Shelly Wifi devices not working well ONLY with OpenWRT

I have a strange problem that may well have nothing to do with OpenWRT, but I am stumped what to try next.

I have started using a variety of Shelly IOT devices. They are in a home with 4 access points, one of which is running OpenWRT. When the Shelly devices (well, most of them) are connected to the OpenWRT AP, they have high ping loss (and various flaky behavior probably related to that loss). Associated with the other AP's, no issues.

The other AP's are two Engenius EAP1300's and a Mikrotik hAP ax2. The OpenWRT AP is a Linksys WRT1900ACS V2 running 23.05.3 (it was .0, I upgraded without any change). The OpenWRT is purely running as an AP, not a router.

I have physically moved the OpenWRT much closer (it was already fairly close) to problematic devices without impact. RSSI and signal reported on each end are very good (45-55db generally).

Everything is on the same subnet, this isn't a DNS or mDNS issue or broadcast issue.

In OpenWRT I think I have very vanilla settings:

Mode N, channel 11, 20MHz
power 20dBm
WPA2-PSK
Management frame protection disabled
KRACK disabled
NOT isolate clients
Short preamble enabled

I think everything else is defaulted, and I've changed a couple of the above back and forth without impact (e.g. management).

Every other device (I have many, mostly esp8266/esp32 running esphome) work fine, with much weaker signals. Indeed, the Shelly devices are new additions, this whole setup has been stable for years.

It's really easy for me to blame shelly but if I shift the shelly's to other AP's, then they work fine also. AP's further away, weaker signals tx and rx. Plus Shelly is a pretty respected brand not known for flaky connectivity. Incidentally it is mostly 2PM's and Plus 1's that are having an issue, I have a natural gas detector from them that works fine.

Is there anything that rings a bell with anyone, some setting to enhance IOT in general if not Shelly in particular (which are ESP32's by the way).

Or maybe it's just this particular Linksys hardware. It's a bit old and maybe the best answer is just to get rid of it, but... any other thoughts?

Linwood

Furthermore "WMM mode" on radio 0 and 1 is said to have issues with certain WiFi client device chips (ESP8266). Disabling "WMM Mode" is said to fix some of these problems.
Radio 2 is not having those issues with ESP8266-based client devices.

Looks like you need something without a Marvell WiFi chipset

Maybe so. I read through that, understood about half of it, but not sure what in there might apply. My radios work, no issue with regions that I see, I tried wmm on and off.

It does sound like reverting to factory firmware may not fix the issue.

I ordered a Mikrotik AP to replace it. Yes, probably Unifi would be better but have a mild preference.

1 Like

Which Mikrotik are you going for?

Few things I saw that Linksys hobbled the V2 firmware in the WiFi chipset breaking stuff that Openwrt cannot fix.

cap ax Gen 6.

I haven't tried any of their pure AP's before, I have a couple of their (what I call) travel routers, hap ac2 and hap ax2. I find their ridiculously complicated GUI a pain, but the features that are exposed are moreso than most brain dead devices the big guys put out, plus so far very reliable. Will find out on the cap I guess.

I did like that it is POE in and out, so I can hang another POE device off of it, apparently (I don't plan to, but there have been cases where I needed that).

I just thought I would add this as a bit of info. Waiting for delivery of new AP.

The packet loss, which I monitor minute-to-minute, is cyclical.

The period is similar (maybe identical) but they are not synchronized. This would seem to imply the issue is related to the device not OpenWRT AP. These devices are just variations on a smart switch, just sitting there waiting for someone to turn them on or off. I'm sure internally they run various loops but I would not think on this slow of an interval.

I think the real answer is get rid of the Netgear device, however I find it interesting that the packet loss is cyclical like this. If anyone has ideas...

This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.