[TowerTalk] Static Discharge Porcupines - great for....
Jim Lux
jimlux at earthlink.net
Sun Aug 9 12:19:16 EDT 2015
On 8/8/15 6:11 PM, Roger (K8RI) on TT wrote:
>
> I'm glad Kim posted that information, but I wish the scientific
> community would go back to averages rather than median. The average is
> much more meaningful than knowing the number that lies half way between
> the highest and lowest figure measured, at least they are for me. Median
> is an interesting number, but average seems to be much more informative
> and typically what is used for design. With a number of samples large
> enough to be statistically valid a single, significant outlier, be it
> high or low can substantially skew the median, but have little effect on
> the average.
Actually, the median is the one you want: half the strokes are higher,
half the strokes are lower. it is not skewed as much by a outlier.
The mean (arithmetic sum divided by the number of entries) would be
skewed by a single very large number.
In practice, what you'd really want is something like the 95th or 99th
percentile. And that's something that's in the standard scientific
literature. I think pretty much any of Uman's books has a figure in it
of stroke current distribution.
http://www.iclp-centre.org/pdf/Invited-Lecture-3.pdf
has a lot of interesting charts..
On page 10, it shows that 50% of the first strokes are around 35 kA or
less, and 90% are <100kA peak.
http://surgelogic.com/documents/technotes/Lightning_Stroke_Tech_Note_1300DB1101.pdf
says 95% of positive strokes <30kA, 98%<60kA
for negative strokes, 82%<30kZ, 98%<60kA
>
> The so called "super strikes", or Positive lightening, which is
> associated with sprites can move the median, but happen so seldom, they
> have little effect on the average and when it come to lightening, do we
> design for the median, average, maximum, or the best we can afford?
Design for what you can afford to lose. I suspect that designing for the
99th percentile is inappropriate for hams..
More information about the TowerTalk
mailing list