Length difference within a LVDS pair at coxial connecter

[LF666]

Hello guys, there is a huge length (more than 10mm) difference within one LVDS pair while I am trying to connect the + and - signal to one coxial connecter.
And I found out it’s mainly because that when I have connected all the ‘‘2’’ (OUT8-) togther, the length of the wire will increase.

And the difference will be much smaller if I only connect the negative signal to one of the four ‘‘2’’ pads and leave the rest of them unconnected:

So should I use the second way of connection ? Or the first one is better and I need to adjust the length of another wire by ‘‘Tune skew’’?

Or maybe two ways are same because in both situation the negative signal will first arrive the pads on the left below ?
Thanks

[craftyjon]

Normally you would not put one half of a differential signal over the outside (shield) of a coax connector. You would either use two different coax connectors (one for + and one for -) or you would use twin-ax cable and connectors.

[LF666]

Hey, thanks for the response.
Actually I am using two connectors for one pair from the begining. But my supervisor told me to make the connection in this way. According to him that one coxial cable could transfer the positive and negative signal in one pair at the same time.

[craftyjon]

I don’t think your supervisor’s plan is not a good one, unfortunately. The whole point of using differential signaling is to reject common-mode disturbances, but the construction of a coax cable means that the inner conductor will almost certainly couple to external disturbances differently from the shield, meaning that the disturbance will no longer be common-mode.

Without knowing more about your design, I’d wonder if a twisted-pair cable might be a better fit for your application than a coax.

[twl]

I second Jon. Maybe your supervisor should get a better understanding of signal propagation delays themselves before giving such assignments to students

One would usually use a transformer matching the impedance of the diff pair to the impedance of the coax cable:

T.

[glenenglish]

supervisors plan is a bad one…WTF.
you only need to connect to one pin anyway because once you pick up the connector, you’ll see it is a chunk of brass.

use two connectors, each side of the diff pair to each inner…

Or use twl’s answer with the transformer, a 2:1 impedance tranny (flux couple mode) use something like a MCL TC2-72T . 2:1 Z ratios dont have super bandwidth. but there are exotic options…

Or use a 1:1 tranny as a common mode choke. (but Z will be wrong , maybe depending on application)
solution is super wideband with correct choice of tranny