Upstream Rate Vs. Range
I had more luck getting the Buffalo Nfiniti to run upstream, so included it in Figure 19. You can see that the Buffalo quits pretty quickly after it drops throughput. Although the Buffalo card driver always reported a 270 Mbps link rate, the plot sure looks like a rate shift took place.
Figure 19: Upstream Rate vs. Range comparison
The same key points we saw in the downstream tests, however, still appear to hold true for upstream. In the end, the RangeMax 240 / Airgo Gen 3 pair maintains higher throughput for longer than any of the draft 11n products.
Does Draft 11n Beat 11g?
Since I had an 11g AP and card handy, I decided to run it through the Azimuth test and compare it to both the draft 11n and Airgo Gen 3 products. Figure 20 is, as they say, the "money shot" and speaks volumes about the current state of draft 11n products based on both Broadcom and Marvell chipsets.
Figure 20: Upstream Rate vs. Range Compared to 802.11g
You can see that the only product that has a rate vs. range curve better than the 11g pair tested is the RangeMax 240 (Airgo Gen 3). Both Broadcom-based products' throughput fall off eariler than the Airgo and fail to transition to lower link rates so that they can stay connected. The Marvell-based Netgear Next does manage to transition to lower link rates and stay connected, but at a throughput lower than that provided by the 11g pair.
The 802.11g products used were a Cisco Series 1200 AP (Air-1232AG-A-K9) and AIR-CB21AG-A-K9 client card. The "Cisco 11g" trace in Figure 20 is actually the average of 20 runs taken overnight to check the repeatability of the Azimuth system. Figure 21 shows an overlay plot of all 20 runs, which shows a repeatability far exceeding anything possible with open-air testing.