Searching \ for '[EE] Graphics card wiggly traces' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: www.piclist.com/techref/index.htm?key=graphics+card+wiggly
Search entire site for: 'Graphics card wiggly traces'.

Exact match. Not showing close matches.
PICList Thread
'[EE] Graphics card wiggly traces'
2005\06\02@054234 by Dave W Turner

picon face
Hi all,

Yesterday I was looking at a broken graphics card, wondering what all
the chips did.  Then, I noticed that a lot of the traces between chips
wiggled around.  There were some spaces between traces, and one of the
traces, wiggled back and forth, filling the space.  It seemed as
though they were either trying to make the trace as long as possible,
or fill the entire board.

Is this something to do with radio interference, or might there be a
reason the trace needs a tiny bit extra resistance that cannot be
provided by a resistor?

Also, there were about 5 empty capacitor spaces.  Does anyone know why
a lot of comercial PCBs end up with unfilled component spaces and
pads, and traces going nowhere?

--

Dave
All us base are belong to you.

2005\06\02@064634 by Michael Rigby-Jones

picon face


{Quote hidden}

Not quite.  At very high clock speeds typical found on modern mainboards
and graphics card, the propagation speed of the signals becomes
significant.  In order to ensure all the signals have the correct
relative timings, the lengths of the pcb traces are critical.  You will
see the traces "wiggled" where the PCB designer has had to make a trace
longer simply to satify signal timing requirements.  This happens a lot
when you have wide busses (e.g. 32-64bit) and routing becomes a
significant problem.

>Also, there were about 5 empty capacitor spaces.  Does anyone
>know why a lot of comercial PCBs end up with unfilled
>component spaces and pads, and traces going nowhere?

Often this is because the PCB is designed to support multiple variants
of a product, the cheaper, less feature rich versions simply have
components removed.  It's also possible that it's been determined that
some components are not necessary to the correct functioning of the
board and have been removed for cost saving purposes.

Regards

Mike

=======================================================================
This e-mail is intended for the person it is addressed to only. The
information contained in it may be confidential and/or protected by
law. If you are not the intended recipient of this message, you must
not make any use of this information, or copy or show it to any
person. Please contact us immediately to tell us that you have
received this e-mail, and return the original to us. Any use,
forwarding, printing or copying of this message is strictly prohibited.
No part of this message can be considered a request for goods or
services.
=======================================================================

2005\06\02@080717 by Alan B. Pearce

face picon face
>>There were some spaces between traces, and one of the
>>traces, wiggled back and forth, filling the space.
>>It seemed as though they were either trying to make the
>>trace as long as possible, or fill the entire board.
>>
>>Is this something to do with radio interference, or might
>>there be a reason the trace needs a tiny bit extra
>>resistance that cannot be provided by a resistor?
>>
>
>Not quite.  At very high clock speeds typical found on modern
>mainboards and graphics card, the propagation speed of the
>signals becomes significant.  In order to ensure all the
>signals have the correct relative timings, the lengths of the
>pcb traces are critical.  You will see the traces "wiggled"
>where the PCB designer has had to make a trace longer simply
>to satify signal timing requirements.  This happens a lot
>when you have wide busses (e.g. 32-64bit) and routing becomes
>a significant problem.

High end layout packages, such as Allegro, will do this automatically for
you. You group signal lines and specify a delay parameter, and then the
layout package ensures that the tracks meet the requirement. These layout
packages also have a lot of smarts in them to ensure that impedance
constraints, and other parameters that depend on layout features are also
maintained.

2005\06\02@093708 by Mike Hord

picon face
> >Not quite.  At very high clock speeds typical found on modern
> >mainboards and graphics card, the propagation speed of the
> >signals becomes significant.  In order to ensure all the
> >signals have the correct relative timings, the lengths of the
> >pcb traces are critical.  You will see the traces "wiggled"
> >where the PCB designer has had to make a trace longer simply
> >to satify signal timing requirements.  This happens a lot
> >when you have wide busses (e.g. 32-64bit) and routing becomes
> >a significant problem.

This is also why we're starting to see a move away from large
busses of parallel signals and a move towards very high speed
serial links (SATA, PCI Express, FireWire).  It's becoming
easier and faster (as I understand it) to do a serial link than
a parallel link.

Mike H.

2005\06\02@111824 by Herbert Graf

flavicon
face
On Thu, 2005-06-02 at 10:41 +0100, dave.w.turnerspamKILLspamgmail.com wrote:
> Hi all,
>
> Yesterday I was looking at a broken graphics card, wondering what all
> the chips did.  Then, I noticed that a lot of the traces between chips
> wiggled around.  There were some spaces between traces, and one of the
> traces, wiggled back and forth, filling the space.  It seemed as
> though they were either trying to make the trace as long as possible,
> or fill the entire board.
>
> Is this something to do with radio interference, or might there be a
> reason the trace needs a tiny bit extra resistance that cannot be
> provided by a resistor?

Neither. It's all about propagation delay. You'll see this sort of
things most commonly with large, high speed parallel buses. On a
graphics card there are several of these: AGP (if your card isn't PCIE),
and multiple memory buses. When you have a large parallel bus it becomes
more and more important that all the signals reach their destination at
the same time. The "wiggles" lengthen some of the wires so that, in the
end, they are all about the same length.

This is getting more and more common, most motherboards these days have
tons of these "wiggles", and it'll become more common as bus speeds
increase.

Now, serial buses tend to get around these problems since each wire (or
pair) is treated separately, but even with something serial like PCIE
(where you have say 16 pairs of wires each directly, in the case of a
16X slot) you'll still see the "wiggles", in fact you'll see PAIRS of
wiggles!


> Also, there were about 5 empty capacitor spaces.  Does anyone know why
> a lot of comercial PCBs end up with unfilled component spaces and
> pads, and traces going nowhere?

Two reasons I can think of for empty spaces:

1. they were put in so that during final testing they could add more
caps if they noticed more noise then expected. Since the noise they did
see was within specs they didn't need the caps.

2. Many manufacturers use the same board for multiple products, so say
your card only has one bank of memory, the pads for the second bank are
there since the next model up has that memory. Strapping is also another
reason for blank pads (i.e. setting a card for NTSC vs. PAL, although
this is almost always done in the BIOS these days). Other possibilities
are hooks for testing (i.e. pads for a JTAG port, or a "disable VGA"
jumper for booting a system with that board not primary display), or
flexibility if a part becomes discontinued or if a cheaper version might
appear (VERY common in the world of serial flash, many boards have pads
for more then one type of flash part, just in case one part disappears).

As for traces going nowhere? Not sure on that one, again could be
related to test, or perhaps it's simply an error in the layout that
wasn't deemed worth fixing.

TTYL



-----------------------------
Herbert's PIC Stuff:
http://repatch.dyndns.org:8383/pic_stuff/

2005\06\03@090320 by alan smith

picon face
Even these high speed links are length matched.  We
are in a design with SATA and PCI-E, and all those
still have the length matching issues.

--- Mike Hord <.....mike.hordKILLspamspam.....gmail.com> wrote:

{Quote hidden}

> --

2005\06\03@091049 by Michael Rigby-Jones

picon face


>-----Original Message-----
>From: EraseMEpiclist-bouncesspam_OUTspamTakeThisOuTmit.edu [piclist-bouncesspamspam_OUTmit.edu]
>Sent: 03 June 2005 14:03
>To: Microcontroller discussion list - Public.
>Subject: Re: [EE] Graphics card wiggly traces
>
>
>Even these high speed links are length matched.  We
>are in a design with SATA and PCI-E, and all those
>still have the length matching issues.
>

Even more critical as the data rates are very high.  However, it's much easier to achieve as there are far fewer high speed lines than a parallel bus.

Regards

Mike

=======================================================================
This e-mail is intended for the person it is addressed to only. The
information contained in it may be confidential and/or protected by
law. If you are not the intended recipient of this message, you must
not make any use of this information, or copy or show it to any
person. Please contact us immediately to tell us that you have
received this e-mail, and return the original to us. Any use,
forwarding, printing or copying of this message is strictly prohibited.
No part of this message can be considered a request for goods or
services.
=======================================================================

2005\06\03@104838 by Herbert Graf

flavicon
face
On Fri, 2005-06-03 at 14:09 +0100, Michael Rigby-Jones wrote:

> >Even these high speed links are length matched.  We
> >are in a design with SATA and PCI-E, and all those
> >still have the length matching issues.
> >
>
> Even more critical as the data rates are very high.  However, it's much easier to achieve as there are far fewer high speed lines than a parallel bus.

Not always. Consider a 16X PCIE connector, it has 32 wires in each
direction, meaning those 32 wires have to be pretty closely length
matched, for each direction.

32X PCIE connectors are as bad as DDR and PCI in number of wires to
route, just much higher frequencies (although 32X connectors haven't
appeared yet it is in the spec). This will only get worse with PCIE Gen
2.0 which will be likely running at 5GHz.

TTYL

-----------------------------
Herbert's PIC Stuff:
http://repatch.dyndns.org:8383/pic_stuff/

More... (looser matching)
- Last day of these posts
- In 2005 , 2006 only
- Today
- New search...