The interaction of the radio frequency identification (RFID) reader and tag is of great interest to the RFID system designer. Important factors to consider that affect system performance include the environment where the RFID will be usedwhich is likely to contain many reflective and absorptive surfaces as well as other tags.
In additionthe increasing use of other RF systems in offices and industrial sites means that other strong RF emitters will be operating in close proximity to the RFID systemalthough in different frequency bands.
An example is an RFID reader that is connected to a network via a WLAN. While there are many system level challengesthe circuit level design of the tagthe integration of the antenna and the rest of the circuit directly affects both the power available to the integrated circuit (IC) and its usable range.
Passive RFID tags
It is also important to recognise that passive RFID tags communicate to the reader or interrogator by modulating the incident RF energy to create a ‘backscatter’ signaland that any losses between the antenna and IC will limit this ability and hence the maximum distance from which the tag can be read.
A basic component of the RFID tag’s IC is the detector diode. The Schottky barrier diode is often used to detect or rectify RF energy. A tag circuit may use such diodes as discretes or in an ICbut for design purposes it is useful to have a model of the Schottky diode suitable for use with various circuit simulators. These models fall into two categories – the so-called lumped equivalent or small-signal modeland the non-linear Spice model.
The former is useful for design using scattering or S parameters and the latter can be used to predict performance under active conditionssuch as displaying a network's output voltage. The lumped equivalent model is shown in Fig.1.
The values shown are for a diode intended for zero-bias operation and that will have a square-law detection characteristic.
Given this model and a desired operating currenta conjugate match can be found that will optimally transfer power into the network. Howeverthis model will not support simulation using non-linear simulation such as harmonic balance.
Using the small-signal diode modelSmith charts can be used to find the optimum match at a given frequency. Our design is focused on tags for the UHF band and we will choose an operating frequency of 915 MHz.
A Smith chart design tool software application may be used to derive a matching network for a single diode. In practiceit is desirable to use two diodes in a voltage doubler configuration to increase the rectifier’s output voltage. The network is derived in terms of inductance and capacitance values.
In order to be builtit would be transformed into equivalent distributed structures and placed between the antenna and the IC.
To verify our design under more realistic conditionsa non-linear circuit simulator may be used. The small-signal diode model is replaced by the SPICE modeland the input power is swept.
The resulting graph of output voltage versus input power indicates that we will obtain about 0.3VDC for a reference input of -6dBm.
To communicate with the readerthe tag modulates the input impedance of the tag IC. A modulating impedance is switched in or out synchronously with the data being transmitted.
The resulting modulation is likely to include both phase and amplitude modulationrequiring the reader to employ complex demodulation.
A poorly matched antenna will limit the degree to which the modulating impedance can affect depth of the modulated backscatter.
Traditionallyantennas are designed in an electromagnetic (EM) simulator and an S-parameter file is brought to the circuit design. The Smith chart method is used to match the antenna port to the IC. The matching circuit transforms the 50ohms at the antenna port to the impedance of the IC.
While this method works finethe matching circuit might cause insertion lossresulting in loss of the signal in both the directions and affecting both the IC’s rail voltage and the ability of the tag to communicate with the reader via modulated backscatter.
Because every fraction of a dB is critical in the operation of a RFID system it is desirable to eliminate the matching circuit and design the antenna to directly match to the IC impedance.
The key to being able to directly match the antenna with the IC is to have an antenna topology that produces variable antenna impedance with respect to one of its geometrical parameters.
A microstrip patch antenna with a well-controlled radiation pattern (see Fig.2) and optimised for narrow band operation around 915MHz was designed using Momentuman EM simulator in the Advanced Design System (ADS). A model is built for this antenna that is a function of the depth of the feed line.
As shown in Fig.3a model of the antennabuilt earlier in an electromagnetic simulatoris brought in direct connection with the IC. Due to the use of this ‘layout component’ which has the appearance of the layout but placed in the schematicit is now possible to tune and optimise the circuit to give a higher output voltage (0.4V at Pin = -6dBm as compared with 0.3V for the conventionally-matched circuit)especially when the tuning process in the circuit simulator responds in real time.
The model of the antenna is built using a new technique that is available in the ADS. The technique relies on adaptively sampling the parameters in a multi-dimensional environment referred to as MAPS.
The response of the antennain terms of its S-parametersis curve fitted to a rational polynomial at every computation until the error is below -60dB.
At this point the modela polynomialis constructed as a function of the parameters. This adaptive sampling contrasts with uniform sampling of the parameters and has two distinct advantageswhich are illustrated in Fig.4a and 4b.
Fig. 4a shows a traditional approach in which results in over sampling (inefficiency) or under sampling with local interpolation (ianaccuracy).
Fig. 4b shows a modified approach with global adaptive modelling (MAPS) and Forsythe interpolation results in better accuracy and higher efficiency.
Firstthe adaptive sampling produces a model having the minimum number of pointseliminating time-consuming EM simulations.
Secondit offers higher accuracy due to Forsythe interpolation that takes into account the global behaviourbecause uniform sampling can result in under or over sampling.
Under sampling results in poor accuracy and over sampling results in unnecessary and time-consuming EM simulations.
The model allows rapid tuning of the depth of the feed pointwhich in turn changes the impedance of the antenna port. This enables the designer to directly match the IC to the antenna port to maximise the rectified output voltage.
In the specific resultsthis design approach yields an improvement of 25percent more output voltage by directly matching the antennacompared with the potentially lossyconventional Smith chart basedindirect matching method.
Traditional Smith chart techniques for matching an antenna to a circuit can result in signal losses as elements are added.
Such lossesas well as the higher costs implied by the added complexityare highly undesirable in radio frequency identification applications due to the need to power the circuit using the incident RF energy and to modulate that signal with the tag’s responses.
An alternative methodbased upon inserting a compact parameterised model or ‘layout component’ into a circuit simulator enables fast tuning of a simple physical characteristic of the antenna to obtain the optimum match without the need for added components or structures.
This method significantly increases the voltage powering the tag IC and the simple design helps to reduce the production cost per tag.
Cory Edelman is an applications engineer for Agilent Technologies in Westlake VillageCalifornia and Murthy Upmaka is a senior application engineer with Agilent’s EEs of Comms EDA organisation. For more informationvisit www.agilent.com or eesof.tm.agilent.com"