Tracewell Systems

The Problem

The Red side of an electronic system deals with the signals in their native, un-encrypted form, and requires the most shielding, filtering, or other isolation/protection techniques to prevent unintended access or snooping. The Black side can be less secure because the data and signals have been encrypted. Besides interception of the unencrypted information, a major concern is preventing the interception of both the Red and Black signals for comparison, which could compromise the encryption algorithm.

The term TEMPEST was coined in the late 60's and early 70's as a codename for the NSA efforts to secure electronic communications equipment from potential eavesdroppers. Officially, it is not an acronym for anything, especially "Tiny ElectroMagnetic Particles Emitting Secret Things." NSTISSAM TEMPEST/1-92 and 2-95 form the public part of what is commonly called "the TEMPEST spec", but finer technical details remain secret. The most stringent non-secret EMI/RFI specification is MIL-STD-461, and TEMPEST can be thought of as 461 on steroids. Both specs state specific limits for electromagnetic field strengths at frequencies out to at least 40 GHz.

The overall problem with Red/Black systems is that two functions usually separated by several feet, multiple shielding barriers (chassis, racks, brick walls), and independent power supplies are now only inches apart, and possibly share a common backplane and/or power system. While both TEMPEST and MIL-STD-461 deal with susceptibility as well as emissions, containing emissions usually is the greater problem. Also, a single power feed for both the Red and Black sides is becoming more common. Signal crosstalk through the power supplies is one of the most difficult problems to solve. In a recent project, crosstalk isolation had to be greater than 120 dB, or 16 times better than the noise floor of a perfect CD recording. Put another way, assuming a 3 volt data stream on the Red side (typical TTL-level signals), crosstalk on the Black side had to be less than 3 microvolts - and that was not the full extent of the TEMPEST spec.

The Approach

There are three separate goals in a Red/Black design: keeping the Red side isolated from the outside world to TEMPEST criteria, keeping the Red side isolated from the Black side to TEMPEST criteria, and keeping the Black side isolated from the outside world, usually to MIL-STD-461 criteria. This three-sided analysis of the problem has driven many of our designs. Each isolation path has both a mechanical and electrical component. Often, the mechanics of the system can be manipulated to handle some aspects of the electrical isolation issues. Also, we can negotiate some of the spec requirements to improve isolation.

For example, one customer had the arrangement of board locations in a backplane optimized for the flow of their signals. But looking at the system with an emphasis on isolation, we suggested rearranging the boards to increase the distance between the most sensitive Red and Black boards, which decreased the amount of physical shielding we needed to add to the system.

Two Solutions

For one particularly difficult system, we applied the three-sided approach to the power line filter. In most Red/Black applications, power comes from the Black side. The system had two separate backplanes, each with a plug-in DC/DC power supply. There was no room for a single, massive, TEMPEST-rated power line filter, so we broke the filter up into three units. One had the power input connector and circuit breaker integrated within a custom, 461-rated filter, and mounted it directly to the rear panel for a perfect RF seal. The chassis had a central tunnel between the Red and Black sides for power distribution and some auxiliary electronics.

In collaboration with our custom filter vendor, we designed a smaller filter that made up the performance difference between 461 and Tempest, and broached both tunnel walls with them, eliminating connectors and RF gaskets. The secondary filters formed the interfaces between the tunnel and the Red and Black sides. No single filter met the full crosstalk requirements, but each of the three crosstalk paths had to pass through two filters in series. Three smaller filters were more expensive than two larger filters, but they also fit in with the mechanical elements of this Red/Black design, and shifted the center of gravity away from the rear of the chassis, closer to the rack mounting flanges at the front, reducing their size and weight. This design is a good example of how taking a systems level approach to the system creates unusual opportunities for optimizing the design; in this case an EMI control scheme reducing mechanical stress and producing a better balanced system.

The second example features an emerging variation of Red/Black, called Red/Grey/Black. This refers to a system with a part that might be either Red or Black, called Grey. We recently had such a project for an Unmanned Aerial Vehicle (UAV). In this conduction-cooled system, the input power could come from either the Red or Black portions of the aircraft, so the isolation from the power input to the Black side had to equal the isolation from the power input to the Red side, an additional isolation burden. While this is a more difficult extension of traditional Red/Black design methodologies, this new trend is a natural fit with the three-sided design approach mentioned above.

Figure 1
An interior compartment for the power supply positioned between the Red and Black signals compartments. In this system, the three compartments share a comFimon backplane and power supply while maintaining Red/Black isolation.

Figure 2
Mechanical features machined into the conduction-cooling sidewalls interact with shield plates and electrical features in the backplane to create three separate Faraday cages within a single, 5-slot chassis.

Figure 3
Critical to the success of the project was a Tracewell designed single-slot, conduction-cooled power supply that maintains the Red/Grey/Black isolation in a single assembly with a common printed circuit board.

The power supply design went from concept to a tested prototype in ten weeks. Having complete in-house capabilities for power supply design and production in the same facility as our system design and manufacturing eliminated risk and facilitated seamless interaction and design integration throughout the project. Tracewell completed the program on time and within budget.