How to debug a reluctant WISP AP association (before I blame the WISP)

I forgot to mention the WISP & laptop equipment involved so I'll list it here:

- WISP AP = Compex Systems WLM54AGP23, 23dBm TX, 802.11b -92 dBm@1Mbps RX

- My radio = 19dBi planar & 28dBm (630mw) radio 1.5 miles away LOS

- My laptop = Lenovo X61 Ubuntu, internal -6dBm antenna & 15dBm (30mW) radio with sensitivity of -90dBm@6Mbps

Given this equipment, what tests can I run to figure out WHY a minor change in the radio's AirOS settings causes a disassociation from the access point - which lasts for hours - and finally corrects itself over time?

How do you know there is no interference?

Incidentally, with radio, you often hack a bit with the antenna and then make the theory match the results. For instance, tilt the antenna back a bit, or twist it a little off vertical. You can't relocate the antenna unfortunately, but sometimes moving it a bit makes a difference.

I'm not sure why the Dell matters. You really should only run the wifi link to the WISP and get that as good as possible before you add a local WAP.

I have no idea if there is interference. I don't even know HOW to test for interference. All I do know is: a) The signal strength is around -66dBm with noise around 88dBm b) The transmit CCQs are in the 100% to about 95% range (usually) c) The connection is good while it's connected d) But the split second I change anything, it loses its association for hours on end.

I did align by signal strength by twisting it and changing the angle to vertical.

Are you saying I should align by some other metric than the best signal strength?

The laptop only matters if someone suggests a Linux tool to debug why the radio won't associate with the WISP AP for hours upon changing any setting in the radio itself.

If you have linux, you could run Kismet to site scan, but not all radios are "promiscuous." Even then, it is possible someone is running wireless video, which Kismet can't detect. But Kismet is a start.

I'm saying move the antenna in whatever way works. If the WISP is at a higher altitude, you probably want to angle the antenna back a little. Or if it is lower, you aim down. Generally quality of signal is what you want rather than signal strength. You can try tilting the antenna a bit to see if being slightly off of vertical improves things. My point is it is OK to hack a little.

Some high gain antennas have significant lobes. It isn't a simple matter of more is more. For the point to point systems I set up, I use the L-com PA24-16, specifically because it is a combination of high gain and has very modest lobes.

If you look at the display, the lobes of the 19dbI antenna are down about 12 dB. The scale makes it look crappier than reality. But I preferred to use the 16dbi version, while has lobes that are further down and much more to the side.

You certainly have enough signal strength, but perhaps the quality of the signal isn't very good.

Doesn't the WISP provide the antenna as part of the package? This might be a case to make it someone else's problem, especially if it is on their dime.

You can view the L-com catalog and see how gain and lobes are not exactly correlated. For instance, those dish type antennas have very small lobes, but the cost is high.

Evey time I mention L-com that troll comes out of the woodwork to diss them. The same with Alfa. Perhaps a pre-emptive diss will keep the poster out. ;-)

Do you know if you have line of sight? Are there tree in the way? I've done 7 mile links, so 7500ft isn't much of a challenge. Actually, if your antenna has lobes, you might not be aimed correctly. That is, you could be receiving off a lobe. If you are aiming with a compass, is it a really compass or a freaking iphone/GPS. I'm totally convinced the Cammenga compass is the only way to aim an antenna if you can't see the transmitter. I've set up satellites with a cammenga, and sat dishes are very directional.

I'm googling for a better external USB radio for the Linux laptop.

Apparently you need a few things in the radio for good packet sniffing:

1. The radio should come with Linux drivers
2. The radio must support "monitor mode"
3. The radio must be promiscuous (that may be the same as monitor mode)
4. The radio must support "packet injection"

So far, by googling, I've found only one radio that does this: Alfa Network AWUS036H ($28), TX=20dBm (100mW), RX=-91dBm@11Mbps802.11b Antenna=2dBi omni, Power = USB, Linux drivers = yes, Monitor mode = yes, Packet injection = yes, Promiscuous=? But if I'm going to buy anything, I prefer a stronger gain such as this 1 Watt USB radio (but I don't know if it is promiscuous): Alfa Network Tube-U(N), TX = 30dBm (1 Watt), RX = -95dBm@ 1Mbps802.11b Antenna Connector=N-Type Male (so you can add any antenna you like Power = USB, Linux drivers = yes, Monitor mode = ?, Packet injection = ?, Promiscuous=?

Interesting. I was going for pure decibels. Is transmit CCQ the quality metric I try to maximize? If not, I don't know what other 'quality' metric to aim for that I can measure at the radio.

I'm currently using the 19dBi Laird Technologies PA24-19 from Streakwave in San Jose, installed on the antenna mast along with the 28dBm (630mW) Bullet M2 radio so I'm familiar with that suggested brand (and I like them).

In addition, if I buy the 1 Watt Alfa Network bullet USB radio, I'll need an additional external antenna. The antenna consumes 100 Watts though, so I'm not sure if that can be handled by the USB port (because 500mA USB x

Of the three Laird antennas on that datasheet, I agree, the lobes of the

One potential issue with the 16 dBi antenna is that, of the three, it's the only one that doesn't support "surface mounting". What is "surface mounting" anyway? They all support 1" to 2" pole mounting, which I'm familiar with. But what is surface mount anyway?

Long story. I prefer to use my own equipment. The WISP charges $400 to install his equipment. I talked him out of that (he was reluctant to agree but I told him I prefer better equipment and better installation). In the end, that's what I got: a) I put the antenna where I wanted it - not where he was gonna put it b) I chose a 630mW (28dBm) radio - he was gonna use one inside the antenna c) I chose a 19 dBi antenna - his was gonna be an all-in-one d) I didn't drill any holes to get into the house - he would have e) I buried the cat5e cable - he would have strung it in the air f) I put the wall plate in the middle of the house - his would have been on an outside wall etc.

In the end, it cost me as much (or more) than it would have for him to do the work - but I have a better installation - and I can control the radio and the settings.

I've never visually 'seen' the WISP antenna but the WISP came to my house and said there was line of sight and he gave me the coordinates to his antenna (which I posted in the picture in the first post).

I'm aiming by signal strength, as read by the radio screwed onto the back of the antenna.

That kind of stuff is EXACTLY why I want to find good sniffer software that works on the Linux laptop (along with a good external antenna & USB radio for that laptop)!

I never use a compass. Too many errors when working with high gain, narrow beamwidth antennas, which need better accuracy.

Instead, I use a road map, some pins, and a piece of plywood. Locate yourself on the map. Insert a pin. Locate a landmark (e.g. mountain top) that you can see. Insert a 2nd pin. Sight between the two pins and rotate the map. Now, the map and the planet are identically oriented. Find your target on the map, insert a 3rd pin, sigh between your location and the target, and point your antenna the same way. I can do about +/- 1 degree this way (if I'm careful or use a theodolite). You can't even get close to that with a compass.

Also, don't aim by signal strength unless you have perfect line of sight. If you happen to pick a reflection, rather than the incident signal, it's not going to be stable. When the wind blows, the signal will change. Best to aim the antenna geographically, tolerate a somewhat lower signal level, and not have it change later.

Not sure that's entirely necessary - I think that's more useful for when you're trying to crack someone's encryption by feeding it the "right" sort of packets to make the AP do the wrong thing!

Monitor mode = promiscuous.

Bear in mind your Bullet is a little linux box so some of the tools you would run on your linux laptop could be persuaded to run on the Bullet. For example you might be able to get Kismet to run in drone mode on the Bullet. I use a little tool called 'wavemon' for simple signal strength [once associated] testing with wireless.

I've seen the same quote attributed to Thomas Jefferson, which is even better than faking it with MLK. Do you know who actually said it?

Karl Rove

Not exactly. See:

Very cool. Works nicely on Ubuntu 11.10. Thanks.

The Cammega is quite a instrument if used carefully. The military spec is 40 mils, but the manufacturer states 2 mils. At that point, the uncertainly in the magnetic compensation is a larger error.

You do need to watch out for stray iron. There is a photo of someone using a lensatic with a freakin' locomotive behind him.

If you do decide to purchase a Cammenga, the phosphorus one is fine. It is not like you are in special ops. If you want to use it at night, just "charge" it with your flashlight.

Obviously for satellite work, you aren't getting reflections. ;-)

The AWUS036H works fine with kismet. They have a 1 watt version.

I have the tube/u. It is less power than the AWUS036H. It works fine in Kismet.

Incidentally, Kismet is totally passive. I'd have to research packet injection, but that sounds like something active.

One you have comms, the firmware will yield a quality signal. I will admit I don't know how they get this figure of merrit, but it seems to be what you want to optimize.

I can see that antenna being a bit tricky to set up due to the lobes.

My point was to hook up the wifi dongle and just monitor. That would be from your existing antenna. You might find that you are fighting with some other source.

I got the impression the whole family of antennas is similar except for dimensions. Mine is pole mounted.

And you are a better person for this! There is a lot to be said for knowing exactly what you have installed and how it was set up. I can't tell you how much money I wasted on a gate control company that was reaming me for repairs on what I finally determined was a PCB with an intermittent bug. [You could push it and signals would change.] I bypassed the service company and arranged to get the board swapped. Problem solved, and the gate guy doesn't get to make a boat payment.

I don't think you want to sniff your local environment. I think sniffing with the antenna is the way to go. I'm going to go back and try to find the coordinates you mentions, but it has been my experience that mountain tops are full of wifi these days. It seems some companies use wifi as a way to remotely read their instrumentation. Maybe all that wifi causes your receiver to think it has a strong signal.

Normally I just do a point to point where there is just one source at each location. Now probably a question for Jeff would be is there a way to see if one direction is working well, but not the other. Perhaps the WISP has interference from your direction, so link that is your transmitter to their receiver is poor, but you can "hear" the WISP just fine.

Yep. Maybe they should call it resolution instead of accuracy? I suppose someone could resolve 2 mils (2 milliradians) or 0.11 degrees, which is about the width of the reticule wire. However, that's useless if the north reference (the compass needle) isn't any better than maybe +/-2 degrees (or worse). With the map table method, there's no magnetic variation, no guessing the center of swing on the needle, and no transfer error. The military standard of 40 milliradians (+/-2.30 degrees) seems more reasonable, but still seems too small.

Amazing. I guess we can do away with all the licensed compass adjusters in the marine and aviation business with such a simple invention.

Unless you're in a canyon, where reflection from the canyon walls cause problems. Same with urban canyons with reflections from buildings.

A quick Google didn't turn up any support for that, nor did I find the origins of the Martin Luther King (or Thomas Jefferson) misattribution. I guess it goes unanswered for now, not to mention being OT, for which I apologize.

Thanks Jeff. That web site makes it clear that both modes capture packets ... but that one mode requires association while the other mode doesn't.

Q: What is the difference between monitor and promiscuous mode?

Monitor mode enables a wireless nic to capture packets without associating with an access point or ad-hoc network.

Promiscuous mode allows you to view all wireless packets on a network to which you have associated.

Wavemon is in the Ubuntu Software Center, which describes it as:

Wavemon allows you to watch signal and noise levels, packet statistics, device configuration and network parameters of your wireless network hardware. It has currently only been tested with the Lucent Orinoco series of cards, although it should work (though with varying features) with all devices supported by the wireless kernel extensions by Jean Tourrilhes.

I will install this wavemon and test it out!

Thanks!

Do what you think you need to do. I'm just saying it is a cinch to set up antennas with a good compass if you know what you are doing.

I'm not sure why you are trying to be snide here. With flux gates, there are ways to compensate for local fields. You can't do that with a handheld compass.

I haven't run into such problems.

There is an iphone app for setting up satellites. Oh, and it has glowing reviews.

Of course, the basis for this program is the crappy iphone compass.

You might want to try it with something other than gnome-terminal as I find that tends to intercept the function keys when switching between screens. Not sure if the GNOME terminal is the default on the Ubuntu desktop at present. xterm will probably be installed, if not I prefer Konsole from KDE as you can split tabs horizontally and vertically and see [say] ping output at the same time as wavemon's histogram.

Sigh. No insult was intended. It's just that if the compass offered the accuracy indicated in the data sheet (0.11 degrees), there would be no need for compass adjusters. The problem is not the compass, but rather that they used and mixed the terms "accuracy" and "resolution". The 0.11 degrees is really the resolution, which means that a fairly normal human can possibly resolve two objects separated by 0.11 degrees. I just tried it on my cheapo hiking compass. The best I can do is about 1.5 degrees, mostly due to the lack of a reticule wire. However, that's resolution, not accuracy. My ability to locate true north would be a measure of accuracy. Comparing the compass with my Droid compass results in about a 5 degree discrepancy (after compensating for magnetic variation). I can't really call that the accuracy, but it's in the ballpark.

Patience, you will run into such problems:

I had an ancient Magellan GPS that would nuts when faced with reflections. My dragon droppings were all over the screen while driving through the local canyons. Also, we have a local Geocaching stash, that's located at the bottom of a highly reflective canyon. Those with WAAS over-ride features can usually find it after some frustration. Those without are usually faced with multiple contradictory positions.

I have both an old Droid A855 phone, and an iPhone 3G. The compass and GPS on the Droid are MUCH better than the iPhone. One Droid app provides a combination on screen compass and bubble level, which really helps with the accuracy.

The satellite pointing apps are kinda nice for determining if there's a possibility of getting DBS reception. I have that problem in the local forest. However, since I'm not using the app for dish alignment, lousy accuracy is just fine. My current setup is a soda straw, taped to an inclinometer and a compass. It works well enough. My only problem is that many of the dish aiming web sites don't bother to distinguish between magnetic north, and true north.