* This is the first attempt at decoding the irda, so is rough around the edges
* isn't correct in a few places. But it is a good resting place for the inital
* scripts that extracted the test data and got me started on just what I was
* looking at. Biggest difference between this and website is... the frames start at different positions, here the checksum is at byte 100, not 110.

* So far the conclusions are:
* 0.  Keying fields(bytes) relative to the model/make Elster A100C, where the E is $1
* 1.  Fields $40 to $44 are WattHours. The hex digits just use the numeric digits
*     to denote the value (check).
* 2.  Fields $70 and $71 are related to power usage, possibly an instantaineous
*     measure of power usage. But, $70 spends most of its time at a large value
*     and $71 spends most time at 0, suggesting they somehow combine to make a 2 byte
*     floating point number. 
* 3.  $78, $83, $84 and probably earlier fields record meter runtime. The finest
*     resolution is the hour, stored in $84 (overflowing to $85) and $75.
# 4.  $100 is a checksum, the sum of all bytes except itself mod 256.


Grabbed a hexdump of the irda output using:
hexdump -C /dev/ttyS1
having already run minicom to set the parameters as 2400 baud, 7E1.
The meter was saying 6824 units used, on 24:04, 5th April, 2009
<hex dump deleted>
As it wasnt clear what, if anything the encoded stuff meant, decided to narrow down whats of interest by looking again when the units had changed.
Coming back +2 units later, is the output any different? Defo each repeat of above is a repeat.
<hex dump deleted>
And again, after a small amount of time has passed while checking over the above.  With meter now reading 6268
<hex dump deleted>

# grabusage.txt used 7 bit codes, but now I know it uses 8 bit which is grabusage2.txt
hexdump -C /dev/ttyS1 | awk '/^[0-9]/ {for(i=2;i<=17;i++){k=strtonum(sprintf("0x%s",$i)); v=k; if(v<32 || v>127)v="?"; printf "%d\t%c\n",k,v;} }' | awk -FA"`printf "\t"`" -- 'BEGIN {m=1;} {if( $2==substr("Elster A100C",m,1) )m++; else m==1; if(m>=8){printf "Restart: %d %d\n",fl,s >> "/dev/stderr"; fl=0; s++; printf "\n"; fflush(); m=1;} printf " %02x",$1; fl++; }' >> grabusage2.txt

# convert grabusage2.txt into lines showing only the values that have changed
# also, only count the line at all if it passes the checksum, which is a sum of whole line but #100, mod 256
tail -n 999999999 -f grabusage2.txt | awk -- 'BEGIN {show[79]=1;} {printf "%d  ",NF; sum=0; for(i=1;i<=NF;i++)if(i!=100)sum+=strtonum(sprintf("0x%s",$i)); if(NF==110 && (sum%256)==strtonum(sprintf("0x%s",$100))){ for(i=1;i<=NF;i++){ noshow[i]=show[i]; if((i in was) && was[i]!=$i)show[i]=1; if(show[i]){ if(!noshow[i] )printf " %d->%s->%s",i,was[i],$i; else printf " %d->%s",i,$i; } was[i]=$i; } } printf "  0x%s?=0x%02x\n",$100,sum%256;}' | less

# convert key columns. #40->x*Wh, .., #43->100*Wh, #44->Wh #70->poss voltage, #78->poss related to unit used, #83 & 84, counters keyed from #78, or maybe #78 is keyed from this?, #100->checksum, sum of whole line but itself
# Suspect the bits have been scattered throughout the bit space, which is why some of it makes no sense.
awk -- < grabusage2.txt 'function h2d(a){return strtonum(sprintf("0x%s",a));} { sum=0; for(i=1;i<=NF;i++)if(i!=100)sum+=strtonum(sprintf("0x%s",$i));  if(NF==110 && (sum%256)==strtonum(sprintf("0x%s",$100))){printf "%d\t%d\t%d",NR,$42*10000+$43*100+$44,h2d($70); } printf "\n"; }' > delme.plot

echo "plot \"delme.plot\" using 1:3" | gnuplot -persist

# average out WattHours counter, seemingly the most high res counter available...
# this converts the units into 'units' per second.
awk < delme.plot '{if($2!=last)printf "%s\n",$0; last=$2;}' | awk -- '{if(0)printf "Original: %d\t%d\n",$1,$2; dtdp=($2-lp)/($1-lt); if(lt>0)for(t=lt+1;t<=$1;t++){printf "%d\t%0.3f\n",t,dtdp;} lt=$1; lp=$2; }' > delme2.plot

# above was a tad noisy, average out of 60 seconds - at least, im thinking the updates
# are per second, running for just over day there is just over a days worth of entries.
awk -- '{ if(sums>=60){printf "%d\t%0.3f\n",NR,avg/sums; sums=0; avg=0; } sums++; avg+=$2;}' < delme2.plot > delme3.plot

echo "plot \"delme3.plot\" using 1:2" | gnuplot -persist


# do the batch processing of fresh results and upload to abatis
bash bin/irdaconv.sh ; find irda/ -name "*.plot3" | xargs cat > delme.plot ; scp delme.plot abatis.flint:.


# extract the unknown rows, #70 and #71, and plot each bit of #70 seperatly, and #71 (which is only ever 4 values)
# this hasn't been converted to the new time appended format - have reproduced this in irdaconv.sh and changed to new format.
awk -- < grabusage2.txt 'function h2d(a){return strtonum(sprintf("0x%s",a));} { sum=0; for(i=1;i<=NF;i++)if(i!=100)sum+=strtonum(sprintf("0x%s",$i));  if(NF==110 && (sum%256)==strtonum(sprintf("0x%s",$100))){n=strtonum(sprintf("0x%s",$70)); printf "%d\t%d\t %d %d %d %d  %d %d %d %d\t%d\t%0.3ff",NR,$42*10000+$43*100+$44,((n/128)%2)+16,((n/64)%2)+14,((n/32)%2)+12,((n/16)%2)+10,((n/8)%2)+8,((n/4)%2)+6,((n/2)%2)+4,((n/1)%2)+2,h2d($71),n*(2^(h2d($71)/32)); } printf "\n"; }' > delme.plot

# plot the 8 individual bits
echo -e "plot [:] [:] \"delme.plot\" using 1:3 with lines, \"delme.plot\" using 1:4 with lines, \"delme.plot\" using 1:5 with lines, \"delme.plot\" using 1:6 with lines, \"delme.plot\" using 1:7 with lines, \"delme.plot\" using 1:8 with lines, \"delme.plot\" using 1:9 with lines, \"delme.plot\" using 1:9 with lines, \"delme.plot\" using 1:10 with lines, \"delme.plot\" using 1:(\$11/32+20) with lines,  \"delme.plot\" using 1:(\$12/32+25) with lines" | gnuplot -persist