Detail for my Mk2_RFdatalog_4b.ino sketch

This is an upgraded version of the original Mk2 sketch for use with my
PCB-based hardware.  It supports two current sensors, CT1 and CT2.  
CT1 is for monitoring the flow of energy at the supply point; CT2 is 
available for monitoring the flow of current to the dump-load.  

Datalog messages are routinely transmitted by the on-board RF facility.  
By using the pin-saving hardware option, the 4-digit display is still 
available for use.

The _2 release includes various changes since the _1 version:

- The transmitted data now only has two integer fields (no message ID):
  . the power at the supply point in Joules (import is +ve, export is -ve)
  . the diverted energy total for today, in kWh
  (note the change in energy sense to match the OEM convention)

- a REQUIRED_EXPORT_IN_WATTS facility has been added.  This acts like a 
  leak in the energy bucket, therefore reducing the amount of surplus power 
  that can be diverted.  When a negative value is entered, this facility
  acts like a PV Simulator, which can be very helful for test purposes.

- The state of the energy bucket is displayed at the Serial Monitor every 
  second in Joules.

- The RFM12B is no longer sent to sleep between RF transmissions, it remains
  awake throughout.  The library call rf12_sendStart() has been replaced by
  rf12_sendNow().  These changes are intended to minimise any disruption
  to the continuous sampling process when RF transmissions are sent.

Changes for version _3:

- For compatibility with other versions of the Mk2 code, the variable cycleCount 
  has been removed.  This variable would have eventually overflowed which 
  could have caused unpredictable effects with other versions of the Mk2 code.

Changes for version _4:

- Code restructured so that all of the main activities are performed within the 
  Interrupt Service Routine.  This includes all of the sampling and power diversion
  functions.  Only the slower activities are dealt with by the main code, in loop().

- A checker mechanism records the minimum number of sample sets per mains cycle.  
  For 50 Hz, this should always be 63 or 64.   {20000us / (104us * 3) = 64.1}
  For 60 Hz, I presume it should always be 53.   {16666us / (104us * 3) = 53.4}

- temperature sensing is now supported using a Dallas OneWire sensor at the "mode" port.
- the output mode is now fixed at compile time (unless the "mode" port switch is available)
- the ADC is now free-running rather than being controlled by a fixed rate timer
- a persistence check had been added for the zero-crossing detector
- the payload + check data are displayed to the screen whenever a datalog message is sent 
 (for more details, please consult the code)

- In version_2, an inversion was introduced so that the reported power at the 
  supply point would be in line with the Open Energy Monitor convention (whereby
  consumption is positive).  Unfortunately, this modification has been lost during 
  the version_3 to version_4 upgrade.  This change can be easily reinstated by 
  adding after Line #525:  tx_data.powerAtSupplyPoint *= -1;

Changes for version _4a:

- During the major restructuring from version _3 to _4, two minor problems were introduced.  
  These have both been fixed by the upgrade to version _4a:

- In version 4, the phaseCal calculation was ineffective because two assignment statements 
  within the restructured ISR routine were in the wrong order.  The order of these statements
  has now been changed so that the phaseCal refinement will again work as intended.    

- The inversion of the reported power at the supply point has been reinstated. The reported 
  power value is now in line with the Open Energy Monitor convention, whereby import is 
  positive and export is negative.

- the display timeout period has been reduced to 8 hours instead of 10.

Changes for version _4b:
- The variables to store copies of ADC data for use by the main code are now declared as 
  "volatile" to remove any possibility of incorrect operation due to optimisation by the 
  compiler.