UHF EPC Tag Performance Evaluation Report #2 in the Series Data and evaluation of nine commercially available EPC UHF tag models, including:
Variation of performance
Read rates in population
Typical performance
Forward/reverse channel impact
Read rates in isolation
Write performance By Daniel D. Deavours A Production of the

2335 Irving Hill Road Lawrence, Kansas May 2005

Copyright 2005, the University of Kansas. All rights reserved. DO NOT COPY.

1. Contents Page 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 2. The Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . .3 2.1 The Readers . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 2.2 The Tags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 3. Populations of Tags . . . . . . . . . . . . . . . . . . . . .8 3.1 The Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8 3.2 The Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 3.2.1 Variations. . . . . . . . . . . . . . . . . . . . . . . . . . . 9 3.2.2 “Dead” and “Quiet”. . . . . . . . . . . . . . . . . . 13 3.3 Secondary Metrics . . . . . . . . . . . . . . . . . . . . . . 14 3.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

Page 6. Read Rates—Multiple Tags . . . . . . . . . 25 6.1 The Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 6.2 The Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 6.2.1 nth Tag Read . . . . . . . . . . . . . . . . . . . . . . . 26 6.2.2 Tags Read Rate . . . . . . . . . . . . . . . . . . . . . 27 6.2.3 Tag Read Rate in Population . . . . . . . . . . 28 6.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

Sidebars: Where Are Avery Strip and Rafsec 458? . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Why 32.5? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Reading with the ThingMagic Mercury4 Reader . . . . . . . . . . . . . . . . . . . . . . 8 Class 0+ and External Capacitor. . . . . . . 13 Are 96-Bit Tags Better? . . . . . . . . . . . . . . . 14

Sidebars: Read Rates. . . . . . . . . . . . . . . . . . . . . . . . . . . 29 What About Vendor Read Rate Claims? . . . . . . . . . . . . . . . . . . . . . . . . . 30

4. Typical Tag Performance . . . . . . . . . . . . . . 16 4.1 The Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 4.2 The Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 4.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Sidebar: Ghost Reads. . . . . . . . . . . . . . . . . . . . . . . . . . 20 Technical Note: How to Compare Tags Using Percent Reads. . . . . . . . . . . . . . . . . . . . . . . . 20

5. Read Rates—Isolation . . . . . . . . . . . . . . . . . 22 5.1 The Experiment . . . . . . . . . . . . . . . . . . . . . . . . 22 5.2 The Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 5.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 Sidebar: What is “Global Scroll”?. . . . . . . . . . . . . 23

Technical Note: Fairness of Read Rates in Population . . . . . . . . . . . . . . . . . . . . . . . . . 28

7. Channel Impact on Performance. . . . . . 31 7.1 The Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 7.2 The Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 7.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Sidebar: What Is Software Attenuation? . . . . . . . 31

8. Write Performance . . . . . . . . . . . . . . . . . . . . . 35 8.1 The Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 8.2 The Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 8.2.1 Alien Squiggle . . . . . . . . . . . . . . . . . . . . . . 36 8.2.2 Alien I2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 8.2.3 Alien M . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 8.2.4 Rafsec 456 . . . . . . . . . . . . . . . . . . . . . . . . . 36 8.2.5 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . 36 8.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 Sidebar: What Do Class Numbers Mean? . . . . . . 35

9. Conclusions and Observations . . . . . . . . 38 Appendix A: Who We Are . . . . . . . . . . . . . . A-1 Appendix B: Acknowledgements . . . . . . B-1

UHF EPC Tag Performance Evaluation

RFID Alliance Lab

i

Tables 2.1 3.1 3.2 3.3 5.1

Page Tags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5 Variations Legend . . . . . . . . . . . . . . . . . . . . . . . . . . .9 Quiet and Dead Tags . . . . . . . . . . . . . . . . . . . . . . . .13 Percent Tags with 70% Read Rates . . . . . . . . . . . .15 Tag Reads per Second . . . . . . . . . . . . . . . . . . . . . . .23

6.1 6.2 6.3 8.1 8.2

Page Tags Read Rates (Population) . . . . . . . . . . . . . . . .28 Tag Read Rates (Population) . . . . . . . . . . . . . . . . .29 Tag Read Rates (Isolation) . . . . . . . . . . . . . . . . . . .29 Write Performance Summary . . . . . . . . . . . . . . . .36 Class Descriptions . . . . . . . . . . . . . . . . . . . . . . . . . .37

Figures Page 2.1 Images of reader equipment (not to scale) . . . . . .4 2.2 Images of tags (not to scale) . . . . . . . . . . . . . . . . . .6 3.1 Alien Squiggle . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10 3.2 Alien I2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10 3.3 Alien M . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10 3.4 Avery DS1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11 3.5 Avery Triflex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11 3.6 Rafsec 456 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11 3.7 Symbol I2010 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .12 3.8 Symbol X2040 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .12 3.9 Symbol I1030 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .12 3.10 External capacitor on a Symbol tag . . . . . . . . . . .14 3.11 Strap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .14 4.1 Squiggle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .16 4.2 I2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .17

UHF EPC Tag Performance Evaluation

4.3 4.4 4.5 4.6 4.7 4.8 6.1 6.2 6.3 6.4 6.5 6.6 6.7 7.1 7.2

Page M . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .17 AveryDS1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .17 Avery Triflex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .18 Rafsec 456 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .18 Symbol I2010 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .18 Symbol X2040 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .19 I2 tags on polystyrene board . . . . . . . . . . . . . . . . .25 DS1 tags on polystyrene board . . . . . . . . . . . . . . .25 I1030 tags on placebo in cardboard box . . . . . . .25 Alien I2 population read rate . . . . . . . . . . . . . . . . .26 Avery DS1 population read rate . . . . . . . . . . . . . . .27 Symbol I1030 population reads . . . . . . . . . . . . . . .27 Tag read rates in population . . . . . . . . . . . . . . . . .28 Symbol I2010 response . . . . . . . . . . . . . . . . . . . . . .32 Alien Squiggle response . . . . . . . . . . . . . . . . . . . . .33

RFID Alliance Lab

ii

1. Introduction FID tag performance matters. Whether you’re trying to meet mandates or considering RFID for internal operations, the ability to read tags is the foundation of any RFID-enabled system. Second-generation products are likely to start appearing late in 2005, but it’s likely that the Generation 1 products that we tested for this report will be in use well into 2006. In this report, we examine a number of aspects of tag performance, including yield (what fraction of tags that you receive will actually work), and variance (the difference in the performance among tags of the same model). We measure how quickly tags can be read, both in isolation and when other tags are present, and report extensively on the results. These tests are thorough, scientific, and repeatable. They should take much of the hype out of advertisements and give users sound information on which to base tag purchase decisions. Similar to our previous report, we chose nine commonly used RFID tags for our tests. These include the Alien Squiggle, I2, and M tag; Avery Dennison DS1 and Triflex; Symbol I2010, X2040, and I1030tag; and Rafsec 457. The tags are described in detail and displayed in Section 2. Due to interest in RFID for pharmaceutical applications, we chose to replace the discontinued Rafsec 458 with the Symbol I1030 tag. We note that Avery Dennison, Alien, and Rafsec have similar itemlevel or pharmaceutical tags available, but they were either not released or not available in sufficient quantities to be included in this report. The complete list of tags and equipment we used for this report is given in Section 2. One unfortunate but real fact about RFID tags is that the quality of tags is currently not consistent, and therefore performance is not consistent. There are considerable variations in performance from one tag to the next, even among tags of the same manufacturer and model. Some models of tags show more variation than others. We performed nearly 1 million read attempts on more than 1,000 tags of the nine models and recorded nearly 15,000 data points. We tracked how many tags were “dead” and how many were “quiet.” (A quiet tag is one that is readable, but only at very short distances.) The results of this effort are given in Section 3. Next, in Section 4, we looked at the performance of a typical tag for each model in much greater detail. We

R

UHF EPC Tag Performance Evaluation

characterized the percent reads vs. distance. This extends results from our first report in a number of ways. First, we chose a “typical” tag from each model. Second, we characterized the performance with much greater accuracy (over one million read attempts were used for each tag). Third, we used the same RFID reader and configuration for all models, so our readers can make direct comparisons among all nine models. One of the themes of this report is “read rates,” or how quickly tags can be read by readers. Historically, read rates have been the subject of much marketing hype. Lately, we have noticed that the hype has toned down (perhaps because users are becoming more educated and recognize hype for what it is). Yet, some of the claims remain out there. Vendors sometimes quote astronomically high read rates. For example, one manufacturer claims “Read rate: up to 1,000 tags/second.” 1 The same vendor also informs potential customers that “Today's RFID tags have read rates varying from as low as 20 tags/second to over 1,000 tags/second.”2 What they don’t tell potential customers is that some tags are much slower than 20 tags/second. Our test results reveal a considerably more conservative range for tags/second. Another vendor claims that one of their readers can read “up to 250 reads per second,”3 but another of their readers only “reads up to 50 tags per second.”4 We don’t deny that these claims might be possible under some conditions and under certain interpretations of “reads per second,” but the information that users need to know is what kind of performance they’re going to get in their conditions, and under definitions that are meaningful to them. In this report, we devote two sections to focusing on carefully measuring read rates. In Section 5 we measure read rates for tags in isolation for all the tags tested in this report. In Section 6, we test read rates of tags in populations. We tested a population of more than 100 Class 0 tags (Avery DS1) and more than 100 Class 1 (Alien I2) tags. We also tested a smaller population of 48 Symbol I1030 tags on a case of pharmaceuti1http://www.symbol.com/products/rfid/data_sheet_tag.html 2http://www.symbol.com/products/whitepapers/rfid_mainstream_s c.html 3http://www.alientechnology.com/products/rfidreaders/alr2750.php 4http://www.alientechnology.com/products/rfid-readers/alr9640.ph

RFID Alliance Lab

1

cal products (placebo). We also explain why some claims made by vendors are misleading. In Section 7, we explain how the “forward channel” and “reverse channel” impacts performance (those terms are explained in Section 7). The results of our extensive tests show that there is little disadvantage to using a “monostatic” (that term is also explained) antenna system with the ThingMagic reader. It also gives insight into how to best use a “bistatic” antenna system. We also started some preliminary tests on write performance of different tags. While we were not able to test enough tags to get statistically strong results, the results that we do have are interesting and revealing. We present our write performance results in Section 8.

UHF EPC Tag Performance Evaluation

We summarize the key findings in our report in Section 9, and give more information about who the RFID Alliance Lab is and what we do at the end of the report. The results shown in this report indicate sharp distinctions between performances of different tags and are immediately useful for consumers choosing what tags are best for their applications. Along the way, we provide insights and parenthetical observations in sidebars. We are pleased with the volume and quality of our testing and results, and we hope that these provide useful insights and save you time, money, and frustration.

RFID Alliance Lab

2

2. The Equipment or this report, we studied nine tags using three different readers. In this section, we describe the equipment that we used. Note that eight of the nine tags we tested were the same as the tags from our previous report. (See sidebar “Where Are Avery Strip and Rafsec 458?”) We chose the Symbol I1030 tag to replace the Rafsec 458 because of the considerable interest in RFID in the pharmaceutical market.

F

2.1 The Readers For the majority of our tests, we used a ThingMagic Mercury4 reader. The Mercury4 ran the operating system Linux rfid 2.4.22-uc0 #203, RadiOS version 2.0.47 (2004-08-30T17:58:51-4:00). This information is important because differences in software and firmware can yield significant differences in performance. (A new firmware was released by ThingMagic during the course of testing for this report. To maintain consistency, we chose to use the “old” firmware for the duration.) The majority of tests were performed using custom software that was executed on the ThingMagic reader. The Mercury4 reader supports a usage model that is not well suited for our testing, so we had to rely on our best engineering judgment at times. All experiments were performed using the default configuration, with the default cables and antenna. Software attenuation was with reference to 32.5 dBm transmission power. (See sidebar “Why 32.5?”) If we used some settings other than the defaults, we described those in the experimental description. For tests involving read tags in isolation, we also used an Alien 9780 and a Symbol AR400 reader with the standard antenna setup. The Alien 9780 was loaded with firmware Version 3.7.3, RxDSP Boot Version 2.2.0, RxDSP Version 2.9.0, and FPGA Version 3.0.0. The Symbol AR400 used software Revision 03.01.09 and firmware Revision 2.2.0. See Figure 2.1 for a picture of the equipment we used.

2.2 The Tags We performed experiments on the tags listed in Table 2.1. Where possible, we listed model number and other information obtained from the vendor data sheets, which are not consistent in terms of the amount of data supplied.

Where Are Avery Strip and Rafsec 458? In the previous report, we included data on the Avery Strip and the Rafsec 458. Both of these tags were advertised as commercially available in production quantity, and this was verified by tag vendors when we obtained the tags. We obtained only a small quantity of Rafsec 458 tags, so when we sought to obtain more, we were surprised to learn that the tag had been discontinued. This was puzzling, especially considering how well the Rafsec 458 performed near water (see our previous report). We do our best to test tags that are commercially available, and we obviously don’t want to waste our time (and yours) testing tags that aren’t going to be commercially available. However, there is always an element of guesswork and judgment that goes into the selection process. We believe Rafsec believed that the Rafsec 458 was, or, before we published, would be available in commercial quantities. In the interim, they apparently decided to discontinue the 458 model. We had a similar experience with the Avery Strip. We completed tests on the Avery Strip as with all the other tags in this report, but found some anomalous results. We approached Avery and their engineers about the issues we observed. In the process, we learned that the Avery Strip was never commercially released, that Avery is currently developing a new Avery Strip based on a 96-bit Class 1 chip (the ones we tested were 64 Continues on page 7...

UHF EPC Tag Performance Evaluation

RFID Alliance Lab

3

ThingMagic Mercury4 Reader

ThingMagic Antenna

Symbol AR400 Reader

Symbol High-Gain Antenna

Alien 9780 Reader

Alien Circular Antenna

Figure 2.1: Images of reader equipment (not to scale)

UHF EPC Tag Performance Evaluation

RFID Alliance Lab

4

Table 2.1: Tags Vendor Model #

Alien

ALL9250

Alien

ALL9254

Alien

ALL9238

Avery

Avery

Common Name

EPC Class

I2

Class 1

M

Class 1

Squiggle

Class 1

DS1

Class 0

Bits

Read/ Write

64

Read/ Write

64

Read/ Write

64

Read/ Write

96

Size (mm)

Vendor’s Description

155 x 16

High gain in a controlled orientation Excellent tag for controlled environment

98 x 35

Very high gain Low environmental dependence Good for general use

100 x 13

Small UHF form factor Very low-cost UHF tag Low-cost 4x6 label solution

Read only

152 x 38

Non-metal/liquid contents (corrugated case) Metal/liquid contents (corrugated case)

Triflex

Class 0

96

Read only

Non-metal/liquid contents (corrugated case) Metal/liquid contents 112 x 102 (corrugated case) Orientation-insensitive (corrugated case)

Symbol

I2010

1x6

Class 0+

256

Read/ Write

163 x 25

General-purpose tag

Symbol

X2040

4x4

Class 0+

256

Read/ Write

90 x 93

Corrugations, chipboard, or wood crates

Symbol

I1030

Pharma

Class 0

96

Read only

30 x 36

Pharmaceutical General purpose

Rafsec

457

457

Class 1

96

Read/ Write

97 x 15

Attached to corrugate/cardboard crates

UHF EPC Tag Performance Evaluation

RFID Alliance Lab

5

Alien I2

Alien M

Alien Squiggle

Avery DS1

Avery Triflex

Rafsec 457

Symbol I1030

Symbol I2010

Symbol X2040

Figure 2.2: Images of tags (not to scale)

UHF EPC Tag Performance Evaluation

RFID Alliance Lab

6

...Continued from page 3 bit), and that even those 96-bit Avery Strip tags were not yet commercially available. Because the Avery Strip tag that we tested was never released and will never be released, we decided not to include the data in this report.

months ago for testing are now considered obsolete. Even though the tags may be obsolete, users may still find them commercially available for some time. With the recent ratification of the Gen 2 EPC standard, we expect new products and new vendors to enter the market, making the current generation of tags obsolete. We’re planning on testing the new Gen 2 tags as soon as they’re available.

These experiences illustrate just how quickly the technology is changing. Some tags that we obtained only four

Why 32.5? This sidebar assumes knowledge of dB, dBi, and dBm, which are discussed in sidebars in our previous report.* For this report, we often attenuate, or reduce, power of the reader from 32.5 dBm. Recall that 30 dBm is 1 watt. In the United States, FCC regulations limit the amount of power that can be transmitted between 902 and 928 MHz to 30 dBm “maximum transmitter power output” and a maximum of 36 dBm “effective radiated power.” Let’s break down those terms. The maximum transmitter power output is the maximum amount of power that can be sent by the transmitter. In this case, the RFID reader is the transmitter, so it is limited to 30 dBm of power. Recall from our previous report when we discussed “antenna gain.” An antenna can amplify the transmitted power in some directions by sacrificing power in others. The maximum amount of amplification is called the gain of the antenna. The effective radiated power, or ERP, is the maximum amount of power that is actually radiated from the antenna. To get the effective radiated power, as measured in dBm, add the transmitter power output, as measured in dBm, and the antenna gain measured in dBi. A 6 dBi gain antenna (typical for RFID antennas), added to 30 dBm transmitter power output, yields 36 dBm ERP.

You’ll notice that the numbers don’t add up. The reader is actually transmitting 32.5 dBm of power, not 30 dBm as required by the FCC. The reason for this is that there is loss in the cables. In fact, ThingMagic has (apparently) determined that there is 2.5 dB of loss in the cables that they provide with the antenna. Starting with 32.5 dBm coming out of the reader, and subtracting a 2.5 dB loss in the cables, means that 30.0 dBm of power arrives at the antenna. That’s why we use 32.5 dBm. By the way, it is possible in the United States to use a higher-gain antenna, such as an 8 dBi antenna, as long as one reduces the transmitter power output by 2 dBm so that the effective radiated power stays under the 36 dBm ERP limit. Generally, an 8 dBi gain antenna has a narrower beam width than a 6 dBi antenna, so doing this may be useful in situations where one wants a longer but narrower read field. In general, people tend to want the largest possible read field; and given FCC constraints, that is accomplished with a 6 dBi antenna. Disclaimer: Do not change the power settings, cabling, or antenna that come with your reader, or you may be in violation of FCC or other local regulations. Check with reader documentation and/or manufacturer for questions about changing power settings, cabling, and antennas that comply with relevant regulations.

*A Performance Analysis of Commercially Available UHF RFID Tags Based on EPCglobal's Class 0 and Class 1 Specifications.

UHF EPC Tag Performance Evaluation

RFID Alliance Lab

7