jackkelly / rfm_edf_ecomanager

Communicate with EDF EcoManager Transmitter Plugs; Whole House TX and CurrentCost Whole House TX using a 433MHz RFM12b and a ATmega328 micro controller

This code runs on a Nanode / Arduino with an HopeRF RFM12b 433MHz wireless module and wirelessly communicates with Current Cost home energy monitors and EDF EcoManager monitors. The main aim of this code is the robust gathering of data from an arbitrary number of EDF EcoManager Wireless Transmitter Plugs and a small number of Current Cost whole-house "sensable" transmitters for tracking the state of the majority of appliances in a building.

Related projects

My RFM_ecomanager_logger code logs data from rfm_edf_ecomanager on a PC.

Current status

The code is "feature complete" in that it implements all of the features listed in the user manual.

Known issues

The code doesn't work especially well with Current Cost Individual Appliance Monitors (CC IAMs). See this issue for more details.

More info

Please see the user manual for more info.

rfm_edf_ecomanager's Issues

Tidy up const variables

All const variables should probably be defined in consts.h rather than in the classes. Then add note to wiki about where to find consts.

Print state of EDF IAMs

Use linked list for rx buffer

Appending needs to be fast. Reading is always done in chronological order. No need to search.

Pairing with TRX crashes Nanode!

Feature: post to Cosm

In the pipeline, im sure, but ability to post to COSM would be useful.

double-check that checksum is checked for TRX packets

Add or modify id lists via serial

User should be able to:

• add single id to list,
• delete entire list
• add multiple ids
• get list of ids.

https://github.com/JackKelly/rfm_edf_ecomanager/wiki#wiki-Serial_commands

Output valid JSON

quotes. state should be 0 or 1 not on or off.

Debug code causes compilation errors

Pulled down the latest code after your updates of last night, but the debug code (to display output etc) now causing IDE (1.0.1 W8 64-bit) compilation errors again:

debug.cpp: In function 'void debug(DebugLevel, const char*, ...)':
debug.cpp:14: error: 'millis' was not declared in this scope
debug.cpp:14: error: 'sprintf' was not declared in this scope
debug.cpp:15: error: 'Serial' was not declared in this scope
debug.cpp:24: error: 'va_list' was not declared in this scope
debug.cpp:24: error: expected `;' before 'vl'
debug.cpp:25: error: 'vl' was not declared in this scope
debug.cpp:25: error: 'va_start' was not declared in this scope
debug.cpp:26: error: 'vsprintf' was not declared in this scope
debug.cpp:27: error: 'va_end' was not declared in this scope

Pairing function

2 modes:

auto_pair = true:
automatically accept all pairing requests (default)

auto_pair = false
require user to instruct pairing.

pair_id defaults to INVALID_ID
Constantly listen for pairing requests.
If request is received {
if pair_with != INVALID_ID then pair() else {
notify user over serial: {pr: } ("pr" is short for pairing request).
If auto_pair==true then set pair_with to .
}

If in auto_pair == false mode then if user wants to accept pairing request then respond with "{p: }" (short for "pair with "). Store a single id to pair with on Nanode in variable pair_with.

pair(){
if TRX {
transmit ACK;
store pair_with in cc_trx_ids;
} else {
store pair_with in cc_txs
}
Serial.print("{p: });
pair_with = INVALID_ID;
}

Why can't we receive one of my TXs even though the EnviR can receive it fine?

Print packet prior to demanchesterisation (just call print packet at start of demanchesterisation)

Script to flash new version

Check that logger isn't running.
Pull latest hex from git
Upload to Nanode

Listen for packets when not doing anything else

Measure mains voltage

To allow us to compensate for voltage variation.

For a guide of how to do the hardware: http://openenergymonitor.org/emon/node/58

Mark sensors as inactive if they go silent

We don't wait for inactive txs and we only poll inactive trxs once per roll call.

Print active status on list command

Occasional eta < millis reports

Learn period between transmissions for CC_TXs

Some CC_TXs transmit every 6100ms, some every 6050ms etc etc. Learn this from data.

Simplify and reduce redundancy in tx and trxs arrays

Use super class with just id and virtual get_id function.

Sub class this for txs.

Also a class for storing multiple sensors, with find function. Should cache last search result.

Why do we sometimes receive insane values for solar thermal?

sometimes:

32767 (= 2^15)
15360
12288
8192 (=2^13)
4096 (=2^12)

Why do we sometimes fail to receive TRXs?

If experiment with sending "ON" command to "prod" TRX works then:

• don't make this the default setting.
  • Instead add a command to en/disable this feature.
  • Modify logger to enable this at init.
  • Modify logger to filter double-responses.
  • Document.
• reduce retries

If experiment fails then try sending ACK instead.

Or try lengthening TRX timout

Check tuning for TRXs

Make absolutely sure that one missing TRX can't affect another TRX

Maybe the problem is fast clock recovery?

Reduce trx timeout time

Re-load wiki once "no block given" Gollum bug has been fixed

https://github.com/github/gollum/issues/601

CcTxs should take rolling av of sample period

Auto pair with TXs

When in auto-pair mode, automatically pair with all heard TXs (and TRXs?)

Move content from readme to wiki

Try making baseband tolerance wider and tinker with carrier frequency

EDF config uses stricter timings than EnviR. This might be why we're having trouble receiving current cost transmitters (issue #13)

Check if pairing with TRX has been successfull by attempting to poll it

Use dynamic, sorted array for trxs and txs

Appending happens very rarely so can be slow. Searching happens very very often so should be fast.

Template which can accept cc_tx and cc_trx. Subclass to add find_next_expect_tx

Update documentation to reflect slight API changes

Use preprocessor directives to switch debug printing on and off

Or create a debug() function, with a small set of priorities (INFO, WARN, ERROR, FATAL). Then, in the code, all I have to do it call debug(msg, WARN).

Use better names for "whole house transmitter"

All Transceivers rename to TRX, all transmitters only rename to TX

IDE compilation errors

Still getting errors (even after Jack's revision) at compile time. IDE 1.0.1 (and 1.0.0) on Windoze 8
Rfm12b.cpp.o: In function Rfm12b::interrupt_handler()': C:\Users\Andy\AppData\Local\Temp\build6281000564122575244.tmp/Rfm12b.cpp:78: undefined reference tospi::select(bool)'
C:\Users\Andy\AppData\Local\Temp\build6281000564122575244.tmp/Rfm12b.cpp:90: undefined reference to spi::select(bool)' C:\Users\Andy\AppData\Local\Temp\build6281000564122575244.tmp/Rfm12b.cpp:99: undefined reference tospi::select(bool)'

Tinker with 1/2-wave aerial

Turn on just a single TX.

Count number of received packets in 10 minutes.

Add 1/2 wave aerial. Count again.

Handle roll-over of millis()

in Manager::wait_for_cc_tx()

Doesn't poll TRXs reliably

{id: 639, t: 1725200, s0: 12, s2: 0} {id: 28, t: 1726609, s0: 502} {id: 639, t: 1731272, s0: 12, s2: 0} <-- 6072 ms since last poll (correct) {id: 28, t: 1732793, s0: 487} {id: 28, t: 1738977, s0: 492} {id: 28, t: 1745161, s0: 493} {id: 28, t: 1751314, s0: 503} {id: 28, t: 1757405, s0: 482} {id: 639, t: 1761426, s0: 12, s2: 0} <-- 30154ms since last poll (5 times too long!) {id: 28, t: 1763494, s0: 504} {id: 28, t: 1769585, s0: 499} {id: 639, t: 1773362, s0: 12, s2: 0} <-- 11936ms since last poll (2 times too long!) {id: 28, t: 1775707, s0: 497} {id: 639, t: 1779434, s0: 12, s2: 0} <-- 6072ms since last poll (correct) {id: 28, t: 1781860, s0: 508}

maybe this is an RF problem?

Don't continually retry IAMs

Optimise for size

Arduino creates much smaller binaries. Supposedly it's optimising for size. Use this setting in my makefile.

Talk to a single emonTX

As rapidly as possible. Allow for multiple parameters, e.g. real, reactive, voltage, 3 CT clamps, temp.

Try using unmodified emonTX code first.

Add option to display raw packets prior to checking

Maybe in post process function. Also print packet after demanchesterisation. And type of packet. Also mark which byte(s) failed demanchesterisation.

CC IAMs not being received.

I Thought Id had a problem with range/location of my Nanode vs IAMs, as not all data was being received.
However, I now believe Im not receiving ANY data from CC IAMs, but EDF IAMs ok, as are CC Optismart and Clamps.

Check compilation on Arduino on Linux and Windows and old Linux

How rapidly can we get new data from EDF IAMs?

currently assuming 6 seconds but need to check. Use different constants in code for tx and trx sample periods.

Tweak RF tuning to individual TXs

Use typedefs in consts.h throughout code

Only report data for sensors in local lists

Add proper header comments to each file

License, disclaimer, description etc

unit tests (running on PC)

Document compilation flags

Don't enable TUNING on a production system.

Guidance on opening the whole house transmitter.

Hi, thanks it's brilliant to have the photo's and that understanding of the wireless code. I'm hacking for an energy monitoring project with Horizon Digital Economy research institute at the University of Nottingham.

I find the transmitter box clicks apart easily except would not part at two points a third of the way down from the top. I can see now that it's due to two legs which fit into the two hollow pillars. By the feel of it I was expecting screws but it was clear that there weren't any so I assumed glue. Did you have any trouble parting the case at that point? Are those legs glued in perhaps and did you use any particular tool or technique to get in?

Fascinating that the black areal is fake - it explains the odd feel of its rotation.

timeout waiting for keyboard input