Installation with E3DC S10 (SunSpec mode) does not work

[SaschaHenning]

Version of the custom_component

0.0.18

Configuration

Not available since installation of intregration does not work.

Describe the bug

System:

• E3DC S10 ( https://www.e3dc.com )
• Modbus Mode: Sunspec ( http://sunspec.org/wp-content/uploads/2016/02/SunSpec-Information-Model-Reference.xlsx )
• Configuration: https://www.loxwiki.eu/download/attachments/9339566/ModuBus-Dokumentation_2016-08-05.pdf?version=1&modificationDate=1471507657000&api=v2 -> see chapter 3.2 for sunspec config

Error occurs when trying to add the integration on "/config/integrations". It cannot install due to Failed to connect to host 192.168.0.245:502 slave 1 - Repeating group count not consistent with model length for model 803,model repeating len = 6, model repeating group len = 32(see logs below)

Did anyone successfully use this integration for an E3DC device?

It works fine when I use the modbus integration for direct register access. Here a part of my configuration (for details compare with excel above):

modbus:
  - name: E3DC
    type: tcp
    host: 192.168.0.245
    port: 502
    sensors:
      - name: e3dc_meter_home_incoming
        unit_of_measurement: kWh
        address: 40248 #40203+46-1
        slave: 1
        device_class: energy
        count: 2
        scan_interval: 15
        data_type: uint32
        scale: 0.001
        precision: 2
      - name: e3dc_meter_home_exporting
        unit_of_measurement: kWh
        address: 40240 #40203+38-1
        slave: 1
        device_class: energy
        count: 2
        scan_interval: 15
        data_type: uint32
        scale: 0.001
        precision: 2
...

Debug log

2021-11-13 21:03:09 DEBUG (MainThread) [custom_components.sunspec] New SunspecApi Client
2021-11-13 21:03:09 DEBUG (SyncWorker_1) [custom_components.sunspec] Client connect to IP 192.168.0.245 port 502 slave id 0 using timeout 60
2021-11-13 21:03:09 DEBUG (SyncWorker_1) [custom_components.sunspec] Client connected, perform initial scan
2021-11-13 21:03:09 ERROR (MainThread) [custom_components.sunspec] Failed to connect to host 192.168.0.245:502 slave 1 - Repeating group count not consistent with model length for model 803,model repeating len = 6, model repeating group len = 32

[CJNE]

Hi @SaschaHenning
There is a good chance that this problem has been fixed by the newly released pySunSpec2 release 1.0.6.
The latest release of this integration uses that version so you might have better luck using that?

[SaschaHenning]

No changes -> tried again today

[CJNE]

OK, then the best chance to get this resolved is to file an issue at pySunSpec 2, the library we rely on for this.
https://github.com/sunspec/pysunspec2/issues
Include that error message about repeating group, it might be a bug in the SunSpec model, or in the inverter firmware.

[CJNE]

Try the latest release, it's using an updated version of pySunSpec2 that might fix your problem.

[SaschaHenning]

Sorry, still no change.

2022-02-07 13:08:40 DEBUG (MainThread) [custom_components.sunspec] New SunspecApi Client
2022-02-07 13:08:40 DEBUG (SyncWorker_8) [custom_components.sunspec] Client connect to IP 192.168.0.245 port 502 slave id 1 using timeout 60
2022-02-07 13:08:40 DEBUG (SyncWorker_8) [custom_components.sunspec] Client connected, perform initial scan
2022-02-07 13:08:40 ERROR (MainThread) [custom_components.sunspec] Failed to connect to host 192.168.0.245:502 slave 1 - Repeating group count not consistent with model length for model 803,model repeating len = 6, model repeating group len = 32

[CJNE]

That's too bad, I had some hope that it would have been fixed. Can you create a bug report over there?

[siku2]

I recently decided to investigate this a bit, because the S10 actually exposes more information in the sunspec mode than in the proprietary E3DC mode.

As it turns out, sunspec isn't really great at backwards-compatibility. The S10 uses the spec from 2016 and the models for 801, 802, and 803 have changed significantly (and in a backwards-incompatible way) since then.
If you manually change the pysunspec2's models submodule all the way back to a commit in 2016 the S10 works out of the box.
I don't think sunspec really considers this a bug or even something that should be supported. The library isn't designed to dynamically swap these model definitions, so I really don't see a way for this to be fixed.

I might create a separate integration specifically for the S10's sunspec mode in the future to handle this.