Compliance with the relevant essential requirements of the standard became mandatory in July 2001 for products sold in the European Union. The problem for TTE manufacturers is that meeting the criteria is technically difficult and potentially costly. Market pressures on the price of the plain old telephone give little, if any, cost margin to play with. Many manufacturers— expecting to make small tweaks to their products in order to comply—have been utterly surprised by the need to virtually redesign telephone products from scratch. This article analyzes the requirements of the standard and attempts to answer two fundamental questions: just how realistic are the requirements, and what techniques can be used to design compliant telephone products?

European Norm (EN) 55024 is the European EMC standard that addresses immunity of IT and telephony products.¹ This immunity standard determines the limits and requirements for both conducted and radiated amplitude-modulated (AM) radio-frequency (RF) disturbances, as well as electrostatic discharges and other transients and surges. In 1998, the standard was updated with the addition of Annex A, which defines the limits for demodulated interference (see Table I).

Demodulated interference occurs when semiconductors within a product encounter AM radio signals. Each semiconductor junction demodulates the RF signal much like an old-fashioned crystal radio set. At radio frequencies, therefore, the modern telephone is nothing more than a complex network of crystal radio sets. The resulting demodulation is heard as an irritating noise superimposed on the wanted speech signal. Annex A provides the acceptable limits for this unwanted noise in various frequency bands.

A good deal of interpretation is required to determine whether all or part of Annex A applies to a given product. The annex presents three clauses defining telephone terminal equipment performance:

Clearly, Annex A is intended to apply to terminal equipment. Current practice, therefore, is to exclude routing equipment such as private branch exchange (PBX) and private automatic branch exchange (PABX). However, if a PBX has a dedicated management terminal with an acoustic interface, then that terminal should be tested. EN 55024:1998 does not apply to cordless terminal equipment such as digital enhanced cordless telecommunications (DECT) and CT0, which have their own product-specific standards. However, it does apply to basic telephones, headsets, speakerphones, voice recorders, digital telephones, fax machines, voice modems, and so on. Clause (i) applies to all of these products.

Demodulated noise is the noise heard by the remote party from local radio interference. In Clause (ii) of the measurement method, the demodulated noise a product presents to the telecommunications port is measured. This clause applies only to devices supporting a telephony service, which EN 55024:1998 defines as: "A service providing users with the ability for real-time two-way speech conversation via a network."^1,2 The phrase "real-time two-way speech" immediately excludes voice recorders and voice modems unless these devices allow a user to make a regular telephone call—as a fax machine does, for example.

For plain old analog telephones, demodulated noise is simply measured in dBm across a 600-W termination, simulating a public switched telephone network (PSTN) line. For digital telephones, such as integrated services digital network (ISDN) terminals, a decoder is necessary to make meaningful measurements.

The term telecommunications port is given a wider definition in EN 55022:1998.³ Curiously, the term is not defined at all in EN 55024:1998. In the 1998 revision of EN 55022, the term's meaning was extended to include indirect connection to a public network via a multiuser network (such as an office local-area network). This revised definition has caused some confusion, so the International Special Committee on Radio Interference (CISPR) has proposed amending it as follows:

Whether this amendment becomes part of the standard is currently under consideration by CISPR. If it does, whether it will end confusion over the issue is debatable. Some manufacturers are confused as to whether products exempt from Clause (ii) must still meet Clause (iii). Purists argue that they do, but it is equally possible that the word "also" in Clauses (ii) and (iii) implies "and." Indeed, there seems little point in testing the acoustic interface of a device that does not support a telephony service.

Clause (iii) of the measurement method discusses the measurement of the demodulated acoustic noise a product presents to the local user due to local radio interference. The clause applies only to devices with an acoustic interface. The sound pressure level (spl) is measured in the presence of a modulated radio disturbance.

Products with multiple acoustic interfaces, such as headset, handset, speakerphones, present even more interpretation problems. It is clearly proper to test the spl at every acoustic interface, but it is unclear as to whether telephone-port demodulated noise should also be tested in each mode. Measuring demodulated noise only in the default operating mode (usually the handset operation) should be sufficient.

Annex A does not specify a measurement method for hands-free phones (speakerphones), even though these products do support a telephony service and do have an acoustic interface. An alternative is to use the ITU-T recommended method.⁴ In addition, aftermarket headsets do support a telephony service and do have an acoustic interface. The difficulty, until very recently, for headset manufacturers has been obtaining a telephone that is compliant with EN 55024:1998 to test with.

Besides the standard's ambiguities, problems with the test method make it very difficult to obtain consistent results from laboratory to laboratory.

Demodulation Mechanism. Every semiconductor junction within an electronic circuit demodulates RF. This includes every diode, transistor, and integrated circuit. Circuits designed to work only at audio frequencies still demodulate RF disturbances, often up to many gigahertz. The demodulation mechanism assumed by the standard for linear circuitry is shown in Figure 1. Semiconductor devices used for audio switching produce wildly different results in practice. The demodulation mechanism for switching circuitry is shown in Figure 2.

Figure 1. Demodulation in linear circuitry.

In Figure 2, a transistor has been used as a microphone mute switch. The transistor will produce no demodulated noise until the peaks of the modulated RF energy in its base circuit are just sufficient to cause it to switch on. Demodulated noise then occurs as shown, and in the case of the simplified microphone circuit, the noise becomes greatly amplified. Demodulated noise power rises disproportionately quickly with extremely small increases in modulated RF power.

Figure 2. Demodulation in switching circuitry.

In a laboratory test situation, small variations in field uniformity from chamber to chamber are inevitable, giving often unrecognizably different results for the same sample.

Repeatability Problems with Cabling. Repeatability problems increase for products with cables. All products to which Annex A applies will have cables of some sort. Differing cable lengths and cable placement (often very slight variations) frequently yield markedly different test results. Figure 3 shows sound-pressure-level measurements taken between 80 and 200 MHz for the same product in the same configuration, in three different chambers at two test laboratories in the United Kingdom.

Figure 3. Comparison of measurements for one product in three chambers.

Poor Correlation Between Conducted and Radiated Tests. CISPR has already recognized a discrepancy between the conducted immunity test above 10 MHz and the equivalent radiated immunity test. Its report concluded that "both test results coincide with each other except a discrepancy approximately above 10 MHz," and "the applied voltage (V_emf) of the RF continuous conducted testing should have frequency dependence." Accordingly, Table A.1 in Annex A has been amended in CISPR 24 to relax the limit above 10 MHz.⁵ It is very likely that this amendment will make it into EN 55024, the only question is when.

With all of the standard's weaknesses and ambiguities, compliance with the relevant essential requirements is now mandatory for TTE sales in Europe. The first question most manufacturers of existing TTE products ask is, "Can I just tweak it to pass?" Experience shows that the answer is a resounding no! Modifications are invariably required to the layout of analog printed circuit boards (PCBs). Although it is impossible in a single article to describe all of the techniques for EMC hardening of a telephony product, a brief introduction to the most important methods is given here.

Isolate Sensitive Circuitry. It is virtually impossible to EMC-harden a sensitive analog circuit, such as a telephone, without proper physical segregation of circuit blocks on a PCB (see Figure 4). Sensitive circuitry, such as microphone circuits, must be confined to a single area—termed the clean zone—well away from noise sources, especially cable entry points that carry in RF disturbances. Experienced designers know roughly which circuit blocks are likely to be sensitive. For existing products, identifying susceptible circuitry by bench testing is recommended prior to reworking the PCB.⁶

Figure 4. Example of telephone circuit segregation.

Less-susceptible circuitry should be placed in another area of the PCB, away from the clean zone. For example, the ringer circuit in a telephone will probably not be particularly sensitive. The positioning of filters to arrest cable-induced RF is also important. Filters need to be as close as possible to cable entry points. Designers should also avoid using flying leads to bring signals onto the analog PCB so that signals can be consistently routed away from sensitive circuitry.

Ground Planes. The clean zone should be entirely covered with a solid copper ground plane. Avoid aperture cuts in the ground plane, and use short ground stubs to connect surface-mount components, especially RF filtering capacitors, to the ground plane. Ground-plane resonance effects (patch antenna) can be avoided at frequencies below 1 GHz by containing all clean-zone circuitry within a 50 x 50 mm square.

Two types of filters are usually required for compliance: conducted-RF filters (applied to cable ports only) and radiated-RF filters (applied to cable ports and elsewhere).

Conducted-RF Filtering. Cable ports need a common-mode filter on all incoming lines. For a two-wire circuit, a PCB-mount choke can be used. For four or more conductors, winding all conductors around a suitable ferrite toroid two or three turns will make an appropriate common-mode choke.

Figure 5 shows a simple PSTN line-cord filter circuit. The part numbers given are from Würth Elektronik, but any equivalent components can be used.⁷ The inductance of the choke and subsequent inductors is not critical. It is important, however, that components have high impedance over the critical frequency range 1–80 MHz. Adding 1-nF capacitors to a chassis is useful only for products with an all-metal chassis such as pay phones. For safety performance, it is critical that these capacitors have a suitable voltage rating (at least 1.5 kV).

Figure 5. Example of a PSTN line filter circuit.

Radiated-RF Filtering. All cable conductors and every track entering the clean zone should also have radiated-RF filters. The best filter is simply an in-line resistor of at least 10 kW. This resistor presents an impedance mismatch at all RF frequencies and thereby reflects RF energy. For cases in which high impedance is not possible, use a 1-kW resistor and a 100-pF capacitor to ground. For very low impedance circuits, a 150-nH inductor and a 100-pF capacitor can be used (see Figure 6).

Figure 6. Radiated RF filters.

Filter Layout. To minimize stray coupling to the ground plane, it is best to place conducted filters on an area of the PCB without a ground plane. Radiated filters must be arranged so that the in-line resistor or inductor sits on the boundary of the ground plane and the shunt capacitor sits immediately inside the ground plane (see Figure 7). In the figure, in-line inductors (0805) and resistors (0603) can clearly be seen placed on the boundary of the clean-zone ground plane.

Figure 7. Radiated-RF filter layout.

Component Limitations. Filter components should be chosen with a suitable RF performance. For example, 0603-format 100-pF capacitors and 0805-format 150-nH chip inductors (such as Coilcraft (Cary, IL) part no. 0805CS-151X_BC) are recommended because both parts have self-resonant frequencies at approximately 1 GHz (the upper limit specified by EN 55024:1998). Physically large electrolytic capacitors should not be used within the clean zone. At high frequencies, these can become significant antennas.

Analog Switches. Designers should avoid using discrete transistors as analog switches. Field-effect transistor–based switches, such as the 4066 devices, have a much better noise margin.

The European EMC immunity standard covering IT and TTE has presented some tough challenges for telephony manufacturers, especially for those with existing products. Compliance with the relevant essential requirements of the standard is technically difficult and potentially costly.

A major shortcoming of the Annex A test method is the inability to obtain consistent test results. Therefore, it is essential that designers plan for a wide pass margin to ensure that the variation in test results do not cause a particular sample to pass in one test lab and fail in another.

It is clear that most products do have to be redesigned to comply with the relevant requirements. Modifications to the layout of analog PCBs is essential in most cases. Understanding some important design techniques for EMC hardening, such as isolating sensitive circuitry and integrating appropriate EMI filtering, is key to designing for compliance to this immunity standard.