Signal Ground Drain Wire

Found a good book!

Alert readers informed me about this book a few weeks ago, after my reply to one reader on the subject of differential transmission line impedance. At the time I said I didn't know of a good reference on the subject -- now I do!

Brian C. Wadell, Transmission Line Design Handbook, Artech House, 1991. ISBN: 0890064369 This book has been out there a while. I don't know how I managed to miss it. It includes handy approximations for transmission line impedance, delay, skin effect loss, dielectric losses, and radiation losses -- very comprehensive. The author addresses most of the popular transmission line formats in use today, including microstrip, buried microstrip, offset stripline, and edge- or broadside-coupled differential striplines.

Brian heavily references the original research articles and measurements. He doesn't pull any punches when the research is fuzzy, or in conflict. He shows you just what we know, as well as what we don't know.

If you're looking for good approximations, this is the source. Now on to this week's technical issue.

Signal Ground Drain Wire

We have a cable phenomenon that has been puzzling us for some time now. The cable carries 4 differential, signal pairs at 600MHz and the connector uses a 2x5, (2mm spacing) chicklet construction as follows:

     ----------------------
     | s1+|s1-|gnd|s3+|s3-|   
     ----------------------
     | s2+|s2-|gnd|s4+|s4-|
     ----------------------
   

The cable itself uses shielded twin axial construction as follows:

     +++++++++++++++++++++++++++++++++++++++++++++++++
     +  -------------------------------------------- +
     +  |    -          -    ||    -          -    | +
     +  |  -   -   G  -   -  ||  -   -   G  -   -  | +
     +  | -  S  -    -  S  - || -  S  -    -  S  - | +
     +  |  -   -      -   -  ||  -   -      -   -  | +
     +  |    -          -    ||    -          -    | +
     +  -------------------------------------------- +       
     +  -------------------------------------------- +
     +  |    -          -    ||    -          -    | +
     +  |  -   -   G  -   -  ||  -   -   G  -   -  | +
     +  | -  S  -    -  S  - || -  S  -    -  S  - | +
     +  |  -   -      -   -  ||  -   -      -   -  | +
     +  |    -          -    ||    -          -    | +
     +  -------------------------------------------- +      
     +++++++++++++++++++++++++++++++++++++++++++++++++
   

The outer box represents an EMI shield and is connected to chassis ground. The four inner boxes represent a foil shield around each pair and is terminated to signal ground. The 'G' represents a "drain wire" and is normally connected to the foil shield at both connectors and terminates into both (signal) gnd pins in the chicklet.

The problem we are encountering is that the outer conductor of the differential pair experiences significant rise/fall time degradation on the order of 2X that of the inner conductors when measured single-ended. However, when we look at a cable which does not have the "drain wire" terminated at either chicklet (i.e. the drain wire is floating) the rise/fall times are much better matched. Since we already have an EMI shield around the entire cable, I wonder if the "drain wire" is necessary.

Are there any caveats we should be aware of when leaving this wire unterminated? Will return currents flow through the foil shield? What if the foil shield is likewise unterminated. Why should disconnecting the "drain wire" at the connectors have such a drastic impact on the rise/fall time of the outer conductors?

One last puzzling item is that cables from another supplier using a similar construction had an inverse response - the degraded r/f times were on the inner conductors?!

Thank you,
Bob Kaminsky

 

Thanks for your interest in High-Speed Digital Design, and for writing to me about this interesting problem.

I am assuming that the drain wire is the ONLY connection between the individual signal shields and the internal ground pin of the chicklet connector, and that the individual signal shields are insulated from the outer EMC shield and from each other.

My guess is that the drain wire connection includes some significant inductance. You can measure this to confirm my suspicions, using a special jerry-rigged connector. There is no cable attached to this connector. On the cable side of the connector, just short together one signal wire (sig+) and the drain wire, providing a length of drain wire comparable to what you believe would be in place in an actual installed cable. Now you are blasting a known amount of dI/dt through the connector on the signal wires, and back through connector on the drain-wire connection to ground. Measure the voltages present on the drain wire on the cable side of the connector.  That should give you a feeling for the significance of the drain wire inductance.

My guess is that it will be significant.  Drain wires don't usually work well for high-speed stuff.

Now, on to your specific problem.  If you drive only one of the wires of the differential pair, the return current for that wire must flow capacitively to the signal shield, and from there back through the shield drain wire to ground. This current, as it passes through the inductance of the drain wire, causes voltages on the shield.  You should be able to measure these voltages. You should also see that the shield voltages so induced differ from one wire to the other, and from cable to cable.  This is because of subtle differences in the positions of the two signal wires with respect to the drain wire within the connector (which affects the inductance) and the separation between the signal shield and the outer cable shield (which affects the capacitance).  You probably will see a funny, wiggly, resonant-looking waveform on the signal shield. My guess is that this funny resonant signal is then coupling back to your signal wires and causing the risetime differences that you observe. In actual operation, only the DIFFERENCE in drive currents will excite the shield voltages, therefore I would expect them to be reduced in amplitude. Please let me know if my conjectures are correct.

Assuming that we have a good model for what is causing the effect, let's move on to a solution.

Apparently, the inductance of the drain wire prevents its being useful as a high-speed ground.  It is causing some resonant behavior between the signal shields and the outer cable ground. Therefore, we would like to eliminate this resonance.  There is a problem, however, with simply opening the drain-wire connection. If we do that we may introduce other weird resonant effects having to do with the floating shield.  What I suggest is that you cipher out the impedance of the transmission line formed by the signal shield and the outer ground, and terminate each drain wire in a resistance equal to that impedance. At that point the drain wires won't be effective as signal shields (they aren't anyway), but neither will the floating shields be able to resonate.

Best regards,
Dr. Howard Johnson