behind the classified state of the art.
• The underlying experimental data and measurement techniques that led to the de- sign of the “Tempest” test standards are currently not accessible for open academic and industrial research. This hinders further development of the field, for exam- ple, towards innovative low-cost solutions suitable for economic mass production by regular office-equipment and consumer-electronics vendors.
• As a consequence of the lack of related public research literature, civilian product designers are not educated in emission security. Opportunities for simple lowest- cost countermeasures that merely require awareness of the nature and mechanisms of emission security risks early in a design process are therefore missed.
Recent requests to have the US “Tempest” standards declassified via the Freedom of In- formation Act have only resulted in the publication of some excerpts. These cover merely standard terminology as well as test and calibration methods that are already widely doc- umented in the electromagnetic compatibility literature. The actual conformance limits and exact test procedures remain unavailable [8, 9, 10, 11, 12].
There are no indications that the community of “Tempest” researchers linked to NATO signals intelligence agencies will release its standards in the foreseeable future. The reasons for this have not been published and we can only speculate. Currently exploited sources of intelligence may involve techniques that are part of the rationale for the emission limits. It is to be expected that signal intelligence agencies are primarily interested in preserving their capabilities, which would be helped by the lowest possible awareness of emission- security risks among users and product designers. If there is such a conflict of interest, it would justify the more independent public emission-security research, as the threat is then not merely a theoretical possibility, but a routinely exploitable vulnerability today. The number of civilian information-processing facilities that have high protection require- ments, and whose infrastructure is based on commercial off-the-shelf technology procured according to open standards, increases continuously, as does our society’s reliance on their security. This calls, in my opinion, for a new generation of emission-security test standards that is based entirely on published data and experimental techniques. Their development should follow the established procedures of international standardization or- ganizations. Any underlying data should be open to scrutiny by academic peer review, in order to prevent that any tradeoffs that have to be made might be influenced by conflict- ing concerns of the signal-intelligence community. A model for such an effort could be the work that led to the international standardization of emission limits for electromagnetic compatibility [90].
5.1
Existing public standards
No public emission-security standards exist today. Two types of electromagnetic-emission limits for information technology have been widely accepted by the market, but neither was designed to reduce the risk of information-carrying emanations, or is even remotely suited to do so.
5.1.1
Ergonomic standards
Shortly after 1990, many manufacturers of CRT computer monitors introduced new “low radiation” models with improved electromagnetic shielding. These products confirm to ergonomic/hygienic standards, aimed at reducing the exposure of humans to electromag- netic fields and their potential biological effects [85, 86, 87]. The TCO’92 specification developed by the Swedish Confederation of Professional Employees (TCO) imposes the following emission limits:
• The electrostatic potential of the screen surface must not exceed 0.5 kV.
• Alternating electric fields must not exceed 10 V/m in a 5–2000 Hz passband and also must not exceed 1.0 V/m in a 2–400 kHz passband, both 30 cm in front of the screen and 50 cm around the unit.
• Alternating magnetic fields must not exceed 200 nT in a 5–2000 Hz passband 30 cm in front of the screen and 50 cm around the unit and also must not exceed 25 nT in a 2–400 kHz passband 50 cm around it.
This standard limits only low-frequency fields below 400 kHz, which are generated by CRT deflection coils. Compromising emanations are typically significantly weaker and occur at much higher frequencies in the HF/VHF/UHF bands (3 MHz–3 GHz). Therefore, a TCO’92 conformance test will not provide any information about the emission-security properties of a device.
5.1.2
Radio-frequency interference standards
The second class of publicly available electromagnetic emanation standards is aimed at minimizing interference with radio communication services. Electromagnetic fields are also generated by devices that were not designed to transmit information. These can cause distortions in radio receivers and this became an increasing concern during the first half of the last century, when a number of countries started to introduce legally binding test standards and limits (e.g., the German VDE standards in 1949 or the US FCC regulations in 1954).
Thanks to the work of the Comit´e International Sp´ecial des Pertubations Radio´electro- niques (CISPR), the various national electromagnetic compatibility standards have been harmonized during the past decade. Manufacturers of information technology equipment now only have to ensure that their products conform to the CISPR 22 specification or the equivalent European standard EN 55022 [89], in order to fulfill the legal requirements on electromagnetic emissions in most countries. This standard imposes the following radiated emission limits:
• Electric fields must not exceed 30 dBµV/m at 10 m distance in any 120 kHz passband in the frequency range 30–230 MHz.
• Electric fields must not exceed 37 dBµV/m at 10 m distance in any 120 kHz passband in the frequency range 230–1000 MHz.
88 5.1. EXISTING PUBLIC STANDARDS The field strength is determined with a special AM measurement receiver with a quasi- peak (QP) detector specified in CISPR 16-1. The output of this detector rises with a time constant of 1 ms, falls with a time constant of 550 ms and is displayed with a critically dampened mechanical indicator with an oscillation period of 100 ms or an equivalent implementation.
In addition, during the measurement, the mains power connector is plugged into a line impedance stabilization network (LISN) which provides a well-defined impedance of both live and neutral against ground, namely 50 Ω in parallel with 50 µH. The voltage measured across this impedance must
• not exceed 66–56 dBµV/m with a quasi-peak detector and 56–46 dBµV/m on av- erage in any 9 kHz passband in the frequency range 150–500 kHz, where the limit decreases linearly with the logarithm of the frequency,
• must not exceed 56 dBµV/m with a quasi-peak detector and 46 dBµV/m on average in any 9 kHz passband in the frequency range 0.5–5 MHz,
• must not exceed 60 dBµV/m with a quasi-peak detector and 50 dBµV/m on average in any 9 kHz passband in the frequency range 5–30 MHz.
On communications lines, the common-mode current measured with a current probe and specified decoupling measures must
• not exceed 40–30 dBµA/m with a quasi-peak detector and 30–20 dBµA/m on av- erage in any 9 kHz passband in the frequency range 150–500 kHz, where the limit decreases linearly with the logarithm of the frequency,
• must not exceed 30 dBµA/m with a quasi-peak detector and 20 dBµV/m on average in any 9 kHz passband in the frequency range 0.5–30 MHz.1
These are the limits for “Class B” devices, which are sold for general domestic use. The standard also specifies less strict limits for “Class A” devices that are for use in industrial environments, where for instance the radiated limits are 10 dB higher, or equivalently measured a factor √10 further away (30 m).
With the European Union’s electromagnetic compatibility directive [88], manufacturers became responsible for guaranteeing conformance to the CISPR 22 interference limits. This led to a significant increase in training and general awareness of radio interference and immunity problems among product designers and it also resulted in the purchase of EMC test and measurement equipment by vendors. As a result, many information technology products sold in Europe became more carefully designed with regard to radio- frequency emission in general, in order to eliminate early in the design phase the risk of failure in EMC compliance tests later. As a side effect, some European post-1990 products show slightly improved emission security as a result of the increased EMC awareness, even though the legally enforced test standards were not designed for that purpose.
A brief look at the motivation and design of the EMC test standards helps to understand why they are not suited for emission security purposes. Radio broadcasters aim at ensuring
1The corresponding voltage limits are 44 dB higher, which is equivalent in the case of a 150 Ω line
a minimum field strength of about 50–60 dBµV/m throughout their primary reception area [126]. The CISPR limits were selected about 20 dB below that level to ensure that, at 10 m distance, the interference from a device will not limit the received signal-to-noise ratio to less than 20 dB.
Radio receivers are also quite sensitive to high-frequency interference signals that reach them via the power-supply cable. Both mains network and communication lines can act as large antennas for the emission of interference signals. Conducted emissions are only limited below 30 MHz, because for higher frequencies, radiated emissions are considered to be the dominant effect and signals on mains and communication cables will show up in measurements as radiated emissions as well.
For frequencies below 30 MHz, the local mains network will become part of the emitting antenna. In the interest of better reproducibility of test results, the signal emitted into the cable is measured directly across a well-defined impedance at the end of the power supply cable provided with the product. The conducted emission limits for mains power lines are lower than those for telecommunication ports, as radio receivers are exposed to power-line noise directly, whereas communication lines just contribute to radiated emissions.
The quasi-peak detector is used as a psychophysical estimation tool. It provides a mea- sure of the approximate annoyance level that impulses of various strengths and repetition frequencies cause for human users of analog audio and television receivers. Strong distur- bance impulses are tolerated if they occur sufficiently rarely, and even weak disturbances can be annoying at high repetition rates. The quasi-peak limits are higher than the aver- age limits, as tests have shown that occasional broadband impulses are less annoying to AM/FM radio and TV users than a continuous narrow-band interference signal.