A 2 years ago I've found that there is a lack of possibility for updating input value when GA has different value than input (changed by other source or telegram failed because of traffic peak).
This is bad but not terrible because it is possibility to read from db what is current state of input (not from knx bus) and send it cyclically to the bus. I am doing this for very important inputs especially for pool heating.
But unfortunately today I've found worst case that probably inputs are event-based. For me input can update its state with some innertion because of CPU usage. But EVERY input should have to update properly its internal state. It is totally unnacceptable that device shows bad state of its internal input. But today I've found that there is a possibility that LM5 can do such thing.
I had no-voltage contact from pool heater controller which close when there is a demand for heating and it switches LM input state. And LM sends telegram to switch KNX output in actuator.
And today I've found that LM had internal bad input state for 3 days and it doeasn't update it in this time. So I've tested if input would change state when I will toggle it so firstly I've disconnected cable from real contact, and nothing changed because contact where opened but on LM it was closed. So then I've simmulated its closing on LM, nothing changed but when I've then again open input then finally it updates properly its internal state and also sends telegram to the bus.
I am not sure what is the direct reason and please tell me how input should work in such edge cases:
- when backup is loaded? Does input get state from backup and not change its value in db untiil some event happen (change input state electronically),
- what happens when input state GA is send by another device? Is it not writable or writable and the value in db will change? (It is easy to check but I cannot do it today)
Another question: is there a possibility to check very low level what is true binary input state? I mean lower level than:
"SELECT name, address, updatetime, value FROM lm5ex_rio"
I very like LM but this part of it has very bad implementation. Reliability of I/O state inside device is foundations of stable system. Of course KNX TP adding some uncertainty but this is understandable but I/O inside device?
The device in which such situation occured has firmware 20180828 and it is LM5 with multiple IOs.
This is bad but not terrible because it is possibility to read from db what is current state of input (not from knx bus) and send it cyclically to the bus. I am doing this for very important inputs especially for pool heating.
But unfortunately today I've found worst case that probably inputs are event-based. For me input can update its state with some innertion because of CPU usage. But EVERY input should have to update properly its internal state. It is totally unnacceptable that device shows bad state of its internal input. But today I've found that there is a possibility that LM5 can do such thing.
I had no-voltage contact from pool heater controller which close when there is a demand for heating and it switches LM input state. And LM sends telegram to switch KNX output in actuator.
And today I've found that LM had internal bad input state for 3 days and it doeasn't update it in this time. So I've tested if input would change state when I will toggle it so firstly I've disconnected cable from real contact, and nothing changed because contact where opened but on LM it was closed. So then I've simmulated its closing on LM, nothing changed but when I've then again open input then finally it updates properly its internal state and also sends telegram to the bus.
I am not sure what is the direct reason and please tell me how input should work in such edge cases:
- when backup is loaded? Does input get state from backup and not change its value in db untiil some event happen (change input state electronically),
- what happens when input state GA is send by another device? Is it not writable or writable and the value in db will change? (It is easy to check but I cannot do it today)
Another question: is there a possibility to check very low level what is true binary input state? I mean lower level than:
"SELECT name, address, updatetime, value FROM lm5ex_rio"
I very like LM but this part of it has very bad implementation. Reliability of I/O state inside device is foundations of stable system. Of course KNX TP adding some uncertainty but this is understandable but I/O inside device?
The device in which such situation occured has firmware 20180828 and it is LM5 with multiple IOs.
Done is better than perfect