
> ------------------------------------------------------------------
> > 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
>   
> 
> 

 


------------------------------------------------------------------
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: