Always one PC dropping out intermittently no matter what I try....

I have a customer with a small office on three floors, with about a dozen PCs and laptops, and a a couple of PCs and printers hard-wired.

The heart of the network is a Netgear FVS114 firewall providing DHCP. There's a hub feeding a Thomson ST180 wireless access point (ground floor), and a cable link to the 2nd floor feeding a Netgear WG602v3 access point. The PCs have AT-WCP200G PCI wireless cards.

Whatever I've tried over the last few months (!) there always seems to be one PC or other which can't stay on the network. Some never seem to have a problem, but the problem seems to shift around the ones which do.

The firewall was replaced recently (identical unit) which improved reliability of the network overall. I've also found that one of the links from the ground to the 2nd floor (note there's a 1st floor in between - I'm in the UK!) has an intermittent open-circuit at one socket; the other seems ok.

Most PCs, downstairs and upstairs, can bind to the Thomson ST180 and keep a stable connection, except for one downstairs (which usually can't see any wireless networks) and one upstairs. I've tried reloading drivers, extension antenna, switching round the WAPs, bringing both downstairs - no improvement.

I thought I was onto something when I bought WirelessMon, which showed about 8 wireless networks, all only intermittently available except for the Thomson ST180 (about -45dB). The Netgear WG602v3 is much stronger upstairs (that's where it lives) at about -15dB, but WirelessMon shows it coming and going several times a minute, seen through the wireless card of one machine which manages a steady connection to the ST180 two floors below. Another nearby machine can't get a steady hold on the downstairs WAP, and can't bind to the upstairs one.

I replaced the Netgear WG602v3 with a US Robotics USR5416. No improvement: the signal in WirelessMon was still shown as coming and going while the downstairs ST180 was steady as a rock.

Changing channel on the downstairs ST180 allowed one PC to get a connection but another one lost it, so I changed back, restoring the previous situation.

I disabled the PCI card on the "unlucky" PC upstairs and fitted a US Robotics USR5420 adapter, which was able to bind to the weak signal from downstairs, but not the Netgear WG602v3 upstairs. Substituting the USR5416 again (configured so that it's possible just to switch the boxes) produced a loss of ability to bind to the ground floor in the PC with the USR5420 USB adapter, and it couldn't bind to the USR5416.

Today I've tried fiddling with the various settings on the (reinstated) Netgear WG602v3. While the downstairs ST180 remains on WEP (128), I switched the WG602v3 to WPA-PSK. No improvement. I've also tried reducing the data rate from "Best" to 24Mb/s, and changed from "g or b" to "g only". No improvement.

Meanwhile, a network in their other office with a single Thomson ST180 and 8 of the same PCI cards soldiers on without a blink. If I could get another ST180 I would!

So, the only thing I can think of is to start replacing the wireless PCI cards, inclining towards US Robotics. However, if you've been generous enough to read this far and can suggest ANYTHING else worth trying, I'll be greatly indebted to you.

Phil, London

PS: I've stumbled on NetStumbler, which is useful. Here's a screnshot:

formatting link

26 networks detected after an hour's running (up to 40 36 hours later!). I've sorted by max SNR (Signal to Noise Ratio) although I'm not certain the "noise" part of that value is being correctly reported. Coincides with max Signal, though.

Just one signal stays steady as a rock, and it isn't the strongest!

Reply to
Philip Herlihy
Loading thread data ...

Philip Herlihy hath wroth:


will not show 2.4GHz devices that cannot be received by your laptop or whatever you're using to run WirelessMon. For example, if somone setup an access point nearby running some MIMO mutation, with all the compatibility options turned off, you won't see it. Same with

108Mbits/sec only access points. So see those, you need spectrum analyzer.

Ok, not stop and think. What would cause the signal to come and go like that? Is there an other access point around on the same SSID? Any possible sources of intereference? The last time I saw something like that, the office had a mess of 2.4GHz wireless cameras (non-WiFi) that everyone conveniniently was ignoring. Start turning things off until you get a steady signal. If you can't do that, take the WG602v3 into a basement or isolated location, and see if you can get a steady signal there.

Ok, you have 3 floors, none of which will pass RF through the floors. My guess for an office building are poured concrete floors, which have that characteristic. The only way this is going to work is one access point per floor. Forget about going through the floors.

Ok, that eliminates the access point as the culprit. What else is belching 2.4Ghz in the area? Note that this includes 2.4GHz desktops and such. See list at:

for probable culprits.

Hmmm... what floor numbers are you on? If you're very high above the ground, with a good view of the city, you're going to pickup plenty of interference. It may not be particularly strong, but it's going to be from many different sources. Your screenprint of the WirelessMon output shows a huge number of other systems, any one of which can cause interference. I suggest that you:

  1. Install a low gain, but directional antenna that keeps the RF pattern inside the building. Something about 8dBi gain with at least a 90 degree beamwidth (AMOS or Franklin Antenna).
  2. Move the access points away from the windows with a view. If difficult, put a piece of sheet metal or aluminium foil behind the antenna in the direction of the outside windows.
  3. Select a channel number (1, 6, or 11) with the fewest *NUMBER* of other systems.

No clue what the signal strength and noise level are doing during these changes. Are you recording numbers as you go along? If not, how will you know you're making an improvement? Juggling adapters (assuming they're not defective) will not have anywhere as big an effect as better antennas. With desktop PCI, the problem is that the supplied antennas are too small, buried behind the PC, on the floor, and in the middle of a mess of cables. In other words, the worst possible location. If you insist on PCI, then get an external antenna. If you insist on shooting through concrete floors, get a

*BIG* external antenna with lots of directional gain.

Have you tried checking to see at what speed your various clients are connecting to the access points? This is usually a good indication of signal "quality". The lower the speed, the more flakey the connection. Do this after passing some traffic as many clients increase their speed back to 54Mbits/sec when the traffic stops. If you're down to 11Mbits/sec or less, you don't have have a good path, enough signal, or are infested with an interference (noise) problem.

Continues on at what speed? If your DSL connection is running at perhaps 1.5Mbits/sec, you could be connected at 5.5Mbit/sec and never really notice the slow connections. Numbers, not prose, please.

I used to have a slogan for such ocassions, where no obvious solution was avaialable. "Change Everything". I even was given a rubber stamp with that inscribed. I think you'll find that doing a "site survey" with WirelessMon or Netstumbler will show that you have crappy signal between floors and lots of interference near windows if you're high up.

I also take issue to using PCI wireless cards in an office environment. The money you save in not having to do wiring is lost in such troubleshooting exercises. You also won't get anywhere near the speed and reliability of a wired network. PC's that do not move should use wired ethernet connections. They should also not try to shoot through floors. The only place I've seen wireless desktops used effectively was on crash carts in a hospital and a laboratory. The others ended up wired. Leave the wireless for the laptops and PDA's which can pick their location for best signal.

Numbers.... give me numbers and I can deduce the problem. Signal strength, variations in signal strength, noise level, variations in noise level, connection speed, variation in connection speed, and results of performance tests.

For a performance test, use IPerf with a wired desktop acting as a temporary server. For destructions, see my rant at:

Notice that I really don't care what speed you get from the internet. It's the local wireless speeds that are important. Therefore, don't bother with an internet based speed test.

Propagation and moving reflections will produce additional systems. Those that do not broadcast their SSID will not be shown. Just eliminate those that have very few beacons heard.

It's also not very nice running continuous probe requests for several days. My IDS system treats that as an active attack in progress and will block any ICMP packets to your MAC address. Run it for perhaps

30 minutes and you'll see all of the strong signal systems.

I also see another possible problem. I don't think that "/" and "#" are valid characters for the SSID. It's suppose to be only alphanumeric characters. Many AP's will accept other characters, but I've seen some clients choke on these.

Actually, the noise values are very important. They usually sit at a fairly low value. However, if they increase above some base value, that means that there was interference present when the probe packets were received.

For the time being, I suggest you ignore interference and concentrate on signal quality and strengths to your own access points. My guess(tm) is that you don't have enough signal through the floors and that even the slightest interference will cause packet loss or disconnects. The IPerf thruput tests should demonstrate that clearly.

Reply to
Jeff Liebermann

Jeff Liebermann hath wroth:

One more idea. Buy or borrow a "hotspot finder". I use an Iogear GWF001 which has proven to be sufficiently sensitive:

It's good for 2 things for your situation:

  1. Turn OFF both of your access points. See any indications on the device? If so, there's your interference. I've found that it detects access points that do NOT broadcast their SSID, which is very handy.
  2. Turn on one of your access points. You should see an indication on the device. Now, take a walk around to the various workstations and see what signal level you get. Repeat with the other access point.

It's crude, doesn't really yield any numbers, but it very effective at catching suprises. I recently had a obvious interference problems that I just couldn't identify. Full scale (5 lights) indication on the pocket "hotspot finder" but Netstumbler and various active scanners didn't find a thing. Kismet might have found something but I misplaced my Linux LiveCD. I eventually found a rather overpowered wireless game server that was somehow setup to be rather well hidden. No clue how it was done as the apparent owner, when confronted, denied all knowledge. However, the system magically disappeared shortly after I left.

Reply to
Jeff Liebermann

Thanks, Jeff, for a very comprehensive reply - it'll take me a while to digest it and follow up the links..

A note on the building: structurally it's a "corner shop" in a predominantly residential suburb, about a mile-and-a-half North of the tip of the 2012 Olympics site (not yet likely to be emitting much except mud). The building is probably late victorian, and the floors are wood.

I'll certainly look into your suggestions. I've also wondered if there is likely to be any merit in adjusting the "Advanced Wireless Settings" mentioned in this (unrelated) link:

formatting link
've also dug out the documentation for the AT wireless cards, and I remember that I hadn't installed the supplied client software, relying instead on the built-in WZC. That's another avenue to explore.

I do take your point about the cost of troubleshooting exceeding the cost of cabling!


Reply to
Philip Herlihy

Just ordered one of these:

formatting link
I've found and downloaded iPerf (neat!). I'll let you know in a few days how I get on...


Reply to
Philip Herlihy

We have a few Victorian (actually Edwardian) dinosaurs masquerading as historical landmarks and student slums. I lived in one like that, complete with exterior plumbing and ungrounded electrical wiring. I had a rope ladder tied to a bed post ready to toss out the window in case of fire. Sure looks nice on the outside, but I don't recommend living in there.

True, the floors were made of wood and compressed termite sawdust. RF should go through that quite nicely. However, one remodeled area included aluminum foil backed fiberglass roll insulation for soundproofing. Similarly, the floor may be wood, but the ceiling underneath turned out to be lath and plaster. You might want to check what is actually in the floor.

If you're really getting interference, reducing the packet size will help statistically, but at the expense of peak thruput. Smaller packets have a better chance of getting through interference than larger packets. However, smaller packets also have a larger percentage of overhead, which will slow things down.

RTS threshold might be useful, but only if you enable CTS/RTS flow control. This also slows things down but prevents packet collisions. This is only really effective if you have a "hidden transmitter" problem, or have a nearby interfering system.

I would turn off 108Mbit/sec mode, afterburner, turbo-G, or any other acronyms that claim to boost speed at the expense of range and reliability.

Ugh. I guess you're desperate. I never read the docs until after I fix the problem.

It's for real. The secret to engineering is knowing when to give up. I won't claim to know the secret as I've spent inordinate amounts of time and customer money chasing problems that could best have been bypassed.

Reply to
Jeff Liebermann

I haven't played with one of those. I would be interested in knowing a few things about it.

  1. How's the sensitivity? Just a guess on how far it can pickup an access point with the stock rubber ducky antennas.
  2. Will it show an access point that is NOT broadcasting its SSID. I need this for interference detection.
  3. Will it pickup anything else (cordless phone, wireless tv camera, etc).

A few days isn't going to work. I'm going to be out of service for a few weeks starting Aug 2. Good luck. IPerf is a very hand tool. Try some of the assorted options.

Reply to
Jeff Liebermann

My reading of the product description is that it will pick up "general" traffic in this band in "roaming" mode. Obviously, like any instrument, it'll need to be "calibrated" by trying it out in various places. I'll post here when I have more to tell you - I'll be away myself for a while from Aug 7.

I'm very grateful for your generous help - thanks!


Reply to
Philip Herlihy

Glad to see you're back online. I've played with the device I mentioned (above, or

formatting link
.. and I'm very pleased with it. It shows how the signal drops at specific points in my house: my workshop is just the wrong side of the hot-water tank and all that plumbing, but I can detect a "corridor" of stronger signal which, um, corresponds with the actual corridor. I can still get 2 LEDs through (or around) the hot water tank and through three walls and a floor. Note the mention of "roaming" mode - in that state (button held down) it monitors all traffic in the band, filtering down to WiFi traffic only when you release the button. Neat. (My access point is a USR 9106 with twin stub antennae.)

Haven't had a chance to try it out on-site yet - other emergencies have had to take priority, and they aren't having so much trouble lately.


Reply to
Philip Herlihy Forums website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.