Lack of foresight can push designers into using cable assemblies that garble signals or are unnecessarily expensive.
By JOHN A. CLARAS
Edited by Leland Teschler
Among design engineers, electrical cables get no respect. They are frequently an afterthought, being among the last items specified during a development effort. This practice can cause difficulties on a variety of levels. Budgets for signal levels are often used up by the time cabling gets considered. There is no such thing as zero-loss cable to bail out engineers who find themselves in such predicaments. Nor is there zero-dollar cable for economically challenged projects. And going with the leastcostly cable/connector assembly exposes the unwary to a special set of land mines.
A few war stories help illustrate common mistakes. Consider the case of one company that left too little space for a connector on the circuit board it was designing. The prayers of engineers there were answered by a newly developed connector that had a tiny footprint. Its manufacturer also quoted an aggressively low price.
Unfortunately, things were not exactly as they seemed. The small connector and its mating piece were sole sourced. Its manufacturer practically gave away the connector to get the cable-assembly business. The result: The board maker took a $100 hosing for every assembly the connector vendor supplied. A more typical price for the 500-MHz, singlecoax cable assembly would have been $15 tops.
As a general guideline, exotic connectors probably kill five projects for every one they save. It is usually better to re-lay-out a circuit board rather than pay for a poor choice on every product produced.
Perhaps the worst example of this phenomenon was a systems company that paid $100,000 for each 25-conductor harness supplied by an "exotic" connector maker. The harness worked fine. It should have cost $2,500. Who made the most profit on this deal?
WHAT DOES IT DO?
There are some basic questions to be answered about any cable application. They may sound obvious, but experience shows many engineers don't really understand their system needs. Among the primary concerns are the amount of signal loss that is acceptable, the degree of shielding required, whether or not the cable must flex, and whether it carries analog or digital signals.
The problem with digital signals is that they can require a cable bandwidth that is deceptively high. Engineers sometimes forget that a square wave has a frequency content that consists of the fundamental frequency plus its odd harmonics. Thus it is a mistake to think that a cable carrying a 100-MHz clock signal needs a bandwidth rating of just 100 MHz.
In reality, cable bandwidth for this case must be wide enough to transmit the energy at 300, 500, 700, 900 MHz, and so forth. Most signal energy is concentrated in the first several harmonics, though the signal spectrum theoretically extends to infinity.
The penalty for insufficient cable bandwidth is pulse rounding. The squarer the square waves the higher the transmitted harmonics. Cables that attenuate high-frequency signals more than lower-frequency terms effectively diminish the rise time of transmitted pulses. Thus there are rules of thumb for specifying cable bandwidth: A cable bandwidth of three times the pulse rate is considered the bare minimum. Cables carrying fast rise-time pulses should use 5 as the goal, 10 for a high safety margin.
Shielding is another often-misunderstood area. Consider the case of a manufacturer that designed a system with RG-316 cable. This low-cost coax provides 50 dB of electromagnetic shielding. Unfortunately, the chassis through which it passed radiated more than that amount. The result was errant clock signals blasting their way into the cable and confusing a microprocessor.
The key to better shielding is distance. EM signal strength falls off with the cube of distance from the radiator, so the farther away from a source, the better. Trouble may arise when cable passes near clock chips, oscillators, or any other type of EMI emitter. This is particularly true for power cables and especially for those of amplifiers: Amplifiers behave like oscillators when they are powered by a signal that is modulated.
Fortunately, shielding does not greatly boost the cost of coax. The price difference between cable shielded to 50 dB and that in the 100-dB range can be 10% or less.
The situation differs for multiconductor cable, however. Twisted pairs provide some protection against differential signals. But adding a foil shield to ordinary wire may double its cost. The best approach is to limit the use of multiconductor power cables to short distances or use shielding for longer cables.
Finally, the coax exhibiting the best shielding is the semirigid type. Most semirigid cable employs either tin-plated copper (TP/CU) or annealed aluminum (TP/AL) jackets and has an appearance resembling that of automotive brake lines. The latter is sometimes called easy bend type and is also inexpensive. Also available is a polyolefin-coated aluminum version that is surprisingly flexible. The polyolefin functions as a mechanism that distributes work hardening in the aluminum to promote pliability.
Semirigid coax has gotten a bad rap for being costly. In reality, an SMA connector on semirigid can be less expensive than conventional cable assemblies, particularly above 2 GHz.
Overspec'ed cable can be as big a problem as underestimating requirements. Some engineers, for example, specify a VSWR (voltage standing wave ratio) at a high level, say 1.03:1 (that is, –36.6-dB return loss). (As a quick review, VSWR in a transmission line is the ratio of maximum to minimum voltage in a standing wave pattern. It is a measure of impedance mismatch between the transmission line and its load. The higher the VSWR, the greater the mismatch, the more signal reflecting back from the load. The minimum VSWR, unity, corresponds to a perfect impedance match and thus zero reflected energy).
The problem is that cables with VSWR close to unity will work great but can be expensive. Experience shows that most stable systems perform well with 1.25:1 cable (–19.1-dB return loss). It's best to check allowable VSWR levels with actual system measurements.
Understand, however, that VSWR tends to be important for frequencies of about 100 MHz and above. It can generally be ignored at lower frequencies and, correspondingly, in multiconductor cable confined to sub-100-MHz use.
Engineers should also think carefully about cables billed as "low loss." The danger is in paying for something you aren't really getting. Two things mainly determine loss qualities no matter who manufactures the cable. Loss is proportional to the surface area of the center conductor, so larger cables will have less loss. Second, cables with a foam dielectric have less loss simply because foam's dielectric constant is close to that of air. So in a nutshell, a large cable with a foam dielectric will have the least loss.
In addition, silver plating on the center conductor can reduce loss somewhat. And solid-center conductors rather than stranded versions will have less loss, perhaps because of the woven nature of the stands.
Ideally, suppliers providing cable assemblies should RF-test the products they ship. Frequently they don't, however. They limit checks to a pull test and a dc test for continuity. The trouble is neither procedure will detect poor ground returns. These arise when crimping and stripping procedures aren't up to par. They manifest themselves as intermittent transients, crosstalk, or random variable attenuation. Small movements can put cables plagued with this problem in and out of specification.
The cure is to have the supplier test the cable assembly at frequencies of interest. The extra cost tends to be minimal.
Want to specify a cable assembly that is hugely expensive and performs poorly? Just put a large connector such as a 7/16 DIN or Nseries on a small RG-178 (0.07-in.-OD) or RG-316 (0.1-in.-OD) cable. Designers sometimes take this approach when a specification spells out use of a large connector, and space constraints dictate use of a small cable.
But large connectors don't mate well with small cable. There are mechanical issues related to getting a good crimp. The small cable tends to shear off easily at the connector. The center pin on an N-Series connector is physically larger than the center conductor of cable such as RG-178, causing a discontinuity at the connection that results in a significant return loss.
There are similar problems with a small connector on a big cable. This sometimes arises when engineers try to terminate a large low-loss cable on a spot too small to hold an appropriate connector. There are both electrical and mechanical challenges when placing a tiny MMCX connector on a big RG-214 (0.40-in.-OD) cable. As in the case of big connector/small cable, the connector center pin is a different diameter than the cable center conductor. Signals reflect at the connection. Getting a mechanically robust crimp is an issue as well.
The irony is that this difficulty is most likely to happen with too-small connectors on large-diameter, low-loss cable. In this case all the advantages of low loss get dissipated in the cable/connector interface.
Loss considerations also enter in when using right-angle connectors. Both VSWR and insertion loss are worse than in swept-straight connectors. Right-angle versions cost more, so they should be specified only when absolutely necessary. In one case, for example, a systems company switched 15 cable assemblies from right-angle SMAs to a sweep style. The change halved costs and improved return loss by 3 dB.
Finally, consider cable phase matching for gigahertz-range systems that measure distance or intervals by examining signal arrival time. Phase matching is a process of building cables having the same or similar electrical lengths. Not all suppliers of cable assemblies can provide phase matching. Dielectric variations don't permit cables to be matched by pure physical likeness. The process generally involves tweaking cable lengths while examining the response at the frequency of interest on a network analyzer.
Bear in mind that the tighter the phase or the higher the frequency, the higher the cost simply because the tuning process takes more time. Tougher phase specifications also sacrifice VSWR.