When I first started testing wireless LAN routers and APs, long, long ago, I used a "standard" client. For 802.11b products, it was an Orinoco Gold CardBus card. And when 802.11g products came onto the scene, I used a Linksys WPC54G. The reason for this was simple: to have only one variable—the router—when testing wireless performance.
Since I started testing draft 802.11n products, however, I have asked manufacturers to provide a "recommended" or "matching" card for the wireless router they sent. Given the draft status of 11n and implementation differences from chipset to chipset, I figured manufacturers would know what would show their products in the best light.
Recently, however, I think this decision is causing more problems than it is solving. Even though most current draft 11n products use "two stream" chipsets (see this article for more about that), all "two stream" chipsets are not equal.
I don't pretend to completely understand the effect that the number of transmit and receive "chains", different chip architectures have on wireless performance. It does seem, however, that the number of antennas on a draft 11n router does affect maximum throughput, and I'm pretty sure that differences in client architectures affect that number, too. Table 1, copied from the Linksys WRT110 review illustrates the point pretty well.
|20 MHz B/W||40 MHz B/W|
|Range Plus||~ 40||~ 60||WRT110||$50-$94|
|Draft 11n - Two Antenna||~ 60||~ 77||WRT160N||$68-$102|
|Draft 11n - Three Antenna||~ 70||~ 94||WRT350N||$129-$180|
Table 1: Linksys Product Line Comparison
I also have had more than a few cases where I strongly suspected that the client adapter was the weak link in wireless performance results. In one of those cases, I actually used a different client than supplied by the manufacturer. But in most others, I just went ahead and used the adapter provided.
But I think the thing that has pushed me toward looking to use a "standard" draft 11n adapter for all testing is that manufacturers are starting to mix chipsets from different manufacturers in their product lines. So if they are ok with not matching chipset families on the router and client sides, I guess then, so am I.
So that leaves me with the decision of which client to use. It turns out that decision isn't as easy as I thought. I went looking for an adapter with the following features:
- Dual-band: This is the future and I don't want to be swapping adapters
- Cardbus or mini-PCIe format: Both these formats support three antennas, while most USB adapter designs support only two. More antennas don't help range, but do determine maximum throughput, in my experience.
- WPS support: I need to be able to run WPS tests. There also is no excuse at this point, for not supporting WPS in all draft 11n products.
- 20 and 40 MHz channel bandwidth modes: Although I'm not a fan of using the channel-hogging 40 MHz bandwidth in the 2.4 GHz band, it's a fact of life. So I need to test it.
- XP and Vista drivers: Note that these must be installable in any notebook and include a client application that supports WPS, since Windows does not have built-in WPS support.
I prefer a built-into-a-notebook solution for its convenience and better antenna orientation, which led me to Intel's 4965AGN mini PCIe adapter. I thought my quest was over as quickly as it started, until I remembered that Intel locks out 40 MHz bandwidth operation in 2.4 GHz. Although I agree with that decision from a neighbor-friendliness view, I need to test both operation modes. So that rules out the Intel adapter.
The only other dual-band product that comes close to meeting my selection criteria is the Linksys WPC600N Cardbus card. But, for some reason, Linksys refuses to add WPS support in its client utility, so that rules out that option, too.
So, for now at least, it looks like I'll continue to test draft 11n products with their manufacturer-recommended clients. And continue on the lookout for a reference client for SmallNetBuilder's draft 11n testing. If you know of any products that meet all my criteria, please drop me a note.