Modified thermal chamber cuts EMI

RF ICs such as GSM system-on-chip (SOC) devices must operate within power and frequency limits over a specified temperature range. Testing requires an environment that attenuates unwanted signals, yet subjects the SOC to required temperatures. Guarding against unwanted signals and testing at specified temperatures are both easy to do, but not at the same time.

Because our RF SOCs contain both RF receivers and transmitters, we needed to eliminate interference from sources such as cellphone repeaters in our building. We also needed to route high-speed digital control signals to the SOCs and transfer test data to a computer. When we couldn’t find a thermal chamber that sufficiently shielded the SOC from unwanted signals and provided a viable digital interface, we modified one with help from the manufacturer.

Engineers have, with limited success, used two methods of testing devices in thermal chambers while shielding against electromagnetic interference (EMI). One method puts the device under test (DUT), RF test equipment, and thermal chamber inside a shielded RF enclosure or screen room. Unfortunately, screen rooms are large and costly, and they require significant resources such as power, heat, and air conditioning. Also, the inside of the screen room can be a relatively noisy EMI environment because of the RF test equipment.

The second method involves placing the DUT inside a semi-shielded metal box and forcing air through the box. The shielding performance of the box depends significantly on the expertise of the person putting the system together. You can’t get tight temperature control unless you place a thermocouple on the DUT and feed the signal back to a temperature controller. Furthermore, the thermocouple can radiate noise from the temperature-control system to the DUT.

Thermal chambers typically let you route RF signals into the box through SMA connections. They let you provide DC power through filtered connectors and low-speed data through filtered D-sub connectors. But they lack entry ports for high-speed data signals.

A more elegant solution would incorporate temperature control into an RF-shielded environment. At a minimum, such a test chamber must

• provide enough space for the DUT,
• provide electric power from the AC mains to reach the DUT,
• provide interfaces for RF signals, DC power, and high-speed digital signals between the DUT and the test equipment,
• provide reliable temperature control, and
• provide a defined shielding capability.

We searched for a commercial system that would meet all of our requirements, but after reviewing several options, we concluded that none could adequately control temperature with the required shielding effectiveness while providing the necessary digital interface. One manufacturer, though, had a history of modifying its chamber to shield against outside EMI. Votsch Industrietechnik’s chamber provided some of the necessary shielding effectiveness but needed further modifications for our application.

Figure 1.  A modified thermal chamber contains a shielded adapter for RF cables and digital control cables.

We found, though, that the chamber lacked the connectors for a high-speed data digital bus. We needed a 38-MHz digital interface to let our boundary-scan (JTAG) test equipment communicate with the processors within the SOC. Our application also required a 34-bit bus to transfer data between the DUT and a PC. The DUT needed clocked-data speeds up to 50 MHz. This requirement created two major challenges:

• The interface could not degrade the integrity of the digital signal. Both the JTAG and data-capture equipment have cables designed to maintain the digital signal integrity. The interface must not add discontinuities into these transmission lines, so the cables must be kept intact between the sources and destinations.
• The interface must not degrade the shielding effectiveness of the thermal chamber. This requirement is extended beyond just the interface. Our goal was to maintain the specified shielding capability with a live system; that is, with all of the connections both internal and external to the chamber terminated and operating according to the application.

Figure 2.  The shielded adapter contains ferrites that surround the cables, providing EMI suppression.

In modifying the chamber, our two main objectives were to shield the digital cabling with the braided tubing and the shielded adapter and to dissipate as much of the conducted RF energy as possible with the various ferrites.

To help us meet these objectives, Votsch added an additional blank entry panel that we could modify to support the 38- and 50-MHz digital signals. We planned to run the digital cables through a shielded adapter attached to the blank entry panel. We encapsulated each digital cable with flexible ferrite tubing, four ferrite cable clamps, and four pieces of ferrite tape. We selected the different ferrite products to provide EMI attenuation at several frequency bands.

We then placed metal-braided tubing over the digital cables on top of the ferrite tubing. Figure 2 shows the cable mounted in the shielded adapter that mounts on the chamber. Half of the shield is removed to show the cables.

To determine the impact of the modifications and to meet our requirement for a defined shielding capability, we brought the modified thermal chamber to a facility equipped with an anechoic room and with equipment that could measure the chamber’s ability to attenuate incoming signals. We placed the chamber on a rotating platform with a transmitting antenna inside and a receiving antenna outside the chamber but still inside the anechoic room.

Figure 3.  The upper trace (red) represents the measured gain of the chamber relative to the equipment noise floor (blue).

We measured the combined energy of the two polarizations (horizontal and vertical) at two angles relative to the horizontal plane of the thermal chamber. We placed a live DUT inside the thermal chamber and connected it in accordance with the application.

Prior to measuring the thermal chamber’s shielding effectiveness, we needed to quantify the transmit-to-receive gains associated with the antennas, cables, amplifiers, and free-space loss. We used gains as the 0-dB reference level in a spectrum analyzer. We then referenced all subsequent measurements to this level.

We wanted to determine how the high-speed digital-signal entry port modified the basic thermal chamber isolation. Thus, we defined a nominal equipment noise-floor level of –50 dBr. Noise floor levels higher than –50 dBr would produce questionable final isolation measurements. There are peaks in the noise profile at 500 MHz, 1750 MHz, and 2500 to 2700 MHz, where the measurements are invalid. These peaks are due to nulls in the transmitting antenna that we numerically added to the measured signal to create the 0-dB reference. These peaks limit the isolation measurements at those frequencies.

We started by performing a frequency sweep between 500 and 3000 MHz over eight combinations of antenna angles and sides of the thermal chamber. We had to determine if the worst-case isolation occurred at frequencies of interest: the 824- to 960-MHz and the 1710- to 1990-MHz cellular frequency bands. Figure 3 shows the worst-case gain of the thermal chamber for all eight conditions listed in Table 1 (four chamber positions and two antenna angles). Figure 3 also shows the ambient noise level where the measurements are invalid.

Discounting the areas where the measurement system was insufficient, we determined that the thermal chamber can achieve better than 35 dB of isolation. Figure 4 compares the chamber manufacturer’s isolation specification versus the worst-case isolation after we modified it.

Figure 4.  After modifications, the chamber still provides better than 35 dB of isolation across the frequency range (blue trace). The red trace indicates the manufacturer’s specification prior to modification.

With the receive antenna at a 0° angle, the front of the chamber was the primary leakage source. With the receive antenna at a 30° angle, the chamber rear (180°) became the primary leakage source. The difference in measured emissions in the 824- to 960-MHz cellular bands is rather significant: 11 dB versus 6.3 dB.

Arriving at precise conclusions from this information was difficult. Because the side faces are peripherally observed from both the front and rear observation points, they don’t appear to contribute to the leakage observed at the front and rear faces. The most physically apparent leakage point(s) are the power transformer ventilation holes on the top of the chamber. Additional effort is required to localize the higher leakage levels observed at the front and rear of the chamber.

While the modified thermal chamber does not provide the 80-dB to 100-dB isolation typical of RF screen rooms, it does shield the test process from the cellular repeaters. We continue to work on this project.

Although we took care to minimize the thermal conductivity of the shielded adapter, we had concerns that it might degrade the thermal chamber’s temperature performance. Subsequent use of the chamber proved that the thermal conductivity was minimal.

Even though we’ve qualified the modified thermal chamber, we still don’t know the level of interference observed by the devices inside the chamber. A site survey will determine the ambient level of interference in the laboratories.

We still need to identify the 11-dB variation in leakage described in Table 1. To do that, we must make localized measurements with a small dipole antenna. Corrective action will depend on the exact cause of the variation.