802.11n Throughput Testing Methodology: Hopeless with Existing EquipmentBy CWNP On 08/16/2007 - 4 Comments
I was a little perturbed at 802.11n when I first started testing it, and now, after discovering the truth, I'm just saddened.
I turned off my entire WLAN network here in the lab. There are no nearby APs, so my Wi-Fi environment is clean as a whistle. I fired up a spectrum analyzer - again, clean. Then I enabled my spiffy new Apple Airport Extreme Basestation (802.11n using an Atheros AR5008-3NX chipset) on 2.4 GHz using channel 6 and 20 MHz channel width (because that's all that is available in the Apple configuration settings). I connected my brand new Buffalo Technologies dual-band 802.11n, 2x3 MIMO CardBus card to a P4 laptop with 1GB of RAM. I started a file transfer, and... it averaged anywhere from 19 to 23 Mbps after several trials. I get roughly 21 Mbps with 802.11a or 802.11g in the same "multipath hell" environment. The laptop was sitting about 3 feet from the Apple basestation. The client utility (I'm using WinXP/sp2 WZC because, if you can believe this, it's better than the Buffalo client utilities for this card) says I'm connecting at 130 Mbps in the 2.4 GHz band. I also tried Windows Vista Business using its integrated wireless client. It also said I connected at 130 Mbps in the 2.4 GHz band.
I decided to check out the 5 GHz band and a 40 MHz channel width, so I configured the Apple basestation for this (though it won't let you choose the channel). My clients (both Vista and XP) connected at 300 Mbps. I ran the same file transfer and got an average of 53 Mbps on both clients. Again, big whoopie. I can't imagine having a more multipath-rich environment than the one I'm in. 802.11n is supposed to smoke in the face of multipath. Where's my throughput! I decided to move my station away from the basestation - no change.
I made sure I didn't have any other bottlenecks by pulling the same file across my Gigabit Ethernet infrastructure. I averaged 105 Mbps over a long period of time. The limitation of 105 Mbps is likely my NAS where the file is stored.
I decided to go for a little stroll around the lab and then around the entire building. In the lab, it was varying from mid 20's to mid 40's using 5GHz with 40MHz wide channels. Walking down the hall with multiple walls in the path between my CardBus card and my basestation, it started going down, down, down just like you'd expect...and just as quickly as 802.11g would have though I'm guessing that at the same distance, my 802.11n throughput is likely better than I would've had with 802.11g. I don't really care to do that kind of comparative testing at the moment.
I fired up AirMagnet Laptop Pro v7.5 (with 802.11n support) to look for clues on this 802.11n slowness. After looking at Association Request frames, Beacons, Data frames, and ACKs, I was even more confused as to what was happening and why. Suddenly I found myself knee deep in the 802.11n draft again...looking at data rates, Tx/Rx STBC support, multiple spatial stream support, MCS support, etc. Most every feature I looked at was disabled in the Apple. All of the data frames in the 2.4 GHz band were moving at 54 Mbps and ACKs at 24 Mbps. This explains the 802.11g-like speeds. Nothing was moving at 802.11n rates (even though my client utility says it's connected at 130 Mbps). My guess is that the existing Apple firmware is basically 802.11a/b/g firmware with minor additions that allow HT (802.11n) clients to operate, but not to take advantage of any of the new features of the HT PHY that give you great range and speed. The client station has a reasonable number of configurable options (through the driver's properties in Windows), though it's not exactly an impressive list. It's certainly more than the Apple AP, but hey, what can you expect for $179, right?
The supported rates for the client were only up to 54 Mbps as shown in the Association Request frame. Tx/Rx STBC isn't supported on the Apple for 2.4 GHz or 5 GHz bands. Beacons also show that the maximum number of spatial streams is 0. Probe Request frames and Association Request frames (both sent by the Buffalo station to the Apple AP), are supposed to have an HT Capabilities information element in them (802.11n, section 188.8.131.52.1), but neither does. This all adds up to, "you get what you pay for, IMHO." Unfortunately, at the moment you can't pay for anything better because it's not available.
I decided to test multiple transfers in 2.4 GHz to see if I would get added throughput. Nope. Combined throughput was still in the 19 Mbps range. In the 5 GHz band, a second data stream boosts overall throughput up to ~70 Mbps (though it's a little jumpy). These results combined with what is (not much) and isn't (most everything) supported in Beacons makes absolutely no sense at all.
Now once again I ask, "Where's my throughput!" 100+ Mbps as advertised? NOT. Right on Apple's website it says, "Blazing fast, it delivers up to five times the performance and up to twice the range compared to 802.11g routers." On Buffalo Technologies' website, it says, "Blazing Speeds (12X Faster than standard 802.11g)." Grrr. I get 21 Mbps out of 802.11a in the same place, and often 19 Mbps on 802.11g. Do the math. 5X? 12X? NOT.