Anyone with IEEE publication access to summarize the following paper for the rest of us? This has relevance for those of us considering and putting up co-located nodes. ...OK, so how "serious" is the cross-channel interference? Did this paper quantify? Did they test with any sort of shielding, etc.? Presumably, newer devices and chipsets have gotten better. This is a 2007 paper.
http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=4142700&url=http%3...
"It has become a widely accepted assumption that multiple IEEE 802.11b/g transmissions in physical proximity can coexist without interfering each other. This is claimed to be the case when using separate channels with a minimum distance of 25 MHz, e.g. channel 1 and 6, which are often referred to as non-overlapping. In contrast we show that in practice cross-channel interference can be present also between non-overlapping channels if the interfering transmitter is in the proximity of the receiver. This phenomenon is known as the "near-far effect" in wireless communications. On IEEE 802.11 this has two main effects: frame corruption due to increased interference noise and channel blocking due to spurious carrier detection. The problem can be particularly serious when using IEEE 802.11 technology to build multi-hop mesh networks. Through an extensive set of experiments with off-the-shelf certified WiFi chipsets we demonstrate the presence and the detrimental effects of cross-channel interference between non-overlapping channels. We adopt an incremental approach: we first consider the case of unacknowledged broadcast packets, then we extend to regular UDP streams, finally we provide preliminary results for multi-hop TCP flows"
Here's another:
http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=4600050&url=http%3...
Sounds like they've rediscovered intermodulation distortion in RF power amplifiers. It's particularly bad with the OFDM modulation schemes used in WiFi that consist of a whole bunch of independently modulated carriers in parallel. Their sum produces a very noise-like waveform (google "central limit theorem") with a high peak-to-average power ratio that's hard to amplify efficiently and cleanly (pick at most one). Remember the standard IM distortion test of a ham SSB linear, the two-tone test? This is the "many-tone test", and it's much worse.
Seems to me the quickest workaround is to put some physical distance between units at a site that serve different sectors. Backing off RF power might also help, but I suspect most people would prefer not to do that.
Oh, another comment about IM distortion of OFDM signals. Remember BPL, Broadband over Power Line, may it hopefully rest in peace? It used OFDM. Even when it omitted RF carrier frequencies in the HF ham bands, QRM to hams was often intractable because of intermodulation distortion in the line amplifiers.