Eng-Tips is the largest engineering community on the Internet

Intelligent Work Forums for Engineering Professionals

GigE thru a military connector

Status
Not open for further replies.

JimRogers

Electrical
Oct 2, 2006
6
0
0
US
I need to run a GigE channel through a military (38999) connector. Does anyone have any experience with using adjacent pins (#22) in a 38999 for high frequency (~250 MHz) signals.

The Cat5e impedance is 100 Ohms and my quick calc shows that the 38999 contact pair will be about 200 Ohms.However the length of the discontinuity is only about 40 mm. The total cable length will be 5 meters and shielded Cat5e will be used.

Will the GigE signal get thru OK?
 
Replies continue below

Recommended for you

Yes. Gigabit Ethernet is being used on more and more military avionics. And those sorts of Mil connectors are still required. Our latest systems used slightly smaller type of mil connectors, but still in the style of 38999.

The pairs were wired to common sense pins (impedance bumps). We had no evidence of problems.

What we did have is every possible miswire in the book.
 
V,

How far did you get between untwisting of one pair and twisting of the mating pair between connectors? (i.e., how long from pin cup to pin cup)

Dan - Owner
Footwell%20Animation%20Tiny.gif
 
A mated pin socket pair seems to be about 40 mm. 1/10 of a wavelength at 250 MHz in the Cat5e cable is about 76 mm. That is the reason that I think that the approach will work.

I will be using the PC pin approach at the source end so I will come onto the board via a standard RJ45, 100 Ohm diff lines to the connector and out. The other end will connect directly to a Cat5e pigtail. So there will be 2 50 mm discontinuities.

I have since found a manufacturer that uses 38999 is the same fashion for their product, which is a good indication.
 
The assemblers maintained the twist as close as possible to the pins. But even still, the untwisted section would have to be about 4 cm +/-. Once we sorted out the miswires, the boxes reliably connected in the 1Gbps mode.

Noise is not an issue with the double shielded cable we used.

 
Our boxes used Glenair Might Mouse series 805 connectors. We didn't have any evidence of problems arising from the non-Ethernet connector (impedance bump, crosstalk).

The problems we DID have were the following:
1) silly wiring errors
2) selecting cable and crimp pins (gauge)
3) ensuring 360° shield termination
4) sorting out the "RJ45" (8P8C) end

Surprisingly, somewhere out there is an alternative Ethernet wiring standard that is different than T-568A and it isn't the expected cross-over nonsense. It's another 8P8C pin/wire colour assignment that results in Channel C being wired backwards ( + <> -). I can't even explain its existence.

Another thing to worry about is scrambling the twisted pairs such that the continuity rings-out 'correct', but the pairs are scrambled up. We worried about this enough to avoid it.

All these silly wiring problems can be avoided if you have one person designing the entire path, end-to-end. And make the associated drawings ultra-detailed (specifying every possible detail, even to the point of providing redundant info as a built-in cross-check).

Obviously select cable to match Cat5/6, and then choose the connector and crimp pins to match the wire gauge.

Wiring inside the box(es) should pay attention to the Cat5/6 requirements.

Terminating the cable shields with neat 360° shield termination, as opposed to a ground wire pigtail, is something our E3 guru demands (and we respect his opinion highly). But it takes some effort to implement it.

And if you have 'RJ45' (8P8C) connectors in the system, then you may be tempted to use fancy RJ45 connectors with an embedded shield that grounds itself at the (for example) laptop end. Make sure the connector actually fits the device before you bother. Same thing with those silly RJ45 boots; they look silly shoved half-way up the cable because they don't fit.

 
Status
Not open for further replies.
Back
Top