Breaking WiMedia-based UWB radios
24 July 2008
When a physical layer radio specification is written, it is intended to govern the design of a transceiver pair, and is often built on years of research and simulation.
However, when it becomes a reality, it must face the challenges of the real world, including those imagined by its authors as well as many not yet conceived.
The WiMedia specification has largely focused on testing to ensure that radios perform to specification. Along with testing, it is sometimes necessary to ascertain whether the design margins built into the specification are still in the radios. It is also important to determine whether the intended margins can be tested with enough accuracy to guarantee a condition will not arise and cause the specification’s intended goals to break down.
If a specification intends a receiver to receive a signal with an EVM (error vector magnitude) of as much as two per cent, it therefore requires the transmitters to transmit no higher than two per cent error. If the test equipment has an uncertainty of one per cent, a transmitter transmitting at three per cent EVM may be measured as passing the two per cent requirement. This ‘false positive’ would allow a condition that will break the communication link. If the margins of the test equipment and radios are not fully understood before the launch of a technology to market, it could spell disaster.
Measuring WiMedia
It is important to be able to make measurements that indicate the quality of the transmitter and to create the best and worst signals possible in a controlled manner. Typically, for radio transmitter measurements, one would utilise a spectrum analyser. These are helpful for WiMediabased multi-band orthogonal frequency division multiplexing (MB-OFDM) power measurements, but do not have the instantaneous bandwidth to capture and demodulate UWB signals.
The typical spectrum analyser in a laboratory can be used to make regulatory measurements where having large spuriousfree dynamic ranges is a key to ensuring the radio is not transmitting in an adjacent regulated band. Some of these measurements need to be made at very low noise floors (-80 to -90dBm). Spectrum analysers accomplish this by sweeping narrow-band front ends. For capture and demodulation of UWB signals, an analysis bandwidth of 500MHz (minimum) is necessary, and for WiMediabased MB-OFDM, UWB signals can be spread to over 1.5GHz.
To capture and analyse these signals, the tool of choice is a digitising oscilloscope combined with vector signal analysis software. Figure 1 depicts the vector signal analysis software and shows a ‘golden waveform’ that is formulated to perform better than the best radios implemented in silicon. This signal is used as a reference point and can help determine important parameters of the transmitter test equipment such as noise floor and repeatability. The software enables a number of measurements, such as constellation and EVM.
With the constellation measurement (upper left hand corner), a tight grouping of points indicates accurate signal generation. The EVM measurement is a RMS compilation of all the errors seen in the constellation and is indicated as a single number either in dB or per cent.
Improvements in modern oscilloscopes allow them to fit perfectly into this application. Focus over the last few generations on low noise, flat magnitude and phase, deep memory, high digitising rates and bandwidth, allow for direct digitisation of UWB signals in the 3GHz to 11GHz frequency ranges approved for this application.
Introducing impairments
With a base understanding of tools and measurements, it is possible to vary from the ideal in a quest to stress the radio implementations to and beyond the specification. Figure 2 depicts a key transmitter parameter that is purposely modulated in a very predictable way. In this example, the carrier is phase modulated with a sine wave and can clearly be seen in the common pilot error measurement in the lower right display.
Breaking receivers is accomplished by distorting the ‘golden waveform’ in ways that coincide with the distortions found in transmitter implementations or that might be introduced by the radio’s operating environment.
Taking these potential errors to the extreme determines the breaking point of the receivers, known as the margin. This is measured using the PER (packet error ratio) measurement; the measure of expected packets received to actual packets received. The specification calls out a percentage at which the dropped, (or packet received with errors), is considered unacceptable. Therefore one may ask how much the transmitted EVM can be allowed to degrade before the PER becomes unacceptable.
When phase distortion is introduced, it causes the constellation (upper left) to ‘rotate’ around the centre of the polar. Phase distortion is one of several impairments that affects a radio’s EVM. The phase distortion is increased until PER of the intended receiver falls below the acceptable percentage called out in the specification. This is then correlated with the EVM measurement, which gives an engineer a method by which to verify that the margins built into the transceiver specification can be tested.
Frequency offset
Another measurement called out in the WiMedia Phy specification is frequency offset. In order to test this parameter, the ‘golden waveform’ is transmitted off its intended channel frequency. For example, if the time frequency code 7 (TFC7) of band group one was being tested, (which by specification should be centred at 4488MHz), the actual centre frequency would then be moved in parts per million away from ideal. While the limit that the receiver should tolerate is well defined in the specification (20 PPM), and all WiMedia registered radios must comply with it, this does not fit well into a “I need to break it” mentality. Therefore, the frequency offset can be pushed away from the specified limit until the design margin has been determined.
It is also useful to understand whether the co-location of WiMedia radios or other radio technologies will break the WiMedia radio, or vice versa. To find out, testing can be done in a very controlled manner by utilising a combination of arbitrary waveform generators and signal sources to produce calibrated RF transmissions.
In the example of multiple WiMedia radios working in close proximity, a useful test is ACPR (adjacent channel power ratio). Specifying ACPR in a radio design is a way of ensuring that the energy leakage from a transmitting radio does not overpower a ‘victim’ radio pair operating in an adjacent channel.
In WiMedia radios, this interference has two major components; ACI (adjacent channel interference) and OCI (occupied channel interference). ACPR measurements are made on transmitters, but the determination of just how much can be tolerated by a receiver returns the subject to breaking radios.
Adjacent channels
This test is set-up with two radio pairs communicating in adjacent channels. Depending on the radio design, ‘images’ of the signal transmitting in one band can also show up at a lower level in another band. If a transceiver pair are communicating on WiMedia channel 13 (3690MHz), with a PER measurement being made to indicate the quality of the pairing, OCI is introduced by creating a channel 14 signal (typically centred at 4488MHz). When generated with test equipment, this can be purposely placed at 3690MHz (on top of the channel 13 signal) to emulate the ‘image’ or leakage from another radio transmitting in the adjacent channel. In this example, the power of the ‘invasive signal’ can be increased until it once again breaks the ‘victim’ receiver pairing as indicated by the PER measurement. Similarly, the interfering signal can be created as a channel 14 signal and injected on channel 14, and then the power level is increased, allowing the receiver’s ACI (or ability to reject just the adjacent channel signal) to be evaluated.
The combination of these tests indicates just how much ACI receivers must be able to tolerate and how much ‘out of band’ power a transmitter can allow. Evaluating these parameters helps the technology meet the goals set out in the original specification.
In the case of other radio technologies, WiMedia radios were intentionally designed to be able to co-exist with other technologies such as WiMAX or 802.11a. This is possible because UWB radios transmit at very low power over wide bandwidths, and can even be designed to detect and null-out some of the sub-carriers; known as DAA (Detect and Avoid).
By using a vector signal generator, it is a matter of coupling in the interfering signal with the desired signal and varying the amplitude and centre frequency to test the performance in this condition. The measure of quality is PER.
The margining of the test equipment and radios provides the specification implementers with confidence that despite measurement errors and process drift, consumers will receive a product that performs as advertised.
MARK LOMBARDI is RF application specialist, Agilent Technologies.
Contact Details and Archive...
Related Articles...
Most Viewed Articles...