HardWare

back to http://scratchpad.wikia.com/wiki/Sasecurity

TableOfContents

Antennas, pc, Wi-Fi cards
* EmbeddedPc * AnteNna * WrapBoard * DcDcConverters * http://www.starnetwifi.com, http://www.miro.co.za are Senao Wi-fi retailers. * AdvertorialCpe * PciCards * ZcoMax Two senoa cards on one Mini-Itx has high bus drain current, might lead to lock-ups. Or is this only if you place two pcmcia cards into one double pcmcia - pci adaptor board. Place copper strip between card to reduce interference

/var /tmp and flash
ON THE LOCUST operating system: /var and /tmp folders gets written to ram not the onboard flash disk. * FlasCardWriteCycle * FlashHdd * SenaoCardsDrawCurrent * RainLeakIntoConnector

DC-DC converters
DcDcConverters

Antennas
AntenNas

[attachment:panelAntenna.jpeg]

http://localhost:8080/HardWare?action=AttachFile&do=get&target=panelAntenna.jpg

Outdoor enclosures
* OutdoorBox

PCengines www.pcengines.ch WRAP boards
Place a 4500uF capacitor on the POE line to smooth out spikes.

Antennas, PCI cards and m-boards
http://www.highgainantennas.com/category_s/87.htm

Wi-fi interference detector
 2.4ghz video transmitters creates blackholes in the channels where Senao cards can't operate. Use directional antennas with video transmitters not omnis

Prism patch
If these are the newer prism3 chipset cards, then you may need to issue the following patch on dev 106: getandverify version23prismpcmcia

Sealing of Pigtails
Seal pigtails with self-amalgamating tape to prevent water from entering. If you do upgrade remeber to work from the outside of your mesh inwards and do your gateway/s last. Lastly, all your nodes in the same mesh are on the same ["ESSID"] and channel otherwise they won't be a mesh (there is also an option to mesh with any ["ESSID"] but this makes for a major security hole so best avoided unless absolutely essential.)

Mesh nodes do several simultaneous jobs: they provide their own internet backhaul,they connect clients wirelessly just like any AP does and they can also act as broadband firewalls for wired LANs. If you had several mesh network in the same area then your clients could be added to the same wiana realm and thus be able to connect to the different mesh clouds, otherwise the typical config is one mesh with the same ["ESSID"] and channel for all nodes and clients.

Weather effects
I'd like to add a little on something Eddie said. He does hit one nail on the head in saying the "toy radios were designed for room temp", but I think there's more... I think that it's important to remember that when we stretch this stuff out over distances - many factors are pushing the hardware & WiFi drivers/software in these "toy radios" well beyond their original design intentions.

Now, does that mean WiFi is not well suited for what all of us are doing? No! In fact, in spite of the original design intentions, it works pretty good... well, most of the time.

We had rain yesterday almost all day long. During the afternoon, a thunderstorm blew in and for no reason, one of my nodes began using a different pathway (marginal route) to reach the gateway. The odd thing is that the segment looks like this:

NODE A  ---   GATEWAY   ---   NODE B BR (The gateway is about halfway between node A to node B, and all 3 are in the same line with each other. Node A just quit meshing with the gateway and jumped clear past it to Node B to repeat back into the gateway.

However, looking at MaxiSpy or SigSpy from Node A, the signal of the GW node was down in the noise. In fact, almost everything was! The gateway could not hear Node A very well either. I was going to swap out radios after church today, as my thoughts were that static from the thunderstorm may have toasted the radio. However, I just got in from church and looked at it and the SNR values are all back where they should be!

It's important to say that there was a power failure in that part of town during the storm and when it came back on, this route changed. I'm thinking there is a lot more that affects this stuff than I once believed. These radios are just short-range products we've stretched with antennas. I think temp changes, static, and even local noise takes more of a toll than we think sometimes. I wonder if an electrical transformer may have began generating more noise after the power came back on? Whatever was the case, it has quit and the optimized route is now being used again.

That makes me wonder if there is something else in your environment causing this? Could static be the culprit? Or, was there a sudden change in temp or humidity right before your problems? Like I said, I think there is a lot in the environment that affects our RF.

The good part is that with mesh, we are well suited to deal with most of it.

Best regards, Kb

> Geez Eddie I did not take that into account. My assumption was the if > the temperature was between 0 -40 C that I would be ok however seeing > the temperature increase on the motherboard then would really concern > me. We have people near the equator having similar systems in place I > would be interested how they cope?

>> this is not something i know for a fact but its just something i >> believe.. i think the actual operating equipment temperature plays >> a big role.. all solid state devices generate more and more noise >> internally the hotter they get.. at the same time the analog >> amplifier circuits on the recieve side and on the transmit side can >> also change enough to actually make the devices move out of the >> analog section of their origional linear bias curves causing even >> more noise and just one fact in this hypothesis the effiency of all >> the amplifier circuits degrade as u move above room temperature.. >> gawd now ya got something to really think about hahahaha  have you >> tried an ice pack on your nodes??? seriously these little toy radios >> were designed for room temperature and most of us have them deployed >> at far greater temperatures..

>>> I normally would put this on my Blog but in this case I am having a >>> problem with the weather. In Toronto we have four distinct season >>> and we are in late spring right now. The leaves are on the trees >>> and we have normally breezy days. Lately we have still days and the >>> air pollution has increased. I keep loosing a number of nodes. >>> They have to physically rebooted. Then they loose connectivity >>> again and then I have to reboot. When the temperature is cooler the >>> connectivity improves. When it rains, believe or not, everything >>> works - just the opposite of what you think. >>> >>> I just wondering if anyone else experience the same thing?

PCI cards
has cheap senao cards and PCI 90degree riser cards for use in mini-itx One has 2 slots and I think is designed to fold over and away from the motherboard. the other is a single slot and folds over the board. I am not sure if there is a dual slot that folds over the board. We have isolated the onboard PCMCIA and the additional riser/slot and card by making a large copper shield that ensures that neither radio is able to see each other, in addition the cable that comes out of each radio is double shielded and has a second grounded shield that is also connected to the copper separation plate. the plate. the cool thing about this was it allowed us to put both radios on the same channel and face their antennas in opposite directions :) (the antennas are our own Borg waveguides).

Copper shield Farraday cage
Yes, we also had a problem with some of them not working. Make sure you get a riser that is only connecting the pci to the pci adaptor, we have had problems with the ones which have capacitor’s inline between the pci adaptor.

Does not running the two cards on the same channel cause the two cards to mesh with each other, even if you screen the cards, surely the antenna's even sectors! will cause some problems?

I am not sure if there is a dual slot that folds over the board. We have isolated the onboard PCMCIA and the additional riser/slot and card by making a large copper shield that ensures that neither radio is able to see each other, in addition the cable that comes out of each radio is double shielded and has a second grounded shield that is also connected to the copper separation plate. The cool thing about this was it allowed us to put both radios on the same channel and face their antennas in opposite directions :) (the antennas are our own Borg waveguides).

The copper seperation plate the idea with is to create a FarradayCage. I have managed to isolate the two sectors prety well, Sigspy shows them but they do not seem to mesh with each other. using sector antennas more than likley helps quite a bit and ensuring clean tuned lengths cables that feed the antennas so that the RF goes where its supposed to go and not reflecting back into the radio/box.

22
Hum interesting, will set-up our test node and have a play. Still I think that having another wifi card on the same channel, shouting down the ear of the other card, it should de-sense the first cell. Here's one for you all, if your client can seen both cells at about equal signal, from time to time the client will swap cells and ip's, can this be stopped. Thinking out load!! If Andrews systems works then what if you make both cells offer the same ip range, surely this will stop the problems of dropping and renewing ip's. This works with standard AP's, but we never have them located next door to each other, only overlapping, this enables you to seamlessly roam. So what I'm saying is if it works for standard Ap install why not Mesh. I have managed to isolate the two sectors prety well, SigSpy shows them but they do not seem to mesh with each other. using sector antennas more than likley helps quite a bit and ensuring clean tuned lengths cables that feed the antennas so that the RF goes where its supposed to go and not reflecting back into the radio/box. Does not running the two cards on the same channel cause the two cards to mesh with each other, even if you screen the cards, surely the antenna's even sectors! will cause some problems? This may be why i had to incrase the sensitivity to 2 in the core nore, each sector spits out a seperate IP range, this is a bit of a pain unless you use autoauth as my living room is just below the antenna and i seem to get a very slight crossover where i sit. With the laptop in the coffee table i have antenna a and if i move it to my knee i get antena b. Further out this is not a problem. Actually i have had 5 AP's inder test all in the same room on the same channel and meshing, i was able o conect a couple of clients to the mesh and did not notice a heck of a lot of diference. (i think it would have got real interesting if i had more clients on) Heres a picture of the array. http://www.borg.net.nz/gallery/album12 Only the top antenna and one side antenna is connected at this stage, i intend (when i get a cherry picker) to remove the top antena and connect the other side as that was the original intention however for testing the centre shaft was there. also these antennas are our own design, they are similar to the normal waveguides that you can find however there are a number of unique changes we have made and ths could be the reason they work in this configuration.

cell ip range
Hum interesting, will set-up our test node and have a play. Still I think that having another wifi card on the same channel, shouting down the ear of the other card, it should de-sense the first cell. Here's one for you all, if your client can seen both cells at about equal signal, from time to time the client will swap cells and ip's, can this be stopped. Thinking out load!! If Andrews systems works then what if you make both cells offer the same ip range, surely this will stop the problems of dropping and renewing ip's.   This works with standard AP's, but we never have them located next door to each other, only overlapping, this enables you to seamlessly roam. So what I'm saying is if it works for standard Ap install why not Mesh.

Does not running the two cards on the same channel cause the two cards to mesh with each other, even if you screen the cards, surely the antenna's even sectors! will cause some problems?

mac
Ok on using mac-auth, most of or clients run mac-auth. The reason I asked about running the same IP/cell id range on both cells is that we have a node in a hotel with a ticked service on it, all works fine as log as the ticket user stays in the same cell id, once the roam to a new cell then they have to log back in, but the the ticket has been used and removed from the realm, and the new cell see's them as a new user. Have you or anyone on the list tried using the same cell id on both cells? what problems do you get and will the node understand what's going on? We run sectors on one of our main sites, but still get trouble sometimes when the cpe can see two cells, it's not a major problem for fixed user's with mac-auth, but it still causes a short drop in contact time as the pc obtains a new ip address.

This may be why i had to incrase the sensitivity to 2 in the core nore, each sector spits out a seperate IP range, this is a bit of a pain unless you use autoauth as my living room is just below the antenna and i seem to get a very slight crossover where i sit. With the laptop in the coffee table i have antenna a and if i move it to my knee i get antena b. Further out this is not a problem. Actually i have had 5 AP's inder test all in the same room on the same channel and meshing, i was able o conect a couple of clients to the mesh and did not notice a heck of a lot of diference. (i think it would have got real interesting if i had more clients on) Only the top antenna and one side antenna is connected at this stage, i intend (when i get a cherry picker) to remove the top antena and connect the other side as that was the original intention however for testing the centre shaft was there. also these antennas are our own design, they are similar to the normal waveguides that you can find however there are a number of unique changes we have made and ths could be the reason they work in this configuration.

Hum interesting, will set-up our test node and have a play. Still I think that having another wifi card on the same channel, shouting down the ear of the other card, it should de-sense the first cell. Here's one for you all, if your client can seen both cells at about equal signal, from time to time the client will swap cells and ip's, can this be stopped. Thinking out load!! If Andrews systems works then what if you make both cells offer the same ip range, surely this will stop the problems of dropping and renewing ip's. This works with standard AP's, but we never have them located next door to each other, only overlapping, this enables you to seamlessly roam. So what I'm saying is if it works for standard Ap install why not Mesh.

I have managed to isolate the two sectors prety well, Sigspy shows them but they do not seem to mesh with each other. using sector antennas more than likley helps quite a bit and ensuring clean tuned lengths cables that feed the antennas so that the RF goes where its supposed to go and not reflecting back into the radio/box.

Does not running the two cards on the same channel cause the two cards to mesh with each other, even if you screen the cards, surely the antenna's even sectors! will cause some problems?

Yea, however there are 2 versions of riser cards. One has 2 slots and I think is designed to fold over and away from the motherboard. the other is a single slot and folds over the board. I am not sure if there is a dual slot that folds over the board. We have isolated the onboard PCMCIA and the additional riser/slot and card by making a large copper shield that ensures that neither radio is able to see each other, in addition the cable that comes out of each radio is double shielded and has a second grounded shield that is also connected to the copper separation plate. the plate. the cool thing about this was it allowed us to put both radios on the same channel and face their antennas in opposite directions :) (the antennas are our own Borg waveguides). Waveguides make excellent roosts for birds.. and no oneelse except myself will ever address this issue on a Wifi discussion but bird crap will sure degrade the performance of a waveguide not to mention having to mess with that in repairs ---no pun intended---  to solve this problem you buy 6 inch tall plastic owls for 3 dollars US funds and put them near the antennas... No Birds land on your antennas or ap's nor do they fly near them. Texas cowboys do some dumb things but they getter done Ok on using mac-auth, most of or clients run mac-auth. The reason I asked about running the same IP/cell id range on both cells is that we have a node in a hotel with a ticked service on it, all works fine as log as the ticket user stays in the same cell id, once the roam to a new cell then they have to log back in, but the the ticket has been used and removed from the realm, and the new cell see's them as a new user. Have you or anyone on the list tried using the same cell id on both cells? what problems do you get and will the node understand what's going on? We run sectors on one of our main sites, but still get trouble sometimes when the cpe can see two cells, it's not a major problem for fixed user's with mac-auth, but it still causes a short drop in contact time as the pc obtains a new ip address.

channels
Important! You will want to use non-overlapping channels, different SSID's, and alternate channels for each hop, as in this example where Qnode1 is a single-radio Qnode, and Qnodes 2 + 3 are dual-radio repeaters, with their radios set alternately to channels 3 and 8:

The reason for this, is that you don't want the two radios in the same Qnode, "talking" to each other, either because their antennas are to closely co-located, their SSID's are the same, or their channel frequencies overlap (less than 5 channels apart). If any one of these conditions exists, you will see your net bandwidth drop by 50 % or more due to mutual interference.

Does not running the two cards on the same channel cause the two cards to mesh with each other, even if you screen the cards, surely the antenna's even sectors! will cause some problems?

802.11b compatable mesh PCI cards
Batch of the discontinued Netgear MA311... but that source is now See http://www.zcomax.co.uk/products.php?product_type=29