Folks:
I'm not quite sure who was doing the calculation but...
<< Theoretical receive sensitivity can be calculated as:
Receive sensitivity = Nt + Ns + SNRmin
where Nt is the thermal noise floor, Ns is the system noise figure, and SNRmin is the minimum SNR required for a given bit-error rate.
802.11b's CCK requires a SNR of about 10dB to decode frames with a FER of 0.08. (We'll leave SIR and/or SINR out of the discussion unless someone insists.) We'll call this SNRmin.
>>
The required channel (S/N) for CCK11 to achieve a packet error rate of 0.08 is about 7 dB, not 10 dB [source: fig 2.16 of "Evolution of 2.4 GHz Wireless LANs", Heegard et. al., in "Wireless Local Area Networks", ed. Bing, Wiley 2003]. The noise effective bandwidth of the 802.11 signal is also closer to 16 MHz rather than the 22 MHz needed to incorporate all the power. The thermal noise floor is thus about -101 dBm. Good noise figures for chipset radios are about 5 dB, with probably 1-2 dB extra board losses, so we obtain a noise floor around -94 dBm; best case receiver is thus (-94+7)=-87 dBm. This is in good agreement with most other published sensitivities: I averaged reported data for about 15 cards and got (-83 +/- 3) dBm at 11 Mbps. So -91 dBm is either a measurement error or the use of a low-noise amplifier (NF < 2 dB at 2.4 GHz) in front of the card receiver. LNA's of that type cost about $1, which is a lot for these inexpensive cards.
Daniel M. Dobkin Enigmatics 1-408-314-2769 [EMAIL PROTECTED]
_______________________________________________ BAWUG's general wireless chat mailing list [unsubscribe] http://lists.bawug.org/mailman/listinfo/wireless
