Dear all, I am following the multi-radio mesh discussion with a lot of interest. Indeed multi-radio environments are tricky. There are few things to remember:
Traditionally, 802.11 counts channel numbers in increments of 5 MHz. Table 15-7 on page 566 of 802.11-2007 <http://standards.ieee.org/getieee802/download/802.11-2007.pdf> lists the center frequencies of each channel in the 2.4 GHz band. This channel numbering is valid for DSSS and OFDM operation. With 802.11, the standard channel bandwidth is 20 MHz (other options are 5, 10, and 40 MHz channels). Although regulatory authorities in Germany allow WLAN operation in the 2.4 GHz in the range from 2400.0 GHz to 2483.5 GHz 802.11 defines the first center frequency to be at 2412 MHz and not at 2410 MHz (± 20/2 MHz). This is because of inevitable out of channel emissions. Neither an OFDM (802.11a/g/j/n) nor a DSSS (802.11b) signal forms a perfect rectangular. For example, 15.4.7.4 in 802.11-2007 explains the transmit spectrum mask for DSSS. Thus, the channels are not orthogonal to each other. So when a device incorporates two radios they will interfere with each other because of their out of band transmissions. Assume that you have two radios transmitting at 20 dBm. Let's assume both antennas are separated by 10 cm. Thus we have ~20 dB path loss between them. So, they mutually receive their signals at 0 dBm. Let's further assume the radios are emitting at 50 dB less in each others frequency band. So we are down to -50 dBm of mutual interference. That's still a lot! Compare this to a free space loss of 60 dB at 10 m distance. The adjacent channel interference is no different in the 5 GHz band. When one radio transmits in channel 36 (center frequency 5.18 GHz) and another radio transmits in channel 40 (center frequency 5.2 GHz), they will experience adjacent channel interference. Yeoh reported that he has 17.8 Mb/s TCP throughput on a single hop connection. Remember that this includes IP and TCP overhead. All in all that's a very good value. With 802.11g the maximum real world throughput is around 22 Mb/s with the 64 QAM 3/4 MCS. Please remember that maximum theoretical throughput calculations often assume MSDUs of 2304 B size. However, with APs connected to an Ethernet you won't see payloads larger than 1500 B. One can squeeze out more throughput with 40 MHz channels, block acknowledgments, frame aggregation etc. However, that usually comes with 802.11n. Another thing to remember is the hardware. Although document <http://www.tti.unipa.it/~ilenia/pub/info07.pdf> discusses 802.11b and not 802.11g NICs it's still extremely interesting to read. There are so many bugs in theses cards that it is very difficult to present reliable results. Best regards, Guido _______________________________________________ Devel mailing list Devel@lists.open80211s.org http://open80211s.com/mailman/listinfo/devel
