Getting 'Magic Switch Mode' to work on sonoff basic R4

[ferbulous]

Hi, i find this new feature great for 2 reasons:

1. Easy to add into existing 2 way wiring without adding new cables.
   Refer to picture 1.

2. Can be installed on the light fixture, and still be controlled even with the wall switch toggled off.

However, with esphome/tasmota installed, it doesn't seem to work.
The light flickers for half a second, and doesn't toggle the relay

IMG_9292.mov

Anyone has any luck with this?
After some discussion in the discord, it could be using zero crossing detection based on pulse detected at pin 5.
I tried assigning zero crossing pulse on pin 5 but that's not working yet.

[sfromis]

No experience with the hardware, but to me the simplest setup for that feature would be a hardware hack to either detect the interruption and signal that as a "button push" (with some software decoupling), or maybe just generate network frequency pulses allowing the firmware to detect the brief pause in this.

[ferbulous]

What would make a lot of sense is to check if it even has an electrical connection to something on the board.

How do i check for that? Not too sure about messing around with the basic when it’s wired to the main

[sfromis]

First step would be looking at the board, to see what's visible.

Electrical checking must be without having it powered. Using a multimeter to check, you can make it beep when you touch two connected points.

[ferbulous]

Anything interesting on the pcb that I need to check? Like tracing pin 5

https://digiblur.com/wiki/devices/relays/sonoff-basic-r4/

[sfromis]

Any gpio pin connected to something may be interesting. And not from such blurry and hardly readable images.

[sfromis]

[lalo-uy]

IN the last image there is a track on GPIO 5, but runs to the bottom of the pic. El mar, 21 nov 2023 a las 9:56, sfromis ***@***.***>) escribió:

…

[image: image] <https://user-images.githubusercontent.com/47082390/284584968-bbbd505f-2557-4697-a050-46c5dac658aa.png> — Reply to this email directly, view it on GitHub <#20067 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACXBW4PD6WCAIPTDAL3ZQHDYFSQIVAVCNFSM6AAAAAA7STJIOWVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TMMZQGM2DA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[barbudor]

Maybe kindly ask @digiblur if he would be able to help by identifying what would be the circuit

[ferbulous]

ssieb has got the 'magic mode' working on esphome, I've tested it and it works.
anyone mind taking a look at the custom component code and implement the change in tasmota?

external_components:
  - source: github://ssieb/custom_components@sswitch
    components: [ magic_switch ]
    refresh: 1min

magic_switch:
  pin: 5
  on_switch:
    - light.toggle: light_1

[sfromis]

The code is here: https://github.com/ssieb/custom_components/tree/sswitch/components/magic_switch
Uses interrupts, and pulse timing.

[lalo-uy]

It looks like they are getting interrups in pin 5 , and considering a trigger is there are a gap > 5ms. El El jue, 23 nov. 2023 a la(s) 07:45, ferbulous ***@***.***> escribió:

…

ssieb has got the 'magic mode' working on esphome, I've tested it and it works. anyone mind taking a look at the custom component code and implement the change in tasmota? external_components: - source: ***@***.*** components: [ magic_switch ] refresh: 1min magic_switch: pin: 5 on_switch: - light.toggle: light_1 — Reply to this email directly, view it on GitHub <#20067 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACXBW4IUTUIVVOPXRZNKC5TYF4SLJAVCNFSM6AAAAAA7STJIOWVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TMNJQGY4TG> . You are receiving this because you commented.Message ID: ***@***.***>

[barbudor]

Actually they trigger a toggle for a positive pulse greater than 12 ms (12000 µs)

A Switch input with proper debounce could work but the minimum supported debounce time is 40ms

[ferbulous]

Do you mind showing example how to add that configuration in tasmota?
Does it require setting up rules or berry (i'm not familar with berry though)

[barbudor]

I explicitly said that Tasmota minimum debounce is 40ms so pretty sure it would not work
But if you want to try, it should be something like:
GPIO5 : Switch_1
Switchmode1 4 but may be Switchmode1 3 as I'm confused by the docs
SwitchDebounce 42

[lalo-uy]

This is the PR: ssieb/esphome_components@72aa9e3 On a later one pulse change from 3000 to 5000. El El jue, 23 nov. 2023 a la(s) 15:10, Barbudor ***@***.***> escribió:

…

@lalo-uy <https://github.com/lalo-uy> could you please point me to that code so I can have a look? — Reply to this email directly, view it on GitHub <#20067 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACXBW4IITFY6JPWFRKOUCPDYF6GSTAVCNFSM6AAAAAA7STJIOWVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TMNJUGIZDM> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[lalo-uy]

I wonder what is in gpio5??. 5mS so not counting 50/60hz cicles……. May be the power supply oscilator falls that fast? No way, the zero crossing will be longer. El El jue, 23 nov. 2023 a la(s) 17:22, Eduardo Spremolla < ***@***.***> escribió:

…

This is the PR: ssieb/esphome_components@72aa9e3 On a later one pulse change from 3000 to 5000. El El jue, 23 nov. 2023 a la(s) 15:10, Barbudor ***@***.***> escribió: > @lalo-uy <https://github.com/lalo-uy> could you please point me to that > code so I can have a look? > > — > Reply to this email directly, view it on GitHub > <#20067 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/ACXBW4IITFY6JPWFRKOUCPDYF6GSTAVCNFSM6AAAAAA7STJIOWVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TMNJUGIZDM> > . > You are receiving this because you were mentioned.Message ID: > ***@***.***> >

[barbudor]

There must be some hardware on GPIO5
My guess would be a dual optocoupler such as SFH620 which acts as a dual-wave-rectifier.
May be something like that:

Or may be something a little more sophisticated with opto, zener to reduce the zero-crossing window

The 12ms threshold is probably to eliminate the recurring pulse when the wave is low

So it still requires the switch to provide a gap long enough while I would consider that a good quality switch would try to minimize the switching time to reduce risk of sparks

[barbudor]

That simplified diagram is unlikely to be the proper schematic as it would require a 25W resistor
And if we increase the resistor to reduce the current, the zero-crossing gap would become larger but not sure how much.
Doing that after a transformer on low voltage could not work as the spike due to the switching would probably be filtered by the transformer.

[Wintespe]

I simply took a "standard" mains switch from Berker (3036) and measured the time behaviour. The switching time is approx. 5 ms. It is easy to see in the oscillogram. I hope it helps with the design.

[barbudor]

Not having a BasicR4 in hand but using a similar setup as @Wintespe, I tried to play with Counter, CounterDebounceHigh.Low but I didn't any good success.

I have started a native driver to define a pin a "MagicSwitch" using similar code as ESPHome
I'm getting some good results but need to refine them.

[Wintespe]

Please test the behaviour of the different switches in a separate test setup as I have shown in an example. Only when the timing behaviour is known, it makes sense to decide to transfer it to Tasmota. A test in the Sonoff Basic can be overloaded with too many parameters.

[barbudor]

@ferbulous I've found a switch that works well and fix the solder on GPIO5 so I can now see the magic switch working and get the waveform
As I was suspecting the circuit is a zero-cross detector which generate a pulse every 10ms (100Hz). The pulse duration is below 1µs.

The pulse is when the signal is crossing zero and when the switch is flipped, as it cuts the power for a short time, a longer pulse occur.
However because of the recurrence, I cannot trigger on that (my very basic 50€ portable scope is not able to trigger on pulse duration, only on edge)

But flipping the switch multiple time I've able to see some different behavior

• pulse length can vary in large amount depending on the flip direction (apprently the switch doesn't flip at the same speed in both direction), depending on the way I'm pressing it (slowly or quickly), .... I have seen pulse that could be around 5ms and some close to 100ms (or more are they were out of screen)
• bouncing : I have seen cases of multiple pulses on a single flipping

So that give me some idea on how to harden the algorithm by adding a rest threshold to avoid multiple trigger
I'll try to move on by the week-end

[barbudor]

Hello @ferbulous ,
I've got a version working pretty well with one of the switches.
The other switch that doesn't work with the original firmware doesn't work neither with my version. I will do some measures on that one to fid out what is its switch time.

PR has been merged so you can get the lastest version from https://ota.tasmota.com/tasmota32/tasmota32c3.bin

• Default minimum pulse length is 4000. Can be changed with MagicSwitchPulse between 1000 and 500000 or by custom compile. If your switch doesn't work with 1000 it's unlikely it will ever work (even with original firmware). It's not possible to go below without risking to be triggered by the zero-cross pulse (arounds 600-650ms on my device).

• Once I have detected a suitable pulse, a masking window of around 300ms is applied (6 x 500ms loop) to avoid bouncing. Should help with bouncing switches. This can only be changed by custom compiling.

Remember that this is triggered by glitches in the mains. If your mains network has glitches, it may be triggered falsely.
Using MagicSwitchPulse to the highest value that still works with your switch could help filtering smallest power glitches. But if at some point your network has glitches in the same range as your switch flipping time, you are screwed with that feature.

[barbudor]

PR : #20247

PR merged, now feature is part of tasmota32c3

[ferbulous]

Huge thanks @barbudor
Will test it once I’m home
On my early test of esphome external component there were some false relay trigger that happens randomly (this was fixed after the author optimized it)
As for the switch not working, might need to contact the sonoff support team for that

[lalo-uy]

The point is how thet looks on 220 ac. El El vie, 24 nov. 2023 a la(s) 18:14, Wintespe ***@***.***> escribió:

…

I simply took a "standard" mains switch from Berker (3036) and measured the time behaviour. The switching time is approx. 5 ms. It is easy to see in the oscillogram. Approx. 5 msec switching time! I hope it helps with the design. Circut.png (view on web) <https://github.com/arendst/Tasmota/assets/121858068/7b88eded-3ee2-49d3-88a6-b6c6e9faf557> Magic.jpg (view on web) <https://github.com/arendst/Tasmota/assets/121858068/ff8266e0-3c73-46eb-ba97-761576c93b0a> Bereker3036.jpg (view on web) <https://github.com/arendst/Tasmota/assets/121858068/755b0c66-964d-4b5b-8c3b-60457c0fcd53> — Reply to this email directly, view it on GitHub <#20067 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACXBW4L3FUAUIBIILYK7SBTYGEE2BAVCNFSM6AAAAAA7STJIOWVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TMNRTGYYTA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[Wintespe]

lalo-uy I think that the time behaviour - the switching characteristic - of the switch is independent of the voltage applied.

[barbudor]

The behavior of the switch no, but the circuit that is behind is likely different than the simplified circuit we used for our test.
So it can impact the waveform and may shrink or stretch the pulse.
Also I would assume that zero crossing is also creating a pulse which length is unknown without having the real hardware in hand.

[Wintespe]

Correct, as long as the real hardware is not known, the input voltage is speculative.. However, this has nothing to do with the timing behaviour of the switch.

[lazy-pete]

Hi, I’ve flashed Tasmota to a BasicR4 and wired it in. Upgraded firmware OTA to the link posted above. Works fine over HAOS, but neither end of the two-way switching triggers the relay. Exactly how to I adjust the MagicSwitchPulse duration ? Code entered in Console ?

My missus is thinking muderous thoughts, and I’d love this to work as I’ve got a bunch of them thinking they’d solve my two-way switching issues that ZBMiniL2Extremes have failed to handle. I’m not very familiar with the insides of these, but can copy paste code if I know where it goes !

[barbudor]

@lazy-pete you can start with the smallest value
MagicSwitchPulse 1000
If that doesn't trigger the magic switch, then it's likely that your wall switch is to fast for that feature. Which is good in terms of security and durability (need to see the positive face of it)

[lazy-pete]

Thanks both - it’s actually stock firmware for the Sonoff ZBBridge I need to use EweLink again.

I tried MagicSwitchPulse 1000 but got the following message -

[ferbulous]

Is pin 05 assigned for magicswitch? Shouldn’t get that error

[lazy-pete]

Maybe not, unless it was default in the bin file further up. How do I add that ?

[ferbulous]

Configuration -> templates -> gpio05 as magicswitch

or use this template ,
Config-> Configure other, paste into template and select activate, save

{"NAME":"Sonoff Basic R4","GPIO":[0,0,0,0,224,10560,544,0,0,32,0,0,0,0,0,0,0,0,0,0,0,0],"FLAG":0,"BASE":1}

[lazy-pete]

Was set as user. Found MagicSwitch. I swear that wasn’t there before !

MagicSwitch 1

Pulse command now works. Switches work occasionally, which unfortunately is not good enough to keep the missus happy.

I presume from chat above that 1000 is the lowest limit ?

Not sure how I find ‘worse’ switches - these will be UK Screwfix standard I suspect (as in, the builders chose them - not fancier ones specified by us).

[barbudor]

@lazy-pete Here is a binary where the minimun pulse has been reduced to 100µs
So you can try values lower than 1000 but beware that starting around 700 it will be triggered by the zero-crossing detection
Let me know
tasmota32c3-magicswitch_low_pulse_limit.zip

[lazy-pete]

Hi @barbudor - I’ll try and give that a go this evening and report back.

[lazy-pete]

had a bash - anything below 1000 seems to result in the state bouncing once or twice after flicking the switch. Stable in between. Working my way back up to find the sweet spot - though it does seem to toggle with every switch now. I don't need to restart the module after each pulse time change do I ?

[barbudor]

You shouldn't restart between each change as the value is not saved to flash.
Once you figured out, make the setting autoset at boot with a boot-rule:

Rule1 on system#boot do MagicSwitchPulse nnnn endon
Rule1 1

That will restore your value in case of restart

[lazy-pete]

Great - thanks. I wonder if I initially inadvertently changed to 1000 then reset to default 4000 by rebooting before actually testing the switch. Would explain what I’m seeing now, which is encouraging as it means somewhere between 1000 and 4000 should work well ! Thanks again. Shall report back once fine tuned / stable.

[arunkandappan]

Does anyone know which type of switch to get? I see that sonoff says double pole rocker, but looking at the diagrams, it appears that what they are looking for is a 3-way switch like this: https://www.homedepot.com/p/Leviton-Decora-15-Amp-3-Way-Switch-White-R62-05603-2WS/100078412
With the two points shorted. Any thoughts?

[sfromis]

What matters is that you need a switch which will connect the input live wire to one of two outputs, with a brief pause while in motion. Then the "magic" trick is that this interruption will be detected.

Your link is dead to me....

[barbudor]

The exact désignation is a SPDT single pôle double throw
Or more commonly a 2-way switch

I think the diagrams are pretty clear

Not that the switch should not be of too high quality (not too fast) as the small loss of connection should still be more than 10ms in normal usage

[mojrs10]

My "magic switch" sometimes power on or off multiple times… it is not good.
What can I try to fix this?
Sonoff firmware use some owerhiting protection with ntc sensor. Is this working on Tasmota and can I do something to see ntc sensor temperature?

[sfromis]

You could try tweaking the setting for MagicSwitchPulse to work better with you physical switch.

If there's a NTC sensor, someone will have to work out what pin and electrical setup it uses. If having that, Tasmota does have support for an over-temperature threshold, SetOption42. Analog input on ESP32-C3 should be on one of gpio 0-4.