It is possible to optimize miniature interconnect systems that are designed to operate at 1 GHz and above in mechanically demanding environments. However, doing so requires the consideration of many issues that wouldn't apply in slower, static applications. Control of issues such as signal integrity and EMI must be established up front in the design cycle, and thereafter maintained over the lifespan of the product.
Applications continue to challenge the designer in terms of size, weight, routing freedom, range of movement, and the number of signal lines. This makes it no easy task to maintain the integrity of these signal lines at the lowest applied cost.
Fortunately, technology improvements, combined with refined design practices, have helped. They've made sure that cabling solutions, optimized for price/performance, can be used well into the 2-GHz frequency range for the lifetimes of many systems. One important design practice involves the use of proper cabling and cable materials.
A cable-interconnection system that's longer than a tenth of the operational wavelength in analog systems is usually treated as a transmission line to ensure adequate signal integrity. For example, a 1-GHz analog signal with a velocity of propagation (VP) of 85% will fit this criteria when the interconnect-system length is 2.55 cm or longer. Digital interconnect systems should also be treated as transmission lines to help the square or trapezoidal data pulses maintain good shape. This typically requires that the tenth harmonic of the fundamental data rate be apparent.
A primary means to preserve signal integrity is through impedance matching of interconnect-system components. Because signal power is reflected by impedance discontinuities, all of the interconnection points in the transmission chain of source, cable, and load need appropriate degrees of matching. When matching of components isn't possible, physical features should be used to avoid abrupt transitions and minimize reflections.
As a rough guide, impedance-controlled cabling provides for data rates in excess of 1 GHz. Coaxial-cable impedances generally range from 50 to 100 Ω. At higher frequencies, coaxial cable approaches the performance of triaxial (double-isolated shielded) cable. That's because the signal tends to flow on the inside of the shield, and the noise on its exterior.
At high speeds, cable-interconnection points require controlled geometries and consistent adherence to processing standards and workmanship. Breakout geometries can significantly affect performance at higher speeds, since shields are typically removed to expose inner signal conductors for termination. This shielding gap creates a local inductive impedance discontinuity.
In addition, hand stripping and termination of ribbons may produce too much length variation, leading to excessive crosstalk and skew variation in timed signals. For this reason, the first-tier cabling houses use lasers and dedicated tooling to provide better length-tolerance control. Field rework and repair should generally not be attempted.
Using matched-impedance connectors and printed circuits is fundamental if they are not to degrade the cable's performance. When multiple signal paths converge on a common connector or printed circuit, sufficient grounding pins or traces must be allocated for signal paths that aren't otherwise separated by ground planes. Generally, increasing the ground-to-signal ratio of connector contacts or interdigitated grounds will support higher speeds by allowing closer impedance matching and decreasing capacitively coupled noise (crosstalk).
Separating terminations by using larger connectors or printed circuits also helps reduce crosstalk by the inverse of the separation squared. But when space is at a premium, this option proves unattractive. Maintaining low-resistance grounding paths is important in avoiding changes in reference voltages that can, like crosstalk, produce inaccurate data interpretation in digital systems.
Overall cabling-system rise time will be relatively unaffected if connectors and transitional pc boards are mismatched by a few ohms. Nor will the rise time be affected if they're very short in comparison to the cable length. The cable rise time will tend to dominate under these conditions.
When choosing the cable to be used, keep in mind that ribbon coax cables are especially suited to single-ended signals. This contrasts with shielded pairs, typically used for differential signals. By using the center conductors of two coaxes for signals and electrically connecting their shields, ribbon coax can be run differentially. But this doesn't take advantage of the coupling effect of a differential pair to provide better side-to-side skew (defined as a difference in delay time from one coax to another).
Ribbon coax can achieve lower skew than round coaxial cables because the length of each coax can be controlled more precisely. Round cables are made by applying a helical twist to the conductors as they're built up in bundles or layers. This increases skew due to the length variations inherent to the cabling process.
Realize that numerous variables affect skew. Since skew is a variation in propagation delay, it can be influenced by any variations in physical or electrical length, as well as changes in signal rise times. In effect, skew is a measure of the quality of all processes used in the assembly's fabrication. For this reason, more companies are contracting for complete assemblies from capable turnkey suppliers. Previously, they'd usually attempt to combine components and assembly labor from various sources into assemblies with low-skew requirements.