crosstalk

Ish. Alternating traces going in opposite directions is a big help, because the coupling doesn't build up that way.

It depends on how far they go together, how fast the signals are, and how vulnerable the receiving trace is.

Cheers

Phil Hobbs

This board has a MicroZed that runs a lot of fast signals to 8 plugin boards. The pinouts were selected (not by me) to minimize crossovers and vias without much thought to signal speeds or directions. The easy fix is to just spread them out.

The traces are, quite by accident, the correct 75 ohms, 8 mil wide microstrips.

Playing with the diff pair calculator in Saturn, one would like a diff pair to be 75 ohms per trace and 150 differential, which implies zero coupling. I get that with about 50 mils between traces, 12 mils of FR4 to the ground plane.

With 8 mil traces and a 16 mil gap, 12 mils down to the plane, I see

75r per trace and 131 differential, which is mediocre isolation.

A 24 mil gap is better, 75 and 141 ohms. 24 is accidentally 2x the 12 mil dielectric thickness.

A ground strip between individual signal-carrying traces provides extra isolation between them. If John Larkin had read enough to become knowledgeable about PC traces. he would have seen that mentioned.

That's not a great fix for ground-loop (inductive) problems. Area inside the loop is an interfering-signal coupler, so 'spread' is a square-law interference increaser.

But not if there is a ground plane underneath as the loop is then the area of the vertical cross section between the track and the ground plane which does not depend on separation between tracks. John

For digital (where density is really needed) within 100-200 MHz we have been doing 4 mil trace 4 mil spacing for ages now. That on the visible (top and bottom layers), referenced to GND planes beneath each. No issue whatsoever. Well, actually I had one, an I2C line was passing too close (probably not 4 mil, may be a whole mm) to a flyback switch (IRF540-ish) for the HV which was doing nice

100V excursions and at times managed to upset the i2c. Pulling the latter up with 1k on each line fixed it (was 2k IIRC).

But spreading the traces is a great fix for crosstalk. Try it.

Some of our traces will be fast 8b10b data streams. At powerup time we "train" the receivers to adapt to the actual data timing. Crosstalk from other signals can wobble the bit edges and potentially make data errors.

I have done something around 100-150 MHz clocked video squeezed into lvds (don't know how much faster the lvds was clocked, some 8 times I believe) using the same routing but not many times and over a short distance, well within an inch. No problems there, either.

How to apply that fix, though, to a cat5 cable's four pairs? In one instance, I found a DC/DC converter that needed its input power through a common-mode bead, because it crosstalked to components a foot and three circuit boards away. Freeze mist on the converter changed the frequency of the interference, after other inspections didn't find the problem.

Ouch... Sounds nightmarish. Many years ago I had a share of nightmarish inductive coupling, a tiny (20mm side IIRC) fan turned out to get into the input signal of a moderately sensitive amp (tens of MHz bandwidth, the interference was well visible at hundreds of mV fullscale at the input). There was nothing I could do other than just get rid of the fan as part of the design. Which worked, luckily it turned out I had enough margin for that.

Crosstalk is negligible between traces on the same plane. It is between traces on different planes running along the same paths that create crosstalk. Very easy to avoid by paying attention when routing... if you don't use an auto-router.

The John Doe troll stated the following in message-id <sdhn7c$pkp$ snipped-for-privacy@dont-email.me:

And the John Doe troll stated the following in message-id <sg3kr7$qt5$ snipped-for-privacy@dont-email.me:

And yet, the clueless John Doe troll has continued to post incorrectly formatted USENET articles that are devoid of content (latest example on Sat, 14 May 2022 17:46:49 -0000 (UTC) in message-id <t5opu8$aeb$ snipped-for-privacy@dont-email.me).

NOBODY likes the John Doe troll's contentless spam.

Further, Troll Doe stated the following in message-id <svsh05$lbh$ snipped-for-privacy@dont-email.me posted Fri, 4 Mar 2022 08:01:09 -0000 (UTC):

Yet, since Wed, 5 Jan 2022 04:10:38 -0000 (UTC) Troll Doe's post ratio to USENET (**) has been 60.5% of its posts contributing "nothing except insults" to USENET.

** Since Wed, 5 Jan 2022 04:10:38 -0000 (UTC) Troll Doe has posted at least 1364 articles to USENET. Of which 148 have been pure insults and 677 have been Troll Doe "troll format" postings.

This posting is a public service announcement for any google groups readers who happen by to point out that the John Doe troll does not even follow the rules it uses to troll other posters.

Fr7RG8qvqnxH

The John Doe troll stated the following in message-id <sdhn7c$pkp$ snipped-for-privacy@dont-email.me:

And the John Doe troll stated the following in message-id <sg3kr7$qt5$ snipped-for-privacy@dont-email.me:

And yet, the clueless John Doe troll has itself posted yet another incorrectly formatted USENET posting on Sat, 14 May 2022 17:53:48 -0000 (UTC) in message-id <t5oqbb$aeb$ snipped-for-privacy@dont-email.me.

Further, Troll Doe stated the following in message-id <svsh05$lbh$ snipped-for-privacy@dont-email.me posted Fri, 4 Mar 2022 08:01:09 -0000 (UTC):

Yet, since Wed, 5 Jan 2022 04:10:38 -0000 (UTC) Troll Doe's post ratio to USENET (**) has been 60.5% of its posts contributing "nothing except insults" to USENET.

** Since Wed, 5 Jan 2022 04:10:38 -0000 (UTC) Troll Doe has posted at least 1367 articles to USENET. Of which 148 have been pure insults and 679 have been Troll Doe "troll format" postings.

This posting is a public service announcement for any google groups readers who happen by to point out that the John Doe troll does not even follow the rules it uses to troll other posters.

LujEfVNcdpEE

The sort of thing that John Doe likes to post. He doesn't have a clue about transmission lines, so imagines he can get away with posting a false and totally unreasonable claim.

In fact what I was proposing was the use of multiple grounded coplanar waveguides on the same board.

If I spent enough time at it I could probably find a specific example, but John Doe wouldn't understand that either, and John Larkin would ignore it.