Measuring throughput of cellular and WiFi MIMO radios, Part 3
In Part 2 of this series, we compared conducted throughput measurement techniques to new OTA (over the air) techniques. We examined the challenges of MIMO OTA test methods, including time-variable wireless environments and antenna field non-uniformity. In this final installment, we discuss the requirements and challenges of constructing a MIMO OTA test bed.
Maximizing MIMO OTA throughput
A MIMO OTA test bed has to demonstrate the upper limit of the DUT performance consistently, repeatedly, and at multiple locations around the world. Referring back to Part 1 Table 3, to maximize MIMO OTA throughput, we need to create a multipath environment, making sure that:
- The signal arrives at the MIMO DUT with a wide angular spread (i.e. from all directions)
- Multipath reflections conform to industry standard PDPs (power delay profiles)
- The test bed guarantees repeatable results
Figure 1. Narrower angular spread due to longer distance to the test antenna array (top left); wider angular spread resulting from proximity to the test antenna array (top right).
The next question is, how to emulate realistic multipath characterized by a standards-based PDP in a small anechoic chamber? Part 2 explained how the delay spread of a given wireless channel is a function of the size of the physical space. Delay spread is shorter for small rooms and longer for larger rooms or outdoor spaces. Channel emulators such as Azimuth ACE, Anite Propsim and Spirent VR5 emulate delay spread, noise, and motion in a wireless channel.
SmallNetBuilder.com uses an octoBox MPE testbed to evaluate and benchmark the performance of 802.11 devices (Figure 2).
Figure 2. A handset (left) under test in a chamber that uses multiple MIMO test antennas (right).
SmallNetBuilder measures throughput using IxChariot software from Ixia, which supports measurements on the TCP/IP or UDP/IP layer. Measurements can also be reported in terms of PER (packet error rate) or BER (bit error rate), but collecting PER or BER statistics requires specialized test software that may not always be available when testing off-the-shelf devices. Thus, reviewers and quality assurance (QA) engineers often use layer 3 traffic for measuring throughput.
Using programmable attenuators, SmallNetBuilder creates plots of throughput versus path loss, as shown in Figure 3.
Figure 3. Example of a throughput vs. attenuation plots produced by www.smallnetbuilder.com.
The IP layer throughput is strongly influenced by the IxChariot test file size. That's because the network stack implementation and the way TCP or UDP frames get queued into buffers and then converted by the 802.11 MAC layer into 802.11 packets on the wireless network. Assuming back-to-back packet transmission with a minimum inter-packet gap, the longer the 802.11 packets, the higher the throughput. Each 802.11 packet contains 802.11 MAC layer and TCP or UDP/IP layer headers, which are overhead. Longer the packets result in a lower portion of a packet being the overhead traffic. Thus, larger test files result in longer packets on the 802.11 airlink medium. SmallNetBuilder uses 2,000,000 to 5,000,000 Byte file size. Higher throughput, as much as double the data rate in some cases, may be achievable using IxChariot high performance throughput script, which transmits 10,000,000 Byte files.