I wondered if it might be possible to replace the total number of minutes with the sum of the errors per minute divided by the sample time interval i.e. the number of 1 minute samples collected. Apologies if my impressions are not well founded.
I'm not quite sure what you are asking.
Could you perhaps give me an example of where you believe the error rate is too low when using your own stats?
I'm not a great fan of using averages as
a) doing so can mask peaks/troughs &
b) if the period is too small, it doesn't give a realistic indication of performance.
e.g. if my connection typically averages say 897 DS error seconds in a given 24 hour period, that would average out as 37.38 errors per hour or 0.62 per minute.
However, in the latest 12 hours of that given period, only 372 errors occurred.
Multiplying that by 2 would equate to only 744 errors per day, 31 per hour & 0.52 per minute
Using the latest 6 hours, 278 errors occurred.
Multiplied by 4, that would give 1112 errors per day, 46.33 per hour & 0.77 per minute.
This might help to explain how my graphing program calculates average error rates:-
Regardless of any time gaps in the data, when 24 hours is chosen, the latest 1440 samples are used
For 12 hours, the latest 720 samples are used
For 1 hour, the latest 60 samples are used.
So, lets say there was a gap of around 4 days between pausing & resuming logging.
The latest 1440 samples would still be used for a 1 day graph, but the start date would be clearly incorrect:-
You entered (or it was determined) a period of 1440 m
*** Using the specified log file ***:-
C:\HG612 Modem Stats with Spaces\Ongoing Stats\TEST\modem_stats.log
The whole log file contains 412740 rows
DSL_MODE = VDSL2
The graphs will be plotted using 1440 samples
From 21/10/2014 14:57 to 26/10/2014 15:58
i.e. 1 Day
Please wait......
* DS & US SNRM
** DS & US SYNC
*** CRC
**** FEC
***** DS HEC
***** US HEC
****** Up Time
******* DS RSCorr.................................
The number of errors per sample are indeed summed for the latest 1440 samples used for the graphs & divided by 24 for the hourly averages & divided by 1440 for the averages per minute.
That method can also be somewhat misleading when attempting to calculate averages etc.
The only realistic method appears to be to continue the logging 24/7, which will include peaks/troughs for day-time/night-time periods & any other periods where error counts can fluctuate due to external influences/interference etc.