How the Lab Tests Were Conducted

How the Tests Were Conducted
For all of our tests, we used a Matrics AR400 reader to perform all reads for the Class 0 tags, and the Alien 9780 reader for all Class 1 tags. With all tests using the Alien 9780, we used the standard circular antennas with the cabling provided by the manufacturer. For all laboratory tests with the AR400, we used Matrics' High Performance antenna, and for the conveyor testing we used Matrics shelf antennas.

For the free-air tests, we placed the reader antenna in a stationary position, facing upward, elevated 3.5 inches off the floor. We suspended the tag three feet above the antenna in free air by attaching the tag to thin plastic filament (50 lb. test fishing line). The two thin filaments of fishing line were attached to a rig that could rotate 360 degrees on poles 12 feet apart.

Research director Deavours (gray shirt) in the lab with Karthik Narayanan, a research assistant

Since the tests were not performed in an anechoic chamber (a chamber used to eliminate all RF "noise" or interference), we used a spectrum analyzer to measure ambient RF energy in the room to ensure that there was no electromagnetic interference that would affect the test results. This was done for all of the tests done in the lab.

We also took measurements to make sure that RF energy from the reader did not bounce off the floor or walls and reach the tag. The term "multipath" is used to describe a situation in which both direct and reflected energy reaches the tag. Because RF energy travels like a wave, the energy traveling on the secondary paths can arrive in phase or out of phase with the primary signal. This can cause constructive interference, which amplifies the received signal, or destructive interference, which decreases the received signal. Or, the signal can be delayed so much that data is received at the wrong time. This is similar to trying to carry on a normal conversation with a loud echo.

We did not read the tags at increasing distances to find the read range or judge performance. Instead, we used a variable, switched attenuator, fully tested and calibrated, to simulate changes in read distance. An attenuator is a device that attaches to a transmission line (a coaxial cable) and reduces the power of a signal as it travels from the reader to the reader antenna through the cable. Attenuators are usually rated in terms of decibels (dB), a logarithmic measurement of the intensity of emitted energy, and the frequency spectrum they are designed for. They work by dissipating the RF energy into heat. The attenuators we used for this report can vary the attenuation by increments of 0.1 dB (factors of 1.023) and are accurate to better than plus or minus 0.01 dB.

By changing the attenuation and reading the tags, we were able to measure the relative performance of the 10 tags. So why not just move the tag? The main reason is to control the conditions of the tests. The attenuator is a precise piece of equipment. It doesn't change with temperature or humidity, which can affect signal attenuation in the real world. This makes our experiments scientifically repeatable and the data more reliable.

Second, it's difficult to accurately place a tag 25 feet from a reader antenna without inviting external interference. Moving the tag horizontally 25 feet away from the reader antenna would cause problems because the floor and ceiling could provide an additional path for the RF signal, which would give artificially high or low values (in other words, RF energy bouncing off the floor and reaching the tag would reach the tag and affect the results of the tests). Access to an anechoic chamber (a chamber that simulates free space with no outside electromagnetic interference) of the necessary size is prohibitively expensive. By suspending the tag only a few feet above the reader antenna and using an attenuator, we were able to achieve accurate, reliable data.

Radiation patterns
To test how much orientation to the reader antenna affects the ability to read the tag, we plotted each tag's radiation pattern—a 3D plot that shows how much RF is radiated in all directions. This information can help end users understand which tags are likely to perform best in applications where the orientation of the tag to the reader antenna cannot be controlled (such as reading airline baggage tags on an airport conveyor).

To generate the radiation pattern of each tag, we kept the distance between the tag and reader antenna constant at 34 inches and rotated the tag 360 degrees at 20-degree increments along the two major axes. At each orientation, we measured the read rate (number of read attempts divided by successful reads) when the transmit-and-receive lines were attenuated in increments of .1 to 4 dB, which simulated increasing the distance between the tag and reader. We were able to simulate changes in distance with relative accuracy of about one inch.

To create the radiation pattern, we need to decide on a point at which the tag is no longer "in field." We decided that at least 50 percent of read attempts had to be successful for the tag to be considered in the read field. We then plotted on a graph (see illustration) the dB of attenuation at which the read rate dropped to 50% percent for each of the 18 different orientations along each axis. (A 50% read rate may seem low, but readers are capable of performing tens to hundreds of reads per second, and a 50% read rate is adequate for most applications.)

Read Distance Test
For the read distance tests, we used the same set up as we used for the radiation pattern tests. We placed the tag 3 feet above the center of the reader antenna in the optimal orientation, which was determined by the radiation patterns. We then set the reader to read the tag 100 to 300 times. To test the performance of the tags, we read each tag using the full output of the reader (in the United States, where we tested, this is 4 watts of radiated power). We recorded the number of reads and divided it by the number of attempts to get the read rate at each attenuation setting. We then stepped up the attenuation in increments of .1 to 4 dB and read the tags again. This process was repeated until the read rate dropped to zero. The report graphs the read rate at various distances for each of the ten tags.

Testing Tags Near a Conductor
While the "free air" tests are useful as baseline metrics of tag performance, few end users will ever use tags that way. The tags are most likely to be used attached to a container, such as a cardboard box, wood or plastic palette, or a product. Furthermore, the container is likely to have a product or material in it. Tracking the contents of the container is the purpose of placing an RFID tag on the container in the first place.

The presence of a material near a tag often changes the performance of the tag in significant ways. Conductors, such as metal, provide some of the greatest challenges for RFID tags. Conductors are everywhere, including places one might least expect them. Boxes of dishwasher detergent (used for our conveyor testing, for example,) are lined with a metal foil. Even a very thin foil of metal is enough to make products difficult to tag.

To assess the performance of tags near metal, we placed each tag varying distances from a large, flat piece of steel. The tags and metal plate were separated by air. The tag was placed 3 feet from the reader antenna. We then used an attenuator to determine the dB attenuation level at which the tag could no longer read. A higher attenuation level, expressed in dB, corresponds to a longer reading distance. We tested each tag at distances from the metal plate ranging from 0 to 2 cm at 2.5 mm increments. At each distance, we increased the reader's dB attenuation level until no reads were observed. This provided us with an approximate maximum read distance for each tag. The RFID Alliance Lab report contains charts of the dB attenuation level for each tag with an approximate conversion into read distance in feet.

Testing Tags Near Water
Pure water is a poor conductor. However, it has a high "dielectric constant"—a measure of a material's ability to store a charge when an electric field is applied—and represents one of the most common, difficult, non-metallic materials for RFID tags to work with. Other dielectrics include virtually everything that isn't metal, including most plastic, paper, cardboard, wood, glass, and ceramics. Most of those items are "low" dielectrics and are much easier to tag than water.

For these tests we used a similar set up to the tests of tags near metal. We placed each tag near a relatively large body of ordinary tap water. We used a standard 10 gallon aquarium to hold the water, because it has a uniformly flat surface. There was a 3 mm thick glass plate between the tag and the water.

We placed the tag various distances from the aquarium, from 0 to 2 cm in 2.5 mm intervals. Note that the closest we could get to the water was about 5.7 mm away, due to the glass and plastic edges of the aquarium. The results we show are the actual distance to the body of water behind the glass (not the distance from the glass), so the smallest measurement is 5.7 mm. At each distance, we increased the reader's dB attenuation level in increments of 3 dB until no reads were observed. This provided us with an approximate maximum read distance for each tag. The RFID Alliance Lab report contains charts of the dB attenuation level for each tag with an approximate conversion into read distance in feet.

Conveyor Testing
Obviously, companies will be using RFID tag in environments that have electromagnetic interference and other conditions that will affect tag performance. The final portion of our testing was to take tags, place them on cases of products, take them into an operational warehouse environment, and test them on a production conveyor system. While this may be the most interesting portion of the tests for end users, it is the least scientific. There are too many unconstrained variables to make the results transferable to other environments.

Since the tests were performed on a production conveyor system, we did not have unlimited access, so we had to limit the number of times we could repeat a test, which negatively impacts the accuracy of our results. Nevertheless, we performed some over 2,400 measurements in the distribution center, and there is enough data to draw some lessons learned, not only about tags but also about interesting interactions between tags and materials.

For these tests, we placed tags on cases of five materials: paper, bottled water, canned food, dishwasher detergent (Cascade), and liquid hand soap. Tags were placed on the cases at a position that gave the highest tag performance. We did not observe any constraints about where the tag could be placed. (Frequently, there are constraints on tag placement, such as printed areas.)

The tests were performed at a large Kansas City area distribution center that distributes products to most major retailers in the region. The conveyor we used had a flat metal sled with a belt and moved at 204 feet per minute (fpm). (It's interesting to note that while some retailers are requiring tags to be read at up to 540 feet per minute, the distribution center we used had nothing that went faster than 250 fpm.) Naturally, faster speeds will reduce tag performance because the tag will be in the read field for less time. If a company needs to meet recent mandates, then it should take that into account when observing our results.

For this test, we placed three different tags on three different cases of the same product. We randomly selected the tags but placed them at the same optimal position. This approach was designed to mimic "real world" conditions as closely as possible.

For Class 0 tags, we used the Matrics AR400 reader with three shelf antennas. For Class 1 tags, we used the Alien 9780 reader with three standard circularly polarized antennas. The three-reader antennas were arranged so that one was on each side of the conveyor, five feet apart, and one was 42 inches above the conveyor pointing down. We chose these reader antennas and configurations because end users told us that these are the ones major retailers are using.

For each tag and product, we varied the power setting on the reader. At each power setting, we ran three cases of identical products through the reader portal five times each to get 15 samples per power setting. The data we present in the report shows the reduction in power setting (in dB) at which we observed a 50 percent read rate. Fifty percent read rate was chosen because it was the easiest to measure. Obviously, everyone wants to achieve 100 percent read rates. Ideally, we would measure the lowest power setting at which 100 percent read rates were achieved. The problem with that approach is that it would require a very large number of experimental samples, which was not practical. We think we chose the best compromise, but it means our results are optimistic. At 540 fpm or 600 fpm, expect lower read rates. To achieve 100% read rates, power settings will need to be significantly higher, perhaps as much as 4 to 5 dB higher.