Download this article in .PDF format
This file type includes high resolution graphics and schematics.

The results of this experiment yield a rough rule of thumb for setting the reference level. The PAPR of the OFDM signals detailed earlier was extracted (Fig. 3) from the complementary cumulative distribution function (CCDF). It might make sense that the optimum reference level would be set at the sum of the expected average packet power and the PAPR, as this is the maximum expected signal level.

3. This plot shows the complementary cumulative distribution function (CCDF) of an 80-MHz-bandwidth, MCS9 VHT signal.

But looking again at Figs. 2 and 3, this is not the case. While the PAPR indeed captures the ratio of the maximum power, as seen in the CCDF, the vast majority of the information is contained in a much lower power level envelope. Setting the reference level at the maximum power level instead takes dynamic range away from the majority of the packets. From the CCDF, it can be seen that the PAPR is around 11 dB. To make good EVM measurements, the reference level should instead be optimized at a delta value closer to 7 dB, or about 4 dB below the sum of the expected average and the PAPR.

In addition to the more stringent SNR requirements for IEEE 802.11ac signals, the larger bandwidths and more complex modulation schemes also require more signal processing to be performed during demodulation measurements. As a result, measurement times for IEEE 802.11ac signals increase over previous revisions of the standard. With measurement test times inherently increasing, it is desirable to intelligently optimize the measurement parameters with respect to measurement quality and duration.

One such parameter in making good modulation accuracy measurements is the number of averages upon which the EVM measurement is based. While IEEE 802.11 standards specify 10 averages for a measurement, this number of overages can result in significantly longer measurement times. Thus, when testing a device, the trick is to identify the number of averages required to gain the desired measurement repeatability. As Fig. 4 shows, measurement repeatability gradually improves as more averages are used to compute the measurement result.

4. These traces show standard deviations of for measured EVM performance levels.

In practical use, measurement repeatability in the range of 0.1 dB is sufficient for most automated test applications. Should a wider variance be acceptable, three averages should yield good results while requiring less test time. Conversely, if repeatability is critical, increasing the number of averages will yield more consistent results at the expense of longer measurement times. When comparing measurement results, both EVM as well as test time, between different applications, it is important to also note the number of averages used in each.

In summary, the latest iteration in IEEE 802.11 specification, the IEEE 802.11ac amendment, continues to push performance forward as it adds an order of magnitude of potential throughput. Adding additional spatial streams and higher order modulation over wider channel bandwidths can be seen to be responsible for much of this greater throughput. And, as with the specification, test and measurement systems similarly require incremental improvements in items such as real-time bandwidth, improved linearity, and greater dynamic range. Optimizing available parameters in the measurements serves to maximize the capabilities of the instrumentation as well as the reliability of the results.

Download this article in .PDF format
This file type includes high resolution graphics and schematics.