K1TTT MOF/LOF Propagation Program


Historical notes on development of K1TTT-MOF algorithms.

History:
About 1987
Start from MINIMUF85 formulas.  These have many obvious shortfalls due
to the limited dataset they were derived from.  They were derived and
tested for a relatively small set of data.  The MINIMUF85 update to 
MINIMUF3.5 was done by curve fits to observed MUF data over several 
paths and by adding fudge factors into the original model equations.
This seemed to obscure the meaning of several of them since the fudge
factors started to override the original meaning of the formulas. 

My original work on my old MUF prediction program was essentially just
basic groundwork for getting the program working.   I developed the
world maps for MUF/LUF display, the method for viewing intermediate
calculations, and the basics of comparing(manually) a prediction with
an actual contest log.
I adapt a LUF program from a SW broadcast predictor for the lower
frequency bands.

1991 or so
After adding my own adjustments to the MUF fudge factors I decide that
the model has been pushed to its limit.  Its time to rip out some of 
the old stuff and replace with cleaned up equations and some different
physical models.  My MUF program is put on the shelf and converted to
MOF.  Detailed reading and thought on the prediction of band openings
for contest operations has shown that the MUF model just isn't useful
for what we need.  It appears from studying several contests and 
comparing to the MINIMUF model and other programs that predict MUF 
using other methods that we are often operating well above the predicted
MUF.  Granted, signal are weak, and by the classic definition of MUF
we could not pass traffic reliably or meet a minimum signal strength
measurement, but we can make contact and exchange contest data.
It becomes obvious that we are actually operating near the MOF, the 
Maximum Observed Frequency, that can be up to 50% higher than the MUF.
Thus I take that name for my program.  The same applies for the Lowest
Observed Frequency, although this seems less common in the literature.

I have been studying the sections of the book "Ionospheric Radio" by
Kenneth Davies that draws heavily on the works of others to present an
overview of radio wave propagation through the Ionosphere.  

Refer to the the MOFCALCS.TXT file for the details of how the parts fit
together, but basically these are the major changes I mad to arrive
at the MOF model.

1. While the shape of the critical frequency plot seemed reasonable the
absolute values of the max and min were way off.  I did extensive 
curve fitting of data taken from plots of the noon time critical 
frequencies at Washington DC (Ionospheric Radio pgs 134-135) vs sunspot
numbers from 1947 to 1986.  My analysis showed the same basic shape as
Davies develops on pg 136 for the maximum values.  I also develop a fit
for the annual variations to account for the "Winter Anomaly" he 
discusses starting on page 135.  These are put into the g2a (f0F2)
model to properly scale the ranges of critical frequency.

2. I add an option to use the 'real' magnetic dip values from Davies
pg 44 world map vs the simplistic magnetic lat/lon formulas in the 
MINIMUF algorithms.

3. The 'Control Point' selection in MIMIMUF was tailored mostly to 
mid-range paths, pretty much 2000-6000km if I remember correctly.  Much
beyond those ranges and it seemed to fall apart.  I generalized the
algorithm to more evenly distribute the Ionospheric Reflection Points
along the path.  While this slowed the calculations by adding more 
points in most cases the predictions seemed much better and adjustments
to the critical frequency produced more predictable changes since it
is sampled at regular intervals along the path.

4. A less obvious problem was the MINIMUF calculation of the Ionosphere
height used to figure hop lengths and reflection angles.  I did some
work to make sure that the results fell in a reasonable range based
on Davies Chapter 5 summaries of height changes due to Sunspot numbers,
diurnal and annual variations, but more work needs to be done on this
area.  I also need to take a better look at how to handle paths where
the height of the ionosphere changes drastically, the "tilted" 
ionosphere problem as discussed in Davies page 182 with  respect to 
trans-equatorial paths.

Starting abt 1992:
Ok, all those formulas are nice, but predictions still not real good
when compared to contest data.  Now I develop method to 'score' the
comparison of the prediction with an actual contest log.  This takes
the actual contacts from a weekend and calculates percentages of the
actual paths worked that were at, above, or below, the predicted MOF
for that path.  I play with that for a while and determine that there
are too many constants to test each one manually... especially since 
after changing one I have to recompile the program.  Now I develope 
the constant 'diddle' process.  This lets the program change each
constant up and down a given percentage, recalculate the prediction, 
rescore against the reference log data, and record change in a log file.
A complete run takes 36-48hrs on a 386dx-24(with 387 coprocessor), the
first few runs show the really weak areas in the formulas, some minor
adjustments raise predictions from 30-40% 'at' value to 50-70%.  I
err on the optimistic side so that only 10-15% of the time we work
paths that are above the predicted MOF, many of these are near the 
edges of the predictions, or could be accounted for by skew paths.

1994:
Offer program on Internet contest reflector in return for contest logs.
Most logs received just confirm that prediction is reasonable.  
Analysis of low flux logs from WPX and ARRL tests shows some weaknesses
in the Transequatorial paths and maybe an over optimistic prediction
for 10 and 15m on high lattitude paths.  But after several attempts to
'diddle' constants again only minor (2-5%) improvements can be made.
The TE path algorithm needs changing, I note that the constants had been
changed in the last analysis to eliminate this correction... a good sign
that it was wrong then also.  Maybe in the high flux situations the TE
correction was masked by other factors, but in low flux it is more 
important.
A windows version is under development, the equations are all the same,
just the presentation is different.

1995
Windows version working, lots of changes.  Calculations common between
Dos and Windows versions but interface all different.  

1996
Add better analysis features to windows version.  Add A/K geomagnetic
activity indices to calculations.  Do lots of adjustments to constants.
Add report for comparison of lots of logs at once, plot results with
MathCAD to find weak points.