Re: An Experiment with Roaming

CrimpOn · ‎2019-10-18

One discussion that comes up over and over is, "Why does my xx device connect to one Orbi unit rather than to the Orbi unit that I think it obviously should?" I began to wonder about a related question: "Do devices really roam (change Orbi devices) as they should? (or not)." And, the answer is clearly, "At least some of them do."

What I had close to hand were my Sony Xperia phone and my Samsung Galaxy Tab A tablet, both over a year old. I brought up the Orbi web interface "Attached Devices" screen while sitting close to the RBR50 router and noted that both were connected to the RBR50 at 5G, as they damn well should be. I then went downstairs and sat close to the RBRS50, and within about one minute, both the phone and the tablet were reported to be connected to the satellite, once again at 5G. Went back close to the router, and they changed back. I have a box of old phones to charge up and see if they do the same thing.

So, "Do at least some devices roam between Orbi units?" Yes, they do. My next experiment will be to see how far away I have to walk to get a device to change from 5G to 2.4G (since 2.4G carries farther, eventually a 2.4G connection will be "better" than a 5G connection).

This experiment does not address the recurring question directly. Most of the time, the devices "at fault" are 2.4G only, relatively cheap, and fixed in position. There is little incentive to program them "to roam" since they are supposed to stay in one place. That's a different topic and a lot harder to perform experiments on. Anyway, this was a simple experiment, cost nothing, and demonstrated that at least one of the things we purchased Orbi for works as expected.

ekhalil · ‎2019-10-19

For a device to be able to do roaming it has to be able to do the following while connected to an existing AP (Orbi Router as example):

Scanning: The client device sends out probe packets to identify AP alternatives. Upon discovery of accessible APs, the device then selects its next AP based on certain criteria, as defined by the device itself (defined RSSSI threasholds in the device).
Authentication: The device sends an authentication request to the new AP and waits for approval
Re-Assosciation to the new AP and de-assisciation from the old AP

The procedure above is only available in mobile devices, that is devices that are supposed to move. This procedure is not available in e.g. security cameras. As long as those devices have an existing AP connected, they will not try to scan, that's why those devices will never roam. They will only do scanning when the connection to an existing AP is completely lost or when the device is power cycled.

CrimpOn · ‎2019-10-19

Experiment #2: "Will a device connected to Orbi at 5G switch automatically to 2.4G when the 5G signal falls below the 2.4G signal?" The answer is: "Not so far." (BIG disappointment. Pride goeth before the fall.)

Connected the Galaxy Tab A at 5G. Brought up the Orbi Attached Devices page. Went downstairs and the tablet switched to the satellite as it did before. Went outside and began to walk farther and farther away. Attached Devices stubbornly displayed "5G", past 100ft. past 150ft. Realized that Attached Devices had ceased updating. Then, the tablet lost connection completely. Opened WiFi Analyzer on my smartphone, which said "No Connection" and showed the Orbi 2.4G at -83dB and the 5G signal not detected at all. (Should have been watching this display the whole time.)

So, (a) I didn't do this experiment very well, and (b) either 5G and 2.4G signals dropped "too far" at the same time, or the mechanism that allows a device to change WiFi access points may not be smart enough to handle 2.4G vs. 5G. (I can only see my screens well at night, so will repeat this experiment - more carefully - another evening.)

I had hoped for "of course, at least one device connected to Orbi will switch between 2.4G and 5G based on which signal is better." What I got was, "not proven."

ekhalil · ‎2019-10-20

@CrimpOn wrote:
...... Then, the tablet lost connection completely. .......

Yes, I think this is not the best way to do such an experiment. I think it will be difficult to execute such an experiment.

The reason is that both the 5 GHz and 2.4 GHz bands will fade down at the same rate, so the difference in signal strength will be almost always the same.

As an example, in iPhone (and I think it should be the same for other mobile devices), for a handover to happen from one band to another there must be a difference of 8 or 12 dB between the signal strenght for the iPhone to choose to start a roaming procedure (When the iPhone sends or receives data, it picks target BSSIDs whose RSSI is 8 dB or greater than the current BSSID’s RSSI. When the iPhone doesn't send or receive data, use a 12 dB differential).

I don't think there is a simple way to do the experiment that you intend to do. :). May be reducing the 5 GHz band transmitt power in Orbi will help.

FURRYe38 · ‎2019-10-20

Signal fade may differ between 5Ghz and 2.4Ghz, dont' forget that 5Ghz is more effected by building materials than 2.4Ghz is. So signal fade may not be the same.