Started to flesh out the MiHome receive pipeline

Spec'ed interface for OOK decoder

Deprecated Messages due to import race with Devices. ...

Template messages (which are cloned with create_message) now in Devices.

Commissioned a tx path for MiHome adaptor Plus. ...

This is a temporary path, but it generates the correct message

LegacyDevice tx path commissioned end to end

Most of the OnAir interface coded or spec'ed now

Implementation checklist added for OnAir

Method signatures defined for OnAir adaptors

Spec'ed out responsibilities of OnAir interfaces

Placeholder added to architecture for radio interface adaptors

More design on the tx/rx and air_interface pipeline

Renamed encoder as TwoBit, as it encodes two bits per byte. ...

This is the encoding protocol used with legacy OOK devices

Started to plan out the tx/rx pipeline for payloads and message buffers

Initial design of a message Router

Capabilities of devices turned into flags, with query method in base class

Mock registry store created, auto_create is working

Documented requirement for a simple persistent data store for the Registry

Started to implement device classes code. ...

Simple mocking interface now working.

Method signatures for Registry sketched out

Initial design for enhanced persistent registry done

Some ideas about device registration added

Method signatures for device classes added

Started to flesh out device classes proposed interfaces

Python 3 now supported. ...

Logger now stores apparent power and real power headings.
Logger now stores temperature from eTRV in csv file

Updated device database comments for MIHO002 device

Now displays receive time of incoming messages

Noted the layering of switching between tx/rx might be sub optimal

Fine tuned settings when sharing MiHome and legacy devices

Renamed new radio2 module as radio, and removed old code

Removed any remaining spi-direct and tidied the build process. ...

Tested monitor.py, switch.py and legacy.py on Raspberry Pi.
All work from the same driver now.

Removed legacy direct-to-SPI interface. ...

This code now does FSK tx/rx and OOK tx from the same driver.

Tested FSK tx and rx on Pi, works fine ...

Tested both monitor.py and switch.py

Fixed compile typos. ...

Build test done on mac

Simplified the choice between len and cbp receive

Updated TODO list for next steps of radio.c

Implemented radio.c/radio_get_payload ...

Not yet tested

Implemented a better HRF_readfifo_burst. ...

This has facilities to read a fixed payload, or a count byte preceeded payload.

Framework for FSK receive in place

Started working through new FSK receiver architecture

Tested with MiHome adaptor plus, new FSK transmitter works fine

FSK Transmit coded, ready for test

Restructured switch.py so we have a tx-only test. ...

This is in preparation for pushing the FSK transmitter into radio.c

Refactored so that we can create a tx-only switch loop tester. ...

This will be useful for integrating a new FSK transmitter without receiver

Tested on Raspberry Pi, switch, monitor, and legacy work with spi scaffold

radio2.py offers temporary spi interface for radio.py ...

Tested on mac only.

Build of radio_rpi.so updated

Removed experimental arduino work

Fixed failed merge

Merge branch 'long-tx'

Improved configurability of message repeat patterns. ...

Tested with different repeat lengths.
Optimum seems to be 8 inner, 1 outer transmits

Turned off TRACE on raspberry pi build

Tested with learning mode. ...

Works with all 4 switch patterns.
Can now independently switch 4 switches.
Some occasional RF reliability problems that need measuring.

Implemented mode switching logic. ...

This is so you can just say radio_standby()
Then radio_transmit() and it pops to standby at end of transmit.
Or, radio_receive()
Then radio_transmit() pops back to receive at the end.
Also massive tidy up and trimming out old comments.

Improved flow of code, specifically radio_transmit() ...

Pushes previous mode and restores at end of transmit.
This should auto-return to standby or receive as appropriate.
This will make it easier to implement mode hopping schemes later.

Simulation gets further now due to gpio_sim detection

Started design of new long payload transmitter

Initial design of new radio transmit scheme

radio_rpi.so now built as part of the pi build process ...

legacy.py now uses the prebuilt radio_rpi.so

encoder now adds preamble. ...

lecacy.py now uses new C radio driver

Tested radio2.py on Raspberry Pi, now works with OOK transmit

Tiny tidy up

Renamed encoder_test to be consistent with other testers

Wrote radio2_test.py as a python OOK tester for new radio.so lib

radio2.py ctypes wrapper ready for testing

Preparations for new radio interface added to the architecture

Tweaks to radio_test.c layout

Tidied up radio_test.c readability

Removed unnecessary spurious radio configures

Fixed payloadlen bug that was introduced in API refactoring ...

Note: Reconfigures as OOK regularly, unneccesarily now

Fixed bug with OOK struct size in hrfm69_test.c

Finished coding radio_test and radio(tx) module

First pass of spec'ing radio.h/c and radio_test.c

Can now reliably turn energenie switches on and off with hrfm69_test.c

Coded a better payload transmit test routine for OOK

Started coding up new burst transmission scheme

Started to write another test case to blurt out OOK energenie messages. ...

This is only a test case.
But it has highlighted a whole load of discrepancies in the original C
code to do with payload length calculation, and the 'magic' zero byte
at the start of the buffer for stuffing the SPI register address in.

Tracked down the 'long sync bit' problem. ...

It's due to the HRF always preceeding each packed with an extra 01 at start.
Existing driver code sends 'chips' to prove this.
Looking like a preamble of zeros will be required to make this a non issue,
and then generate the real 80000000 preamble as part of the payload.

Got the radio chirping again after a long period of not working at all

hrfm69_test.c now sending a tone. ...

Discovered that 1/5 times it sticks waiting for TXREADY.
This might suggest an unresolved sequencing error in startup.

spi_test.c can now read back the version number of the HRF69 chip

Simple GPIO tester works on Raspberry Pi (flashes LEDs on radio)

Built a simple C test harness to start testing the radio with

Baud rate for OOK increased to 4800bps to bring in line with latest Energenie builds. ...

Sync pattern now recognised correctly as 0x80,0x00,0x00,0x00
Switching is eratic, but learning always learns.

Added green LED around transmit and receive transactions. ...

This is useful for measuring total on-air time on logic analyser

Experiments on Arduino with larger register writes and FIFO write ...

Getting some corruption on the Arduino RAM buffers.
Logic analyser traces show that the tables are getting corrupted
and the wrong data is being written to the radio registers.

Managed to write/read the mode register from real RFM69 with Arduino

Reset, select, write working on Arduino

SPI being driven on Arduino

Got simple 5 pin interface with GPIO working on Arduino Pro Micro

Preparations for wiring up to Arduino

A 'go for broke' OOK transmitter added to hrf69_test

Implemented most of hrf69.c write and read functions

HRF_write_reg written

Factored out all stdio from all testers and code

Mock read and write test working for radio register

Proved that we can read a register back from simulation

First pass at writing a test in hrf69_test.c

Marked the interface queries in HRF69.h

First attempt at specing out implementation of HRF69 based on python

spi_test working on Mac

Refactored delay into a platform module

gpio_sim working on Mac

Wrote GPIO simulator in a mostly platform independent way.

Framework in place for Arduino GPIO and simulated GPIO

GPIO is now tagged as RPi specific in build process

Created file architecture for arduino build

Started refactoring

Moved all C into a drv folder.

First pass at defining hrf69.h and radio.h interfaces

Started to refactor code into C

Pre-factoring work done prior to pushing more code into the C domain.

Tweaked payload format further. ...

Reset all devices and re-learnt the house code
Can now switch all devices on and off, together, and independently.
legacy.py test code now demonstrates this in a loop

Added a test mode to test all 16 patterns of legacy control messages. ...

Run legacy.py and try numbers 0 to F in learning mode.
In theory you should be able to select 4 different switch addresses
plus an 'all addresses', but I can't get this to work on my 7 plugs,
they all respond as switch number 1 regardless of what code I send
in learning mode.

Attempts to get multiple device addresses working. ...

Seems to be a discrepancy between documentation and sample C code.
Stopping here until I get this clarified.

Renamed modulator as encoder

7 legacy adaptors now turning on and off with this trimmed code

Improved reliability of syncing of OOK messages

Got old code working again. ...

Looks like the extra bytes inserted by the buffer bug actually make it work.
This is probably related to sync bytes

Merge branch 'master' of https://github.com/whaleygeek/pyenergenie

Brought back old test code that was working

More experiments with payload timing

Corrected sync preamble. ...

Brought in original legacy C code for reference

Tweaks to legacy learn feature

Rewrote legacy.py to use new modulation code. ...

Not yet tested on a real Raspberry Pi

Refactored modulator into a standalone module

Modulator tested and working

Implemented a standalone switch exerciser demo. ...

Note this only really works with Adaptor Plus
(unless you seed the directory with your device id's first)

Started work on OOK support. ...

Ported C code into radio.py
Split out monitor and switch. Monitor.py is now just a monitor.
Switch.py needs more work to make it just a switch.
legacy.py will be a demo app for legacy (green button) devices.

Updated device database

Added rudimentary support for MIHO005 (HOUSE MONITOR) and MIHO013 (ETRV) ...

Fixed bug where eTRV sends an empty JOIN message that crashed the code
Added some extra fields to the CSV to support apparent power for House Monitor
Note: Voltage scaling on house monitor is incorrect.

Removed and tweaked a lot of trace messages to make it less noisy

Added RESET at start to reset the radio to a known clean state. ...

Added a SWITCH ON/OFF message every 5 seconds.
Re-tested and reports now seem to be a reliable 10 seconds.

Improved message decoding. ...

Now builds a dictionary of seen devices.
Monitor FFFFFF messages removed, not needed.
Catches bad payloads and prints a warning rather than crashing.
Accepts and responds to a JOIN command.
Message and value encoders written and working.
Message and value decoders written and working.

Message decoder added. ...

Messages polled at a faster rate.
Now see unsolicited switch change messages from MiHome AdapterPlus.
Removed a lot of debug messages.
Real value decoding now working.

Improved encoding and decoding of OpenHEMS payloads. ...

Now uses pydict structures in the user app.
Encoding, encryption, decryption, and decoding done inside OpenHEMS.
This simplifies the application interface.

Added source files from first working tests