23 replies to this topic

[40H3X]

I've noticed that in the device list of my AP there was a client with a low (IMHO) RSSI -84dBm. As soon as this AP was switched off the client moved to the other AP and it went to -74dBm not that great but better.

 

I read the following:

 

Quote </ Unless you use an Enterprise Grade WiFi system such as from Cisco/Aironet or Aruba, with managed APs, any other scheme will have the consumer-grade problems. As IEEE 802.11 does not require a client device to choose the best-signal (strongest for which the client has decryption keys). So  a client device (laptop, tablet, smartphone), will choose first-heard during a channel-scan. A scant few take the time to sit for 100mSec (the default beacon interval) on each channel. Few do, as it takes too much time and battery usage.

So non-Enterprise WiFi burdens the human user of the Client to choose best-AP if their device does not. /> Quote

 

is this true...?

Edited by 40H3X, 18 November 2015 - 16:08.

[WanWizard]

Well... as always it isn't that simple.

 

An AP is a shared medium, and it's radio is therefore shared amongst all it's clients. And the client with the slowest connection determines the speed of the radio. This means that even if you're sitting next to the AP and your Windows applet says you have a 150Mbps connection, the effective speed could be 1Mbps, because of your daughters iPhone on the first floor that has a lousy connection. Especially smartphones are a problem, because of their weak radio transmitter.

 

Enterprise solutions have several things to counter this.

 

The first is that in an enterprise network, cells are much smaller. An average AP in an office building runs at 5mW, 20 times weaker than the maximum allowed. So devices get out of range quicker, and switch therefore quicker to another AP. Second is that all AP's broadcast the same SSID, and most devices list AP's per SSID with the most powerful on top, so due to the coverage there is less chance a weak AP is selected. And most have a sofisticated management platform that constantly monitors all AP's, and some have the ability to bump clients that get below a configured threshold, forcing the client to do a reconnect, which most likely will be to a closer and stronger AP.

 

Home users often have only one AP, and if they have more, they are either too far apart (no cell overlap), or have a different SSID, so a client will not automatically roam from one AP to the next.

[Erik Slagter]

I have a setup here, comparable to what 40H3X desribes. Two independent access points, non-enterprise. If you have SSID and password and some other traits exactly the same, the client can and will roam whenever necessary. That "necessary" is kind of debatable. Phones and tables try to be idle as much as possible, to save battery. But even these will scan the whole band every now and then. If they find a suitably equal SSID, they WILL step over.

 

There are few considerations though. Just the presence of another access point with the same SSID will not make the client switch, they will take other factors in consideration. Exactly which is implementation-dependent. In my experience, phones will delay roaming until necessary, that is e.g. when the signal strength becomes below -75 dB or even worse. Laptops will generally roam more aggressively.

 

The text which is quoted is correct, but only to a certain extent. Only on enterprise WLAN you have a feature called "seamless roaming" which means that when a client roams from one access to another, the client doesn't need to re-authenticate. This means there will be no suspension of service which you otherwise would experience for a few seconds, during re-authentication. On the other hand, many clients don't even support seamless roaming and also a few seconds interruption is not always a problem. So in my opinion the text is too black/white.

 

This ...

So  a client device (laptop, tablet, smartphone), will choose first-heard during a channel-scan

... is definitly not true. Even my very basic esp8266 client considers signal strength when associating. I know because this feature was implemented recently

 

Also my phones and tables will re-associate when I move to another floor, eventually (because they typically will try to use the current one as long as possible). I have a setup with two access points on two bands (both 2.4 GHz and 5 GHz), the 2.4 Ghz access points share one channel, the 5 GHz access points use different ones. For roaming it doesn't matter much, although a "same" SSID on the same channel is typically preferred.

 

A common misconception about WLAN is that the access points determine which will be used to have a client associated. This is not true. The client determines the access point (and so also when to roam). There are some extensions to allow an access point to "suggest" to the client to roam, but the client's word is always final.

 

So to conclude, the above citation is not really true.

Edited by Erik Slagter, 18 November 2015 - 17:39.

[theparasol]

i can confirm the story of Erik

Edited by theparasol, 18 November 2015 - 17:47.

[40H3X]

Thanks for for the response guys @Erik I have a similar setup 2 AP's both 2.4 and 5 all with the same SSID and key. Yet, here the 2.4(6 and 11) and 5 (5,8 and 5,7) are on different channels, you choose not to, why?

[theparasol]

I rather would not share the same channel on the 2.4 Ghz

It can work very good but not if 2 clients are active on different floors. Then accesspoints and clients have to take turns who may send.

But this same channel setup is far better than using overlapping channels that would create in this case interference for sure.

But nowadays the 2.4 Ghz is rather crowed. Lucky are those that dont have to share the frequencies with their neighbors.

Since not always accesspoint and clients can hear eachother. The so called hidden node problem.

Its rather complicated but most of the time @home it sorts out ok but not in enterprise setups where there are 20 ~ 30 active clients or more in each room.

Luckily most wireless access is rather burst traffic. Worst case is 20~30 clients streaming a youtube HD compliant videostream and all connected by 2.4 Ghz.

[Erik Slagter]

I use 2.4 GHz for legacy. In my case that means guests with "legacy" phones and my esp8266's. I am using channel 1 for both access points because all of the spectrum from channel 1 up to channel 14 is equally crowded. In that case it pays to use a lower channel, because they are less interfered by magnetron radiation. So that is why both are on the same channel. My 5 GHz access points are using separate channels, both of them not interfered by anything else, no others use them.

 

Considerations here: sharing a channel means more often waiting for idle slots and more often colisions + interference BUT also it makes roaming smoother because the client instantly sees the other access point on the same SSID, without having to scan the band, which mobile devices don't do very often (power saving).

 

Another consideration, especially for the 5 GHz band: only the first four "slots" are released for general use. All other slots have additional restrictions as to in/outdoors use, power restrictions (TPC) and DFS (avoiding frequencies that are used by weather radars). Many mobile clients avoid these slots for that reason, only a handful clients can use the higher slots. Leaving only four slots, which is quite disappointing. Newer 802.11ac compliant clients tend to support some more slots, but not necessarily all of them.

 

What I find most disappointing is that WLAN in the 5 GHz band has existed since 2001. I have had several access points and laptops with 802.11a support, no problem. Somewhere from about 2007 on, no 5 GHz supports available, either access points, laptops and mobile devices. Even when 802.11n became common, which does support channels in 5 GHz, most access points and clients would only support channels in the 2.4 GHz band. Unbelievable. Only now, with 802.11ac, which REQUIRES implementations to support 5 GHz we can use 5 GHz again.

 

Most people have never used a WiFi scanner, obviously. The people I show it, understand what I am talking about. The 2.4 GHz band is a mess.

[Erik Slagter]

Oh BTW at my work we do 2.4 GHz for guest internet access, all serious business is over 5 GHz.

[theparasol]

Its a pity to see that even nowadays for example most low en midrange notebooks still have 2.4 ghz only wifi chips installed.

Even from bad manufacturers like mediatek/ralink. A comparable "good" chip from Qualcomm Atheros, Broadcom or Intel is not used.

If a company decided to buy laptops they tend to buy most value for money and they still get 2.4 Ghz WiFi.

If they even consider 5 Ghz the amount of models decreases significant and price goes steep up.

Manufacturers like HP tend to whitelist the wlan hardware in bios so its not doable to "upgrade" or replace the WiFi chip with a better model unless you buy the 3 times more

expensive HP branded optional part.

 

Hope 2.4 Ghz gets abandoned soon, AC WiFi chips do dualband "standard" so lets hope the crowd cries for "AC WiFi" soon!

[Erik Slagter]

I replaced the wlan card on my laptop (2.4 GHz only), but only with one out of two "supported" cards. Any other card will probably be blocked by the BIOS, it's not only HP that does that I was lucky I could get my hands on such a "supported" card and it even existed for 5 GHz. Most laptop owners won't have this luxury. And I must say, reception is not great. I guess the card can't really use the antenna's very well, they're probably optimised for 2.4 GHz.

 

So that way, we will never get to 5 GHz.

 

Anyway, at least I started with the access points, they work at both bands and most of the phones and tablets in house can use the 5 GHz band. I was quite not amused when I discovered that the Samsung Tab 10.1 I bought two years ago, was advertised as 5 GHz capable, but was restricted by a driver update to 2.4 GHz. I should have sent it right back then. On the other hand the display resolution sucks too, so I bought myself another one and passed this one on to my son. He won't mind

Edited by Erik Slagter, 19 November 2015 - 19:25.

[theparasol]

Yup, many manufacturers do lockdown the wlan adapters that can be used.

 

The more daring users "hack" and reflash their bios, the whitelisted wlan adapter get replaced by the pnp id of the make/model that the user wants to place.

Works for home situations but not for commercial usage.

Customer requested 80 laptops with dualband wifi chip but the accountmanager "selected" the wrong one from the list, or the manufacturer silently replace the Quallcom Atheros chip with a inferior type...

 

What a sick thing to do!

Such kind of actions you dont expect from a manufacturer like Samsung. I would for sure return the crippled device to let them know I wont accept these kind of updates.

[Erik Slagter]

Mind you, Samsung are just the same bunch of villains like any other tech company. The only difference is that they often have the "highest end" phones (newest features) compared to others. And they allow you freely to flash a custom image. That is why I like them, but nothing more than that...

[Trial]

Hi,

"And they allow you freely to flash a custom image."

not true. They make it relative easy but you loose waranty if you do so.

 

I like them to but not because Samsung is so nice instead because the 3rd party support like cyanogenmod.

 

ciao

[theparasol]

Just like I want to have openpli on my receivers, I like to have cyanogenmod on my phone and tablet.

Besides that my accesspoints for my personal usage I want to have openwrt on them.

Before buying any new gear I check what current hardware can be used with them.

Too bad I'm a rather geeky person doing that otherwise all manufacturers where already far more supportive towards opensource software alternatives for their products.

[Erik Slagter]

"And they allow you freely to flash a custom image."

not true. They make it relative easy but you loose waranty if you do so.

Not completely. Depends on the shop.

 

Always better than other manufacturers where you need to actually crack the phone before you can install your own image.

[40H3X]

I replaced the wlan card on my laptop (2.4 GHz only), but only with one out of two "supported" cards. Any other card will probably be blocked by the BIOS, it's not only HP that does that I was lucky I could get my hands on such a "supported" card and it even existed for 5 GHz. Most laptop owners won't have this luxury. And I must say, reception is not great. I guess the card can't really use the antenna's very well, they're probably optimised for 2.4 GHz.

 

So that way, we will never get to 5 GHz.

 

Anyway, at least I started with the access points, they work at both bands and most of the phones and tablets in house can use the 5 GHz band. I was quite not amused when I discovered that the Samsung Tab 10.1 I bought two years ago, was advertised as 5 GHz capable, but was restricted by a driver update to 2.4 GHz. I should have sent it right back then. On the other hand the display resolution sucks too, so I bought myself another one and passed this one on to my son. He won't mind

In the early days I used Realtek (cheap), but after numerous problems I invested in better NIC's and it really pays out. Here (non-enterprise) my router (BSD) uses, on all 4 subnets, an Intel Pro/1000 and I experience no in/our errors or collisions. My laptops use Intel and Atheros (it's one of my criteria when I buy stuff) alas with tablets there's not much to choose

 

With WIFI it's all Broadcom or Atheros. Yet when using both in the same network I’ve experienced problems in receiving IP on MAC address with transparent client bridging, but with ARP-NAT this was solved. I tried WDS, but ran into problems with DHCP on MAC (again). https://wiki.openwrt...owto/clientmode

 

Btw I can confirm that there is a difference (it’s minimal) between channel 1 (-46dBm) or 11 (-49dBm) on 2,4GHz, here I did experience that the higher channels on 5GHz have a slightly higher RSSI.

[theparasol]

A good brand isn't always a guarantee their wireless adapters work flawless. Intel for example had their pain in the but 6235 model. They suffered almost 2 years from random disconnects / timeouts, the only thing that could solve it was using 802.11g instead of 802.11n. Fixing it in their drivers took far to much time.

In generic the lower the frequency the further the signal travels / more penetrating using exact same transmit power.

On some frequencies you may use more transmit power so the signal can achieve higher/better RSSI values.

 

http://www.air802.co...egulations.html

 

But it seems not every piece of hardware can use the higher channels of the 5 Ghz. Even worse some cheap stuff even doesnt like to work with channel 12 and 13 in the 2.4 Ghz spectrum. If the wlan chip thinks its in America (or seems to be designed for it) it wont detect a network nor communicate on channel 12 and 13.

[Erik Slagter]

In generic the lower the frequency the further the signal travels / more penetrating using exact same transmit power.

On some frequencies you may use more transmit power so the signal can achieve higher/better RSSI values.

That is actually an advantage of 5 GHz. 2.4 GHz channel reach way too far and cause interference all over the place.

 

There is not really a difference in reach between individual channels on the 2.4 GHz band nor on the 5 GHz band. If you HAVE to use 2.4 GHz band channel, look for a channel that suffers least interference and if that doesn't matter because all channels are crowded (and that's mostly the case), use channel 1, because it's least bothered by stray magnetron radiation. We did an experiment with an access point in spectrum analyzer mode (no, you can't buy these at the Dynabyte ), turned on a magnetron oven about 20 meters from it and watched. It appeared that channels 3-14 were heavily distorted by the stray radiation, channel was influenced less. The access points we're using are about 900 euro's a piece

[theparasol]

Well the kind of accesspoints we use have directional antennas and beamforming. They will create less interference since they dont send the signal in all directions around and the power it uses to transmit will be optimal to reach the client that needs it. Its very nice stuff to work with, but for home usage with small amount of clients too expensive.

[openpli4me]

Just increase the output power until the neighbors have changed channels, perfect range, no signal drops, simpels.