Tiny twisted-pair transmission line solves test-fixture woes
Commonly available wire fits in tight spots.
Glen Chenier, Allen, TX; Edited by Brad Thompson and Fran Granville -- EDN, December 16, 2005
Engineers often construct test fixtures that include high-speed differential signals. Although miniature coaxial cable is widely available, there's no commercial off-the-shelf source for small-gauge twisted-pair cable that's suitable for differential signals. Although Category 5 Ethernet cable contains four twisted pairs, it's too large for crowded fixtures and for attachment to the Amp Z-Pack connectors some fixtures require. Many engineers are unaware that they can twist together two lengths of AWG #30 Kynar-insulated wire—garden-variety wire-wrap and prototype cut-and-jumper wire—to make a 102Ω differential-transmission line. If you use Kynar's dielectric constant and the insulation's thickness to compute its properties, the line's calculated differential impedance works out to 110Ω. In practice, differential TDR (time-domain-reflectometer) measurements show that the line's actual impedance consistently measures 102Ω—only 2% away from the target impedance and thus close enough for most practical purposes.
To make your own twisted pair, start with a long AWG #30 Kynar-insulated wire and fold it in half. Enlist a co-worker to hold the cable's closed end by slipping the loop around a screwdriver's blade. If you're working alone, slip the loop around a doorknob. Tightly twist the two wires' free ends together and insert the twisted ends into the chuck of a Dremel (www.dremel.com) rotary tool. Tighten the chuck and hold the Dremel tool so that the wires are stretched tightly, are of the same length, and lie parallel with each other.
Apply a slight amount of tension to the wires and start the tool. As the wires twist together, the pair shortens and pulls the tool's operator toward the loop support. A variable-speed Dremel tool works best when you operate it at its slowest setting. If you have only a fixed-speed Dremel tool, avoid overtwisting the wires by preparing a length of 10 to 20 ft of cable at a time. The extra length allows time to switch the tool off and avoid overtwisting the wires. Cut off and discard the cable's nonuniformly twisted end sections.
The amount of twist in the wires is not critical, but the wires should be firmly twisted together. Using approximately eight to 10 twists/in. works well. To count the twists, hold a portion of the cable against a ruler or measuring scale under a magnifier and count 16 to 20 "bumps," or half-twists, per inch. Using too many twists per inch uses excess wire and increases losses and propagation delay. For the lengths in a test fixture, losses are insignificant except at extremely high frequencies.
You can also use a variable-speed hand drill with a ¼- or 3/8-in. chuck to twist the wires, but you need to fold the wires' free ends several times and wrap them in duct tape to ensure a snug fit in the drill's chuck. When using any power tool, wear safety glasses or other eye protection during the procedure.
-
I learned the bench vise / eggbeater-drill variation of the technique as a co-op student in the 70s. We were builing homebrew switching supplies for a low current 300 Vdc and needed nicely coupled bifilar windings.
Frank Caherty - 2008-22-5 15:09:00 PDT -
This was standard practice back in the early 70's at Marconi....Except we didn't use a friend we would put the spare ends into a vice and pull hard using a hand drill to spin the wires together....An extra tug at the end of twisting the wires ensures that the wires 'set' and stay twisted.
John Marjoram - 2006-10-2 02:43:00 PST -
I use a piece of #10 or 12 steel wire [a la coat hanger]bent into a shape similar to a cup hook. Put one end of the wires into the hook in a cordless drill and the other ends into a bench vise. Drill away!
The whole thing seems to work better with solid wires when the wires have been pre-stretched a bit. Stranded doesn't seem to care, it is unruly. Hot melt glue is marvellous for keeping the twisted pair from unravelling all over the bench. I just dot some on every 3-6 inches while it is still in the "fixture".
Jim Keyes - 2006-8-2 19:31:00 PST -
Let's be nice to those poor production guys. With SOX (Sarbannes Oxley) auditors breathing down their necks, they don't need engineers 'stealing' production inventory, and messing up their counts. Most places have a relatively simple form to fill out to get production supplies into the engineering lab.
I've used that drill technique for twisting all sorts of wire pairs, even decent signal level channels. It lets you keep good differential signals, while allowing you access to one side if you need to monitor currents in the line.
Kevin Aylward - 2006-6-2 10:55:00 PST -
I use this technique using both wire wrap and magnet wire. For magnet wire, the characteristic impedance of the twisted pair is a strong function of the number of turns per inch, and I presume it is for wire-wrap wire as well. I will often attach a square-wave signal source through the desired impedance to the twisted pair at one end, terminate the other end in the same impedance, then twist with the drill while watching a scope attached at the source until the reflections disappear. With 24 AWG enameled magnet wire, about 4 turns per inch gives a 32 ohm Zo, while 50 ohms needs about two and a half turns per inch.
Glenn Dixon - 2006-11-1 07:06:00 PST
