ECN0024 (GPS ground and layout)

9 messages Options
Embed this post
Permalink
Werner Almesberger

ECN0024 (GPS ground and layout)

Reply Threaded More More options
Print post
Permalink
I'm scratching my head at ECN0024 :) Alvaro, is there any part of
it that's already covered by the schematics or is it all really
just a reminder for things to do in the layout ?

Also in the latter case, it would be good to mention the key
points of what should be changed/observed.

- Werner

_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core
AlvieBoy

Re: ECN0024 (GPS ground and layout)

Reply Threaded More More options
Print post
Permalink
Werner Almesberger wrote:
> I'm scratching my head at ECN0024 :) Alvaro, is there any part of
> it that's already covered by the schematics or is it all really
> just a reminder for things to do in the layout ?

That was mostly a reminder, but we should discuss that a bit more.

The GPS chip has two grounds, digital and RF. GTA02 does not distinguish between one and other, nor in schematics nor in layout.

Digital ground is very noisy. Also we need to avoid ground current to flow near the RF part. One idea is to have a separate ground for the RF part, and connect
it to digital ground at one point, forming some sort of ground peninsula, thus avoiding digital current flow to interfere with RF. We should also have these
ground planes (RF) filled with through-hole vias, to avoid EMI emission (and maybe mitigate interferences).

Comments ?

Álvaro

>
> Also in the latter case, it would be good to mention the key
> points of what should be changed/observed.
>
> - Werner
>
> _______________________________________________
> gta02-core mailing list
> [hidden email]
> https://lists.openmoko.org/mailman/listinfo/gta02-core

_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core
Werner Almesberger

Re: ECN0024 (GPS ground and layout)

Reply Threaded More More options
Print post
Permalink
?lvaro Lopes wrote:
> One idea is to have a separate ground for the RF part, and connect
> it to digital ground at one point, forming some sort of ground peninsula,

Whether to split ground planes or not seems to vary mainly with the
designer, less with the design.

To take random examples from a quick google search, the pro-split side
has a quite convincing example here:
http://www.signalintegrity.com/Pubs/news/9_04.htm

And so does the contra-split side (also check out the link to the 2001
article):
http://www.hottconsultants.com/techtips/split-gnd-plane.html

My interpretation is that splitting ground planes gives you more
control, but at the cost of increasing complexity and the risk of
unintended consequences.

> We should also have these
> ground planes (RF) filled with through-hole vias, to avoid EMI emission
> (and maybe mitigate interferences).

Putting that RF into a "cage" sounds good to me. This should also help
to reduce ground impedance.

- Werner

_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core
Werner Almesberger

Re: ECN0024 (GPS ground and layout)

Reply Threaded More More options
Print post
Permalink
In reply to this post by AlvieBoy
?lvaro Lopes wrote:
> Digital ground is very noisy.

Speaking of interference, would it make sense to bead/filter the
digital signals that cross into GPS land, i.e., TX, RX, and INT ?

- Werner

_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core
AlvieBoy

Re: ECN0024 (GPS ground and layout)

Reply Threaded More More options
Print post
Permalink
In reply to this post by Werner Almesberger
Werner Almesberger wrote:

> ?lvaro Lopes wrote:
>> One idea is to have a separate ground for the RF part, and connect
>> it to digital ground at one point, forming some sort of ground peninsula,
>
> Whether to split ground planes or not seems to vary mainly with the
> designer, less with the design.
>
> To take random examples from a quick google search, the pro-split side
> has a quite convincing example here:
> http://www.signalintegrity.com/Pubs/news/9_04.htm
>
> And so does the contra-split side (also check out the link to the 2001
> article):
> http://www.hottconsultants.com/techtips/split-gnd-plane.html

Same guy. :)

"You must also arrange the common connection so that it does not encourage ground currents circulating between the analog and digital regions (or else you
defeat the whole point of isolating the grounds). That's the tricky part about having separate DGND and AGND plane regions: It doesn't help to separate them
unless you also prevent high-speed currents from passing from one to the other. "

That's why junction of the two ground planes should be as close as possible to the PMU. Remember, ground actually "feeds" electrons onto the system.

These analysis are however only for common mode. in RF, most of the signal does not actually travel on the copper, but in the dielectric. Currents flowing
nearby cause electromagnetic fields, which disturb the transmission.

Personally I'd like to not split grounds, but to achieve the same "effect" we might have to change layout a bit. This was primarily intended to avoid SD card
interference.

>
> My interpretation is that splitting ground planes gives you more
> control, but at the cost of increasing complexity and the risk of
> unintended consequences.
>
>> We should also have these
>> ground planes (RF) filled with through-hole vias, to avoid EMI emission
>> (and maybe mitigate interferences).
>
> Putting that RF into a "cage" sounds good to me. This should also help
> to reduce ground impedance.
>
> - Werner


_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core
Werner Almesberger

Re: ECN0024 (GPS ground and layout)

Reply Threaded More More options
Print post
Permalink
?lvaro Lopes wrote:
> Same guy. :)

For a moment, you had me worry that the same guy was arguing both
sides :-)

> That's why junction of the two ground planes should be as close as possible
> to the PMU. Remember, ground actually "feeds" electrons onto the system.

Yup. So would you move the PMU to a move central location ? I
suppose, codec and PMU could trade places (with respect to the
original GTA02 layout) without causing too much trouble.

> Personally I'd like to not split grounds, but to achieve the same
> "effect" we might have to change layout a bit.

What kind of change would that be ?

- Werner

_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core
AlvieBoy

Re: ECN0024 (GPS ground and layout)

Reply Threaded More More options
Print post
Permalink
Werner Almesberger wrote:
> ?lvaro Lopes wrote:
>> Same guy. :)
>
> For a moment, you had me worry that the same guy was arguing both
> sides :-)

He might aswell be :)

>> Personally I'd like to not split grounds, but to achieve the same
>> "effect" we might have to change layout a bit.
>
> What kind of change would that be ?

Avoiding low-frequency return paths (least resistance path) crossing RF areas. Ensuring high-freq returns are not split (least impedance path).

Some references regarding this problem I found interesting:

http://www.cvel.clemson.edu/emc/tutorials/Current_Paths/Current_Paths.html
http://www.elmac.co.uk/pdfs/planes.pdf
http://www.edn.com/ednmag/archives/1995/011995/02df3.htm
http://www.pcb.ddiglobal.com/media/PWB.pdf
http://books.google.com/books?id=H4P3hZ0HKDQC&pg=PA124&lpg=PA124&dq=current+flow+%22copper+plane%22&source=bl&ots=ukWfCiGCiL&sig=-C0RD8ZCNKfdPAS1TyLE8h4bNVQ&hl=en&ei=K6PNSqbpONOgjAeTmqz2Bw&sa=X&oi=book_result&ct=result&resnum=2#v=onepage&q=current%20flow%20%22copper%20plane%22&f=false
http://pcdandf.com/cms/magazine/192/3413-effects-of-plane-splits-on-high-speed-signals-parts-1-and-2
http://emcesd.com/tt2007/tt060207.htm
http://www.commsdesign.com/showArticle.jhtml?articleID=49900228
http://www.cvel.clemson.edu/pdf/EMCS07-05.pdf

Álvaro

_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core
Werner Almesberger

Re: ECN0024 (GPS ground and layout)

Reply Threaded More More options
Print post
Permalink
?lvaro Lopes wrote:
> Some references regarding this problem I found interesting:

Wow, nice collection !

> http://pcdandf.com/cms/magazine/192/3413-effects-of-plane-splits-on-high-speed-signals-parts-1-and-2

I get just "{mosimage}" for the images in this one, using Konqueror
or Firefox. In fact, the HTML says "<p>{mosimage}<br />". Is this
the expected result ?

- Werner

_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core
AlvieBoy

Re: ECN0024 (GPS ground and layout)

Reply Threaded More More options
Print post
Permalink
Werner Almesberger wrote:

> ?lvaro Lopes wrote:
>> Some references regarding this problem I found interesting:
>
> Wow, nice collection !
>
>> http://pcdandf.com/cms/magazine/192/3413-effects-of-plane-splits-on-high-speed-signals-parts-1-and-2
>
> I get just "{mosimage}" for the images in this one, using Konqueror
> or Firefox. In fact, the HTML says "<p>{mosimage}<br />". Is this
> the expected result ?

Unfortunately yes. But text is quite readable. I might try to buy those magazine issues.

Álvaro

_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core