Page 1 of 3

Channelvelocity thoughts

PostPosted: Sun Apr 23, 2006 9:09 am
by Mark57
I've been rereading this thread on a more efficient Kismet channel hopping idea and was wondering/considering in light of that thread what's the best velocity setting to use and why. The default is 5 channels per second. I use 7 with a sourcechannel scheme of 6,6,1,1,11,3,4,6,11,10,6,1,5,7,11,6,1,8,6,11,9,6,2,6,6. I've played with the velocity number but have not been able to see much difference in results whereas changing to the custom sourcechannel does make a difference in my area. Just wondering what velocity setting you use and if you have strong feelings about that setting.

I'm using a 1 watt pre amp so I'm usually able to see the signal for a longer period of time and therefore have increased the velocity setting based on that. CPU is 1.7 GHz and card is Senao 2511-CD PLUS EXT2 with a 5.5 dB omni.

PS, I've been running Netstumbler side by side with Kismet and it's interesting to see that Netstumbler still finds stuff that Kismet doesn't have time to get. I know multiple cards, etc would help but I'm trying to optimize what I have. Netstumbler finds from 2 to 18 additional networks that Kismet missed per 2 hour run so it's not a huge number. That's with Kismet finding 2,500 to Netstumbler's 900 networks on that same run.

PostPosted: Sun Apr 23, 2006 9:55 am
by Dutch
Mark57 wrote:I've been rereading this thread on a more efficient Kismet channel hopping idea and was wondering/considering in light of that thread what's the best velocity setting to use and why. The default is 5 channels per second. I use 7 with a sourcechannel scheme of 6,6,1,1,11,4,6,11,10,6,1,7,11,6,1,8,6,11,9,6,2,6,6. I've played with the velocity number but have not been able to see much difference in results whereas changing to the custom sourcechannel does make a difference in my area. Just wondering what velocity setting you use and if you have strong feelings about that setting.

I'm using a 1 watt pre amp so I'm usually able to see the signal for a longer period of time and therefore have increased the velocity setting based on that. CPU is 1.7 GHz and card is Senao 2511-CD PLUS EXT2 with a 5.5 dB omni.

PS, I've been running Netstumbler side by side with Kismet and it's interesting to see that Netstumbler still finds stuff that Kismet doesn't have time to get. I know multiple cards, etc would help but I'm trying to optimize what I have. Netstumbler finds from 2 to 18 additional networks per 2 hour run so it's not a huge number. That's with Kismet finding 2,500 to Netstumbler's 900 networks on that same run.


It's far better to fifle with the defaultchannels, than the channelvelocity IMHO.

I'm using a channelvelocity of 5, due to the following reason : Most AP's are setup to send a beacon every 100 msec, i.e. 10 beacons pr second. With a channelvelocity of 5, you have a chance of receiving up to 2 beacons pr channel during that 1 second timeslot, thereby increasing the chance of receiving a complete unmangled beacon or datapacket on each of those channels that are scanned in the timeslot.
If anything I wouldn't increase the channelvelocity, but decrease it. But it all boils down to how fast you are moving ofcourse.

With regards to the defaultchannels, I'm using this order : 1,6,11,7,13,2,8,1,6,11,3,4,1,6,11,9,10,11,6,1,1,6,11,5,12. The observant reader will notice that I've included 1,6,11 in each pair of 5 channels, so each "timeslot" contains the 3 channels most often used. This means that a scan where all channels has been listened to, would take 5 seconds.
With the speedlimit in the city being 50 km/h, and the actual attainable driving speed in the city being around 35-40 km/h during the day, it would mean that I'd at maximum would move 69.44 meters, but normally around 48.61 meters between each complete allchannel scan, giving me a fair chance for being inside the range of AP's no matter what channel they are using, during that 5 second timecycle.

This setup works fine for me, but as usual YMMV..

Dutch

P.S. For the metric challenged among you colonials, 50 km/h = 31.1 mph, 35 km/h = 21.7 mph and 40 km/h = 24.9 mph.

PostPosted: Sun Apr 23, 2006 10:08 am
by Mark57
That's good stuff. I like your logic. I've tried settings between 1 and 10 at speeds averaging 25 mph but did not see any measurable difference. The vast majority of my stumbling is well under 40mph and closer to 25mph.

PS, Channel 4 is missing in your scheme or a typo.

As always, more testing seems to be needed

PostPosted: Sun Apr 23, 2006 10:42 am
by Dutch
Mark57 wrote:That's good stuff. I like your logic. I've tried settings between 1 and 10 at speeds averaging 25 mph but did not see any measurable difference. The vast majority of my stumbling is well under 40mph and closer to 25mph.

PS, Channel 4 is missing in your scheme or a typo.

As always, more testing seems to be needed

Typo... I'm in the ETSI regulatory domain, so I'm not scanning channel 14.. That should be channel 4 instead.

Dutch

PostPosted: Sun Apr 23, 2006 11:19 am
by King_Ice_Flash
You would not belive how many people are using ch 14 in my area, especially in the major cities where the wireless traffic can be quite bad. I should get a job for the FCC trying to detect out of regulation signals :D Then I can drive on company time.

PostPosted: Sun Apr 23, 2006 8:46 pm
by Airstreamer
Mark57 wrote:I've been rereading this thread on a more efficient Kismet channel hopping idea and was wondering/considering in light of that thread what's the best velocity setting to use and why. The default is 5 channels per second. I use 7 with a sourcechannel scheme of 6,6,1,1,11,4,6,11,10,6,1,7,11,6,1,8,6,11,9,6,2,6,6. I've played with the velocity number but have not been able to see much difference in results whereas changing to the custom sourcechannel does make a difference in my area. Just wondering what velocity setting you use and if you have strong feelings about that setting.

I'm using a [color="Red"]1 watt pre amp [/color]so I'm usually able to see the signal for a longer period of time and therefore have increased the velocity setting based on that. CPU is 1.7 GHz and card is Senao 2511-CD PLUS EXT2 with a 5.5 dB omni.

PS, I've been running Netstumbler side by side with Kismet and it's interesting to see that Netstumbler still finds stuff that Kismet doesn't have time to get. I know multiple cards, etc would help but I'm trying to optimize what I have. Netstumbler finds from 2 to 18 additional networks that Kismet missed per 2 hour run so it's not a huge number. That's with Kismet finding 2,500 to Netstumbler's 900 networks on that same run.

Just being "nit picky". How much gain does the pre-amp have? (As that is what would matter to Kismet since it doesn't transmit, whereas N/S does.)

PostPosted: Mon Apr 24, 2006 4:19 am
by Mark57
Airstreamer wrote:Just being "nit picky". How much gain does the pre-amp have? (As that is what would matter to Kismet since it doesn't transmit, whereas N/S does.)


It's 18 dB gain but the antenna is an Mobil Mark 5.5 dBi mag mount omni so the pattern is not as flat as a high gain antenna if that's where you're headed. Yes, this is receive only. There is no transmit. The amp can increase noise levels but it also increases networks found.;)

PostPosted: Mon Apr 24, 2006 4:39 am
by Mark57
I did some testing last night but I'm not ready to post results just yet. I want to do all the tests again tonight if the forecasted severe weather will allow. I think you'll find the results interesting. I can say without a doubt, both sourcechannel schemes that Dutch and I use are superior to the default 1,6,11,2,7,3,8,4,9,5,10. As to channelvelocity, that's more interesting.;)

Due to the nature of 2.4 GHz signals, it's hard to find a good test lab out in the real world environment that results can be duplicated. What I did was drive a 1 mile by 1 mile square section (4 mile route total) that has residential, business, and a major hospital complex along the way. My start and stop place on this route is in an area where there are few if any WiFi signals so I can get a clean cut off to record the results of each drive. The tough part in a real world test lab is controlling variables like speed, what lane you're forced to drive in to maintain that speed, are 4 major stop lights green or red and how long you wait at each, etc.

I'd "like" to run all tests 3 times but it takes almost 2 hours to do all the tests just once. Depending on the time of day you start, your environment can change a lot in those two hours. I'll post the results when I finish testing.

PostPosted: Mon Apr 24, 2006 4:44 am
by King_Ice_Flash
The thing that I am wondering is if the increased time it takes to cycle through all of the extra hops will be affected by higher speed.

PostPosted: Mon Apr 24, 2006 4:51 am
by Mark57
King_Ice_Flash wrote:The thing that I am wondering is if the increased time it takes to cycle through all of the extra hops will be affected by higher speed.


That's what Dutch was trying to minimize with his scheme. The key is to have a hop scheme that is adequate to not miss the most prevalent channels in use and yet scan the others often enough to not miss those either. I think at some point the increase in travel speed will degrade the results more than the hop or velocity settings could handle. What is that speed? Good question, and hard to measure.

Again, we know more than one card can fix a lot of this but that's not possible for everyone.

PostPosted: Mon Apr 24, 2006 6:13 am
by Airstreamer
[quote="Mark57"]It's 18 dB gain but the antenna is an Mobil Mark 5.5 dBi mag mount omni so the pattern is not as flat as a high gain antenna if that's where you're headed. Yes, this is receive only. There is no transmit. The amp can increase noise levels but it also increases networks found.]
Ok, thanks, just curious. NS is triggering the 1 watt transmit amp, right?

PostPosted: Mon Apr 24, 2006 6:28 am
by Mark57
Airstreamer wrote:Ok, thanks, just curious. NS is triggering the 1 watt transmit amp, right?


Noooooo, I never use the amp with Netstumbler. I don't want to microwave my eyeballs.:eek: I never use it in transmit mode although it's capable. I only use it with Kismet. I run Netstumbler on a second laptop and separate antenna.

PostPosted: Tue Apr 25, 2006 6:14 am
by Mark57
[quote="Mark57"]I did some testing last night but I'm not ready to post results just yet. I want to do all the tests again tonight if the forecasted severe weather will allow. I think you'll find the results interesting. I can say without a doubt, both sourcechannel schemes that Dutch and I use are superior to the default 1,6,11,2,7,3,8,4,9,5,10. As to channelvelocity, that's more interesting.]

Test update: I'm still accumulating data and it's looking pretty interesting. I want to make sure I have enough data to be able to make a valid conclusion. Some of the people on the route I'm driving are getting VERY curious/suspicious. I had to stop and tell some construction guys working next to my start-stop point what I was doing after driving by 14 times in two hours yesterday. They laughed and said they were about to block the street to ask me what the heck I was doing. :D One of the guys uses WiFi and was intrigued with Kismet and NetStumbler. I cut the testing short when 3 tornado's dropped and started doing damage in El Reno just a few miles west of me and heading my way. Ahhh, Spring time in Oklahoma.;)

PS, I want everyone to have access to the raw data to make up your own minds, but I'm having trouble deciding the best way to present it.

PostPosted: Tue Apr 25, 2006 6:22 am
by Dutch
[quote="Mark57"]Test update: I'm still accumulating data and it's looking pretty interesting. I want to make sure I have enough data to be able to make a valid conclusion. Some of the people on the route I'm driving are getting VERY curious/suspicious. I had to stop and tell some construction guys working next to my start-stop point what I was doing after driving by 14 times in two hours yesterday. They laughed and said they were about to block the street to ask me what the heck I was doing. :D One of the guys uses WiFi and was intrigued with Kismet and NetStumbler. I cut the testing short when 3 tornado's dropped and started doing damage in El Reno just a few miles west of me and heading my way. Ahhh, Spring time in Oklahoma.]
Time to hit the Vinylcutting sign store and make the "CoWF Network Research Vehicle" side stickers...

Looking forward to your presentation and conclusions, Mark.

Dutch

PostPosted: Tue Apr 25, 2006 6:27 am
by streaker69
[quote="Mark57"]Test update: I'm still accumulating data and it's looking pretty interesting. I want to make sure I have enough data to be able to make a valid conclusion. Some of the people on the route I'm driving are getting VERY curious/suspicious. I had to stop and tell some construction guys working next to my start-stop point what I was doing after driving by 14 times in two hours yesterday. They laughed and said they were about to block the street to ask me what the heck I was doing. :D One of the guys uses WiFi and was intrigued with Kismet and NetStumbler. I cut the testing short when 3 tornado's dropped and started doing damage in El Reno just a few miles west of me and heading my way. Ahhh, Spring time in Oklahoma.]

What is wrong with you? You should be combining Netstumbling with Storm Chasing. I expect live video of upcoming tornados from you in the future.