|When||Sun 01 Feb 2004 7:00pm|
|Who||Interested members of the Software and Communications and Professor Jim Binkley|
|Topic||802.11b lower layers & Linux drivers.|
- Is there any error correction in 802.11b (I believe not)
- Is there any automatic ping or gain control on 802.11b (I believe in Ad hoc mode there is neither, but in managed mode there is)
- what will the driver cough up to upper layers if the reciever gain is set too low? too high?
- suggestions on testing
- multipath behavior effects on testing?
- is it possible to get RSSI data out of the hardware when in Ad hoc mode? The current driver doesn't provide any
- interference from other wifi or microwave equipment - experiences? we have these high-gain antennas that will tend to select out local interferers when the rocket is close, but what about at maximum height?
From Prof. Binkley:
"Actually I have lecture materials, but the entire morass needs to be printed out and I don't have time at the moment. Said morass is at: http://www.cs.pdx.edu/~jrb/netsec/lectures/80211
theory.txt - overview/rant
associate.txt - packet trace of association
beacon.txt - packet trace of AP beacons
might make a good starting point."
So if you have time, please go over these notes before we meet Sunday, at which point we can batter the poor fellow with questions.
We met with Prof. Binkley for an hour and a half and discussed the ins and outs of the lower levels of IEEE 802.11b and how we might best use it for LV2's telemetry channel. He started with an intro to the technology, including 802.11b's many modes, and the talk changed into a barrage of question and answers. Prof. Binkley did a fine job of dealing with all of our ridiculous questions.
We can't get access to the MAC layer; $25,000 development kit for the firmware, and all existing firmware is insanely proprietary. A few have "leaked": Wavelan, Orinoco, etc.
Note that Proxim bought Lucent which was AT&T which bought NCR's original WAVELAN folks. So it's changed a bunch of times.
802.11b spec is available for free download from IEEE.
Some "magic" features available, but it's a tricky business; you need this card version with that driver and this patch, etc etc. Probably not worth the time and effort.
- Managed mode - bad idea for us.
- Lots of control packets + Data
Ad Hoc mode
- little bit of control + mostly data
- Lucent has a "Demo" ad hoc mode which came out before the 802.11b standard
- RTS-CTS - bad idea for us.
- Usual unicast data mode: A sends, B acks
- RTS/CTS mode: A RTS, B CTS, A sends, B acks (for hidden node problem). Takes a chunk of bandwidth away for no apparently good reason.
- Power savings mode (TDMA like coordination for power savings.) - bad idea for us.
- RTS-CTS - bad idea for us.
- No client probe, just AP beacons, auth, auth, assoc. request, assoc. response.
- Insanely complex.
- Not deterministic.
- Lots of hidden weirdnesses.
- No reason for us to use it (no advantages).
Ad Hoc Mode:
- ad hoc = IBSS mode - independent basic service set.
- Still uses 100Hz beacons. Probably <= 64 bytes per beacon, but wow, what a waste. Might not even serve a useful purpose!
- Still uses SSID
- Some kind of Master/Slave dynamic (?)
- We might want demo mode because we lose the 100Hz beacon!
Physical Layer Stuff:
- DSSS, 8QPSK
- Lucent Orinoco silver = 56bit wep, gold = 128bit wep, otherwise exactly the same.
- Orinoco card = Hermes chipset
Channels 1, 6, and 11 theoretically don't interfere with each other... right now we're using Channel 1 for the rocket, so maybe channel 11 for LTC-LM link. Or if we need to tune to the WiFi cylindrical patch antenna, we can go to channel 6 and go +/- depending on resonant frequency.
MAC Layer Stuff:
Because there's no FEC, any bit flip or drop on a wireless packet will cause it to disapear because of the FCS (Frame Check Sequence (checksum)). So FEC on the data isn't worth it. Keeping packets small = good.
At 11Mbps, why do we only get 5Mbps of throuput?
- "ping-ponging": waiting for the ack wastes time (for unicast ONLY - multicast, broadcast don't use acks!)
- Preamble (header): said to be at 2mbps, then right afterwards a declaration of the actual speed
- Relatively conservative fairness protocol: the back off is more conservative than ethernet, as well as listening for other nodes.
It looks like we are going down the right track. In "demo" ad hoc mode using broadcast packets, we have our one-way non-acked comm system. But no FEC, so we should try to keep packets relative small. With the right software, we should even be able to get some kind of signal strength measurement.