EAP 245 + Omada, connectivity issues
This thread has been locked for further replies. You can start a new thread to share your ideas or ask questions.
Configuration
Two EAP245-v3, latest firmware (2.3.0 Build 20190731 Rel. 51932), connected over Gigabit Ethernet to Cisco 3750G PoE switch, located at a fair distance between each other, configured with non-overlapping channels (1,11 for example).
Omada controller v 3.2.1 running on a Linux box, connected over Ethernet to the same switch.
Four SSIDs at both 5Ghz and 2.4Ghz.
Fast roaming, dual band 11k report enabled. Force disassociation disabled (but same results with Enabled)
Airtime fairness disabled (but same results with Enabled)
Band steering disabled, Mesh disabled.
QOS: WMM, Unscheduled Automatic Power Save Delivery - enabled. No Acknowledgement disabled.
Cloud access enabled.
Multiple (~12) mobile devices connected. Phones, laptops, some IoT devices (connect to a separate SSID).
Interference (as reported by Omada) is marginal on both APs and both 2.4 and 5 Ghz.
The Problem:
The network is unpredictable. At times I get very high performance, and seconds later (without changing location), speed drops to close to zero. Changing from one SSID to another usually solves the problem (but these two SSIDs are equivalent - only the name is different).
Sometimes my devices are connected (as reported by both the device itself, RSSI, and the Clients tab in the Omada controller), yet data doesn't get through.
I don't see a lot of (or any) Tx, Rx errors / dropped packets, even when this issue persists.
At times, Omada may report high utilization of an access point (both Tx and Rx) with very few, and sometimes no clients are connected (and definitely none of the clients is creating this pressure on the AP).
At times, my devices report ‘incorrect password’ to the WiFi, even though it is correct (and in fact unchanged). RSSI is pretty good, -52 dBm for instance (this is not a reception issue - These issues happen even if I am a few meters of the AP, and sometimes even standing right next to it). Some permutation of resetting everything multiple times helps - I am yet to figure out what is the winning sequence.
Reset for the router sometimes seems to fix the issue, sometimes not.
I’ve done some root cause analysis:
- Changed the Switch
- Powered-on the APs using the supplied PoE power supply
- Changed various configurations (WiFi bands, transmit power, airtime fairness, see above)
- I’ve tried disabling Omada controller
- I’ve changed the cables
- I stopped short of spraying the APs with holy water. Other than that I think I tried everything.
I’m seeing excellent reviews for this product (that's why I bought it), yet in my environment, I’m seeing the performance that’s worse than a consumer-grade AP from 7 years ago..
I appreciate your help
No. Interferences are WiFi frames which got corrupted due to collisions, for example if two devices send data at the very same time or if there are strong reflections of EM waves caused by metal structures in a room or if there are more signals from DECT devices, bluetooth devices or even microwave ovens all working in the 2.4 GHz frequency band.
Valid frames sent by other devices which have successfully acquired AirTime in the same channel are recognized by the EAP as Rx frames, thus registering it as utilization (of the WiFi channel, not of the EAP!). Every device – not only APs – is always listening to the channel to find out whether the channel has been acquired already by another device. This method is called Listen-Before-Talk (LBT) and it's part of collision avoidance measures (CSMA/CA) on a shared medium such as a wireless network.
For example, see the following screenshot. The EAP isn't used by any device. Nevertheless it shows 16% utilization of the WiFi channel. Laptops, smartphones and some IoT devices in the same room are not only transmitting data to nearby APs using the same channel, but also receive data (the SSID beacon, see Tx frames!) from this otherwise completely unused EAP:
Thus, high channel utiliziation but low EAP data throughput is an indicator for an overcrowded wireless channel.
Yes, I know, you wrote that your EAPs use different channels. But can you really assure that no other wireless devices (DECT, bluetooth, remote controls etc.) use the same channel the EAP is using? What's with the legitimate clients?
So, some more questions:
How many WiFi clients are in the room the EAP is in? How many do use the EAP actively?
Where do you see packet drops in Omada Controller? Can you post a screenshot please?
How many foreign APs does the EAP see? Please perform a wireless survey to find out and post the result.
So to make sure I understand, any transmission from any neighboring device, connected to other APs that use the same WiFi channel, would register as Rx ?
And Tx, in the case of no or very few clients, would be what, the SSDI beacon?
And Interference would be invalid frames - garbled transmissions that are impossible to decipher (like collisions) ?
> can you really assure that no other wireless devices (DECT, bluetooth, remote controls etc.) use the same channel the EAP is using?
I don't know, could I? Is there a tool I could use besides a Wi-Fi analyzer ?
> How many WiFi clients are in the room the EAP is in? How many do use the EAP actively?
In total, about 12 max. 10-2 or 8-4 split. 3 IoT devices (on a separate 'Guest' SSID), and the rest are either mobile phones (which are idle most of the time) or laptops (one is using an Ethernet connection, and so is idle most of the time, the other is just browsing the web for work - not a traffic hog)
> Where do you see packet drops in Omada Controller? Can you post a screenshot please?
I actually see very few dropped packets, that was my point. See screenshot.
> How many foreign APs does the EAP see? Please perform a wireless survey to find out and post the result.
See below, survey taken at three different floors. EAP is on floor 0 and 2.
Floor -1:
Floor 0:
R1D2 wrote
Floor 2
That's correct.
Without a spectrum analyzer, locating the exact source of interference would be unreasonably difficult. Software spectrum analyzers depend on the WiFi adapter of the host system and recognize only WiFi devices, but there are low-price spectrum analyzers on the market which can discover interference over a broad range of frequency bands including non-WiFi devices (beware, low-cost still means a thousands bucks). For example, this is a quick spectral analysis of the 2.4 GHz band in my office:
I see, looks fine. 16dBm TX power is intentional, yes?
Anyway, I would recommend 20 MHz channel width for the 2.4 GHz band. 40 MHz requires 8 channels, thus overlapping with other channels.
As for the WiFi survey I meant a survey done by the EAP itself so you can see what concurrent devices the EAP sees. Start the scan under »Rogue AP Detection«, find results under »Insight → Untrusted Rogue APs«.
Changing the WLAN involves re-negotiation of frequency band, WiFi speed and channel width between the AP and the client device.
As I can see from your WiFi survey above your device connected to the ESSID »Rosenbaum2« was one time using the 2.4 GHz radio, channel 11 @20MHz channel width and the other time it did use the 5 GHz band, channel 161 @80 MHz channel width. I would not call those two basic service sets »equivalent«, albeit both are in the same extended service set (»Rosenbaum2«). What makes you sure you were really using the same radio when switching to the second equivalent ESSID? Did you check the frequency band, too?
Again, I suggest to set 20 MHz channel width for the 2.4 GHz radio and to try increasing Tx power if using non-overlapping 2.4 GHz channels. If this works better then, the effect you observe might be related to interferences.
Next is speedtest.net. I just did a quick test with an AP (not an EAP, but one on the same floor I am currently).
Test setup: old MacBook, 2.4 GHz, 802.11g, channel 6, negotiated WiFi speed 54 Mbps, expected data throughput ~22 Mbps.
See the results below. How would you explain the different results between the two test servers selected automatically by speedtest.net?
Let's try dslreports.com:
Not much interferences during the test:
