|
|
Posted by Smarty on 02/25/07 04:27
Charles,
This is a really surprising and clever design technique, and one which I
never knew existed. The benefit will degrade from any lack of symmetry in
rise times and fall times such as to not exactly cancel one-another, but
apparently this is not an issue in the Dell designs. Also, the approach
works well in the digital domain, but it would be unavailable as a design
option when doing mixed or analog-only PCB or backplane layout.
As another hardware designer, you will probably appreciate the single most
impressive piece of analog shielding I have personally ever seen,. which was
in a fairly recent ICOM wide band receiver I purchased which continuously
tunes from VLF to microwave frequencies. You can literally tune the receiver
to any of the frequencies where the internal local oscillators of the radio
reside / operate including the BFO, and 3 IF/mixer stages, and not hear even
a faint hint of these signals even though they are being generated a few
inches away from a super-sensitive FET front-end with .1 microvolt
sensitivity. In this case, the shields were literally double copper Faraday
cages, soldered along their entire seams and to the ground plane of the PCB.
The isolation was at least 100 dB, probably more.
This is off-topic for this newsgroup, so I will end it here.
Smarty
"Charles Marslett" <c-deletethis-marslett@yahoo.com> wrote in message
news:57d1u2pigavvmec08m0dde7roueb5h9331@4ax.com...
> On Mon, 19 Feb 2007 08:56:07 +0100, Martin Heffels
> <goofie@flikken.net> wrote:
>
>>On Mon, 19 Feb 2007 01:57:39 -0500, "Smarty" <nobody@nobody.com> wrote:
>>Yeah, the software radios are quite nice. It's amazing how they can keep
>>out the RF generated by the computer, while some of them are crammed
>>inside
>>the computer :-)
>>
>>-m-
>
> At first glance it's true, but then you think a bit about the way
> those cheap boxes are designed (at least at Dell, I work there!) -- to
> minimize the shielding (cost) you do a lot to balance the radiation.
>
> In a previous life (before VLSI was bought by Philips) I worked on a
> project to eliminate most of the shielding requirements entirely --
> just make sure that every signal was balanced with one very nearby. It
> works really well at GHz frequencies and it saved on wires as well.
> You just run 20 wires from here to there (< 1/2 inch of board width
> even with a 4 layer board) and make sure that any H/L transition is
> balanced with a L/H transition, make sure there are enough to power an
> A/C to D/C converter in the second chip and you wind up with no power
> plane (or power plane radiation) and no signal RFI. And very little
> conducted EMI.
>
> At a couple of inches you could do decent RF without much (or any)
> shielding. Of course, computers are getting smaller every year, so it
> might be a bit more challenging now to find any spot 2 inches from the
> CPU and memory.....
>
> --Charles
Navigation:
[Reply to this message]
|