More questions on tuning my wireless network...

I compiled some other questions I can't get any answer to. (either I am not asking them right, or no one really knows)

Reply to
lrtherond
Loading thread data ...

snipped-for-privacy@gmail.com hath wroth:

Or, those that know are too busy to answer questions. I have time today, so draw your own conclusions.

Yep. It's a feature. As the error rate increases, the access point will reduce the speed in order to improve the receiver sensitivity, which automagically improves the error rate.

No. Your WRT54G(???) is not unstable. Having it change speed is perfectly normal. You'll get a 54Mbit/sec connection up to a range of about 10ft using the stock antenna. Any farther, and it will slow down. I can did some graphs and curves out of SmallNetBuilder.com if you want details.

Fixing the wireless speed at 24Mbits/sec will result in a different behavior. You'll have a usable 24Mbit/sec link out to about 25ft at which point it will just quit working due to a sky high BER (bit error rate). However, there are places where you might want a fixed data rate. I do it often, but at much slower speeds, such as 6 or

12Mbits/sec.

Sure it can if you mean variations in latency. For example, if you lose a packet due to a noise hit or collision, your latency will be double (or more) of what it appears normally. That's the way I detect collisions, errors, noise, interference, and sometimes reflections. Anything that interferes with packet delivery will cause a change in latency.

Actually, they're a well guarded secret in the tradition of "You don't need to know". Besides, too much information tends to confuse the consumer.

Tweak which parameters? What are you optimizing for? Range? Thruput? Packet loss? Interference? I'll see what I can find. There are documents covering various timing parameters and settings, but no single document that covers everything.

Notice that my list did not include signal strength. That's handled by antenna design and is generally unaffected by anything you can set in your router (except tx power). Latency is affected by literally everything. If you want to minimize latency, you basically have to have a good strong signal, with execellent signal to noise ratio, and few reflections, or the retransmissions will increase latency spectacularly. What numbers for latency were you expecting? For what application? Video, games, etc? Also realize that latency is most commonly affected by your broadband connection. If you're measuring your latency to include what's added by the internet, you fighting the wrong battle.

Are the number of walls variable or does "some walls" mean that you can't count? How far apart? How many walls? What are they made from? How transparent do they appear to your existing hardware (signal levels)? You're trying to solve a specific problem, with a specific location, using a specific number of walls. Being intentionally vague doesn't help.

I have no idea what you have to work with and therefore cannot recommend anything specific in the way of antennas. Since this is an indoor affair, a large yagi or dish are out due to size considerations. Same with a monster panel antenna. Basically, you're limited to small panel, patch, and biquad antennas.

If wireless can't be made to work well, there are wired altenatives that use the phone lines, power lines, CATV coax, CAT5, and fiber that might work.

Sigh. Neither are particularly applicable or useful for an indoor system. Your existing WRT54GL diversity system is currently spacial diversity. However, when converting it to a point to point link, I usually switch to no diversity receive. If you are also using either WRT54GL as an access point (for wireless client connections), you will have to deal with a compromise antenna system. Diversity reception does not work with two different antennas. Be sure to read the Golf Course article in the Cisco URL below:

Give up now, while you're still sane. You won't find an empty channel and you will get lots of interference. You're best bet is to move to

5.8GHz (802.11a) or minimize your dependence on wireless by using wired networking.

Incidentally, WDS and other forms of store and forward repeaters do not work well in the presense of interference. No wonder you're having a latency problem. Every time some other station transmits, your WDS system loses a packet. If you're going through the remote WDS end as an access point, you lose two packets.

My various DD-WRT v23 SP2 boxes seems to report signal strengths on the low side. It's like that on all my boxes.

Now, where does it show signal strength in percent?

Yeah, that's the result of all that interference which creates errors, which causes the AP to slow down in order to compensate.

A few paragraphs ago, it was 35. You must live in one of those fast growing areas.

Speed of DSL? Make and model of workstation? Make and model of server? Operating system and versions? You may know all these but anyone offering to help will need to guess.

Make and model of game console? What type of wireless client? Same as before for the apartment? Range, construction, signal strength?

Make and model of laptop? Type of wireless card? Operating system version? Which wireless card in the Playstation?

You may have signal strength but I'll bet your thruput sucks. Download and install IPerf:

Install it one a desktop that is connected to your main wireless router via CAT5. Run the server as: iperf -s On various clients, run: iperf -c ip_address_of_the_server and get TCP speed statistics. Note that this test does NOT use any internet benchmarking sites or apps. I suspect you're going to see some rather miserable thruput.

If you plug into your remote WRT54GL (the one that does NOT have the DSL modem connected), you'll get full speed, as if you were using it as an ethernet client radio. However, if you connect to it via wireless, you're using it as a store and forward repeater. Your maximum thruput will be cut in half because in store and forward, only one transmitter can be on at a time.

The surest sign of success is pollution. Wireless is successful.

You might want to look at this list of interference sources that Netstrumbler and such cannot see:

Reply to
Jeff Liebermann

Jeff, you do well to correct all my sloppy questions.

I need to take as much time to answer your various questions as you took to provide your initial answers.

In other words, it's going to take some time...

Reply to
lrtherond

Google apparently misplaced that answer I spend 1 hour typing!

Oh boy...

Reply to
lrtherond

After the blow Google gave me yesterday, I will summarize the situation...

I don't do any file transfers over my wireless network. So, my throughput requirements are limited to the capacity of my DSL. (3000-6000 Kbps / 512-768 Kbps)

Latency is a bigger concern to me, since I use Skype-like applications and online gaming. I certainly do not want my wireless laptops (connected to the WDS client AP) to increase overall latency.

I ran Iperf and I confirmed I was getting 2.3 Mbps of effective throughput. My eyes bled. During these tests, the same pattern could be observed each time: Transmission speed would rapidly degrade from 54 Mbps to 11 Mbps.

So, I converted the WDS client AP to Wireless Bridge mode. Probably, it's not a transparent bridge, but it's no issue in my case.

Since this change, my latency has improved significantly when all clients are active. (6.2 Mbps, which is > the capacity of my DSL)

The drawback is that my wireless laptops and PSP must now struggle to get a good link from the remaining AP. It works well enough for Web access.

I will now go on and see if I can improve the strength of the signal between the AP and the Wireless Bridge.

Thanks for all your good advice.

Reply to
lrtherond

snipped-for-privacy@gmail.com hath wroth:

Some of my postings in other groups are not appearing on Google Groups. Methinks something might be wrong with Google, but I don't want to burn the time finding out.

My DSL is 1500/256Kbits/sec. I'm jealous.

Have you tried any of the online VoIP testers? Unlike the more conventional bandwidth testers, they show jitter. You can also download the Java applet and run it locally to identify jitter, which is one manifestation of packet loss. (free) (expensive) Try it:

Lots of others (but I can't find the free local tester I was thinking of). Note the IPerf can also report jitter.

They will increase latency. The question is by how much. A wireless direct link should add 2-3 msec additional delays. However, if you have packet loss from interference, it will climb quickly. WDS, when used as a store and forward repeater will make any delays 2 or more times worse (as you just discovered).

Yep. That sucks. You should be doing much better. However, the interference is probably the culprit. What you might find is that the performance won't be much better if you move in close to the access points. That's the because the interference will appear in the 802.11 delay spread (between packets) causing just as much interference as with a weak signal. However, that's still no guarantee of success. One of the fun demonstrations that I've done is to place a repeater or WDS bridge in a closed room along with the main WDS router and a few wireless laptop clients. With all these radios in a small area, there is quite a bit of mutual inteference. Using IPerf, thruput sucks and interference effects are obvious. So are reflections and multipath. Basically repeater and WDS do not work if the end points can hear each other. If they can, they simply add to the interference. It's possible that your WDS system is creating its own interference. Try it in an RF isolated environment such as a basement.

That's normal. When moving no traffic, the SNR is good, so the speed creeps back up to 54Mbits/sec. As soon as traffic starts, the errors start to appear along with the traffic. The AP tries to compensate by slowing down. The delay in changing speed varies with manufactory, but my guess is about 30 seconds of no traffic.

Actually, DD-WRT v23 sp2 is a transparent bridge. There are some articles on the subject in the DD-WRT wiki and blog.

What do you mean "all clients are active"? Are you moving traffic simultaneously with other wireless clients? That will just split the bandwidth in half. Try to setup an ideal test, with one AP and one client, turning everything off. Then add clients to see what breaks.

It's not going to work for gaming or VoIP. Web access is not time critical. Gaming and VoIP are.

That will help, but I think doing something to prevent external and self-generated intereference would have a bigger effect.

Ye're welcome.

Reply to
Jeff Liebermann

I'll check this out.

WDS is gone. I'll avoid keeping the wireless laptops on while gaming or using Skype.

Yes, I must say that in my conditions, this technology barely works.

Thanks, I just upraded to that version.

I meant "powered up". If I have them produce traffic, things get worse, as expected.

I don't believe there is any solution to that when you live across from Cisco's campus and 80% of your neighbors are Cisco engineers running wireless networks.

Reply to
lrtherond

snipped-for-privacy@gmail.com hath wroth:

That's the down side of license-free technology.

I've been having problems with v23 sp2 only on my WRT54GS installs. The remote admin access page seems to hang after about a days operation. It doesn't happen with WRT54G or various Buffalo products. Weird.

It varies. The small amount of traffic generated by an idle wireless computer should not affect your connection very much. It's only when the idle laptop starts to move traffic, that problems begin. However, that assumes that most of automagic update, polling features, and "ET call home" features of Windoze XP are disarmed or blocked. Even an idle Windoze box generates quite a bit of traffic allegedly doing nothing.

Sure, there are solutions.

  1. Switch to 802.11a at 5.8GHz. I'm sure Cisco has some systems working on 802.11a, but they will have less range, and don't penetrate as badly. In your circumstances, I would borrow a spectrum analyzer and check first.
  2. 900MHz OFDM is coming. See:

Finding commerical products is rough, but it's possible to throw together something based solely on their MiniPCI products. Note that this is currently FCC type certified.

  1. Switch to power line, phone line, coax cable, or fiber optic wired networking.
  2. Wrap the house in sheet metal and foil.
Reply to
Jeff Liebermann

Cabling-Design.com 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.