#Nanoleaf beta firmware discussion
1 messages · Page 2 of 1
Isn’t it possible for you to opt-out of beta to get the firmware?
no, devices don't support firmware "downgrades" (you can only upgrade to a firmware with a higher version number)
Ok, how many bulbs are we talking about?
all of them :) but I only own 3.
should have left 1 out of this beta if i was thinking ahead, but oh well
You should be their personal hero and they should shower you with gifts. 😃
Was the bug with the custom cluster their main problem all along or just one of many?
They said in their Discord that bulbs that break due to beta firmware will be replaced in any case.
If you're in the US I'd be willing to ship you some of mine
So, you are completely lost as Nanoleaf customer? 😃 How do your WiFi bulbs work?
I took all my bulbs out but have yet to replace the strips (2). They work pretty good. The matter implementation has some issues, but I actually just use their proprietary protocol (via HA integration) to control them anyways
I'm actually looking to get some other NL wifi products (like the skylight) but def won't be getting their thread stuff
Which WiFi bulbs do you use instead of the Nanoleaf Essentials now?
Is there a hardware difference between the native and the migrated bulbs?
How do they work for you?
Native and migrated?
I thought it is with WIZ like with EVE. You can migrate old WIZ bulbs to the Matter firmware or you can buy native WIZ Matter bulbs. Sorry, maybe language barrier… 😃
Ah yes that's right. I've found the migrated ones to have a lot more problems... Native are very stable. They have some issues when turning on with a specified color and brightness (via matter) which I'm trying to work with support to fix.
The integration (local protocol) has no issues so I'm using that for now
Plus you can access animated scenes via local protocol which is a plus
Ok, so you are more or less happy with WIZ now.
@digital star Another question: Didn’t you know how to remove Matter ghost nodes? Can you describe that? I have 5 ghosts now. I have 48 Matter over Thread devices/bulbs paired to HA.
I have now arrived at Node ID 214. Since I've been using the Matter beta server (for over a year now), I've resetted/readded 2 of my 32 EVE devices. The other 180 Node IDs are the result of resetting/readding 16 Nanoleaf bulbs.... The 5 ghosts are definitely the result of incomplete pairing attempts of the Nanoleaf bulbs. I see these ghosts every 30min in my log and I want to get rid of them.
Yeah I'm happy with them, no complaints
You'll have to use the websocket (maybe the dashboard supports it?) let me look
yeah youll have to use the websocket:
from matter_server.client.client import MatterClient
import aiohttp
from aiorun import run
url = f"http://192..../ws"
async def func():
async with aiohttp.ClientSession() as session:
async with MatterClient(url, session) as client:
await client.remove_node(node_id=5)
return 0
run(func())
Ok, thanks. But I do not know how to get there. What do you mean with Dashboard?
The URL is the IP of my HAOS?
Yeah but it's not in the dashboard unfortunately
Yeah and you have to have the websocket turned on in the addon
Hopefully Marcel forgives me for teaching you this 🙂
Thank you. Will check it tomorrow
This update doesn't make you want to try the bulbs again? They should finally be stable... Also I remember Nanoleaf bulbs are brighter than most WiZ bulbs.
You can also wait a bit, I'm planning on adding the remove to the dashboard. So we can have some more advanced commands on the builtin dashboard.
Ok, I am going to delete one via the websocket. So we have 4 ghosts to test that new functionality. Thanks
No because I still don't have high confidence in them to be stable long term. The issue with the custom cluster issue is still lurking there in the background.
You're somewhat right about the brightness. Nanoleaf is 1100 lm white and 800lm color and wiz is 800 lm for both, but for my use case that doesn't matter
What is the custom cluster issue?
you gotta scroll up in the nanoleaf discord about it, was a massive pin point about it all
The short of it is there's some custom attributes for the NL bulb, that when read, essentially cause the bulbs to hang and/or crash. Despite the attributes being only zeros for some reason. NL simply disabled it as a temporary fix but they've yet to figure out the real issue.
EVE has actual custom attributes on their plugs for power monitoring and HA actually uses them to show power usage... Those work fine
Hi, here is how it looks like with 16 Nanoleaf Matter over Thread bulbs for 3 days of uptime with the latest 3.6.173 firmware: https://imgur.com/a/soZ89EP
The result is not perfect, but its also not that bad. There are only short unavailability time slots (black slots in the screenshot). The devices always come back to HA by themselves, but sometimes it takes 5 to 7 minutes.
All my Matter over Thread bulbs/devices are paired to HA and Apple Home. Maybe I will unpair all Nanoleaf devices from Apple Home and bridge them via HomeKit Bridge from HA to Apple Home to reduce the Thread traffic in my Thread network.
How does it look for you guys?
Interesting... I'm not sure what custom attributes Nanoleaf has? I know Matter has some missing features for lights such as no Scenes, but hopefully these features will get added to Matter 1.3 or a future version spec.
Nobody knows what it's for, like I mentioned, it's something like 256 bytes of zeros
If it's not really used then I'm not sure what the problem is with it getting disabled.
There's no loss in functionality but my concern is 1) why is it there at all and 2) why don't they know why it's breaking things
point 2 is my day to day at work sometimes lol
note that reading the attributes did actually work fine - the problem was specifically performing a wildcard subscription to get notified of all attributes, something that home assistant does but apparently google and apple don't do. During the subscription process, the device reports back the current values of all subscribed attributes to make sure everything is in sync, but when nanoleaf bulbs hit the custom attributes, something happened that stopped it from completing the data reports, and so the flow couldn't continue onto the next step where the subscription became active.
and in my testing, this only happened when multi admin was in use (i.e. the bulb was joined to multiple matter fabrics). the problem is that because of how matter devices get added to home assistant using the matter apis provided by google or apple on a mobile device, the devices are usually first joined to a temporary device vendor fabric, and then shared with home assistant, so the bulbs were usually in the multi admin state when HA was trying to add them.
Nope, Apple and Google are also doing wildcard - we actually werent doing that until a few weeks ago and switch to wildcard on request of NL because the manual attribute path subs were running them into memory constrains.
So everybody is doing a wildcard sub on devices, at least all major Matter controllers do that
The different MAY be (but I am not 100% sure) that we are on a newer code base of the sdk
And ofc we use the python wrapper of the sdk which has quirks on its own
Yeah, I think THIS is why it popped so big on HA users
yeah, I suspect this is purely on the bulb side and not (directly) related to the Matter SDK version. Would be interesting to get some thread traffic captures of the problem to poke at in wireshark, but they just released an updated to the firmware branch i'm on which ports the "fix" from the release branch forwards, so I don't see the problem any more :)
Can you Wireshark production firmwares? (as in, would you be able to decrypt the traffic?) 🤔
This seems weird to me. So it is not that they have a generic problem with that particular custom attribute, but it needs some other circumstances.
I think it should be possible to decrypt the traffic if there's a way to pull the key for the matter fabric out of home assistant (ha can decrypt it, after all)
hmm, so yeah - would need to set up the matter sdk on the controller side to log ("leak") the established session keys
i was able to reproduce with chip-tool, which makes doing that a bit easier
Shouldn’t this also work?
that just describes how to set up wireshark to do live captures using a chip with thread rcp firmware.
Ok, I thought this is what you are looking for. 😂
decrypting the matter traffic is a separate issue :)
it's basically the same thing as "ok, i've got wireshark set up to capture ip packets over ethernet, now i need to read the https traffic" :)
Ok, I didn’t know. I thought being able to read the thread traffic is enough to also be able to read the Matter traffic.
ok, so it's actually more like "i've got wireshark set up to capture ip packets on a secure wifi network, now i want to read the https traffic" ;)
there's multiple levels of security going on here
This is what makes me think that they might have the attribute storage set up incorrectly in the sdk; like it's set to use per-fabric storage but they didn't actually allocate memory for more than one fabric or something.
(but then why would it break only on subscriptions, i could still read the attributes fine)
all around just weird :)
and the controller needs “leak” the send packages too. Unless they use symmetric encryption
i need to review the matter specs to be sure, but basically everyone uses symmetric encryption for all data transfer. asymmetric encryption is just too slow.
the trick is that most modern protocols use something like diffie-hellman key exchange to generate a unique symmetric key for each connection.
that's the key that has to be leaked in order to decrypt the data.
Ahh, that’s definitely an advantage and makes everything easier. Didn’t know that asymmetric encryption is significantly slower than symmetric
HK does similar. Though it generates multiple keys for different purposes (eg in thread iirc there are at least 3 - tx, rx and events). And there’s a counter so each packet is encrypted differently.
A bit annoying because UDP and packet loss means you de-sync a lot and end up starting new sessions.
it looks like matter added the ability to retry sending a packet with the same sequence number a few times to help with that
oh, huh, it looks like there is actually a way to remove the google play services fabric from matter devices from the google side. Open the Google settings, All services, Devices & sharing, Matter devices, and in the dropdown in the top right pick "This Android device". It'll then show all the devices it has added the setup fabric to, and let you remove it from those devices.
So just a small update, ever since updating my Nanoleaf bulbs to the latest firmware they have been working flawlessly. I had constant issues before but now it's perfect. It's kind of boring, having my bulbs just work.
Got that latest firmware actually promoted to stable?
28 days ago:
They have been dead silent since for the most part
but seems like they have a ton of matter/thread hardware betas going on, most of which people here are in
and they all seem to be using the same sillabs MG24 SOC, so fingers cross they fixed the upstream issues
They sent around a survey… about a week ago, i think? to everyone who opted into the public firmware betas asking about reliability and any concerns, so they're not completely silent on the topic :)
Isn't that the same chip that the SkyConnect uses?
No actually SkyConnect, Sonoff E and Conbee III all use MG21.
It's kind of ironic if the Thread Border Router uses the less powerful chip.