Deconz-rest-plugin: Scene transition firmware issues in IKEA trådfri lights

[lbschenkel]

I have done some experimentation on my IKEA lights. This is my setup:

• deCONZ 2.05.42, firmware 0x26280500
• TRÅDFRI bulb GU10 WS 400lm (003.182.92), firmware 1.2.217
• TRÅDFRI bulb E14 opal 400lm (703.182.84), firmware 1.2.217
• TRÅDFRI bulb E27 CWS opal 600lm (703.389.51), firmware 1.3.002

As far as I'm aware, this is the latest firmware available for each light.

I have noticed the following behaviour: if you create a scene with a transition, and later recall that scene, during the period of that transition the light will fail to respond to new commands (from the REST perspective, not sure what is happening at the Zigbee layer), and the state in deCONZ will be out of sync with the state of the light.

For example, let's say I recall a scene with a 10 minute transition. During this period the light will appear "hung" and the following commands will not work on the light (but will affect the deCONZ state):

• setting on to either true or false
• setting bri to any value, with or without transitiontime (not even 0)
• setting bri_inc with or without transitiontime (but see caveat below)

But to my surprise, I noticed that the following do work and "unfreeze" the light, making it able to accept commands again:

• recalling a different scene
• setting ct (but for IKEA lights deCONZ sends xy to work around a different bug, so not sure what ct would actually do)
• setting bri_inc to 0 (I believe that this gets translated to a "stop" command)

My testing was with direct commands, but I did a quick check and it looks like group commands have the exact same effect on the light as the direct ones.

I got aware of this when I tried to implement a wake-up light by transitioning between two scenes with a 30 minute transition time, and then the lamp "hangs" and I tried different things until I figured out that some commands actually managed to make it respond again.

Curiously, the lights don't freeze in the same way when not recalling scenes. If I send bri=255 transitiontime=6000 (or bri_inc) the light does respond to subsequent commands during the transition time. I have seen some reported issues that seem related (but don't explicitly mention scenes): https://github.com/dresden-elektronik/deconz-rest-plugin/issues/450, https://github.com/dresden-elektronik/deconz-rest-plugin/issues/384, https://github.com/dresden-elektronik/deconz-rest-plugin/issues/66

I wonder if with the knowledge above, this firmware issue could be worked around on the deCONZ side? For example, if the last command sent to the light was a scene recall any new command could be preceded by a "stop" command to make sure the light will respond. Just brainstorming here. Depending on what is deemed an acceptable solution, I might be able to submit a PR with the implementation (if not too complex, I'm a beginner in this codebase with limited Zigbee knowledge).

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[lbschenkel]

Hi @manup, any comments on the above?

[manup]

Sorry for the late response, great findings about the transition time.

I wonder if with the knowledge above, this firmware issue could be worked around on the deCONZ side? For example, if the last command sent to the light was a scene recall any new command could be preceded by a "stop" command to make sure the light will respond. Just brainstorming here. Depending on what is deemed an acceptable solution, I might be able to submit a PR with the implementation (if not too complex, I'm a beginner in this codebase with limited Zigbee knowledge).

In theory yes but this would open the window for new issues which are hard to handle reliably. There are already networks with > 50 IKEA lights out there, I'm afraid this approach doesn't scale well.

I think this issue with your discoveries should be forwarded to IKEA developers so they can recreate and hopefully fix the transition time bugs (shouldn't be that hard).

Part of the IKEA Tradfri team is active on Reddit, might be a good way to get the message through:

https://www.reddit.com/r/tradfri/

[lbschenkel]

As suggested, reported in https://www.reddit.com/r/tradfri/comments/au903n/firmware_bugs_in_ikea_bulbs/

[manup]

Nice very good writeup hope it will reach the devs.

There seems to be new firmware available now:
https://www.reddit.com/r/tradfri/comments/avqc73/new_firmware_out_this_day_ota_is_coming/

The changelogs don't mention anything related to the problems mentioned here, but might be worth a try.

[ebaauw]

Looks like new firmware for the colour and for the white spectrum bulbs. My colour bulb upgraded without any issues (image type 0x2801). Looks like the new white spectrum firmware is for a different (newer?) type than my bulb. Mine wants image type 0x2202, but the only 0x2201 and 0x2203 are new.

After upgrading, the colour bulb still ignores _OnOff_ or _Level Control_ commands while a _Color Control_ transition is in progress. Sending a _Color Control_ commands while a _Level Control_ transition is in progress, seems to cancel the level transition. _Current Level_ seems to be updated only once the transition has completed.

[lbschenkel]

It looks like the new official firmware is simply the test firmware which got promoted to official (however, the binaries are not exactly the same even though the versions are). The only exception that I can see is that for 0x11C1 the release version (12214572) is still behind the test one (12222572).

I had previously upgraded all my devices to the newest test firmware, and as far as I can see there has been no perceptible difference in behaviour regarding those bugs. Actually I didn't notice any change of behaviour at all.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[pvizeli]

Can we reopen this issue?

[Phontana]

Yes, we should reopen the issue since the scene transition issues are still present.