Selection Criteria of Filtered RJ Modular Jack Connectors

INTRODUCTION
Since the development of the RJ jack/plug modular interconnect system by AT&T in 1974-5 for residential application, the sheer volume of these devices has swelled to surpass all other types of connectors combined. Due to its small size, high reliability, low cost, and ease of use, applications have expanded beyond basic residential telephone use to include MODEM connections, intra-office work-station interconnects, and LAN transceiver to twisted pair wiring. A common example of the last application is Ethernet (10BASE T). At the other extreme, these devices have been adapted to connect low current DC power sources to various loads. The standard method of connection uses twisted pair wiring for economy and in many instances, is already installed and available for use. Twisted pair wiring acts both as an emitter and a receptor for line-to-ground (L-G) noise signals and low level surges. L-G noise present in some high density modern offices, and increasing in some instances, can create electronic havoc in the operation of L-G noise sensitive equipment. Even equipment emissions within allowable FCC standards (FCC Part 15 Subpart J) have interference potential. The remedial use of valuable PCB space to contain filter elements becomes increasingly less viable due to shrinking equipment size. Higher data rates (e.g. 100BASE T) and associated harmonics are on the horizon and promises to challenge us all. To suppress unwanted noise and low level surges, an effective solution is to filter at the inter- connect point where RJ jacks/plugs are used. Filtering can be passive and may consist of inductors (ferrite cores) and/or capacitors. Appreciable attenuation from 30mHz-1000mHz can be demonstrated and depending on the application, with minimal desired signal disturbance. Analogous to bulkhead mounted power line filtering, this "last chance" signal line filtering at the PCB perimeter (or "first chance" f or incoming noise) is especially suitable for retrofitting situations, when PCB discrete filter space allocation is non-existent. This retrofitting is especially beneficial in the common situation where a noise problem is discovered during full system testing, usually near production start-up dates. A speedy solution is needed with minimal effect on other parts. Corcom's "Signal Sentry" series of filtered RJ jacks was designed with these features in mind, and represents a compact, cost-effective, and performance-oriented noise problem solver.

TYPES OF MEANINGFUL TESTS
To avoid or to solve data line noise emission or susceptibility problems using filtered jacks, predictive methods to determine performance are certainly worthwhile goals. Experience has taught power line filter users that, although far from perfect, rough estimates as to filter component circuitry and values, are more time-saving than "shots-in-the-dark" in order to solve noise problems. To achieve this rough estimate of component values and performance, consider the following three methods of evaluation: -GTEM Cells -Impedance -Insertion Loss (IL)

GTEM Cells
In the EMI world, the use of this equipment has gained a substantial following. Lack of reflected waves, repeatability, conditional FCC acceptability, and lab compatibility are the main virtues. On the downside are high cost (for equipment not generally in high use) and limited equipment-under-test physical size. Testing would be functional, and could consist of a well shielded noise generator or equipment whose circuitry is similar to the final version, placed inside the GTEM cell. Various output jacks, terminated in an unshielded wire pair, would be measured for wire pair radiation, and compared for output level.

Impedance
Most filtered jacks utilize ferrites' for the inductive, or more correctly, the series element. The selection of high resistivity types exhibit increasing resistance with increasing frequency; ferrite manufacturers characterize by impedance at a specified frequency. Although this permits efficient component testing, it doesn't mesh with standard filter/EMI terminology and test equipment readout. Also, impedance doesn't address other filter elements (shunt).

Insertion Loss (IL)
IL, usually expressed in a 50 ohm environment, is the familiar connector, cable, and test equipment standard. Although IL and equipment performance do not always correlate well, estimates of required attenuation and insertion loss are still very helpful in achieving the desired emissions level. Magnitudes and locations of peaks and valleys in IL versus frequency plots, as well as the frequency range of high/low IL, can aid in the selection process. Since the characteristic impedance of unshielded twisted pair data lines is a known 100 ohms, prediction should be more realizable than with power line filters, since power lines are difficult to quantify. For these reasons, the IL method will be the focus of this discussion. Subsequent IL measurements were all performed using a custom test fixture, HP Model 8568A Spectrum Analyzer, and a shielded room. The test fixture consisted of a small rectangular metal enclosure with a vertical center partition for mounting various filtered jacks. Shielded test cables were brought through the enclosure and terminated with very small leads. Both the common and differential modes (CM and DM) setups are illustrated in Figure 1. The CM configuration is a direct 50 ohm, which agrees with the generally accepted 50 ohms from each line to ground. The DM setup matches twisted pair 100 ohm characteristic Z, via 2 to 1 balanced to unbalanced transformers. To demonstrate the effectiveness of the test set-ups, Figure 1 graphs illustrate the effects of short unshielded test wires in these setups - note the less than 2dB loss up to 100MHz. This is attributed to transformer and unshielded mis-match losses , and is considered acceptable. The IL data generated used an unfiltered jack as a reference (0dB) in order for users to easily compare the effects of adding filtering, with all other aspects constant.

PERFORMANCE COMPARISONS
Performance levels of three (3) categories of filtering will be compared: -Ferrite blocks (1 piece with multiple holes) or individual ferrite sleeves. -Capacitors only (82 & 820pF) -Combinations of ferrites and capacitors

Ferrites Only (Figures 2A-2C)
A comparison of low permeability (LO u) and medium permeability (MED u) blocks revealed a slightly broader CM response and lower DM loss associated with the LO u materials. The lower DM losses may be especially significant for data transmission rates at the standard 10 Megabyte rate: 3dB versus 0.5dB. Ferrite sleeves yielded a flat 2dB CM IL from 30-800MHz. The sleeves should only be considered where minimum "cross-talk" is desired, as some sacrifice in attenuation occurs as illustrated in comparing Figs. 2B and 2C.

Figure 2A - Low-u Ferrite Block Only (Fair-Rite 61) 125 perm:

Figure 2B - Standard Ferrite Block Only (Steward 28) 850 perm:

Ferrites and Capacitors (Figures 3A and 3B)
Immediately apparent in direct comparison with previous ferrite only levels is a magnitude in CM improvement (note difference in vertical axis scales). Above resonance, the CM response is a nearly flat 25dB out to 1GHz. If predominant system noise is above 200MHz, the best choice would be the 82pF version, as this would minimize capacitive effects on data wave shape and twisted pair characteristics. The two grounded 82pF caps in series across each line pair, or 4lpF, is insignificant to category 3 mutual capacitance limit of 20,000pFd per 1000 feet, as specified in TSB-36 of EIA/TIA related to twisted pair cables. Also DM loss below 50MHz is inconsequential for the 82pF model. If system noise is concentrated at or near either resonance point (90 or 27OMHz), that cap type should be selected, as this represents internal capacitor attributes, and its resonant frequency location may be affected by external ground inductance.

Figure 3A - Capacitors only - no ferrites 820pF

Figure 3B - Capacitors only - no ferrites 82pF

Ferrites and Capacitors (Figs. 4A, 4B, 5A, 5B, 6)
Figures 4A and 4B illustrate the effects of adding ferrite blocks or sleeves (MED u), respectively, to the 820pF capacitive models; 30 to 200MHz CM IL improvement is especially noteworthy (8-15dB), while a 3 or 4dB increase occurs beyond that, for the ferrite block. Improvement associated with the sleeve model is roughly one-half that of the blocks. Figures 5A and 5B explain 82pF type enhancements, with the inclusion of the same type of ferrite blocks/sleeves. The block CM increases 5-l0dB from 30-500MHz, but only a two dB (except for the resonance area) with sleeves. For DM comparisons, note the vertical axis scale change between 82 and 820pF. If absolute minimum DM losses are desired with very little high frequency CM sacrifice, refer to Fig. 6. The LO u ferrite block with 82pF caps exhibit near zero DM losses and about 25dB CM IL from 150-900 MHz. In order to assist users in model selection for the various ferrite/ capacitor combination styles, the attribute table below may be useful.

Figure 4A - Standard Ferrite Block + Capacitors 850 perm (Steward 28) 820pF

Figure 4B - Standard Ferrite Sleeves + Capacitors (Fair-Rite 43) 850 perm 820pF

Figure 5A - Standard Ferrite Block + Capacitors (Steward 28) 850 perm 82pF

Figure 5B - Standard Ferrite Sleeves + Capacitors (Fair-Rite 43) 850 perm 82pF

Figure 6 - LO-u Ferrite Block + Capacitors (Fair-Rite 61) 125 perm 82pF

FERRITE/CAPACITOR COMBINATION ATTRIBUTE TABLE

 * or LO u block for even lower loss

CONCLUSION
It is hoped that this test method background, various style attributes, and selection criteria will aid Circuit Designers, Network Analysts, EMI Engineers, etc., and will narrow the choices to that particular model to best suit the application. Other ferrite materials (HI u) were investigated but no benefits were discovered; in fact, DM losses generally increased, while CM IL decreased.

BIOGRAPHY - Don Talend, Telecommunications Project Engineer
A native of Chicago, Illinois, Don received his BSEE degree from the University of Illinois, and has 33 years of experience in Component Engineering - the last 7 in telecommunications. He received a patent for Corcom's filtered modular jack. Don is happily married to Kathy, and has 4 grown offspring, (one is an EE). He is convinced these are great times to live in, and will get even better.

Figure 1:

By Don Talend, Corcom Engineer